kayg/infraternal
Watch
1
0
Fork
You've already forked infraternal
 0
Code Issues 356 Pull requests Projects Releases Packages Wiki Activity Actions

Relay mode (TURN in the binary) and the WebRTC direct path, side by side (D128, D131, D132, D185) #103

New issue
Open
opened 2026-10-08 13:09:03 +00:00 by kayg · 233 comments
kayg commented 2026-10-08 13:09:03 +00:00
Owner
Copy link

Build relay mode (infraternal relay, the server relaying by default, the 'Run a relay here' checkbox backend) and the WebRTC direct path side by side, so both can be benchmarked at the India-Germany latency. The client measures each relay's round trip, picks the lowest, re-checks on network changes, switches mid-session without lost or duplicated bytes, and lets the user override. Direct sessions fail closed after a ~60 s lease if the agent loses the server (D128). Benchmarks: keystroke p50/p99 and throughput per path, switch time, relay CPU and RSS per session at the D170 scale.

Build relay mode (`infraternal relay`, the server relaying by default, the 'Run a relay here' checkbox backend) and the WebRTC direct path side by side, so both can be benchmarked at the India-Germany latency. The client measures each relay's round trip, picks the lowest, re-checks on network changes, switches mid-session without lost or duplicated bytes, and lets the user override. Direct sessions fail closed after a ~60 s lease if the agent loses the server (D128). Benchmarks: keystroke p50/p99 and throughput per path, switch time, relay CPU and RSS per session at the D170 scale.
kayg commented 2026-10-08 18:37:54 +00:00
Author
Owner
Copy link

Launched relay-direct (Codex GPT 6.1 Sol Medium, web search on; Sol because it is transport + trust/security integration). Branch relay-direct from remote-agents 1f6e39b. Owner priority tonight: network first.

Launched relay-direct (Codex GPT 6.1 Sol Medium, web search on; Sol because it is transport + trust/security integration). Branch relay-direct from remote-agents 1f6e39b. Owner priority tonight: network first.
kayg commented 2026-10-08 18:39:46 +00:00
Author
Owner
Copy link

Started relay-direct as sole writer at base/head 1f6e39b8cb. Read #103 and #78 history, parent #5, local/main decisions, process and wire/remote contracts. Initial merge remote-agents: already up to date, no conflicts. Work order: authenticated relay and measured route selection/resume; candidate WebRTC adapters behind a trait; switching. No foundational stack choice, push/deploy or perf VM. Tests/benchmarks use infraternal-dev CPUWeight10000 and unique browser port4803. Existing remote lease is not shipping enforcement per #78; relay/direct must enforce its own bounded server authority.

Started relay-direct as sole writer at base/head 1f6e39b8cb11c3af39b0173e6b2c69f059fa9dc9. Read #103 and #78 history, parent #5, local/main decisions, process and wire/remote contracts. Initial merge remote-agents: already up to date, no conflicts. Work order: authenticated relay and measured route selection/resume; candidate WebRTC adapters behind a trait; switching. No foundational stack choice, push/deploy or perf VM. Tests/benchmarks use infraternal-dev CPUWeight10000 and unique browser port4803. Existing remote lease is not shipping enforcement per #78; relay/direct must enforce its own bounded server authority.
kayg commented 2026-10-08 18:41:33 +00:00
Author
Owner
Copy link

Additive relay/direct contract proposal before implementation [D64/D128/D131/D132/D170/D185]: preserve wire/shim layouts and durable server writer ids; route grants bind session, user/device writer, enrolled host, permitted relay/path, server generation and <=60s monotonic authority. Connection loss cannot renew a grant; audit-required sessions admit server relay only. TURN allocations/permissions require server-issued short-lived authority and bounded admission; enrollment tokens never become TURN credentials. No unrestricted proxy or browser-selected destination. Client RTT probes are nonce-correlated, bounded, measured on each complete terminal route (not only browser->relay); stale replies after network changes are discarded. Persist only opaque route preference, scoped to server+host; unavailable override falls back without erasing it. Switching keeps shim cursors/input replay and remaining credit. Direct engine remains behind an adapter pending owner stack selection. Sources checked: https://www.rfc-editor.org/rfc/rfc8656.html (TURN allocation/permission authentication), https://github.com/algesten/str0m (sans-I/O, no internal tasks), https://github.com/webrtc-rs/webrtc and https://github.com/webrtc-rs/rtc (current async wrapper/sans-I/O core). These are candidates, not measured recommendations. Initial modules will expose policy/lease/adapter boundaries without enabling unauthenticated network listeners.

Additive relay/direct contract proposal before implementation [D64/D128/D131/D132/D170/D185]: preserve wire/shim layouts and durable server writer ids; route grants bind session, user/device writer, enrolled host, permitted relay/path, server generation and <=60s monotonic authority. Connection loss cannot renew a grant; audit-required sessions admit server relay only. TURN allocations/permissions require server-issued short-lived authority and bounded admission; enrollment tokens never become TURN credentials. No unrestricted proxy or browser-selected destination. Client RTT probes are nonce-correlated, bounded, measured on each complete terminal route (not only browser->relay); stale replies after network changes are discarded. Persist only opaque route preference, scoped to server+host; unavailable override falls back without erasing it. Switching keeps shim cursors/input replay and remaining credit. Direct engine remains behind an adapter pending owner stack selection. Sources checked: https://www.rfc-editor.org/rfc/rfc8656.html (TURN allocation/permission authentication), https://github.com/algesten/str0m (sans-I/O, no internal tasks), https://github.com/webrtc-rs/webrtc and https://github.com/webrtc-rs/rtc (current async wrapper/sans-I/O core). These are candidates, not measured recommendations. Initial modules will expose policy/lease/adapter boundaries without enabling unauthenticated network listeners.
kayg commented 2026-10-08 18:52:29 +00:00
Author
Owner
Copy link

Route-selection preview: packages/protocol/src/routes.ts (not yet installed) bounds 32 routes, 96 pending nonce-correlated probes and five RTT samples/path; eight policy tests pass (147 assertions). Samples use a single local monotonic clock and a network generation, default to server relay, preserve slower explicit overrides through failure, and audit policy excludes direct/extra relays. Cache the median when a reply arrives so selecting performs no per-path sorting. This is library/policy evidence only: browser wiring, actual route RTT/PTY benchmarks and switching remain unimplemented. Candidate metadata verified with explicit versions: str0m0.24.1 supports existing OpenSSL; webrtc0.21.0 wraps its sans-I/O rtc core with selectable ring/AWS-LC; turn0.17.2. Isolated candidate build is a spike, not a production dependency. Sources: https://docs.rs/str0m/0.24.1/str0m/ and https://github.com/webrtc-rs/webrtc. Initial fast gate cold-build still running; infraternal-dev load has reached55.55/37.28/28.51 on8cores. No threshold changes.

Route-selection preview: packages/protocol/src/routes.ts (not yet installed) bounds 32 routes, 96 pending nonce-correlated probes and five RTT samples/path; eight policy tests pass (147 assertions). Samples use a single local monotonic clock and a network generation, default to server relay, preserve slower explicit overrides through failure, and audit policy excludes direct/extra relays. Cache the median when a reply arrives so selecting performs no per-path sorting. This is library/policy evidence only: browser wiring, actual route RTT/PTY benchmarks and switching remain unimplemented. Candidate metadata verified with explicit versions: str0m0.24.1 supports existing OpenSSL; webrtc0.21.0 wraps its sans-I/O rtc core with selectable ring/AWS-LC; turn0.17.2. Isolated candidate build is a spike, not a production dependency. Sources: https://docs.rs/str0m/0.24.1/str0m/ and https://github.com/webrtc-rs/webrtc. Initial fast gate cold-build still running; infraternal-dev load has reached55.55/37.28/28.51 on8cores. No threshold changes.
kayg commented 2026-10-08 18:56:51 +00:00
Author
Owner
Copy link

Baseline fast failed the existing initial_attach_requires_explicit_credit_before_replay fixture before any relay source was installed: scripts/verify-agent-attach-credit.py:17 creates target/ac without creating its parent (FileNotFoundError); external CARGO_TARGET_DIR leaves no workspace target directory. Added a worktree-local target symlink to this job's HOME-cache target, preserving all source/assertions and keeping build/state off /tmp. Will repeat the exact gate. This failure is a harness setup issue, not a transport timing conclusion. No remote source/threshold changes.

Baseline fast failed the existing initial_attach_requires_explicit_credit_before_replay fixture before any relay source was installed: scripts/verify-agent-attach-credit.py:17 creates target/ac<PID> without creating its parent (FileNotFoundError); external CARGO_TARGET_DIR leaves no workspace target directory. Added a worktree-local target symlink to this job's HOME-cache target, preserving all source/assertions and keeping build/state off /tmp. Will repeat the exact gate. This failure is a harness setup issue, not a transport timing conclusion. No remote source/threshold changes.
kayg commented 2026-10-08 19:17:33 +00:00
Author
Owner
Copy link

Failing-first direct-wrapper preview proof: after two real Unix-socket Input frames coalesce, revoke lease authority after the first frame. Existing Receive returned the buffered second Input(seq5,"second") instead of PermissionDenied (0pass/1fail,0.00s; .scratch/relay-direct/direct-buffered-red.log). This is a new direct-feature integration defect, not a claim that inherited trusted sockets need leases. Proposed narrow shared seam: Channel::check_authority defaults to a no-op for existing channels; Receive rechecks before extracting any frame, including buffered suffixes; the direct wrapper supplies its current server lease. No copied parser or per-keystroke SQLite lookup. Actual source remains staged until the contract gate is green. New wrapper also refuses renewal across account/device/path/host/session/writer/generation changes, drains admitted writes on revocation, and retires cancelled partial writes. Peer fingerprint and server-grant verification remain the signaling adapter's responsibility; no listener or direct admission enabled.

Failing-first direct-wrapper preview proof: after two real Unix-socket Input frames coalesce, revoke lease authority after the first frame. Existing Receive returned the buffered second Input(seq5,"second") instead of PermissionDenied (0pass/1fail,0.00s; .scratch/relay-direct/direct-buffered-red.log). This is a new direct-feature integration defect, not a claim that inherited trusted sockets need leases. Proposed narrow shared seam: Channel::check_authority defaults to a no-op for existing channels; Receive rechecks before extracting any frame, including buffered suffixes; the direct wrapper supplies its current server lease. No copied parser or per-keystroke SQLite lookup. Actual source remains staged until the contract gate is green. New wrapper also refuses renewal across account/device/path/host/session/writer/generation changes, drains admitted writes on revocation, and retires cancelled partial writes. Peer fingerprint and server-grant verification remain the signaling adapter's responsibility; no listener or direct admission enabled.
kayg commented 2026-10-08 19:18:31 +00:00
Author
Owner
Copy link

Completed additive contract step e43590b5ce: contracts/relay.md specifies trust, bounded leases, whole-path probes and byte/credit continuity before code. No engine/listener/admission enabled. Initial setup failure and unchanged passing fixture are recorded above. Exact CPUWeight10000/port4803/RUST_TEST_THREADS1 scripts/verify fast exit0; raw .scratch/relay-direct/contract-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19182544 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376)
7 pass
0 fail
34390 expect() calls
Ran 7 tests across 2 files. [5.60s]

Completed additive contract step e43590b5ce760c7a5d5df01117e2a9aa76d87c10: contracts/relay.md specifies trust, bounded leases, whole-path probes and byte/credit continuity before code. No engine/listener/admission enabled. Initial setup failure and unchanged passing fixture are recorded above. Exact CPUWeight10000/port4803/RUST_TEST_THREADS1 scripts/verify fast exit0; raw .scratch/relay-direct/contract-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19182544 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376) 7 pass 0 fail 34390 expect() calls Ran 7 tests across 2 files. [5.60s]
kayg commented 2026-10-08 19:32:00 +00:00
Author
Owner
Copy link

Completed route policy source step 1f257bc2f0: packages/protocol/src/routes.ts. Bounded correlated local-clock probes (32 routes/96 outstanding, five cached samples), audit precedence, failed-override recovery, server/host-scoped storage. Preview eight tests/147 assertions; production probe/attach wiring remains pending. No dependency/engine chosen. scripts/verify fast exit0, raw .scratch/relay-direct/routes-source-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19182544 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376)
7 pass
0 fail
34390 expect() calls
Ran 7 tests across 2 files. [4.33s]

Completed route policy source step 1f257bc2f076a2b913eac39059a89630dacca98d: packages/protocol/src/routes.ts. Bounded correlated local-clock probes (32 routes/96 outstanding, five cached samples), audit precedence, failed-override recovery, server/host-scoped storage. Preview eight tests/147 assertions; production probe/attach wiring remains pending. No dependency/engine chosen. scripts/verify fast exit0, raw .scratch/relay-direct/routes-source-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19182544 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376) 7 pass 0 fail 34390 expect() calls Ran 7 tests across 2 files. [4.33s]
kayg commented 2026-10-08 19:35:12 +00:00
Author
Owner
Copy link

Route test slice full gate first attempt failed in unchanged crates/infraternal/tests/vt_credit.rs:22: credit probe failed: t1-gate-approximate-1 (0 pass/1 fail, 11.66s). No Rust source diff; machine load 41.26/39.50/33.77. Raw .scratch/relay-direct/routes-tests-fast.log. This job also ran candidate diagnostics concurrently; pausing all its diagnostics and rerunning the complete gate unchanged, without threshold/test edits. No commit made for this slice yet.

Route test slice full gate first attempt failed in unchanged crates/infraternal/tests/vt_credit.rs:22: credit probe failed: t1-gate-approximate-1 (0 pass/1 fail, 11.66s). No Rust source diff; machine load 41.26/39.50/33.77. Raw .scratch/relay-direct/routes-tests-fast.log. This job also ran candidate diagnostics concurrently; pausing all its diagnostics and rerunning the complete gate unchanged, without threshold/test edits. No commit made for this slice yet.
kayg commented 2026-10-08 19:37:40 +00:00
Author
Owner
Copy link

Candidate diagnostics only, no engine chosen: infraternal-dev, CPUWeight=10000, Rust1.99 release, 8 visible cores, shared load42–48. str0m0.24.1 OpenSSL constructed unconnected RSS deltas: 200=12,816KiB, 1,000=64,288KiB, 10,000=643,088KiB. webrtc0.21.0 ring constructed unconnected WITH real loopback UDP sockets/runtime drivers: 200=58,348KiB, 1,000=291,960KiB, 10,000=2,940,144KiB. Actual retained cohorts, not multiplication, but differing socket/driver costs and NOT connected terminal-session memory. Init totals (str0m/webrtc us): 200=264339/101428; 1000=697992/522890; 10000=6872414/6872628. Native webrtc loopback reliable ordered data channel:31 one-byte echoes p50=67us/p99=108us; exact verified 50MiB echoed at2.484MiB/s; establishment2.030180s. Excludes browser/PTY/auth/TURN, no network-delay profile yet. str0m pair fixture incorrectly assumed echo writes always accept under256KiB backpressure; fixed spike to bounded echo queue and outstanding-frame window before rerun. Raw .scratch/relay-direct/candidate-metrics.log, spike sources .scratch/relay-direct/candidates (never production dependencies). Isolated candidate graph cargo-deny licenses/sources/advisories all OK under existing deny.toml, raw candidate-deny.log. Neither production binary size nor browser interoperability is proven. Whole route acceptance remains pending.

Candidate diagnostics only, no engine chosen: infraternal-dev, CPUWeight=10000, Rust1.99 release, 8 visible cores, shared load42–48. str0m0.24.1 OpenSSL constructed unconnected RSS deltas: 200=12,816KiB, 1,000=64,288KiB, 10,000=643,088KiB. webrtc0.21.0 ring constructed unconnected WITH real loopback UDP sockets/runtime drivers: 200=58,348KiB, 1,000=291,960KiB, 10,000=2,940,144KiB. Actual retained cohorts, not multiplication, but differing socket/driver costs and NOT connected terminal-session memory. Init totals (str0m/webrtc us): 200=264339/101428; 1000=697992/522890; 10000=6872414/6872628. Native webrtc loopback reliable ordered data channel:31 one-byte echoes p50=67us/p99=108us; exact verified 50MiB echoed at2.484MiB/s; establishment2.030180s. Excludes browser/PTY/auth/TURN, no network-delay profile yet. str0m pair fixture incorrectly assumed echo writes always accept under256KiB backpressure; fixed spike to bounded echo queue and outstanding-frame window before rerun. Raw .scratch/relay-direct/candidate-metrics.log, spike sources .scratch/relay-direct/candidates (never production dependencies). Isolated candidate graph cargo-deny licenses/sources/advisories all OK under existing deny.toml, raw candidate-deny.log. Neither production binary size nor browser interoperability is proven. Whole route acceptance remains pending.
kayg commented 2026-10-08 19:44:43 +00:00
Author
Owner
Copy link

Unchanged full gate retry also failed VT credit fixture (approximate grant16384) at scripts/verify-vt-credit.py:42: producer marker absent within1.5s. First failure was initial marker absent within5s. No Rust source diff; this job's measured diagnostics were paused during retry. Shared load then33.25/37.73/34.46, now falling to16.82/29.11/31.71. Logs routes-tests-retry-fast.log and vt-credit-retry-stderr.log. A third complete gate is running at reduced load; no budget/fixture edits. Route tests remain uncommitted until green. str0m bounded delayed datagram diagnostic verified50MiB exactly atRTT300:31 echo p50=302306us/p99=331871us, throughput0.776MiB/s, encrypted datagram payload115054006bytes for104857600echo application bytes. RTT0:22us/41us,16.219MiB/s,115047112datagram payload bytes. Both cap outstanding16×16KiB and echo backlog256KiB. Delay diagnostic excludes real socket scheduler/browser/PTY/TURN/auth; flood jitter0, echo jitter±30ms. Raw str0m-direct-metrics.log/str0m-delayed-metrics.log.

Unchanged full gate retry also failed VT credit fixture (approximate grant16384) at scripts/verify-vt-credit.py:42: producer marker absent within1.5s. First failure was initial marker absent within5s. No Rust source diff; this job's measured diagnostics were paused during retry. Shared load then33.25/37.73/34.46, now falling to16.82/29.11/31.71. Logs routes-tests-retry-fast.log and vt-credit-retry-stderr.log. A third complete gate is running at reduced load; no budget/fixture edits. Route tests remain uncommitted until green. str0m bounded delayed datagram diagnostic verified50MiB exactly atRTT300:31 echo p50=302306us/p99=331871us, throughput0.776MiB/s, encrypted datagram payload115054006bytes for104857600echo application bytes. RTT0:22us/41us,16.219MiB/s,115047112datagram payload bytes. Both cap outstanding16×16KiB and echo backlog256KiB. Delay diagnostic excludes real socket scheduler/browser/PTY/TURN/auth; flood jitter0, echo jitter±30ms. Raw str0m-direct-metrics.log/str0m-delayed-metrics.log.
kayg commented 2026-10-08 19:52:48 +00:00
Author
Owner
Copy link

Completed step f8034d46c8: packages/protocol/src/routes.test.ts: eight new policy tests/147 assertions; total protocol suite15 tests/34537 assertions. Third unchanged full gate green following the two recorded existing VT credit timing failures. No deadlines/budgets changed. No live path or terminal switch enabled. scripts/verify fast exit0; raw .scratch/relay-direct/routes-tests-third-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19182544 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.97s]

Completed step f8034d46c8f40c2e01842e599a7f923ed78e2fd1: packages/protocol/src/routes.test.ts: eight new policy tests/147 assertions; total protocol suite15 tests/34537 assertions. Third unchanged full gate green following the two recorded existing VT credit timing failures. No deadlines/budgets changed. No live path or terminal switch enabled. scripts/verify fast exit0; raw .scratch/relay-direct/routes-tests-third-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19182544 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.97s]
kayg commented 2026-10-08 20:03:47 +00:00
Author
Owner
Copy link

Real Chromium149.0.7827.55 interoperability completed for both candidates, negotiated reliable ordered binary channel, actual native loopback UDP sockets,16 outstanding16KiB frames/256KiB cap. Each verified50MiB with exact sequence/content. str0m0.24.1/OpenSSL:31 one-byte echoes p50=0.300ms/p99=4.800ms,2.669585MiB/s, answer-to-open23.8ms, load33.04/24.38/26.94. webrtc0.21.0/ring:0.600ms/7.200ms,3.168889MiB/s,answer-to-open409.1ms,load28.70/23.04/26.57. Scope CPUWeight10000/port4803, no own builds ran during these measurements. Signaling passed only through private inherited pipes; no SDP/grant/credential logged or persisted. Excludes PTY, authenticated server signaling, TURN and network delay;31-sample p99 is max, shared-load runs are diagnostic, not acceptance or a winner. Standalone native browser-fixture binaries: str0m8280792bytes/webrtc4528040bytes, dynamic Linux builds, NOT production static-binary additions. Sources .scratch/relay-direct/browser-candidates.mjs and candidates/src/bin/*-browser.rs; raw *-browser-metrics.log. No candidate dependency adopted in the production workspace. Engine selection remains owner-owned; public generic direct adapter and tests continue independently. Preparing isolated TURN0.17.2 server allocation/CPU/RSS measurement at200 allocations; production TURN trust/permission/lifetime bounds must not be inferred from a stock server fixture.

Real Chromium149.0.7827.55 interoperability completed for both candidates, negotiated reliable ordered binary channel, actual native loopback UDP sockets,16 outstanding16KiB frames/256KiB cap. Each verified50MiB with exact sequence/content. str0m0.24.1/OpenSSL:31 one-byte echoes p50=0.300ms/p99=4.800ms,2.669585MiB/s, answer-to-open23.8ms, load33.04/24.38/26.94. webrtc0.21.0/ring:0.600ms/7.200ms,3.168889MiB/s,answer-to-open409.1ms,load28.70/23.04/26.57. Scope CPUWeight10000/port4803, no own builds ran during these measurements. Signaling passed only through private inherited pipes; no SDP/grant/credential logged or persisted. Excludes PTY, authenticated server signaling, TURN and network delay;31-sample p99 is max, shared-load runs are diagnostic, not acceptance or a winner. Standalone native browser-fixture binaries: str0m8280792bytes/webrtc4528040bytes, dynamic Linux builds, NOT production static-binary additions. Sources .scratch/relay-direct/browser-candidates.mjs and candidates/src/bin/*-browser.rs; raw *-browser-metrics.log. No candidate dependency adopted in the production workspace. Engine selection remains owner-owned; public generic direct adapter and tests continue independently. Preparing isolated TURN0.17.2 server allocation/CPU/RSS measurement at200 allocations; production TURN trust/permission/lifetime bounds must not be inferred from a stock server fixture.
kayg commented 2026-10-08 20:22:00 +00:00
Author
Owner
Copy link

Hook gate first attempt failed existing scripts/verify-agent.py:34 idle RSS<15360KiB; no numerical reading emitted before assert. Both static builds PASS (x86 size19182352bytes, ARM16118176). Measured identical workspace debug builds, only channel/transport hook differs, CPUWeight10000,31 fresh daemon startups each (RTT0): HEAD without hook min14972/median15284/max15464KiB,5/31 exceed; hook min14908/median15292/max15512KiB,7/31 exceed. Overlapping ranges and8KiB median difference do not establish a hook regression; existing baseline itself violates the budget intermittently. Raw baseline-idle-rss.log/hook-idle-rss-31.log/hook-rss-summary.json. No thresholds or unrelated source changed; retrying full gate. Isolated real TURN0.17.2 server (separate process) retained200 allocations: baseline3628/retained5188KiB, delta1560KiB=7.8KiB/allocation; allocation cohort startup99,988us. Idle10s server CPU0 ticks at100Hz (not proof of zero wakeups). One active plaintext loopback allocation among199 idle,31 echoes p50=48us/p99=97us,exact52,428,000bytes echoed11.466MiB/s,server floodCPU213ticks=2.13CPU-seconds. CPU0.0426s/MiB;200 sessions at1MiB/s would require8.52cores as an ESTIMATE, not concurrent-flood proof. Prototype permits one ephemeral in-memory fixture principal over loopback; excludes WebRTC/PTY/production authorization/delay/transport-byte accounting. No credential in files/logs/argv; raw turn-scale-metrics.log. TURN candidate still needs bounded nonce/permission/allocation authority and ≤60s control-loss retirement before adoption. No production TURN/WebRTC dependency chosen.

Hook gate first attempt failed existing scripts/verify-agent.py:34 idle RSS<15360KiB; no numerical reading emitted before assert. Both static builds PASS (x86 size19182352bytes, ARM16118176). Measured identical workspace debug builds, only channel/transport hook differs, CPUWeight10000,31 fresh daemon startups each (RTT0): HEAD without hook min14972/median15284/max15464KiB,5/31 exceed; hook min14908/median15292/max15512KiB,7/31 exceed. Overlapping ranges and8KiB median difference do not establish a hook regression; existing baseline itself violates the budget intermittently. Raw baseline-idle-rss.log/hook-idle-rss-31.log/hook-rss-summary.json. No thresholds or unrelated source changed; retrying full gate. Isolated real TURN0.17.2 server (separate process) retained200 allocations: baseline3628/retained5188KiB, delta1560KiB=7.8KiB/allocation; allocation cohort startup99,988us. Idle10s server CPU0 ticks at100Hz (not proof of zero wakeups). One active plaintext loopback allocation among199 idle,31 echoes p50=48us/p99=97us,exact52,428,000bytes echoed11.466MiB/s,server floodCPU213ticks=2.13CPU-seconds. CPU0.0426s/MiB;200 sessions at1MiB/s would require8.52cores as an ESTIMATE, not concurrent-flood proof. Prototype permits one ephemeral in-memory fixture principal over loopback; excludes WebRTC/PTY/production authorization/delay/transport-byte accounting. No credential in files/logs/argv; raw turn-scale-metrics.log. TURN candidate still needs bounded nonce/permission/allocation authority and ≤60s control-loss retirement before adoption. No production TURN/WebRTC dependency chosen.
kayg commented 2026-10-08 20:26:11 +00:00
Author
Owner
Copy link

Candidate measurement correction: current str0m0.24.1 primary docs require every mutation followed by poll_output drain to Timeout, including refused writes. Reviewing spike drivers found initial browser offer/channel setup and bulk echo writes violated that rule. Earlier str0m browser/datagram measurements are superseded; fixing fixtures and rerunning before the owner comparison. No production stack adopted. Source https://docs.rs/str0m/0.24.1/str0m/#the-single-mutation-invariant. Generic direct wrapper RELEASE real Unix-socket diagnostic (1000 exact16-byte echoes,31 warmup): baseline p50/p99=7422/36469ns,guarded9196/35991ns,inline wrapper152bytes×200=30400bytes excluding engine/watch allocations; load15.57/20.96/25.33. Raw direct-wrapper-metrics.log. Core hook full gate retry passed daemon RSS/native tests and is rebuilding static artifacts after controlled baseline comparison.

Candidate measurement correction: current str0m0.24.1 primary docs require every mutation followed by poll_output drain to Timeout, including refused writes. Reviewing spike drivers found initial browser offer/channel setup and bulk echo writes violated that rule. Earlier str0m browser/datagram measurements are superseded; fixing fixtures and rerunning before the owner comparison. No production stack adopted. Source https://docs.rs/str0m/0.24.1/str0m/#the-single-mutation-invariant. Generic direct wrapper RELEASE real Unix-socket diagnostic (1000 exact16-byte echoes,31 warmup): baseline p50/p99=7422/36469ns,guarded9196/35991ns,inline wrapper152bytes×200=30400bytes excluding engine/watch allocations; load15.57/20.96/25.33. Raw direct-wrapper-metrics.log. Core hook full gate retry passed daemon RSS/native tests and is rebuilding static artifacts after controlled baseline comparison.
kayg commented 2026-10-08 20:34:05 +00:00
Author
Owner
Copy link

Hook full-gate retry exit1: daemon RSS passed, but unchanged agent_review::repeated_tiny_credit_pauses_preserve_shell_progress failed scripts/verify-agent-snapshot.py:48 (snapshot count/cursor invariant after1-byte credit),5 passed/1 failed,63.20s. Cause not established; no fixture/budget edit. Raw authority-hook-retry-fast.log. Static x86/ARM bothPASS, sizes19182352/16118176bytes. Hook remains uncommitted. Earlier progress statement concerned individual passing tests, not the full native gate; this summary supersedes it. Running corrected str0m fixture measurements serially before third unchanged full gate.

Hook full-gate retry exit1: daemon RSS passed, but unchanged agent_review::repeated_tiny_credit_pauses_preserve_shell_progress failed scripts/verify-agent-snapshot.py:48 (snapshot count/cursor invariant after1-byte credit),5 passed/1 failed,63.20s. Cause not established; no fixture/budget edit. Raw authority-hook-retry-fast.log. Static x86/ARM bothPASS, sizes19182352/16118176bytes. Hook remains uncommitted. Earlier progress statement concerned individual passing tests, not the full native gate; this summary supersedes it. Running corrected str0m fixture measurements serially before third unchanged full gate.
kayg commented 2026-10-08 20:37:31 +00:00
Author
Owner
Copy link

Owner foundation decision remains needed before production TURN/WebRTC dependencies or listeners are enabled [PROCESS3/D185]. Measured candidates, no selection: str0m0.24.1/OpenSSL exposes a sansIO mutable engine; adapter must drive/drain it and supply TURN/socket/clock handling (https://docs.rs/str0m/0.24.1/str0m/). It reuses the already-present OpenSSL family. webrtc0.21.0/ring supplies Tokio socket/driver orchestration around rtc0.21; compatible with existing Tokio/ring but requires explicit send backpressure (256KiB in fixtures), adding driver/socket state. Both Chromium149 interoperated and verified50MiB. CORRECTED str0m results supersede earlier invalid-drain fixture: Chromium loopback31 echoes p50/p990.400/0.800ms,7.747734MiB/s,answer-to-open8.6ms,load12.28/17.30/22.47; webrtc prior comparable loopback0.600/7.200ms,3.168889MiB/s,409.1ms,load28.70/23.04/26.57. Unequal load/preparation means these are diagnostics, not a winner. str0m delayed-encrypted-datagram profile300ms±30ms echo p50/p99=300336/330390us,50MiB exact,0.776MiB/s,flood jitter0,load6.28/14.12/20.89; native socket scheduler/browser/PTY/TURN excluded. RTT0=23/48us,15.911MiB/s. Datagrams115054006bytes for104857600echo application bytes atRTT300 (UDP/IP headers excluded). Real retained unconnected cohorts200/1000/10000 were reported above; str0m64KiB/peer vs webrtc292KiB/peer including its UDP/drivers. At200 those deltas12.5/57.0MiB, not connected terminal costs. Both isolated graphs passed licenses/advisories/sources before TURN was added; rechecking full graph. TURN0.17.2 separate-server200 allocation numbers reported above; its public server config does not alone implement scoped participant permissions/control-loss expiry or nonce-resource bounds. These are required adapter work, not permission to expose a stock relay (https://github.com/webrtc-rs/webrtc/tree/master/turn). Owner must record the adopted WebRTC and TURN foundation; job continues generic lease/route modules and gated tests meanwhile. Remaining acceptance: production joins/toggle/default relay, trusted signaling/fingerprint/admission, route probing/byte-offset switch/credit fence, real PTY+browser per-path RTT300/jitter, connected memory/concurrent CPU, TURN/TCP/TLS/ICE/loss/fuzz/overhead/static delta. Raw corrected logs: str0m-{browser,direct,delayed}-corrected-metrics.log. Spike sources remain .scratch/relay-direct/candidates, never merged; no other hosts used.

Owner foundation decision remains needed before production TURN/WebRTC dependencies or listeners are enabled [PROCESS3/D185]. Measured candidates, no selection: str0m0.24.1/OpenSSL exposes a sansIO mutable engine; adapter must drive/drain it and supply TURN/socket/clock handling (https://docs.rs/str0m/0.24.1/str0m/). It reuses the already-present OpenSSL family. webrtc0.21.0/ring supplies Tokio socket/driver orchestration around rtc0.21; compatible with existing Tokio/ring but requires explicit send backpressure (256KiB in fixtures), adding driver/socket state. Both Chromium149 interoperated and verified50MiB. CORRECTED str0m results supersede earlier invalid-drain fixture: Chromium loopback31 echoes p50/p990.400/0.800ms,7.747734MiB/s,answer-to-open8.6ms,load12.28/17.30/22.47; webrtc prior comparable loopback0.600/7.200ms,3.168889MiB/s,409.1ms,load28.70/23.04/26.57. Unequal load/preparation means these are diagnostics, not a winner. str0m delayed-encrypted-datagram profile300ms±30ms echo p50/p99=300336/330390us,50MiB exact,0.776MiB/s,flood jitter0,load6.28/14.12/20.89; native socket scheduler/browser/PTY/TURN excluded. RTT0=23/48us,15.911MiB/s. Datagrams115054006bytes for104857600echo application bytes atRTT300 (UDP/IP headers excluded). Real retained unconnected cohorts200/1000/10000 were reported above; str0m~64KiB/peer vs webrtc~292KiB/peer including its UDP/drivers. At200 those deltas12.5/57.0MiB, not connected terminal costs. Both isolated graphs passed licenses/advisories/sources before TURN was added; rechecking full graph. TURN0.17.2 separate-server200 allocation numbers reported above; its public server config does not alone implement scoped participant permissions/control-loss expiry or nonce-resource bounds. These are required adapter work, not permission to expose a stock relay (https://github.com/webrtc-rs/webrtc/tree/master/turn). Owner must record the adopted WebRTC and TURN foundation; job continues generic lease/route modules and gated tests meanwhile. Remaining acceptance: production joins/toggle/default relay, trusted signaling/fingerprint/admission, route probing/byte-offset switch/credit fence, real PTY+browser per-path RTT300/jitter, connected memory/concurrent CPU, TURN/TCP/TLS/ICE/loss/fuzz/overhead/static delta. Raw corrected logs: str0m-{browser,direct,delayed}-corrected-metrics.log. Spike sources remain .scratch/relay-direct/candidates, never merged; no other hosts used.
kayg commented 2026-10-08 20:45:06 +00:00
Author
Owner
Copy link

Completed step ead2c04f4c: Shared channel/transport authority seam (14 added lines/one doc correction). Existing borrowed codec and inherited no-op retained; prototype failing-first buffered Input now refuses revocation. Third unchanged complete gate green after recorded RSS/snapshot failures, no budget changes. Both static sizes within existing allowance. scripts/verify fast exit0; raw .scratch/relay-direct/authority-hook-third-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19182352 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.59s]

Completed step ead2c04f4c43c6c689ad3dc939d55612e008c021: Shared channel/transport authority seam (14 added lines/one doc correction). Existing borrowed codec and inherited no-op retained; prototype failing-first buffered Input now refuses revocation. Third unchanged complete gate green after recorded RSS/snapshot failures, no budget changes. Both static sizes within existing allowance. scripts/verify fast exit0; raw .scratch/relay-direct/authority-hook-third-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19182352 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16118176 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.59s]
kayg commented 2026-10-08 20:52:32 +00:00
Author
Owner
Copy link

Generic direct preview failing-first: valid encoded2MiB Output with peer reading64KiB every80ms is retired by the arbitrary1s active-write limit (TimedOut,FAIL1.01s). Raw direct-slow-frame-red.log. Proposed/implemented preview correction reuses the existing5s partial-frame budget for writes; expiry/revocation cancellation remains immediate and partial-write cancellation still retires both directions. This permits bounded progressing repaint traffic without changing any repository gate or selecting an engine. Public-API real-socket suite/strict clippy running before installation. Source direct.rs; test direct-integration.rs. Merged remote-agents5f7b701 into relay-direct at692263b (no conflicts), full integration gate still running. Read all new #78 comments, including its independently observed intermittent15MiB debugRSS assertion and pending shared-TCP credit fix; no uncommitted upstream files copied.

Generic direct preview failing-first: valid encoded2MiB Output with peer reading64KiB every80ms is retired by the arbitrary1s active-write limit (TimedOut,FAIL1.01s). Raw direct-slow-frame-red.log. Proposed/implemented preview correction reuses the existing5s partial-frame budget for writes; expiry/revocation cancellation remains immediate and partial-write cancellation still retires both directions. This permits bounded progressing repaint traffic without changing any repository gate or selecting an engine. Public-API real-socket suite/strict clippy running before installation. Source direct.rs; test direct-integration.rs. Merged remote-agents5f7b701 into relay-direct at692263b (no conflicts), full integration gate still running. Read all new #78 comments, including its independently observed intermittent15MiB debugRSS assertion and pending shared-TCP credit fix; no uncommitted upstream files copied.
kayg commented 2026-10-08 21:03:00 +00:00
Author
Owner
Copy link

Integration692263b full fast exit1: unchanged VT credit case t1-gate-faithful-1 failed at crates/infraternal/tests/vt_credit.rs:22 (0pass/1fail,19.56s). Raw upstream-credit-merge-fast.log. No conflicting files or threshold edits. Other Rust rows and both static buildsPASS (x8619184400/ARM16120224bytes). Repeating unchanged complete gate before installing direct source. Upstream now2f6ed99 adds an isolation/input-overload regression; read new#78 comments and will merge at next step boundary. Direct public-API preview9pass/0fail/1ignored2.72s, strictClippyPASS after moving test64KiB read scratch off async stack; slow legal-frame failing-first/correction recorded above. No foundation selected.

Integration692263b full fast exit1: unchanged VT credit case t1-gate-faithful-1 failed at crates/infraternal/tests/vt_credit.rs:22 (0pass/1fail,19.56s). Raw upstream-credit-merge-fast.log. No conflicting files or threshold edits. Other Rust rows and both static buildsPASS (x8619184400/ARM16120224bytes). Repeating unchanged complete gate before installing direct source. Upstream now2f6ed99 adds an isolation/input-overload regression; read new#78 comments and will merge at next step boundary. Direct public-API preview9pass/0fail/1ignored2.72s, strictClippyPASS after moving test64KiB read scratch off async stack; slow legal-frame failing-first/correction recorded above. No foundation selected.
kayg commented 2026-10-08 21:12:30 +00:00
Author
Owner
Copy link

Completed step 692263b4c4: Integration of remote-agents5f7b701 at692263b, no conflicts, no upstream edits. Native credit backpressure and its real QUIC/Unix test inherited. Full unchanged retry green after recorded faithful VT producer failure; budgets unchanged. scripts/verify fast exit0; raw .scratch/relay-direct/upstream-credit-merge-retry-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19184400 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120224 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.36s]

Completed step 692263b4c4bbac3d402a02bcf59c6abd2172d712: Integration of remote-agents5f7b701 at692263b, no conflicts, no upstream edits. Native credit backpressure and its real QUIC/Unix test inherited. Full unchanged retry green after recorded faithful VT producer failure; budgets unchanged. scripts/verify fast exit0; raw .scratch/relay-direct/upstream-credit-merge-retry-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184400 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120224 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.36s]
kayg commented 2026-10-08 21:21:37 +00:00
Author
Owner
Copy link

Integration 1ab9f0319d: unchanged full scripts/verify fast retry failed existing repeated_tiny_credit_pauses_preserve_shell_progress / scripts/verify-agent-snapshot.py assertion. Evidence: .scratch/relay-direct/upstream-isolation-merge-retry-fast.log. fmt, native fmt, clippy, cargo-deny and both static builds PASS; full gate remains red. Retrying unchanged; no fixture or budget changes.

Integration 1ab9f0319d170bc402d26bc509492729fda1be9e: unchanged full scripts/verify fast retry failed existing repeated_tiny_credit_pauses_preserve_shell_progress / scripts/verify-agent-snapshot.py assertion. Evidence: .scratch/relay-direct/upstream-isolation-merge-retry-fast.log. fmt, native fmt, clippy, cargo-deny and both static builds PASS; full gate remains red. Retrying unchanged; no fixture or budget changes.
kayg commented 2026-10-08 21:24:59 +00:00
Author
Owner
Copy link

Blocked checkpoint on relay-direct, HEAD 1ab9f0319d; remote-agents 2f6ed99 merged without conflicts. Three complete unchanged integration fast runs failed: first VT producer initial marker; second and third repeated_tiny_credit_pauses_preserve_shell_progress at scripts/verify-agent-snapshot.py:48 (snapshot count/output cursor). No budget/fixture changes. Latest raw log .scratch/relay-direct/upstream-isolation-merge-third-fast.log, exit1. Independent source work remains outside production until a green integration checkpoint. #78 also records existing gate instability; no unrelated upstream fix was attempted.

Prepared restart points: .scratch/relay-direct/direct-source.rs, .scratch/relay-direct/direct-integration.rs, .scratch/relay-direct/evidence-draft.md. Generic direct source has not landed; its public-API real-socket preview has9 passing tests/1 ignored cost diagnostic and strict clippy green. Failing-first tests cover cached Input after revocation and a legal slow2MiB Output; generic authority hook and bounded RTT/preference policy already committed. Source/tests must be separate atomic gated commits after merging remote-agents at each step.

Foundation owner record still needed before adopting TURN/WebRTC. Corrected Chromium diagnostics: str0m0.24.1 echo0.400/0.800ms and50MiB7.747734MiB/s; webrtc0.21.0 echo0.600/7.200ms and50MiB3.168889MiB/s. Unequal shared load; no winner inferred. These exclude production TURN/PTY/auth/delay. Real200-allocation TURN fixture delta1560KiB, one active among199 idle11.466MiB/s and2.13CPU seconds for52,428,000bytes. Full engine/session CPU/RSS atD170 remains absent. All detailed evidence and exclusions are in preceding comments and the scratch evidence draft.

Task NOT done: production relay CLI/join/server default/host toggle, trusted admission/signaling, live inventory/probes/preferences, offset handover/mid-session fallback and full latency-proxy/PTy/browser acceptance remain pending. Next worker needs green integration and the owner foundation record; never close #103. Scratch and job-local build caches are retained for this unfinished checkpoint; no candidate processes, push or deployment.

Latest scripts/verify fast output verbatim:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | FAIL (101) |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

Blocked checkpoint on relay-direct, HEAD 1ab9f0319d170bc402d26bc509492729fda1be9e; remote-agents 2f6ed99 merged without conflicts. Three complete unchanged integration fast runs failed: first VT producer initial marker; second and third repeated_tiny_credit_pauses_preserve_shell_progress at scripts/verify-agent-snapshot.py:48 (snapshot count/output cursor). No budget/fixture changes. Latest raw log .scratch/relay-direct/upstream-isolation-merge-third-fast.log, exit1. Independent source work remains outside production until a green integration checkpoint. #78 also records existing gate instability; no unrelated upstream fix was attempted. Prepared restart points: .scratch/relay-direct/direct-source.rs, .scratch/relay-direct/direct-integration.rs, .scratch/relay-direct/evidence-draft.md. Generic direct source has not landed; its public-API real-socket preview has9 passing tests/1 ignored cost diagnostic and strict clippy green. Failing-first tests cover cached Input after revocation and a legal slow2MiB Output; generic authority hook and bounded RTT/preference policy already committed. Source/tests must be separate atomic gated commits after merging remote-agents at each step. Foundation owner record still needed before adopting TURN/WebRTC. Corrected Chromium diagnostics: str0m0.24.1 echo0.400/0.800ms and50MiB7.747734MiB/s; webrtc0.21.0 echo0.600/7.200ms and50MiB3.168889MiB/s. Unequal shared load; no winner inferred. These exclude production TURN/PTY/auth/delay. Real200-allocation TURN fixture delta1560KiB, one active among199 idle11.466MiB/s and2.13CPU seconds for52,428,000bytes. Full engine/session CPU/RSS atD170 remains absent. All detailed evidence and exclusions are in preceding comments and the scratch evidence draft. Task NOT done: production relay CLI/join/server default/host toggle, trusted admission/signaling, live inventory/probes/preferences, offset handover/mid-session fallback and full latency-proxy/PTy/browser acceptance remain pending. Next worker needs green integration and the owner foundation record; never close #103. Scratch and job-local build caches are retained for this unfinished checkpoint; no candidate processes, push or deployment. Latest scripts/verify fast output verbatim: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | FAIL (101) | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+
kayg commented 2026-10-08 21:36:44 +00:00
Author
Owner
Copy link

Run 2 started as sole writer on relay-direct at base/head 1ab9f0319d. Read all 23 #103 comments and all parent #5 comments, CLAUDE.md, local/main decisions, architecture/process, and restart checkpoint. Build order: QUIC relay CLI/join/default/host setting; production inventory/probes/preferences; byte-offset switching; feature-disabled direct adapters/equal-load comparison. No foundational engine choice, push/deploy or other hosts. Per orchestrator, only named VT credit/snapshot timing failures may be rerun alone and accepted if passing; every other gate failure blocks. Benchmarks local CPUWeight=10000, browser port4803, scratch/build targets under HOME cache.

Run 2 started as sole writer on relay-direct at base/head 1ab9f0319d170bc402d26bc509492729fda1be9e. Read all 23 #103 comments and all parent #5 comments, CLAUDE.md, local/main decisions, architecture/process, and restart checkpoint. Build order: QUIC relay CLI/join/default/host setting; production inventory/probes/preferences; byte-offset switching; feature-disabled direct adapters/equal-load comparison. No foundational engine choice, push/deploy or other hosts. Per orchestrator, only named VT credit/snapshot timing failures may be rerun alone and accepted if passing; every other gate failure blocks. Benchmarks local CPUWeight=10000, browser port4803, scratch/build targets under HOME cache.
kayg commented 2026-10-08 21:36:49 +00:00
Author
Owner
Copy link

Independent verification started on review-relay-direct, detached target 1ab9f0319d; merge base with dev f63307e8f6. Read all #103 comments and contracts. The specified relay-direct/result.md is absent; the final #103 checkpoint explicitly reports unfinished production relay/direct support and a red integration gate. Reviewing landed policy/authority seam and inherited transport against the contract; no implementation fixes, engine adoption, push, or deployment.

Independent verification started on review-relay-direct, detached target 1ab9f0319d170bc402d26bc509492729fda1be9e; merge base with dev f63307e8f668cd349246a5848ed2c006d0bc3c92. Read all #103 comments and contracts. The specified relay-direct/result.md is absent; the final #103 checkpoint explicitly reports unfinished production relay/direct support and a red integration gate. Reviewing landed policy/authority seam and inherited transport against the contract; no implementation fixes, engine adoption, push, or deployment.
kayg commented 2026-10-08 21:37:48 +00:00
Author
Owner
Copy link

Additive QUIC relay proposal [D131/D132/D176]: keep the server as the end-to-end host identity verifier. Extra relay enrollment is a separate host-bound role and one-use hashed invitation, not an ordinary host join token. Persist public relay coordinates/role generation only; no private material. Relay endpoint certificates use a server-selected relay..internal SAN and serverAuth, never CSR-requested extensions. QUIC tunnel allocation is authorized by current server control, fixed to the enrolled server endpoint, bounded to 200 sessions and <=60s monotonic grants; it cannot choose arbitrary destinations. Relay termination must preserve the inner server/host authentication and durable writer/byte cursors. First logical step is the additive role/invitation schema; native/browser path wiring follows independently of the pending WebRTC engine decision.

Additive QUIC relay proposal [D131/D132/D176]: keep the server as the end-to-end host identity verifier. Extra relay enrollment is a separate host-bound role and one-use hashed invitation, not an ordinary host join token. Persist public relay coordinates/role generation only; no private material. Relay endpoint certificates use a server-selected relay.<host UUID>.internal SAN and serverAuth, never CSR-requested extensions. QUIC tunnel allocation is authorized by current server control, fixed to the enrolled server endpoint, bounded to 200 sessions and <=60s monotonic grants; it cannot choose arbitrary destinations. Relay termination must preserve the inner server/host authentication and durable writer/byte cursors. First logical step is the additive role/invitation schema; native/browser path wiring follows independently of the pending WebRTC engine decision.
kayg commented 2026-10-08 21:42:49 +00:00
Author
Owner
Copy link

Run 2 unchanged baseline native gate failed ONLY crates/infraternal/tests/vt_credit.rs:22, t1-gate-approximate-1 (0 pass/1 fail, 7.85s). All six agent_review tests including repeated_tiny_credit_pauses_preserve_shell_progress passed. Load 42.89/38.90/32.90 on 8 cores. Raw HOME-cache relay-direct/run2-baseline-fast.log. Per orchestrator run-2 exception, rerunning the exact compiled VT-credit test alone under CPUWeight=10000; no source/deadline/budget changes. Full gate is still completing static builds; no new commit yet. Relay schema preview applies all 14 migrations and rejects six invalid role updates.

Run 2 unchanged baseline native gate failed ONLY crates/infraternal/tests/vt_credit.rs:22, t1-gate-approximate-1 (0 pass/1 fail, 7.85s). All six agent_review tests including repeated_tiny_credit_pauses_preserve_shell_progress passed. Load 42.89/38.90/32.90 on 8 cores. Raw HOME-cache relay-direct/run2-baseline-fast.log. Per orchestrator run-2 exception, rerunning the exact compiled VT-credit test alone under CPUWeight=10000; no source/deadline/budget changes. Full gate is still completing static builds; no new commit yet. Relay schema preview applies all 14 migrations and rejects six invalid role updates.
kayg commented 2026-10-08 21:43:55 +00:00
Author
Owner
Copy link

Isolated unchanged VT-credit rerun failed t1-gate-approximate-1 again (0 pass/1 fail, 11.41s): scripts/verify-vt-credit.py:30, detached initial producer must finish. Raw HOME-cache relay-direct/run2-baseline-vt-alone.log and target/review-vt3-fixes/t1-gate-approximate-1.stderr.log. Load 42.57/39.33/33.32. The orchestrator exception requires a passing isolated rerun and is NOT satisfied yet. No new commit/listener, no fixture/deadline/budget change. Completing the full baseline static builds before another isolated check; #118 owns the unstable fixture. Relay role/store/signer/tunnel drafts remain uninstalled in .scratch/relay-direct.

Isolated unchanged VT-credit rerun failed t1-gate-approximate-1 again (0 pass/1 fail, 11.41s): scripts/verify-vt-credit.py:30, detached initial producer must finish. Raw HOME-cache relay-direct/run2-baseline-vt-alone.log and target/review-vt3-fixes/t1-gate-approximate-1.stderr.log. Load 42.57/39.33/33.32. The orchestrator exception requires a passing isolated rerun and is NOT satisfied yet. No new commit/listener, no fixture/deadline/budget change. Completing the full baseline static builds before another isolated check; #118 owns the unstable fixture. Relay role/store/signer/tunnel drafts remain uninstalled in .scratch/relay-direct.
kayg commented 2026-10-08 21:50:47 +00:00
Author
Owner
Copy link

Independent relay authority preview (not installed): five real SQLite tests PASS in 0.18s using the worktree native dependencies. Evidence HOME-cache relay-direct/relay-store-preview-tests.log; source .scratch/relay-direct/relay_store.rs and relay_store_tests.rs. Covers unlocked/live owner-only role creation; distinct host/relay tokens; replacement generation; exactly one concurrent redemption; failed certificate insertion rolling back token consumption; disable/removal after store restart; 31 extra-relay cap plus the default server route. No listener or engine enabled. Full baseline x86 static size 19,184,496 bytes, within 19,244,272 maximum; ARM build ongoing. The gate exception still awaits a passing unchanged isolated VT-credit rerun; second isolated attempt started as load fell to 26.03/35.70/34.45.

Independent relay authority preview (not installed): five real SQLite tests PASS in 0.18s using the worktree native dependencies. Evidence HOME-cache relay-direct/relay-store-preview-tests.log; source .scratch/relay-direct/relay_store.rs and relay_store_tests.rs. Covers unlocked/live owner-only role creation; distinct host/relay tokens; replacement generation; exactly one concurrent redemption; failed certificate insertion rolling back token consumption; disable/removal after store restart; 31 extra-relay cap plus the default server route. No listener or engine enabled. Full baseline x86 static size 19,184,496 bytes, within 19,244,272 maximum; ARM build ongoing. The gate exception still awaits a passing unchanged isolated VT-credit rerun; second isolated attempt started as load fell to 26.03/35.70/34.45.
kayg commented 2026-10-08 21:52:18 +00:00
Author
Owner
Copy link

QUIC relay transport preview: nine real-socket tests PASS, 5.87s, CPUWeight=10000, HOME-cache relay-direct/relay-transport-preview-tests.log. Four new tests: exact 1,048,576-byte forward/echo to fixed upstream; wrong peer/61s lease opens no upstream socket; control revocation closes an idle/active allocation; new role generation cannot renew an existing allocation. Five inherited TLS tests also pass after genericizing its IO for an inner authenticated tunnel. No production source/listener installed; browser/PTY/RTT300/at-scale evidence and server-control wiring remain absent. Source .scratch/relay-direct/{relay_forward,relay_forward_tests,relay_tunnel,remote_tls_tunnel}.rs. Third VT check failed faithful-1 (17.47s); it overlapped the 5.87s preview, so do not use it as isolated acceptance. First two isolated checks failed. Important gate-snapshot correction: the first fast run began before schema creation, so only its static builds embedded the draft migration. It is not a complete schema gate. A second full fast run now checks the unchanged schema-only tree; no source drafts will be installed while it runs.

QUIC relay transport preview: nine real-socket tests PASS, 5.87s, CPUWeight=10000, HOME-cache relay-direct/relay-transport-preview-tests.log. Four new tests: exact 1,048,576-byte forward/echo to fixed upstream; wrong peer/61s lease opens no upstream socket; control revocation closes an idle/active allocation; new role generation cannot renew an existing allocation. Five inherited TLS tests also pass after genericizing its IO for an inner authenticated tunnel. No production source/listener installed; browser/PTY/RTT300/at-scale evidence and server-control wiring remain absent. Source .scratch/relay-direct/{relay_forward,relay_forward_tests,relay_tunnel,remote_tls_tunnel}.rs. Third VT check failed faithful-1 (17.47s); it overlapped the 5.87s preview, so do not use it as isolated acceptance. First two isolated checks failed. Important gate-snapshot correction: the first fast run began before schema creation, so only its static builds embedded the draft migration. It is not a complete schema gate. A second full fast run now checks the unchanged schema-only tree; no source drafts will be installed while it runs.
kayg commented 2026-10-08 21:54:36 +00:00
Author
Owner
Copy link

Relay endpoint signer preview now compiles and passes existing host-CA tests plus a new real-OpenSSL purpose/SAN check; raw HOME-cache relay-direct/relay-ca-preview-tests.log. Endpoint leaf uses only serverAuth and server-chosen relay..internal/relay URN; SSL-server verification succeeds, SSL-client verification fails, hostile CSR names/CA privileges are ignored. Draft .scratch/relay-direct/ca-relay.rs only; no production signer installed. Self-review restart checklist at .scratch/relay-direct/run2-review-checkpoint.md records remaining server-control/renewal/enrollment/fuzz/browser/scale gaps and the need to measure or revise the QUIC-to-relay/inner-TLS-TCP tunnel before adoption. Schema-only full gate ongoing; head remains 1ab9f03.

Relay endpoint signer preview now compiles and passes existing host-CA tests plus a new real-OpenSSL purpose/SAN check; raw HOME-cache relay-direct/relay-ca-preview-tests.log. Endpoint leaf uses only serverAuth and server-chosen relay.<host UUID>.internal/relay URN; SSL-server verification succeeds, SSL-client verification fails, hostile CSR names/CA privileges are ignored. Draft .scratch/relay-direct/ca-relay.rs only; no production signer installed. Self-review restart checklist at .scratch/relay-direct/run2-review-checkpoint.md records remaining server-control/renewal/enrollment/fuzz/browser/scale gaps and the need to measure or revise the QUIC-to-relay/inner-TLS-TCP tunnel before adoption. Schema-only full gate ongoing; head remains 1ab9f03.
kayg commented 2026-10-08 21:56:55 +00:00
Author
Owner
Copy link

Independent review RD1 — P1 — #103 production acceptance is unfinished at target 1ab9f031. The real target CLI refuses relay --help: exit 2, unrecognized subcommand 'relay'. Failing assertion relay_operator_command_exposes_udp_port (crates/infraternal/tests/relay_review.rs), 0 passed/1 failed, 0.02s, committed immediately on review-relay-direct at f40f1711d9. This is a completion blocker against D131/D185, consistent with the author's explicit unfinished checkpoint; it is not evidence of an unauthenticated listener. Default TURN, host toggle, trusted direct signaling/admission and live route handover remain absent. Raw relay-cli-red.log is under HOME-cache relay-direct-review. No implementation fix or engine adoption.

Independent review RD1 — P1 — #103 production acceptance is unfinished at target 1ab9f031. The real target CLI refuses relay --help: exit 2, unrecognized subcommand 'relay'. Failing assertion relay_operator_command_exposes_udp_port (crates/infraternal/tests/relay_review.rs), 0 passed/1 failed, 0.02s, committed immediately on review-relay-direct at f40f1711d92272c9f2ff18627bd97d7e52020f24. This is a completion blocker against D131/D185, consistent with the author's explicit unfinished checkpoint; it is not evidence of an unauthenticated listener. Default TURN, host toggle, trusted direct signaling/admission and live route handover remain absent. Raw relay-cli-red.log is under HOME-cache relay-direct-review. No implementation fix or engine adoption.
kayg commented 2026-10-08 21:57:43 +00:00
Author
Owner
Copy link

Stable schema-only scripts/verify fast exit1: only failed test is vt_credit::unaffordable_repaints_park_until_sufficient_credit, t1-gate-approximate-1, 10.02s. All agent_review tests, including snapshot, passed. fmt/native fmt/clippy/cargo-deny/static rows PASS. Sizes x86 19,184,496; ARM 16,120,304 bytes, both within existing budgets. Raw HOME-cache relay-direct/run2-schema-fast.log. Final isolated exact compiled VT check started with no other own build/test running; exception remains conditional on its success. #118 latest controlled-load evidence is 1/10 approximate/1, 1/10 approximate/16384, 0/10 faithful cases under 16 CPU workers and load30.79–46.79. No new commit or deadline edits; head1ab9f03. Preparing 10 rounds each, strictly alternating str0m0.24.1 and webrtc0.21.0, same Chromium/payload16x16KiB window/50MiB exact echo, under one CPUWeight10000 scope; raw echo samples/load will be retained. Diagnostic comparison excludes PTY/auth/TURN/delay and makes no foundational selection.

Stable schema-only scripts/verify fast exit1: only failed test is vt_credit::unaffordable_repaints_park_until_sufficient_credit, t1-gate-approximate-1, 10.02s. All agent_review tests, including snapshot, passed. fmt/native fmt/clippy/cargo-deny/static rows PASS. Sizes x86 19,184,496; ARM 16,120,304 bytes, both within existing budgets. Raw HOME-cache relay-direct/run2-schema-fast.log. Final isolated exact compiled VT check started with no other own build/test running; exception remains conditional on its success. #118 latest controlled-load evidence is 1/10 approximate/1, 1/10 approximate/16384, 0/10 faithful cases under 16 CPU workers and load30.79–46.79. No new commit or deadline edits; head1ab9f03. Preparing 10 rounds each, strictly alternating str0m0.24.1 and webrtc0.21.0, same Chromium/payload16x16KiB window/50MiB exact echo, under one CPUWeight10000 scope; raw echo samples/load will be retained. Diagnostic comparison excludes PTY/auth/TURN/delay and makes no foundational selection.
kayg commented 2026-10-08 21:58:51 +00:00
Author
Owner
Copy link

Schema gate exception NOT satisfied: exact compiled VT-credit isolated check, with no other own build/test, failed t1-gate-approximate-1 again (9.45s, 0 pass/1 fail), detached initial producer scripts/verify-vt-credit.py:30. Raw HOME-cache relay-direct/run2-schema-vt-alone.log. Stable schema-only full fast raw run2-schema-fast.log failed only that fixture; all other Rust rows PASS. #118 owns its event-based readiness fix (controlled-load failures now recorded there). Need that fix or an unchanged passing isolated/full gate before any commit; no approval question or budget change. Head remains1ab9f03; untracked migration plus source previews retained. Alternating 10-round/engine comparison is now running serially under CPUWeight10000 with no own build/test; it can produce owner evidence independently of installing production source.

Schema gate exception NOT satisfied: exact compiled VT-credit isolated check, with no other own build/test, failed t1-gate-approximate-1 again (9.45s, 0 pass/1 fail), detached initial producer scripts/verify-vt-credit.py:30. Raw HOME-cache relay-direct/run2-schema-vt-alone.log. Stable schema-only full fast raw run2-schema-fast.log failed only that fixture; all other Rust rows PASS. #118 owns its event-based readiness fix (controlled-load failures now recorded there). Need that fix or an unchanged passing isolated/full gate before any commit; no approval question or budget change. Head remains1ab9f03; untracked migration plus source previews retained. Alternating 10-round/engine comparison is now running serially under CPUWeight10000 with no own build/test; it can produce owner evidence independently of installing production source.
kayg commented 2026-10-08 22:00:42 +00:00
Author
Owner
Copy link

Equal-window browser comparison attempt1 stopped in round2/webrtc on a 10s echo timeout. Three runs completed, each exact50MiB: r1 str0m p50/p99 .5/4.8ms throughput2.406994MiB/s; r1 webrtc .6/4.0ms 4.238042MiB/s; r2 str0m .3/7.9ms 3.323010MiB/s. Load38.87 rising to52.37/42.73/37.91 on8cores. Raw HOME-cache relay-direct/run2-equal-load.log. Not a complete10-round comparison; no engine selection. Added content-free timeout diagnostics (sample/channel state/native process exit) without changing 31 echoes,50MiB payload,256KiB window or10s echo deadline. Will restart a complete alternating window and retain the failure. Current polling/send usage checked against the pinned local0.21.0 source and primary docs https://docs.rs/webrtc/latest/webrtc/data_channel/ ; cause not established, no upstream/candidate code patched.

Equal-window browser comparison attempt1 stopped in round2/webrtc on a 10s echo timeout. Three runs completed, each exact50MiB: r1 str0m p50/p99 .5/4.8ms throughput2.406994MiB/s; r1 webrtc .6/4.0ms 4.238042MiB/s; r2 str0m .3/7.9ms 3.323010MiB/s. Load38.87 rising to52.37/42.73/37.91 on8cores. Raw HOME-cache relay-direct/run2-equal-load.log. Not a complete10-round comparison; no engine selection. Added content-free timeout diagnostics (sample/channel state/native process exit) without changing 31 echoes,50MiB payload,256KiB window or10s echo deadline. Will restart a complete alternating window and retain the failure. Current polling/send usage checked against the pinned local0.21.0 source and primary docs https://docs.rs/webrtc/latest/webrtc/data_channel/ ; cause not established, no upstream/candidate code patched.
kayg commented 2026-10-08 22:02:24 +00:00
Author
Owner
Copy link

Comparison attempt2 stopped in r1/webrtc: first echo (sample0) timed out at10s, browser peer connected/channel open/buffered0, native process still alive. Shared load63.25/49.70/41.09 on8cores. Raw HOME-cache relay-direct/run2-equal-load-second-failure.log (retained under that name); cause not established, no candidate patched/chosen. Restarting exactly10 alternating attempts per engine in one scope, now recording each failed attempt and continuing the schedule. Echo10s/flood120s deadlines and31 samples/50MiB/16x16KiB remain unchanged; partial timing samples retained on failure. Aggregate will state success/failure counts and distinguish successful-run distributions from censored >10s timeouts; no invented throughput for failed rounds.

Comparison attempt2 stopped in r1/webrtc: first echo (sample0) timed out at10s, browser peer connected/channel open/buffered0, native process still alive. Shared load63.25/49.70/41.09 on8cores. Raw HOME-cache relay-direct/run2-equal-load-second-failure.log (retained under that name); cause not established, no candidate patched/chosen. Restarting exactly10 alternating attempts per engine in one scope, now recording each failed attempt and continuing the schedule. Echo10s/flood120s deadlines and31 samples/50MiB/16x16KiB remain unchanged; partial timing samples retained on failure. Aggregate will state success/failure counts and distinguish successful-run distributions from censored >10s timeouts; no invented throughput for failed rounds.
kayg commented 2026-10-08 22:08:16 +00:00
Author
Owner
Copy link

Independent review RD2 — P2 — Retired transports expose cached Input. Two real coalesced frames; consume first, retire channel, raw read returns BrokenPipe, but Receive::next returns second Input(seq5, six bytes). Confirmed on real quinn/mTLS (0 pass/1 fail, 0.07s; 11e2dd1670) and existing real TCP/TLS fixture (0 pass/1 fail, 0.04s; 58097d86b8), both committed on review-relay-direct. Receive's new check_authority calls use NativeChannel/TcpTls's inherited no-op and miss their retirement watch. Locations: transport.rs:153, channel.rs:15, remote_native.rs:74, remote_tls.rs:53. This proves the retirement/cancellation invariant gap, not a bypass of the separate server AuthorityGuard or an implemented direct lease. Raw native-retired-red2.log/tls-retired-red.log under HOME-cache relay-direct-review. Target fast also reproduced existing snapshot count/cursor failure at verify-agent-snapshot.py:48; static verification continues.

Independent review RD2 — P2 — Retired transports expose cached Input. Two real coalesced frames; consume first, retire channel, raw read returns BrokenPipe, but Receive::next returns second Input(seq5, six bytes). Confirmed on real quinn/mTLS (0 pass/1 fail, 0.07s; 11e2dd1670db031481634981de91735311e28a19) and existing real TCP/TLS fixture (0 pass/1 fail, 0.04s; 58097d86b850f385d577aeddfbaf6b28775a59fc), both committed on review-relay-direct. Receive's new check_authority calls use NativeChannel/TcpTls's inherited no-op and miss their retirement watch. Locations: transport.rs:153, channel.rs:15, remote_native.rs:74, remote_tls.rs:53. This proves the retirement/cancellation invariant gap, not a bypass of the separate server AuthorityGuard or an implemented direct lease. Raw native-retired-red2.log/tls-retired-red.log under HOME-cache relay-direct-review. Target fast also reproduced existing snapshot count/cursor failure at verify-agent-snapshot.py:48; static verification continues.
kayg commented 2026-10-08 22:10:28 +00:00
Author
Owner
Copy link

Completed the requested alternating comparison diagnostic at unchanged head 1ab9f0319d: 10 rounds each in one CPUWeight10000 scope on infraternal-dev/8cores, Chromium149.0.7827.55, 16x16KiB window, exact50MiB/round. Final window 10/10pass each,310echoes/engine,524,288,000exactechoed bytes each. Pooled echo p50/p99 ms: str0m0.24.1=.45/7.90; webrtc0.21.0=.50/7.30. Median throughput MiB/s:2.381 (1.832–2.964) vs4.780 (3.487–7.667); establishment medians16.90/212.15ms. One-minute load64.49–75.34; no concurrent own build/test. Earlier aborted windows had webrtc10s echo timeouts (r2 unknown sample; r1/sample0 with connected/open channel/native alive); retained, unresolved, NOT erased by successful window. No engine chosen/patched. Excludes PTY/auth/TURN/RTT300/render/connected RSS at D170. Raw HOME-cache relay-direct/run2-equal-load{,-first-failure,-second-failure}.log; full report run2-engine-comparison.{json,md}, rounds CSV run2-engine-rounds.csv. Schema/source commits still blocked by failed isolated VT fixture; no production relay/direct shipping claim.

Completed the requested alternating comparison diagnostic at unchanged head 1ab9f0319d170bc402d26bc509492729fda1be9e: 10 rounds each in one CPUWeight10000 scope on infraternal-dev/8cores, Chromium149.0.7827.55, 16x16KiB window, exact50MiB/round. Final window 10/10pass each,310echoes/engine,524,288,000exactechoed bytes each. Pooled echo p50/p99 ms: str0m0.24.1=.45/7.90; webrtc0.21.0=.50/7.30. Median throughput MiB/s:2.381 (1.832–2.964) vs4.780 (3.487–7.667); establishment medians16.90/212.15ms. One-minute load64.49–75.34; no concurrent own build/test. Earlier aborted windows had webrtc10s echo timeouts (r2 unknown sample; r1/sample0 with connected/open channel/native alive); retained, unresolved, NOT erased by successful window. No engine chosen/patched. Excludes PTY/auth/TURN/RTT300/render/connected RSS at D170. Raw HOME-cache relay-direct/run2-equal-load{,-first-failure,-second-failure}.log; full report run2-engine-comparison.{json,md}, rounds CSV run2-engine-rounds.csv. Schema/source commits still blocked by failed isolated VT fixture; no production relay/direct shipping claim.
kayg commented 2026-10-08 22:11:28 +00:00
Author
Owner
Copy link

Run 2 blocked checkpoint; task NOT complete. Branch relay-direct, base/head 1ab9f0319d; no new commits, push/deploy, or enabled production listeners. The stable schema-only fast gate failed only vt_credit approximate-1; exact compiled isolated rerun also failed, so the orchestrator exception is not satisfied. Need #118's fixture fix or an unchanged passing isolated/full gate; no owner approval question or budget change.

Prepared role schema/store/signer/tunnel drafts:17 native-preview tests PASS total (5 SQLite,9 mTLS/QUIC,3 CA;1 diagnostic ignored). Not production acceptance. Only production-tree change is untracked crates/infraternal-server/migrations/0014_relay_roles.sql. All implementation drafts/reviewer gaps/restart instructions in .scratch/relay-direct/run2-checkpoint.md and run2-review-checkpoint.md. Server-control authorizer, CLI/join/toggle, live route/switching, feature-disabled engines, fuzz and real RTT300/D170 acceptance remain pending.

Completed10 alternating browser rounds/engine; numbers and two earlier unresolved webrtc timeouts recorded in preceding comment. Artifacts HOME-cache relay-direct/run2-engine-comparison.{json,md}, run2-engine-rounds.csv, run2-equal-load*.log. This measures candidates only; no engine decision. Retaining job scratch/build caches for this unfinished restart; removing temporary preview executables, no candidate/browser processes left by this job.

Stable scripts/verify fast output verbatim (exit1; HOME-cache relay-direct/run2-schema-fast.log):

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | FAIL (101) |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)

Exact compiled isolated VT-credit output verbatim (exit101; run2-schema-vt-alone.log):

running 1 test
test unaffordable_repaints_park_until_sufficient_credit ...
thread 'unaffordable_repaints_park_until_sufficient_credit' (302) panicked at crates/infraternal/tests/vt_credit.rs:22:13:
credit probe failed: t1-gate-approximate-1
note: run with RUST_BACKTRACE=1 environment variable to display a backtrace
FAILED

failures:

failures:
unaffordable_repaints_park_until_sufficient_credit

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 9.45s

Run 2 blocked checkpoint; task NOT complete. Branch relay-direct, base/head 1ab9f0319d170bc402d26bc509492729fda1be9e; no new commits, push/deploy, or enabled production listeners. The stable schema-only fast gate failed only vt_credit approximate-1; exact compiled isolated rerun also failed, so the orchestrator exception is not satisfied. Need #118's fixture fix or an unchanged passing isolated/full gate; no owner approval question or budget change. Prepared role schema/store/signer/tunnel drafts:17 native-preview tests PASS total (5 SQLite,9 mTLS/QUIC,3 CA;1 diagnostic ignored). Not production acceptance. Only production-tree change is untracked crates/infraternal-server/migrations/0014_relay_roles.sql. All implementation drafts/reviewer gaps/restart instructions in .scratch/relay-direct/run2-checkpoint.md and run2-review-checkpoint.md. Server-control authorizer, CLI/join/toggle, live route/switching, feature-disabled engines, fuzz and real RTT300/D170 acceptance remain pending. Completed10 alternating browser rounds/engine; numbers and two earlier unresolved webrtc timeouts recorded in preceding comment. Artifacts HOME-cache relay-direct/run2-engine-comparison.{json,md}, run2-engine-rounds.csv, run2-equal-load*.log. This measures candidates only; no engine decision. Retaining job scratch/build caches for this unfinished restart; removing temporary preview executables, no candidate/browser processes left by this job. Stable scripts/verify fast output verbatim (exit1; HOME-cache relay-direct/run2-schema-fast.log): +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | FAIL (101) | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) Exact compiled isolated VT-credit output verbatim (exit101; run2-schema-vt-alone.log): running 1 test test unaffordable_repaints_park_until_sufficient_credit ... thread 'unaffordable_repaints_park_until_sufficient_credit' (302) panicked at crates/infraternal/tests/vt_credit.rs:22:13: credit probe failed: t1-gate-approximate-1 note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace FAILED failures: failures: unaffordable_repaints_park_until_sufficient_credit test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 9.45s
kayg commented 2026-10-08 22:14:37 +00:00
Author
Owner
Copy link

Independent verification step 309758b: added renewal JSON-bound fuzz target using the existing workspace engine. CSR, writer binding and renewal each completed 100,000 native debug smoke executions, exit0/no crash; no coverage instrumentation or sanitizer, so not nightly fuzz acceptance. Signed CSR tests6/6, enrollment7/7 and remote-server27/27 active tests pass (7 diagnostics ignored). Strict workspace Clippy and fmt pass with all review assertions present. Review changes remain tests only; target fast reproduced snapshot count/cursor failure at verify-agent-snapshot.py:48, static rows still running. Machine infraternal-dev, CPUWeight10000; shared load rose above60 on8cores, no latency/scale certification.

Independent verification step 309758b: added renewal JSON-bound fuzz target using the existing workspace engine. CSR, writer binding and renewal each completed 100,000 native debug smoke executions, exit0/no crash; no coverage instrumentation or sanitizer, so not nightly fuzz acceptance. Signed CSR tests6/6, enrollment7/7 and remote-server27/27 active tests pass (7 diagnostics ignored). Strict workspace Clippy and fmt pass with all review assertions present. Review changes remain tests only; target fast reproduced snapshot count/cursor failure at verify-agent-snapshot.py:48, static rows still running. Machine infraternal-dev, CPUWeight10000; shared load rose above60 on8cores, no latency/scale certification.
kayg commented 2026-10-08 22:42:14 +00:00
Author
Owner
Copy link

Original target 1ab9f031 fast gate completed, exit 1. Test executables were compiled before review edits. Existing repeated_tiny_credit_pauses_preserve_shell_progress failed at scripts/verify-agent-snapshot.py:48 (snapshot count/output cursor); no implementation was modified. Both static executables pass (x86 19,188,752 bytes; ARM 16,120,224 bytes). Latency gate started locally with CPUWeight=10000, port 4893; native phase load 69.68/74.63/71.47 on 8 CPUs.

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | FAIL (101) |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

Original target 1ab9f031 fast gate completed, exit 1. Test executables were compiled before review edits. Existing repeated_tiny_credit_pauses_preserve_shell_progress failed at scripts/verify-agent-snapshot.py:48 (snapshot count/output cursor); no implementation was modified. Both static executables pass (x86 19,188,752 bytes; ARM 16,120,224 bytes). Latency gate started locally with CPUWeight=10000, port 4893; native phase load 69.68/74.63/71.47 on 8 CPUs. +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | FAIL (101) | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+
kayg commented 2026-10-08 22:46:10 +00:00
Author
Owner
Copy link

Run 3 started as sole writer on relay-direct at base/head 1ab9f0319d. Read all 41 #103 comments, parent #5 comments, repository rules/process and run-2 restart points. Continue atomic schema/module/test commits for the 17 passing previews first; named native instability will be compared with an own detached dev baseline in the same load window per the run-3 exception. Then QUIC relay production wiring, route choice, offset switching and feature-disabled direct candidates. No foundation selection, push/deploy, other hosts or credential inspection; CPUWeight10000 local benchmarks, port4803, HOME-cache scratch.

Run 3 started as sole writer on relay-direct at base/head 1ab9f0319d170bc402d26bc509492729fda1be9e. Read all 41 #103 comments, parent #5 comments, repository rules/process and run-2 restart points. Continue atomic schema/module/test commits for the 17 passing previews first; named native instability will be compared with an own detached dev baseline in the same load window per the run-3 exception. Then QUIC relay production wiring, route choice, offset switching and feature-disabled direct candidates. No foundation selection, push/deploy, other hosts or credential inspection; CPUWeight10000 local benchmarks, port4803, HOME-cache scratch.
kayg commented 2026-10-08 22:46:29 +00:00
Author
Owner
Copy link

Verification step at review head 309758b7: both retirement tests fail in 5/5 additional exact-test repetitions each (10/10 total). scripts/verify latency native phase is red: n=100/profile, PTY echo p99 430.18 ms baseline, 511.33 ms other-session flood, 3933.63 ms same-session flood; echo budget 334 ms, flood-growth budget 16.67 ms. Local scope CPUWeight=10000, 300ms RTT +/-30ms, port 4893. Native start load 69.68/74.63/71.47 on 8 CPUs; auth-browser phase load 82.15/77.37/73.07. These are budget failures under shared overload, not evidence attributing a regression to relay/direct; its production paths remain absent. Browser gate still running.

Verification step at review head 309758b7: both retirement tests fail in 5/5 additional exact-test repetitions each (10/10 total). scripts/verify latency native phase is red: n=100/profile, PTY echo p99 430.18 ms baseline, 511.33 ms other-session flood, 3933.63 ms same-session flood; echo budget 334 ms, flood-growth budget 16.67 ms. Local scope CPUWeight=10000, 300ms RTT +/-30ms, port 4893. Native start load 69.68/74.63/71.47 on 8 CPUs; auth-browser phase load 82.15/77.37/73.07. These are budget failures under shared overload, not evidence attributing a regression to relay/direct; its production paths remain absent. Browser gate still running.
kayg commented 2026-10-08 22:54:32 +00:00
Author
Owner
Copy link

Schema-only run-3 fast exit1: repeated_tiny_credit_pauses_preserve_shell_progress failed at verify-agent-snapshot.py:17, 90s shell-progress marker completed-0 absent (agent_review 5pass/1fail,126.44s). All other Rust gate rows PASS; static sizes x86 19,184,496 and ARM 16,120,304 bytes. Load during native tests 79.87–92.50 on8cores. Raw HOME-cache relay-direct/run3-schema-fast.log. Own detached dev baseline 1c3ae5f40b at HOME-cache relay-direct/baseline is compiling the exact same test; fixture SHA256 afdd8faf20d6ee9c39e537e67217a052b1c22ad1d707f48f1791689065d0a9d5 matches both trees. Will compare failure evidence per run-3 exception, then run native remainder and TS checks skipped by fast short-circuit. No commit or deadline/budget changes yet.

Schema-only run-3 fast exit1: repeated_tiny_credit_pauses_preserve_shell_progress failed at verify-agent-snapshot.py:17, 90s shell-progress marker completed-0 absent (agent_review 5pass/1fail,126.44s). All other Rust gate rows PASS; static sizes x86 19,184,496 and ARM 16,120,304 bytes. Load during native tests 79.87–92.50 on8cores. Raw HOME-cache relay-direct/run3-schema-fast.log. Own detached dev baseline 1c3ae5f40b578a6bc25b8ec83a9a13c8220dcd53 at HOME-cache relay-direct/baseline is compiling the exact same test; fixture SHA256 afdd8faf20d6ee9c39e537e67217a052b1c22ad1d707f48f1791689065d0a9d5 matches both trees. Will compare failure evidence per run-3 exception, then run native remainder and TS checks skipped by fast short-circuit. No commit or deadline/budget changes yet.
kayg commented 2026-10-08 22:59:58 +00:00
Author
Owner
Copy link

Run-3 exception reproduced on own detached dev 1c3ae5f4: same verify-agent-snapshot.py:17 90s completed-0 shell marker failure. Branch agent_review 5pass/1fail126.44s; exact baseline test0pass/1fail98.78s, no source/deadline changes. Baseline compiled from its checkout; native GNU debug cache reused only after exact snapshot/build-recipe SHA256 equality (ABI x86_64-linux-gnu, ReleaseSafe). Load82.98/79.80/77.55 on8cores at baseline completion. Raw HOME-cache relay-direct/run3-schema-{fast,baseline-snapshot-cached}.log. Native libraries and remaining integration targets still need gates before schema commit. Failure output verbatim, branch then baseline:

---- repeated_tiny_credit_pauses_preserve_shell_progress stdout ----

thread 'repeated_tiny_credit_pauses_preserve_shell_progress' (328208) panicked at crates/infraternal/tests/agent_review.rs:13:5:
scripts/verify-agent-snapshot.py:

Traceback (most recent call last):
File "/home/kayg/Developer/infraternal/.worktrees/relay-direct/scripts/verify-agent-snapshot.py", line 81, in
run(parser.parse_args().binary)
~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/kayg/Developer/infraternal/.worktrees/relay-direct/scripts/verify-agent-snapshot.py", line 37, in run
completed(marker)
~~~~~~~~~^^^^^^^^
File "/home/kayg/Developer/infraternal/.worktrees/relay-direct/scripts/verify-agent-snapshot.py", line 17, in completed
assert time.monotonic() < deadline, f'shell blocked: {marker}'
^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: shell blocked: /home/kayg/.cache/codex-scratch/infraternal/ts328209/completed-0

note: run with RUST_BACKTRACE=1 environment variable to display a backtrace

failures:
repeated_tiny_credit_pauses_preserve_shell_progress

test result: FAILED. 5 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 126.44s

error: test failed, to rerun pass -p infraternal --test agent_review

running 1 test
test repeated_tiny_credit_pauses_preserve_shell_progress ...
thread 'repeated_tiny_credit_pauses_preserve_shell_progress' (369618) panicked at crates/infraternal/tests/agent_review.rs:13:5:
scripts/verify-agent-snapshot.py:

Traceback (most recent call last):
File "/home/kayg/.cache/codex-scratch/relay-direct/baseline/scripts/verify-agent-snapshot.py", line 81, in
run(parser.parse_args().binary)
~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/kayg/.cache/codex-scratch/relay-direct/baseline/scripts/verify-agent-snapshot.py", line 37, in run
completed(marker)
~~~~~~~~~^^^^^^^^
File "/home/kayg/.cache/codex-scratch/relay-direct/baseline/scripts/verify-agent-snapshot.py", line 17, in completed
assert time.monotonic() < deadline, f'shell blocked: {marker}'
^^^^^^^^^^^^^^^^^^^^^^^^^^^
AssertionError: shell blocked: /home/kayg/.cache/codex-scratch/infraternal/ts369619/completed-0

note: run with RUST_BACKTRACE=1 environment variable to display a backtrace
FAILED

failures:

failures:
repeated_tiny_credit_pauses_preserve_shell_progress

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 5 filtered out; finished in 98.78s

error: test failed, to rerun pass -p infraternal --test agent_review

Run-3 exception reproduced on own detached dev 1c3ae5f4: same verify-agent-snapshot.py:17 90s completed-0 shell marker failure. Branch agent_review 5pass/1fail126.44s; exact baseline test0pass/1fail98.78s, no source/deadline changes. Baseline compiled from its checkout; native GNU debug cache reused only after exact snapshot/build-recipe SHA256 equality (ABI x86_64-linux-gnu, ReleaseSafe). Load82.98/79.80/77.55 on8cores at baseline completion. Raw HOME-cache relay-direct/run3-schema-{fast,baseline-snapshot-cached}.log. Native libraries and remaining integration targets still need gates before schema commit. Failure output verbatim, branch then baseline: ---- repeated_tiny_credit_pauses_preserve_shell_progress stdout ---- thread 'repeated_tiny_credit_pauses_preserve_shell_progress' (328208) panicked at crates/infraternal/tests/agent_review.rs:13:5: scripts/verify-agent-snapshot.py: Traceback (most recent call last): File "/home/kayg/Developer/infraternal/.worktrees/relay-direct/scripts/verify-agent-snapshot.py", line 81, in <module> run(parser.parse_args().binary) ~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/kayg/Developer/infraternal/.worktrees/relay-direct/scripts/verify-agent-snapshot.py", line 37, in run completed(marker) ~~~~~~~~~^^^^^^^^ File "/home/kayg/Developer/infraternal/.worktrees/relay-direct/scripts/verify-agent-snapshot.py", line 17, in completed assert time.monotonic() < deadline, f'shell blocked: {marker}' ^^^^^^^^^^^^^^^^^^^^^^^^^^^ AssertionError: shell blocked: /home/kayg/.cache/codex-scratch/infraternal/ts328209/completed-0 note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace failures: repeated_tiny_credit_pauses_preserve_shell_progress test result: FAILED. 5 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 126.44s error: test failed, to rerun pass `-p infraternal --test agent_review` running 1 test test repeated_tiny_credit_pauses_preserve_shell_progress ... thread 'repeated_tiny_credit_pauses_preserve_shell_progress' (369618) panicked at crates/infraternal/tests/agent_review.rs:13:5: scripts/verify-agent-snapshot.py: Traceback (most recent call last): File "/home/kayg/.cache/codex-scratch/relay-direct/baseline/scripts/verify-agent-snapshot.py", line 81, in <module> run(parser.parse_args().binary) ~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^ File "/home/kayg/.cache/codex-scratch/relay-direct/baseline/scripts/verify-agent-snapshot.py", line 37, in run completed(marker) ~~~~~~~~~^^^^^^^^ File "/home/kayg/.cache/codex-scratch/relay-direct/baseline/scripts/verify-agent-snapshot.py", line 17, in completed assert time.monotonic() < deadline, f'shell blocked: {marker}' ^^^^^^^^^^^^^^^^^^^^^^^^^^^ AssertionError: shell blocked: /home/kayg/.cache/codex-scratch/infraternal/ts369619/completed-0 note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace FAILED failures: failures: repeated_tiny_credit_pauses_preserve_shell_progress test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 5 filtered out; finished in 98.78s error: test failed, to rerun pass `-p infraternal --test agent_review`
kayg commented 2026-10-08 23:09:05 +00:00
Author
Owner
Copy link

Schema native remainder: libraries/docs PASS (real native/TLS/SQLite/auth/VT corpus); integration agent_sessions1/1, idle_canary1/1, shim7/7 PASS. vt_credit approximate-1 failed initial producer marker (12.30s). Additional non-excepted vt_reconnect failed its unchanged20s detached-emitter deadline at verify-vt-reconnect.py:43 (27.87s total), after retained replay and direct approximate repaint PASS. Load58.38/64.19/70.81 on8cores. Raw HOME-cache relay-direct/run3-schema-native-{libraries,integration-remainder}.log, target/review-vt2-fixes/reconnect.stderr.log. Cannot accept this failure under run-3 exception. Testing upstream434d4ff6 fixture CPU-weight inheritance in an isolated driver against the unchanged script/binary/deadline, without modifying production scheduling or tests. If it resolves the deadline, inherit that existing upstream fix as a separate gated integration step; no job-local timing fix.

Schema native remainder: libraries/docs PASS (real native/TLS/SQLite/auth/VT corpus); integration agent_sessions1/1, idle_canary1/1, shim7/7 PASS. vt_credit approximate-1 failed initial producer marker (12.30s). Additional non-excepted vt_reconnect failed its unchanged20s detached-emitter deadline at verify-vt-reconnect.py:43 (27.87s total), after retained replay and direct approximate repaint PASS. Load58.38/64.19/70.81 on8cores. Raw HOME-cache relay-direct/run3-schema-native-{libraries,integration-remainder}.log, target/review-vt2-fixes/reconnect.stderr.log. Cannot accept this failure under run-3 exception. Testing upstream434d4ff6 fixture CPU-weight inheritance in an isolated driver against the unchanged script/binary/deadline, without modifying production scheduling or tests. If it resolves the deadline, inherit that existing upstream fix as a separate gated integration step; no job-local timing fix.
kayg commented 2026-10-08 23:22:42 +00:00
Author
Owner
Copy link

Completed step 4ca298a814: Inherited remote-agents434d4ff6 fixture-only CPUWeight propagation to its own UUID-named detached shim scope; no production scheduling or deadline change. Failing-first unchanged reconnect20s emitter deadline, isolated weighted fixture passes both wrapped/less cases. Full gate snapshot6/6 agent_review, VT credit1/1(14.12s), reconnect1/1(16.60s) PASS. Load33.25–52.07 on8cores during this run; raw weighted-reconnect proof run3-weighted-reconnect.log. Relay schema held in HOME-cache until this prerequisite integration is committed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-priority-integration-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19184400 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120224 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.78s]

Completed step 4ca298a814afefc6ab58747adb7701ff6093f891: Inherited remote-agents434d4ff6 fixture-only CPUWeight propagation to its own UUID-named detached shim scope; no production scheduling or deadline change. Failing-first unchanged reconnect20s emitter deadline, isolated weighted fixture passes both wrapped/less cases. Full gate snapshot6/6 agent_review, VT credit1/1(14.12s), reconnect1/1(16.60s) PASS. Load33.25–52.07 on8cores during this run; raw weighted-reconnect proof run3-weighted-reconnect.log. Relay schema held in HOME-cache until this prerequisite integration is committed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-priority-integration-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184400 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120224 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.78s]
kayg commented 2026-10-08 23:27:59 +00:00
Author
Owner
Copy link

Schema restored on head4ca298a8: full fast exit1, ONLY executed test failure is existing verify-agent.py:34 idle RSS<15360KiB, agent_daemon0pass/1fail3.15s. The fixture emitted no reading; no native gate exception claimed for RSS. Other Rust rows and static builds PASS. Raw HOME-cache relay-direct/run3-schema-weighted-fast.log. Load5.88/14.33/34.96 at completion. This supersedes the chat progress statement that its native tests passed, which misread the phase. No budget/source changes; rerunning the exact unchanged full gate. The same debug RSS threshold variation was measured on both baseline and hook in run1 comments (baseline median15284/max15464KiB), but that is diagnostic history, not acceptance of this run.

Schema restored on head4ca298a8: full fast exit1, ONLY executed test failure is existing verify-agent.py:34 idle RSS<15360KiB, agent_daemon0pass/1fail3.15s. The fixture emitted no reading; no native gate exception claimed for RSS. Other Rust rows and static builds PASS. Raw HOME-cache relay-direct/run3-schema-weighted-fast.log. Load5.88/14.33/34.96 at completion. This supersedes the chat progress statement that its native tests passed, which misread the phase. No budget/source changes; rerunning the exact unchanged full gate. The same debug RSS threshold variation was measured on both baseline and hook in run1 comments (baseline median15284/max15464KiB), but that is diagnostic history, not acceptance of this run.
kayg commented 2026-10-08 23:31:15 +00:00
Author
Owner
Copy link

Fresh-start diagnostic on infraternal-dev under CPUWeight10000, no own build/test during measurement, load7.06–7.76 on8cores. schema: n=10, min/median/max KiB=15124/15334.0/15428, >=15360 KiB=4/10; dev: n=10, min/median/max KiB=11628/11720.0/11916, >=15360 KiB=0/10. Dev artifact compiled from own detached1c3ae5f4 and hash9902a8a3950eb856da286014fc5b36664c3dd0cee9e1a1ecebb972b16ae50fd3. Dev lacks inherited remote transport, so this does NOT isolate relay-schema overhead or permit ignoring the RSS gate. Raw HOME-cache relay-direct/run3-{schema,dev}-idle-rss.log. Current schema debug starts straddle the strict budget; two full gates failed. No budget/source change, running a third unchanged complete gate; every non-excepted row must pass.

Fresh-start diagnostic on infraternal-dev under CPUWeight10000, no own build/test during measurement, load7.06–7.76 on8cores. schema: n=10, min/median/max KiB=15124/15334.0/15428, >=15360 KiB=4/10; dev: n=10, min/median/max KiB=11628/11720.0/11916, >=15360 KiB=0/10. Dev artifact compiled from own detached1c3ae5f4 and hash9902a8a3950eb856da286014fc5b36664c3dd0cee9e1a1ecebb972b16ae50fd3. Dev lacks inherited remote transport, so this does NOT isolate relay-schema overhead or permit ignoring the RSS gate. Raw HOME-cache relay-direct/run3-{schema,dev}-idle-rss.log. Current schema debug starts straddle the strict budget; two full gates failed. No budget/source change, running a third unchanged complete gate; every non-excepted row must pass.
kayg commented 2026-10-08 23:40:17 +00:00
Author
Owner
Copy link

Completed step 0d2ddc4cb7: Additive migration0014 separates relay role/generation, one-use hashed invitations and public relay endpoint certificates from ordinary host identity. No listener/engine enabled. Existing real SQLite tests apply the complete migration chain; next slices install the five drafted role-store tests and signer/tunnel modules separately. Static sizes and exact gate output follow. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-schema-weighted-third-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.91s]

Completed step 0d2ddc4cb770cb0b33b6fa2cb049a6369cc8187e: Additive migration0014 separates relay role/generation, one-use hashed invitations and public relay endpoint certificates from ordinary host identity. No listener/engine enabled. Existing real SQLite tests apply the complete migration chain; next slices install the five drafted role-store tests and signer/tunnel modules separately. Static sizes and exact gate output follow. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-schema-weighted-third-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.91s]
kayg commented 2026-10-08 23:52:05 +00:00
Author
Owner
Copy link

Completed step 97d07ee6c9: Persist explicit owner-authorized relay roles with canonical one-use capabilities. Serialize inventory limits and authorization inside writes; consume the invitation and publish public certificate authority atomically. Reject separate authorization reads or early token consumption because revocation and storage failure can race. This module opens no listener. D131, D132 and D176 require distinct relay trust. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-store-source-fast-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.47s]

Completed step 97d07ee6c90fde20514594355a0f08b37d415730: Persist explicit owner-authorized relay roles with canonical one-use capabilities. Serialize inventory limits and authorization inside writes; consume the invitation and publish public certificate authority atomically. Reject separate authorization reads or early token consumption because revocation and storage failure can race. This module opens no listener. D131, D132 and D176 require distinct relay trust. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-store-source-fast-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.47s]
kayg commented 2026-10-08 23:53:02 +00:00
Author
Owner
Copy link

Production-routing finding: RemoteHosts is keyed only by host UUID (crates/infraternal-server/src/remote_runtime.rs:35), and run_actor owns one AgentConnection; replacement calls connection_lost for every attachment, then reattach_all (embedded.rs:523,541,561). Mounting a second relay route as an ordinary host reconnect would therefore replace all attachments on that host. Add explicit route identity and scoped handover before production route selection; the pure selector and forwarding previews do not establish this integration. Continuing the individually gated preview slices; no listener is enabled.

Production-routing finding: RemoteHosts is keyed only by host UUID (crates/infraternal-server/src/remote_runtime.rs:35), and run_actor owns one AgentConnection; replacement calls connection_lost for every attachment, then reattach_all (embedded.rs:523,541,561). Mounting a second relay route as an ordinary host reconnect would therefore replace all attachments on that host. Add explicit route identity and scoped handover before production route selection; the pure selector and forwarding previews do not establish this integration. Continuing the individually gated preview slices; no listener is enabled.
kayg commented 2026-10-09 00:03:18 +00:00
Author
Owner
Copy link

Additive relay admission proposal for approved step (a), D131/D132/D176: one private relay invitation signs two purpose-separated leaves from one local CSR proof: client-only enrolled control identity and server-only relay endpoint. Consume the token and publish both public records in one SQLite transaction; insert failure rolls back both identities and consumption. Reuse existing host signing, mTLS, Call/Result/Event framing and owner-session authorization; no new foundation or production WebRTC engine. Relay port/destination come from server role/public endpoint metadata; a CLI port mismatch or bind failure is refused. Signing completion rechecks role authority and conservatively advances the initial Unix-second timestamp by elapsed monotonic time, so queued signing cannot freeze invitation validity. Scratch prototypes only so far; listener remains unmounted.

Additive relay admission proposal for approved step (a), D131/D132/D176: one private relay invitation signs two purpose-separated leaves from one local CSR proof: client-only enrolled control identity and server-only relay endpoint. Consume the token and publish both public records in one SQLite transaction; insert failure rolls back both identities and consumption. Reuse existing host signing, mTLS, Call/Result/Event framing and owner-session authorization; no new foundation or production WebRTC engine. Relay port/destination come from server role/public endpoint metadata; a CLI port mismatch or bind failure is refused. Signing completion rechecks role authority and conservatively advances the initial Unix-second timestamp by elapsed monotonic time, so queued signing cannot freeze invitation validity. Scratch prototypes only so far; listener remains unmounted.
kayg commented 2026-10-09 00:06:24 +00:00
Author
Owner
Copy link

Completed step b5a9975ad7: Add five real SQLite tests for owner-only role creation, host/relay token separation, generation replacement, concurrent single redemption, rollback, durable retirement and the 31-extra-relay inventory bound. Reject mocked transaction outcomes: the actual migration and writer behavior must enforce these invariants. D131, D132 and D170 define trust and scale. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-store-tests-fast-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.58s]

Completed step b5a9975ad7be80cab25e8543127ba42b6ec2e353: Add five real SQLite tests for owner-only role creation, host/relay token separation, generation replacement, concurrent single redemption, rollback, durable retirement and the 31-extra-relay inventory bound. Reject mocked transaction outcomes: the actual migration and writer behavior must enforce these invariants. D131, D132 and D170 define trust and scale. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-store-tests-fast-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184496 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.58s]
kayg commented 2026-10-09 00:17:26 +00:00
Author
Owner
Copy link

Completed step 5ff1faa3b4: Keep host client identities separate from relay server identities. Derive relay SANs from the server-authorized host UUID and retain seven-day validity and administrator-unlocked CA policy. Reject CSR-requested names/usages and reuse of a host or server leaf, which would cross role authority. D131, D133 and D168 define the identity boundary. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-source-fast-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19184432 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.41s]

Completed step 5ff1faa3b41b279210ef87f678260f835ef339db: Keep host client identities separate from relay server identities. Derive relay SANs from the server-authorized host UUID and retain seven-day validity and administrator-unlocked CA policy. Reject CSR-requested names/usages and reuse of a host or server leaf, which would cross role authority. D131, D133 and D168 define the identity boundary. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-source-fast-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184432 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.41s]
kayg commented 2026-10-09 00:23:17 +00:00
Author
Owner
Copy link

ca-tests full fast attempt1 failed only the existing verify-agent.py idle RSS assertion; every other Rust row passed. No budget/source change or RSS exception claimed. Repeat the complete unchanged gate before any commit. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-tests-fast-attempt1.log. Load 53.53 45.93 35.87 54/1830 838494.

ca-tests full fast attempt1 failed only the existing verify-agent.py idle RSS assertion; every other Rust row passed. No budget/source change or RSS exception claimed. Repeat the complete unchanged gate before any commit. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-tests-fast-attempt1.log. Load 53.53 45.93 35.87 54/1830 838494.
kayg commented 2026-10-09 00:23:30 +00:00
Author
Owner
Copy link

ca-tests full fast attempt2 failed only the existing verify-agent.py idle RSS assertion; every other Rust row passed. No budget/source change or RSS exception claimed. Repeat the complete unchanged gate before any commit. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-tests-fast-attempt2.log. Load 53.17 46.22 36.13 55/1933 839817.

ca-tests full fast attempt2 failed only the existing verify-agent.py idle RSS assertion; every other Rust row passed. No budget/source change or RSS exception claimed. Repeat the complete unchanged gate before any commit. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-tests-fast-attempt2.log. Load 53.17 46.22 36.13 55/1933 839817.
kayg commented 2026-10-09 00:24:55 +00:00
Author
Owner
Copy link

CA-purpose test slice remains uncommitted at head5ff1faa3: three unchanged scripts/verify fast attempts failed only verify-agent.py:34 idle RSS <15360KiB (3.02s,3.17s,3.06s). Rust fmt/native fmt/clippy/deny and both static rows passed. Cargo stopped at agent_daemon before server tests; no claim that the new CA test ran. Raw run3-ca-tests-fast-attempt{1,2,3}.log in HOME-cache/relay-direct. RSS has no allowed baseline exception. Stopped the sequential preview writer and continuing fresh-start measurement before another gate.

CA-purpose test slice remains uncommitted at head5ff1faa3: three unchanged scripts/verify fast attempts failed only verify-agent.py:34 idle RSS <15360KiB (3.02s,3.17s,3.06s). Rust fmt/native fmt/clippy/deny and both static rows passed. Cargo stopped at agent_daemon before server tests; no claim that the new CA test ran. Raw run3-ca-tests-fast-attempt{1,2,3}.log in HOME-cache/relay-direct. RSS has no allowed baseline exception. Stopped the sequential preview writer and continuing fresh-start measurement before another gate.
kayg commented 2026-10-09 00:26:36 +00:00
Author
Owner
Copy link

Fresh-start measurement after the three RSS gate failures: same debug binary, ten starts, no own build/test running, local CPUWeight10000 scope. RSS min15124/median15344/max15512KiB;3/10 >=15360KiB. Earlier schema-only median15334/max15428,4/10 over limit; samples establish variability, not a passing budget or causal module regression. Anonymous memory was ~2244KiB and PSS ~12.5-12.9MiB. Load42.45-43.85. Raw run3-ca-idle-measure.log and measurement script in HOME-cache/relay-direct. Keeping the assertion unchanged; recording the purpose test directly, then one unchanged complete gate with warmed build artifacts.

Fresh-start measurement after the three RSS gate failures: same debug binary, ten starts, no own build/test running, local CPUWeight10000 scope. RSS min15124/median15344/max15512KiB;3/10 >=15360KiB. Earlier schema-only median15334/max15428,4/10 over limit; samples establish variability, not a passing budget or causal module regression. Anonymous memory was ~2244KiB and PSS ~12.5-12.9MiB. Load42.45-43.85. Raw run3-ca-idle-measure.log and measurement script in HOME-cache/relay-direct. Keeping the assertion unchanged; recording the purpose test directly, then one unchanged complete gate with warmed build artifacts.
kayg commented 2026-10-09 00:29:05 +00:00
Author
Owner
Copy link

Direct real-OpenSSL purpose test passed: ca::tests::relay_leaf_is_server_only_and_cannot_replace_a_host_or_the_server,1passed/0failed/163filtered,0.09s. It verifies server-selected relay SANs despite hostile CSR and accepts SSL_SERVER while refusing SSL_CLIENT chain purpose. Raw run3-ca-purpose-alone.log. This does not override the three native idle-RSS failures; CA-test slice remains uncommitted pending a complete passing fast gate.

Direct real-OpenSSL purpose test passed: ca::tests::relay_leaf_is_server_only_and_cannot_replace_a_host_or_the_server,1passed/0failed/163filtered,0.09s. It verifies server-selected relay SANs despite hostile CSR and accepts SSL_SERVER while refusing SSL_CLIENT chain purpose. Raw run3-ca-purpose-alone.log. This does not override the three native idle-RSS failures; CA-test slice remains uncommitted pending a complete passing fast gate.
kayg commented 2026-10-09 00:37:26 +00:00
Author
Owner
Copy link

Fourth CA-test fast attempt: native daemon/snapshot/VTcredit/reconnect and CA purpose test passed; only auth::tests::socket_tests::recovery_subscription_updates_every_signed_in_socket failed at auth/tests.rs:1711 (server closed websocket unexpectedly:None).146server tests passed/1failed/17ignored,143.82s. All other Rust gate rows passed, statics19184432/16120304bytes; TS was skipped by dispatcher. Exact test on own detached dev1c3ae5f4 passed1/0,1.46s (build23.13s); no baseline exception accepted. socket_tests.rs identical SHA256da40d945c48a19f8e028cf70f5bc1b67cbf3cd2b0dfe8e23ef515f2b1e1ca1ba. Load fell from41.42 to11.93 across comparison. Logs run3-ca-tests-fast-attempt4.log and run3-ca-baseline-socket.log. Continuing exact branch recheck and full gate, without production changes or native threshold edits.

Fourth CA-test fast attempt: native daemon/snapshot/VTcredit/reconnect and CA purpose test passed; only auth::tests::socket_tests::recovery_subscription_updates_every_signed_in_socket failed at auth/tests.rs:1711 (server closed websocket unexpectedly:None).146server tests passed/1failed/17ignored,143.82s. All other Rust gate rows passed, statics19184432/16120304bytes; TS was skipped by dispatcher. Exact test on own detached dev1c3ae5f4 passed1/0,1.46s (build23.13s); no baseline exception accepted. socket_tests.rs identical SHA256da40d945c48a19f8e028cf70f5bc1b67cbf3cd2b0dfe8e23ef515f2b1e1ca1ba. Load fell from41.42 to11.93 across comparison. Logs run3-ca-tests-fast-attempt4.log and run3-ca-baseline-socket.log. Continuing exact branch recheck and full gate, without production changes or native threshold edits.
kayg commented 2026-10-09 00:44:54 +00:00
Author
Owner
Copy link

Completed step 459524ae1c: Verify the real signed relay leaf accepts SSL server purpose and refuses SSL client purpose, using only server-selected SANs despite a hostile CSR. Keep inherited host signer tests. Reject inspecting certificate text alone: OpenSSL chain-purpose verification exercises the actual authority boundary. D131 and D168 require separate roles. Direct OpenSSL purpose check1pass/0fail,0.09s; three prior native idle-RSS failures and ten fresh-start measurements remain recorded on #103. No listener or production admission is enabled. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-tests-fast-attempt5.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 19184432 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.97s]

Completed step 459524ae1cb081d52a589b174e7a24812dd8b160: Verify the real signed relay leaf accepts SSL server purpose and refuses SSL client purpose, using only server-selected SANs despite a hostile CSR. Keep inherited host signer tests. Reject inspecting certificate text alone: OpenSSL chain-purpose verification exercises the actual authority boundary. D131 and D168 require separate roles. Direct OpenSSL purpose check1pass/0fail,0.09s; three prior native idle-RSS failures and ten fresh-start measurements remain recorded on #103. No listener or production admission is enabled. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-tests-fast-attempt5.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 19184432 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.97s]
kayg commented 2026-10-09 01:05:39 +00:00
Author
Owner
Copy link

Generic-TLS first gate failed native idle RSS before transport tests. x86static passed19183920bytes; ARM then stalled waiting through the configured RUSTC_WRAPPER=sccache. Owned scope run-p976758-i17751688.scope remained active,CPUweight10000/no quota/no throttling/noOOM/frozen0;CPU usage stayed ~228.292s across inspections. Its only build processes were cargo sleeping onfutex and an sccache client(wait_woken), no compiler. Stopped ONLY this owned gate scope, not the shared cache daemon or another job. Restarting the complete unchanged gate with RUSTC_WRAPPER empty so actual compiler children run inside our scope. No source, budget, dependency or toolchain choice changed; generic TLS remains uncommitted. Raw run3-tls-source-fast-attempt1.log; currentHEAD459524ae.

Generic-TLS first gate failed native idle RSS before transport tests. x86static passed19183920bytes; ARM then stalled waiting through the configured RUSTC_WRAPPER=sccache. Owned scope run-p976758-i17751688.scope remained active,CPUweight10000/no quota/no throttling/noOOM/frozen0;CPU usage stayed ~228.292s across inspections. Its only build processes were cargo sleeping onfutex and an sccache client(wait_woken), no compiler. Stopped ONLY this owned gate scope, not the shared cache daemon or another job. Restarting the complete unchanged gate with RUSTC_WRAPPER empty so actual compiler children run inside our scope. No source, budget, dependency or toolchain choice changed; generic TLS remains uncommitted. Raw run3-tls-source-fast-attempt1.log; currentHEAD459524ae.
kayg commented 2026-10-09 01:14:16 +00:00
Author
Owner
Copy link

Upstream prerequisites now available: dev459070fe merged #118 fixture stability, including CPUWeight applied at shim launch (scripts/agent_fixture.py), credit/readiness fences and virtual auth test time; remote-agents cf82af31 packs Linux x86 CLI relative relocations (owner-measured medianRSS14780->11408KiB, existing GNU/LLD loaders, no threshold/profile/engine change). Our direct-compiler genericTLS gate had idleRSS15344KiB pass then terminal.open refused at verify-agent.py:65, outside allowed exceptions; otherRust rows/staticpass19183920/16120224bytes. Preserved TLS patch inHOME-cache/run3-tls-pending.patch and restored it to isolate prerequisites. Integrating only the proven relocation change first, then #118 atomic fixture commits; no native product deadlines or global cache daemon changes. CurrentHEAD459524ae.

Upstream prerequisites now available: dev459070fe merged #118 fixture stability, including CPUWeight applied at shim launch (scripts/agent_fixture.py), credit/readiness fences and virtual auth test time; remote-agents cf82af31 packs Linux x86 CLI relative relocations (owner-measured medianRSS14780->11408KiB, existing GNU/LLD loaders, no threshold/profile/engine change). Our direct-compiler genericTLS gate had idleRSS15344KiB pass then terminal.open refused at verify-agent.py:65, outside allowed exceptions; otherRust rows/staticpass19183920/16120224bytes. Preserved TLS patch inHOME-cache/run3-tls-pending.patch and restored it to isolate prerequisites. Integrating only the proven relocation change first, then #118 atomic fixture commits; no native product deadlines or global cache daemon changes. CurrentHEAD459524ae.
kayg commented 2026-10-09 01:15:03 +00:00
Author
Owner
Copy link

Correction to the previous prerequisite finding: genericTLS direct-compiler ARMstatic size was16119440bytes, not16120224; x86was19183920. Exact PT_INTERP lines are in run3-tls-direct-compiler-fast.log. Both passed the unchanged static budget.

Correction to the previous prerequisite finding: genericTLS direct-compiler ARMstatic size was16119440bytes, not16120224; x86was19183920. Exact PT_INTERP lines are in run3-tls-direct-compiler-fast.log. Both passed the unchanged static budget.
kayg commented 2026-10-09 01:27:43 +00:00
Author
Owner
Copy link

Completed step 4c1766cc43: Imported owner build prerequisite cf82af31: packed Linux x86_64 relative relocations; ARM unchanged. Full gate measurements below are from this branch. Owner upstream paired RSS/control measurements are provenance, not newly measured here. Relay production wiring remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relocation-prerequisite-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.76s]

Completed step 4c1766cc437bf36a084b38ed1fdd157172d42ac3: Imported owner build prerequisite cf82af31: packed Linux x86_64 relative relocations; ARM unchanged. Full gate measurements below are from this branch. Owner upstream paired RSS/control measurements are provenance, not newly measured here. Relay production wiring remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relocation-prerequisite-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.76s]
kayg commented 2026-10-09 01:28:08 +00:00
Author
Owner
Copy link

At head4c1766cc, ten fresh idle daemon starts after full fast, with no own build/test overlap, local CPUWeight10000: min11752/median11808/max11936KiB;0/10 at or above15360KiB. Load10.79-11.10 (exact samples in run3-relocation-idle-measure.log). Prior unchanged CA-tree median15344/max15512,3/10 over; different load windows, so no equal-load latency claim. Per-daemon idle11808KiB x D170 fleet200=2.25GiB across hosts; this is not relay/session RSS or a connected-scale proof. Full fast static sizes18082576/16120304bytes. Proceeding owner #118 fixture prerequisite imports; relay CLI, browser path and per-session scale remain pending.

At head4c1766cc, ten fresh idle daemon starts after full fast, with no own build/test overlap, local CPUWeight10000: min11752/median11808/max11936KiB;0/10 at or above15360KiB. Load10.79-11.10 (exact samples in run3-relocation-idle-measure.log). Prior unchanged CA-tree median15344/max15512,3/10 over; different load windows, so no equal-load latency claim. Per-daemon idle11808KiB x D170 fleet200=2.25GiB across hosts; this is not relay/session RSS or a connected-scale proof. Full fast static sizes18082576/16120304bytes. Proceeding owner #118 fixture prerequisite imports; relay CLI, browser path and per-session scale remain pending.
kayg commented 2026-10-09 01:28:24 +00:00
Author
Owner
Copy link

Correction: the fresh-start load range was 11.10-11.89. D170 specifies 1,000 nodes (not200;200 is concurrent terminal sessions), so idle-daemon11808KiB x1,000=11.26GiB distributed across node hosts, and x10,000headroom=112.61GiB. These arithmetic fleet projections are not measured connected relay/server memory. Ten-start RSS values and0/10budget failures in the prior comment remain correct.

Correction: the fresh-start load range was 11.10-11.89. D170 specifies 1,000 nodes (not200;200 is concurrent terminal sessions), so idle-daemon11808KiB x1,000=11.26GiB distributed across node hosts, and x10,000headroom=112.61GiB. These arithmetic fleet projections are not measured connected relay/server memory. Ten-start RSS values and0/10budget failures in the prior comment remain correct.
kayg commented 2026-10-09 01:34:19 +00:00
Author
Owner
Copy link

First full fast for fixture import31bb5b06 failed ONLY auth review3_backend_reply_has_a_deadline at review3_tests.rs662: elapsed1.998452063s below2s;146serverpass/1fail/17ignored,198.24s. Every other Rust row/static passed. Exact test in owned detached dev1c3ae5f4 passed1/0,3.44s (deadline2004ms), same load window30.98. No exception accepted and no commit. Raw run3-prerequisite-31bb5b06-fast.log and run3-prerequisite-31bb5b06-baseline-deadline.log. Repeating full unchanged gate, preserving all timeout/budget assertions.

First full fast for fixture import31bb5b06 failed ONLY auth review3_backend_reply_has_a_deadline at review3_tests.rs662: elapsed1.998452063s below2s;146serverpass/1fail/17ignored,198.24s. Every other Rust row/static passed. Exact test in owned detached dev1c3ae5f4 passed1/0,3.44s (deadline2004ms), same load window30.98. No exception accepted and no commit. Raw run3-prerequisite-31bb5b06-fast.log and run3-prerequisite-31bb5b06-baseline-deadline.log. Repeating full unchanged gate, preserving all timeout/budget assertions.
kayg commented 2026-10-09 01:41:52 +00:00
Author
Owner
Copy link

Completed step f68555d5bb: Replace the earlier post-open scope adjustment with owner #118 fixture admission from 31bb5b06. Apply caller CPU weight before readiness, authenticate fixture-owned shim identity before scope updates, and expose exact credit and ordered Pong observations. Reject production scheduling or timeout changes and duplicate proxy implementations. Decisions 13, 41 and 75. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-31bb5b06-fast-attempt2.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.37s]

Completed step f68555d5bbc5ecae672b2f73a48214132f7509cd: Replace the earlier post-open scope adjustment with owner #118 fixture admission from 31bb5b06. Apply caller CPU weight before readiness, authenticate fixture-owned shim identity before scope updates, and expose exact credit and ordered Pong observations. Reject production scheduling or timeout changes and duplicate proxy implementations. Decisions 13, 41 and 75. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-31bb5b06-fast-attempt2.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.37s]
kayg commented 2026-10-09 01:44:09 +00:00
Author
Owner
Copy link

Security review finding in uninstalled scratch proposal only: run3-relay-store-pair.rs would put the relay control leaf in remote_certificates, which ordinary host admission reads; that would give relay-control credentials host execution authority. Reject that design before installation. Revised additive proposal: a separate relay_control_certificates public ledger and client-only relay-control SAN/URI, alongside the committed server-only endpoint ledger; one invitation atomically publishes both scoped records. Server dispatch chooses relay-control admission from its durable ledger, never host authority. Production host certificates, wire layouts and pending WebRTC foundation stay unchanged. Existing committed relay_store/CA code has no paired-control insertion or listener. Paths .scratch/relay-direct/run3-relay-store-pair.rs, crates/infraternal-server/src/relay_store.rs and enrollment.rs. New tests must prove the control fingerprint is refused by ordinary host admission before any production listener.

Security review finding in uninstalled scratch proposal only: run3-relay-store-pair.rs would put the relay control leaf in remote_certificates, which ordinary host admission reads; that would give relay-control credentials host execution authority. Reject that design before installation. Revised additive proposal: a separate relay_control_certificates public ledger and client-only relay-control SAN/URI, alongside the committed server-only endpoint ledger; one invitation atomically publishes both scoped records. Server dispatch chooses relay-control admission from its durable ledger, never host authority. Production host certificates, wire layouts and pending WebRTC foundation stay unchanged. Existing committed relay_store/CA code has no paired-control insertion or listener. Paths .scratch/relay-direct/run3-relay-store-pair.rs, crates/infraternal-server/src/relay_store.rs and enrollment.rs. New tests must prove the control fingerprint is refused by ordinary host admission before any production listener.
kayg commented 2026-10-09 01:49:40 +00:00
Author
Owner
Copy link

Completed step 1d53735a72: Import owner #118 partial-credit test from 3e291f2e. Observe actual EOF, repaint cost and producer markers instead of silence and elapsed scheduling time. Preserve payloads and exact 249/250 ms production unit coverage. Reject longer assertion sleeps or relaxed budgets. Decisions 13 and 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-3e291f2e-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.07s]

Completed step 1d53735a722e224cef16b18a3d1d0d4281b5370b: Import owner #118 partial-credit test from 3e291f2e. Observe actual EOF, repaint cost and producer markers instead of silence and elapsed scheduling time. Preserve payloads and exact 249/250 ms production unit coverage. Reject longer assertion sleeps or relaxed budgets. Decisions 13 and 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-3e291f2e-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.07s]
kayg commented 2026-10-09 01:56:45 +00:00
Author
Owner
Copy link

Coverage finding: original17 previews cover relay_forward_tests outer QUIC forwarding and remote_tls five inherited TCP/TLS cases separately; no forward test invokes dial_over(Tunnel,...). Prepared additional independent-inner-CA/outer-CA test, both peer fingerprints and exact1MiBecho, at .scratch/relay-direct/run3-nested-tls-test.rs. Uninstalled/uncompiled; no passing nested transport claim. Run after original previews are committed and before production relay listener. CurrentHEAD1d53735a; snapshot fixture274f7942 full gate running without failures so far.

Coverage finding: original17 previews cover relay_forward_tests outer QUIC forwarding and remote_tls five inherited TCP/TLS cases separately; no forward test invokes dial_over(Tunnel,...). Prepared additional independent-inner-CA/outer-CA test, both peer fingerprints and exact1MiBecho, at .scratch/relay-direct/run3-nested-tls-test.rs. Uninstalled/uncompiled; no passing nested transport claim. Run after original previews are committed and before production relay listener. CurrentHEAD1d53735a; snapshot fixture274f7942 full gate running without failures so far.
kayg commented 2026-10-09 01:57:11 +00:00
Author
Owner
Copy link

Completed step 0f963dbb4c: Import owner #118 snapshot fixture from 274f7942. Drain consumed credit and observe actual shim EOF before fencing delivery or resetting cursors. Preserve all four 12 MiB floods and ring-wrap checks. Reject shortened payloads and silence-based waits. Decisions 13, 75 and 81. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-274f7942-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.60s]

Completed step 0f963dbb4c4f79b0aafa3327980b15fa5a231b08: Import owner #118 snapshot fixture from 274f7942. Drain consumed credit and observe actual shim EOF before fencing delivery or resetting cursors. Preserve all four 12 MiB floods and ring-wrap checks. Reject shortened payloads and silence-based waits. Decisions 13, 75 and 81. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-274f7942-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.60s]
kayg commented 2026-10-09 02:06:27 +00:00
Author
Owner
Copy link

Completed step 9a4c0b45f5: Import owner #118 test-only Tokio feature from b60f19f8. Enable configured deadline control without changing production dependency features or the lockfile. Reject a new clock engine or production timeout changes. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-b60f19f8-fast-resolved.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.31s]

Completed step 9a4c0b45f519514dcbf9f22c254d3254248199b8: Import owner #118 test-only Tokio feature from b60f19f8. Enable configured deadline control without changing production dependency features or the lockfile. Reject a new clock engine or production timeout changes. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-b60f19f8-fast-resolved.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.31s]
kayg commented 2026-10-09 02:12:29 +00:00
Author
Owner
Copy link

Completed step 703b6824b6: Import owner #118 observed-readiness fixtures from 64ed588f. Freeze correctness clocks during actual SQLite and socket IO and cross exact configured deadline boundaries after Hello or Pong. Preserve sixteen sockets, real delivery and release measurements. Reject enlarged production deadlines or relaxed admission. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-64ed588f-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.20s]

Completed step 703b6824b66cfa19141afc9dd2c1e06a36b8ed53: Import owner #118 observed-readiness fixtures from 64ed588f. Freeze correctness clocks during actual SQLite and socket IO and cross exact configured deadline boundaries after Hello or Pong. Preserve sixteen sockets, real delivery and release measurements. Reject enlarged production deadlines or relaxed admission. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-64ed588f-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.20s]
kayg commented 2026-10-09 02:18:40 +00:00
Author
Owner
Copy link

Completed step 71d7921532: Import owner #118 review cases from 8b7cccde. Use stationary time and exact advances after observing Call or Welcome while retaining revocation, late-result and transaction checks. Reject retries and extra production deadline allowances. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-8b7cccde-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.53s]

Completed step 71d7921532892d109e199f6ebe2083aaeab6d318: Import owner #118 review cases from 8b7cccde. Use stationary time and exact advances after observing Call or Welcome while retaining revocation, late-result and transaction checks. Reject retries and extra production deadline allowances. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-8b7cccde-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.53s]
kayg commented 2026-10-09 02:25:23 +00:00
Author
Owner
Copy link

Completed step 30f37c7892: Import owner #118 negative shutdown fixture from 3b51a63c. Release the controlled clock after committed logout and queued Resume so the bounded writer fallback can close; every Welcome remains a failure. Preserve the original read hang bound. Reject production authority or timeout changes. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-3b51a63c-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.22s]

Completed step 30f37c7892134e6a95414ce847696dcc664fbbda: Import owner #118 negative shutdown fixture from 3b51a63c. Release the controlled clock after committed logout and queued Resume so the bounded writer fallback can close; every Welcome remains a failure. Preserve the original read hang bound. Reject production authority or timeout changes. Decision 41. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prerequisite-3b51a63c-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120304 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.22s]
kayg commented 2026-10-09 02:44:11 +00:00
Author
Owner
Copy link

GenericTLS full fast after all #118 imports failed ONLY agent_queues saturated_terminal_queues_preserve_every_byte_and_daemon_liveness: verify-agent-queues.py25/client.open refused (generic error),16.50s. AllotherRustrows passed; statics18082064/16119440bytes. scripts/agent_fixture.py SHA256f62b49f924b0dcbd3eca7d8a6ccc170c8ddd0efc93f040564d167f74c524bb85 exactlymatchesdev. No allowedexception andnoTLScommit. Stablecopy raw run3-tls-after-prerequisites-queue-failure.log. Runningexactqueuefixturealone afterbuilds tocheckstartup. Loghygiene correction: resumer reused old tls-source attempt1 path, overwriting the earlier interrupted RSS/sccache raw log; its recorded issue evidence remains, and run3-tls-direct-compiler-fast.log is preserved. Future gate logs now use unique run timestamps, preventing overwrite. CurrentHEAD30f37c78; onlytrackedTLS30linegenericIOdelta.

GenericTLS full fast after all #118 imports failed ONLY agent_queues saturated_terminal_queues_preserve_every_byte_and_daemon_liveness: verify-agent-queues.py25/client.open refused (generic error),16.50s. AllotherRustrows passed; statics18082064/16119440bytes. scripts/agent_fixture.py SHA256f62b49f924b0dcbd3eca7d8a6ccc170c8ddd0efc93f040564d167f74c524bb85 exactlymatchesdev. No allowedexception andnoTLScommit. Stablecopy raw run3-tls-after-prerequisites-queue-failure.log. Runningexactqueuefixturealone afterbuilds tocheckstartup. Loghygiene correction: resumer reused old tls-source attempt1 path, overwriting the earlier interrupted RSS/sccache raw log; its recorded issue evidence remains, and run3-tls-direct-compiler-fast.log is preserved. Future gate logs now use unique run timestamps, preventing overwrite. CurrentHEAD30f37c78; onlytrackedTLS30linegenericIOdelta.
kayg commented 2026-10-09 02:44:35 +00:00
Author
Owner
Copy link

Exact unchanged queue fixture afterbuilds exited0: "queue overload: 12 streams retired then replayed x 65 Input frames; 2640000 payload bytes exactly once; dead shim isolated; daemon Ping: PASS". Raw run3-tls-queue-alone.log, localCPUWeight10000,load13.76. This does notoverride firstfullgatefailure. RepeatingcompleteunchangedTLSgate with uniqueimmutablelognames; commitonlyifallrowsgreen.

Exact unchanged queue fixture afterbuilds exited0: "queue overload: 12 streams retired then replayed x 65 Input frames; 2640000 payload bytes exactly once; dead shim isolated; daemon Ping: PASS". Raw run3-tls-queue-alone.log, localCPUWeight10000,load13.76. This does notoverride firstfullgatefailure. RepeatingcompleteunchangedTLSgate with uniqueimmutablelognames; commitonlyifallrowsgreen.
kayg commented 2026-10-09 02:50:42 +00:00
Author
Owner
Copy link

Completed step 68f19a1744: Generalize the existing TLS channel over AsyncRead/AsyncWrite while preserving mutual identity verification, ALPN, partial-write retirement and default TCP behavior. Add dial_over for an already-admitted tunnel. Reject copying a second TLS parser or treating relay identity as the inner server identity. D64, D131 and D176 require end-to-end host authentication. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tls-source-fast-1791513875410819123-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.52s]

Completed step 68f19a1744f1a1e75de30553ad4ca9de041042a8: Generalize the existing TLS channel over AsyncRead/AsyncWrite while preserving mutual identity verification, ALPN, partial-write retirement and default TCP behavior. Add dial_over for an already-admitted tunnel. Reject copying a second TLS parser or treating relay identity as the inner server identity. D64, D131 and D176 require end-to-end host authentication. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tls-source-fast-1791513875410819123-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.52s]
kayg commented 2026-10-09 03:04:09 +00:00
Author
Owner
Copy link

Completed step f7d4a2cf19: Adapt one paired QUIC bidirectional stream to existing asynchronous byte IO so the inner authenticated channel is reused. Reject inventing a new frame parser or using the outer stream ID as durable writer identity. No listener or destination is selected here. D64, D128 and D176 preserve identity and transport separation. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-source-fast-1791513875410819123-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.29s]

Completed step f7d4a2cf19c80d454c4b020c759f4995fd537f3e: Adapt one paired QUIC bidirectional stream to existing asynchronous byte IO so the inner authenticated channel is reused. Reject inventing a new frame parser or using the outer stream ID as durable writer identity. No listener or destination is selected here. D64, D128 and D176 preserve identity and transport separation. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-source-fast-1791513875410819123-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.29s]
kayg commented 2026-10-09 03:15:42 +00:00
Author
Owner
Copy link

Completed step 0dc178453a: Forward authenticated QUIC allocations only to one configured server endpoint after current-control authorization. Bound pending handshakes, allocations and monotonic leases; control revocation and generation replacement retire idle and active forwards. Reject arbitrary destinations, traffic-based renewal and unbounded admission. This reusable module is not mounted by startup. D64, D128, D131 and D170 define the limits. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-forward-source-fast-1791513875410819123-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.48s]

Completed step 0dc178453ace12284c656480dfdcc9379ead91f1: Forward authenticated QUIC allocations only to one configured server endpoint after current-control authorization. Bound pending handshakes, allocations and monotonic leases; control revocation and generation replacement retire idle and active forwards. Reject arbitrary destinations, traffic-based renewal and unbounded admission. This reusable module is not mounted by startup. D64, D128, D131 and D170 define the limits. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-forward-source-fast-1791513875410819123-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.48s]
kayg commented 2026-10-09 03:31:06 +00:00
Author
Owner
Copy link

Forwarding tests full fast failed only clippy unused_async_trait_impl in the immediate test Authorizer (relay_forward_tests.rs:9). Test and all other gate rows passed; static binaries 18082064/16119440 bytes. Replace the test method with std::future::ready; no production change. Raw run3-forward-tests-fast-1791513875410819123-attempt1.log. Re-run the full gate before committing.

Forwarding tests full fast failed only clippy unused_async_trait_impl in the immediate test Authorizer (relay_forward_tests.rs:9). Test and all other gate rows passed; static binaries 18082064/16119440 bytes. Replace the test method with std::future::ready; no production change. Raw run3-forward-tests-fast-1791513875410819123-attempt1.log. Re-run the full gate before committing.
kayg commented 2026-10-09 03:33:32 +00:00
Author
Owner
Copy link

Owner 2026-10-09: no Mac Studio exists; the str0m vs webrtc comparison runs on the shared perf VM (perf-test-69-63) under a lock arranged with calternal. Owner also approved the per-chunk on-demand budget (D188, c1f66be).

Owner 2026-10-09: no Mac Studio exists; the str0m vs webrtc comparison runs on the shared perf VM (perf-test-69-63) under a lock arranged with calternal. Owner also approved the per-chunk on-demand budget (D188, c1f66be).
kayg commented 2026-10-09 03:37:30 +00:00
Author
Owner
Copy link

Completed step 886ac2fb4b: Add four real QUIC/mTLS tests for exact 1 MiB echo, refusal before opening upstream on wrong peer or an overlong lease, control revocation and generation retirement. The five inherited TLS tests continue exercising the generic inner channel. Reject mocked socket forwarding or a transport echo as browser/PTY production acceptance. D128, D131 and D176 require current authority and byte preservation. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-forward-tests-fast-1791516666904713898-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.13s]

Completed step 886ac2fb4b2b4ea8c3e010355119f0a413c2c1c6: Add four real QUIC/mTLS tests for exact 1 MiB echo, refusal before opening upstream on wrong peer or an overlong lease, control revocation and generation retirement. The five inherited TLS tests continue exercising the generic inner channel. Reject mocked socket forwarding or a transport echo as browser/PTY production acceptance. D128, D131 and D176 require current authority and byte preservation. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-forward-tests-fast-1791516666904713898-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.13s]
kayg commented 2026-10-09 03:44:55 +00:00
Author
Owner
Copy link

Completed step 75ee0a7c29: Exercise dial_over with a real admitted QUIC tunnel rather than only testing QUIC and TCP/TLS independently. Use independent inner and outer issuers, verify both inner peer fingerprints, and deliver an exact 1 MiB echo. Reuse the existing socket and certificate fixtures. Reject treating separate handshake tests as end-to-end tunnel identity evidence. Decisions 131 and 176. Targeted test raw /home/kayg/.cache/codex-scratch/relay-direct/run3-nested-tls-1791517063342517301-unit.log. No production listener or browser/PTY acceptance is implied. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-nested-tls-1791517063342517301-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.47s]

Completed step 75ee0a7c29d6c95c0ae790d093b58867d82095c7: Exercise dial_over with a real admitted QUIC tunnel rather than only testing QUIC and TCP/TLS independently. Use independent inner and outer issuers, verify both inner peer fingerprints, and deliver an exact 1 MiB echo. Reuse the existing socket and certificate fixtures. Reject treating separate handshake tests as end-to-end tunnel identity evidence. Decisions 131 and 176. Targeted test raw /home/kayg/.cache/codex-scratch/relay-direct/run3-nested-tls-1791517063342517301-unit.log. No production listener or browser/PTY acceptance is implied. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-nested-tls-1791517063342517301-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.47s]
kayg commented 2026-10-09 03:46:10 +00:00
Author
Owner
Copy link

At HEAD 75ee0a7c, two new assertion tests fail on unchanged relay_forward.rs: late renewal revives an expired holder (0 passed/1 failed, 0.08s), and an older issued grant remains live (0/1, 0.02s). Evidence run3-lease-negative-1791517512750284139-{renewal_after_unpolled_expiry_cannot_revive_a_holder,older_control_grant_cannot_replace_the_admitted_deadline}.log under HOME cache/relay-direct. Current retirement loop replaces its old deadline before checking expiry; it also lacks issuance ordering. Fix will retain the admitted deadline and refuse regressed updates. D128.

At HEAD 75ee0a7c, two new assertion tests fail on unchanged relay_forward.rs: late renewal revives an expired holder (0 passed/1 failed, 0.08s), and an older issued grant remains live (0/1, 0.02s). Evidence run3-lease-negative-1791517512750284139-{renewal_after_unpolled_expiry_cannot_revive_a_holder,older_control_grant_cannot_replace_the_admitted_deadline}.log under HOME cache/relay-direct. Current retirement loop replaces its old deadline before checking expiry; it also lacks issuance ordering. Fix will retain the admitted deadline and refuse regressed updates. D128.
kayg commented 2026-10-09 03:54:13 +00:00
Author
Owner
Copy link

Production route integration finding: remote_link.rs QuicLink already keeps terminal streams by durable writer ID; embedded.rs run_actor routes by writer. Use that existing per-writer transport map for alternate terminal paths, preserving the primary host control connection. Reject RemoteHosts::admit for path selection: its Source replacement invokes connection_lost/reattach_all and affects every attachment (embedded.rs:523,561). Add route generation fencing to queued terminal delivery so retired-path frames cannot change offsets or credit. This is transport composition using D176, not a new foundational engine choice. Implements scoped D128/D185; no source wiring or browser acceptance yet.

Production route integration finding: remote_link.rs QuicLink already keeps terminal streams by durable writer ID; embedded.rs run_actor routes by writer. Use that existing per-writer transport map for alternate terminal paths, preserving the primary host control connection. Reject RemoteHosts::admit for path selection: its Source replacement invokes connection_lost/reattach_all and affects every attachment (embedded.rs:523,561). Add route generation fencing to queued terminal delivery so retired-path frames cannot change offsets or credit. This is transport composition using D176, not a new foundational engine choice. Implements scoped D128/D185; no source wiring or browser acceptance yet.
kayg commented 2026-10-09 04:02:16 +00:00
Author
Owner
Copy link

Completed step 08ce6141b3: Three assertion regressions fail on unchanged 75ee0a7c: an expired holder revives after suspension, expiry before first poll disappears, and a backdated grant remains live. Carry the admitted lease into retirement; check its expiry before consuming any queued renewal, reject regressed issuance/deadlines, and prioritize expiry. Reject replacing the old deadline first or relying on watch select order. Clarify that trusted control senders must seal after revocation because updates coalesce. D128, D131 and D176. Negative raw logs run3-lease-negative-1791517512750284139-*.log and run3-lease-first-poll-negative.log; positive run3-lease-positive.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-lease-source-fast-1791517685150475809-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.60s]

Completed step 08ce6141b3dde0ee65d14fb0e40fab26d47fd8de: Three assertion regressions fail on unchanged 75ee0a7c: an expired holder revives after suspension, expiry before first poll disappears, and a backdated grant remains live. Carry the admitted lease into retirement; check its expiry before consuming any queued renewal, reject regressed issuance/deadlines, and prioritize expiry. Reject replacing the old deadline first or relying on watch select order. Clarify that trusted control senders must seal after revocation because updates coalesce. D128, D131 and D176. Negative raw logs run3-lease-negative-1791517512750284139-*.log and run3-lease-first-poll-negative.log; positive run3-lease-positive.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-lease-source-fast-1791517685150475809-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.60s]
kayg commented 2026-10-09 04:08:02 +00:00
Author
Owner
Copy link

Completed step 717db44066: Add four holder tests covering late renewal after suspension, expiry before the first poll, regressed issuance, and a timely same-generation renewal followed by control loss. The first three failed on the unchanged implementation before its fix. Reject handshake-only retirement coverage: scheduler suspension must not revive spent authority. D128 and D131. Raw negative and positive evidence is retained under HOME cache/relay-direct. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-lease-tests-fast-1791517685150475809-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.28s]

Completed step 717db44066d0459fcfeed08110a1eec7c6949dec: Add four holder tests covering late renewal after suspension, expiry before the first poll, regressed issuance, and a timely same-generation renewal followed by control loss. The first three failed on the unchanged implementation before its fix. Reject handshake-only retirement coverage: scheduler suspension must not revive spent authority. D128 and D131. Raw negative and positive evidence is retained under HOME cache/relay-direct. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-lease-tests-fast-1791517685150475809-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082064 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119440 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.28s]
kayg commented 2026-10-09 04:09:04 +00:00
Author
Owner
Copy link

At 717db440, both independent-review regressions fail on unchanged source: retired native QUIC and TCP/TLS channels reject raw IO with BrokenPipe yet Receive.next returns buffered Input {in_seq:5, bytes:second}. Each test0 passed/1 failed. Logs run3-cached-retirement-negative-1791518911036506049-{remote_native::tests::retired_native_channel_refuses_coalesced_input,daemon::remote_tests::retired_tls_channel_refuses_coalesced_input}.log under HOME cache/relay-direct. Shared Receive already checks Channel::check_authority before exposing frames; those channels leave the default hook unconditional. Land each channel hook with its own gate. TLS test is retained in cache and temporarily removed while the independent native slice is gated; no failing test is committed. D64/D128/D130.

At 717db440, both independent-review regressions fail on unchanged source: retired native QUIC and TCP/TLS channels reject raw IO with BrokenPipe yet Receive.next returns buffered Input {in_seq:5, bytes:second}. Each test0 passed/1 failed. Logs run3-cached-retirement-negative-1791518911036506049-{remote_native::tests::retired_native_channel_refuses_coalesced_input,daemon::remote_tests::retired_tls_channel_refuses_coalesced_input}.log under HOME cache/relay-direct. Shared Receive already checks Channel::check_authority before exposing frames; those channels leave the default hook unconditional. Land each channel hook with its own gate. TLS test is retained in cache and temporarily removed while the independent native slice is gated; no failing test is committed. D64/D128/D130.
kayg commented 2026-10-09 04:09:38 +00:00
Author
Owner
Copy link

Engine comparison on the quiet perf VM (perf-test-69-63, 4 cores, run by the orchestrator under the shared lock, 2026-10-09 05:58-06:09 CEST, 1-min load 0.5-3.8). Same harness as run 2 (headless Chromium 149 ↔ static musl candidate binary, loopback, 31 one-byte echoes + 50 MiB exact echo per round), 20 strictly alternating rounds per engine:

str0m 0.24.1 webrtc 0.21.0
rounds passed 20/20 20/20
echo p50 / p99 0.40 / 0.90 ms 0.50 / 2.90 ms
connection established, median 8.5 ms 208.9 ms
50 MiB echo throughput, median (min-max) 11.3 MiB/s (7.2-18.4) 12.8 MiB/s (6.4-19.6)

Throughput ranges overlap (no difference shown). Excludes PTY, auth, TURN and network delay. Raw log and harness: ~/.local/state/codex-jobs/infraternal/relay-direct/perf-vm-engines/. Owner decides (foundation).

Engine comparison on the quiet perf VM (perf-test-69-63, 4 cores, run by the orchestrator under the shared lock, 2026-10-09 05:58-06:09 CEST, 1-min load 0.5-3.8). Same harness as run 2 (headless Chromium 149 ↔ static musl candidate binary, loopback, 31 one-byte echoes + 50 MiB exact echo per round), 20 strictly alternating rounds per engine: | | str0m 0.24.1 | webrtc 0.21.0 | |---|---|---| | rounds passed | 20/20 | 20/20 | | echo p50 / p99 | 0.40 / 0.90 ms | 0.50 / 2.90 ms | | connection established, median | 8.5 ms | 208.9 ms | | 50 MiB echo throughput, median (min-max) | 11.3 MiB/s (7.2-18.4) | 12.8 MiB/s (6.4-19.6) | Throughput ranges overlap (no difference shown). Excludes PTY, auth, TURN and network delay. Raw log and harness: ~/.local/state/codex-jobs/infraternal/relay-direct/perf-vm-engines/. Owner decides (foundation).
kayg commented 2026-10-09 04:11:08 +00:00
Author
Owner
Copy link

Owner decision D192 (main 793e4a31): the direct path uses str0m 0.24.x behind the existing trait; remove the webrtc-rs candidate. Next relay-direct run: wire str0m as the direct engine (still off by default until relay + switching pass acceptance), and benchmark the direct path at the India–Germany profile.

Owner decision D192 (main 793e4a31): the direct path uses str0m 0.24.x behind the existing trait; remove the webrtc-rs candidate. Next relay-direct run: wire str0m as the direct engine (still off by default until relay + switching pass acceptance), and benchmark the direct path at the India–Germany profile.
kayg commented 2026-10-09 04:24:40 +00:00
Author
Owner
Copy link

Completed step be603cbb01: The reviewer real-QUIC test failed on 717db440: raw IO rejected retirement while Receive returned buffered Input sequence5. Implement the existing Channel authority hook using the native retirement signal, fencing coalesced bytes before dispatch without copying or adding a parser. Reject relying on the next socket read to observe retirement. D64, D128 and D130. Negative run3-cached-retirement-negative-1791518911036506049-remote_native::tests::retired_native_channel_refuses_coalesced_input.log; targeted positive5 passed/1 diagnostic ignored in0.14s, run3-native-retirement-positive.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-native-retirement-source-fast-1791519011820829195-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082640 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119952 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.57s]

Completed step be603cbb01f3fbfe700a6ac14ea97408c8924128: The reviewer real-QUIC test failed on 717db440: raw IO rejected retirement while Receive returned buffered Input sequence5. Implement the existing Channel authority hook using the native retirement signal, fencing coalesced bytes before dispatch without copying or adding a parser. Reject relying on the next socket read to observe retirement. D64, D128 and D130. Negative run3-cached-retirement-negative-1791518911036506049-remote_native::tests::retired_native_channel_refuses_coalesced_input.log; targeted positive5 passed/1 diagnostic ignored in0.14s, run3-native-retirement-positive.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-native-retirement-source-fast-1791519011820829195-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082640 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119952 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.57s]
kayg commented 2026-10-09 04:31:07 +00:00
Author
Owner
Copy link

Completed step e12f63199f: Retain the independent-review real-QUIC regression after reproducing its assertion failure locally. Write two Input frames together, consume the first, retire, and require the raw read and cached second frame to fail. Reject fake byte channels as proof of native transport retirement. D64, D128 and D130. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-native-retirement-tests-fast-1791519011820829195-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082640 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119952 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.79s]

Completed step e12f63199ffcd88faa05de5de0e15824b0f377c1: Retain the independent-review real-QUIC regression after reproducing its assertion failure locally. Write two Input frames together, consume the first, retire, and require the raw read and cached second frame to fail. Reject fake byte channels as proof of native transport retirement. D64, D128 and D130. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-native-retirement-tests-fast-1791519011820829195-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082640 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119952 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.79s]
kayg commented 2026-10-09 04:46:04 +00:00
Author
Owner
Copy link

Completed step 4f2cbcfa36: The independent-review real mutual-TLS test failed on717db440: raw IO rejected retirement while Receive returned buffered Input sequence5. Implement the existing Channel authority hook over the TLS retirement signal for default TCP and generic tunnel IO. Reject duplicating the decoder or relying on another socket read. D64, D128 and D176. Negative run3-cached-retirement-negative-1791518911036506049-daemon::remote_tests::retired_tls_channel_refuses_coalesced_input.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tls-retirement-source-fast-1791519011820829195-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18083280 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120592 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.31s]

Completed step 4f2cbcfa367e814c3747a13006b3a150b250652b: The independent-review real mutual-TLS test failed on717db440: raw IO rejected retirement while Receive returned buffered Input sequence5. Implement the existing Channel authority hook over the TLS retirement signal for default TCP and generic tunnel IO. Reject duplicating the decoder or relying on another socket read. D64, D128 and D176. Negative run3-cached-retirement-negative-1791518911036506049-daemon::remote_tests::retired_tls_channel_refuses_coalesced_input.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tls-retirement-source-fast-1791519011820829195-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18083280 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120592 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.31s]
kayg commented 2026-10-09 04:52:10 +00:00
Author
Owner
Copy link

Completed step 5f2d6c4559: Retain the independent-review two-frame retirement regression using the existing real mutually authenticated socket fixture. The original assertion failed locally before the hook fix. Reject raw IO failure alone as evidence that buffered terminal input is fenced. D64, D128 and D176. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tls-retirement-tests-fast-1791519011820829195-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18083280 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16120592 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.41s]

Completed step 5f2d6c455902c18999798e402633f6d96f13eb4a: Retain the independent-review two-frame retirement regression using the existing real mutually authenticated socket fixture. The original assertion failed locally before the hook fix. Reject raw IO failure alone as evidence that buffered terminal input is fenced. D64, D128 and D176. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tls-retirement-tests-fast-1791519011820829195-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18083280 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16120592 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.41s]
kayg commented 2026-10-09 04:56:18 +00:00
Author
Owner
Copy link

At5f2d6c45, the existing real-QUIC prefixed-control fixture fails0/1 in0.02s after adding retirement assertions: authority=false,prefix=false,greeting=false. The wrapper leaves check_authority unconditional and serves its retained prefix/once-only greeting without calling the retired native channel. Raw run3-prefixed-retirement-negative-1791521713036301555.log under HOME cache/relay-direct. Extend the wrapper hook and check around awaited cache locks before reading/suppressing bytes. No new fixture or parser. D64/D128/D130.

At5f2d6c45, the existing real-QUIC prefixed-control fixture fails0/1 in0.02s after adding retirement assertions: authority=false,prefix=false,greeting=false. The wrapper leaves check_authority unconditional and serves its retained prefix/once-only greeting without calling the retired native channel. Raw run3-prefixed-retirement-negative-1791521713036301555.log under HOME cache/relay-direct. Extend the wrapper hook and check around awaited cache locks before reading/suppressing bytes. No new fixture or parser. D64/D128/D130.
kayg commented 2026-10-09 05:14:13 +00:00
Author
Owner
Copy link

Completed step 1ea2aeaed6: The real admitted QUIC wrapper fixture failed on5f2d6c45: its authority hook, retained-prefix read and suppressed greeting write all bypassed underlying retirement. Delegate the existing authority hook and recheck after awaited cache locks and before exposing retained bytes. Reject considering the underlying native hook sufficient when a wrapper can serve bytes without invoking IO. D64, D128 and D130. Negative run3-prefixed-retirement-negative-1791521713036301555.log; targeted positive9 passed/1 diagnostic ignored in0.67s, run3-prefixed-retirement-positive.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prefixed-retirement-source-fast-1791521867214618713-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119184 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.63s]

Completed step 1ea2aeaed6b618a802cbcba251688e6a81d405a7: The real admitted QUIC wrapper fixture failed on5f2d6c45: its authority hook, retained-prefix read and suppressed greeting write all bypassed underlying retirement. Delegate the existing authority hook and recheck after awaited cache locks and before exposing retained bytes. Reject considering the underlying native hook sufficient when a wrapper can serve bytes without invoking IO. D64, D128 and D130. Negative run3-prefixed-retirement-negative-1791521713036301555.log; targeted positive9 passed/1 diagnostic ignored in0.67s, run3-prefixed-retirement-positive.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prefixed-retirement-source-fast-1791521867214618713-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119184 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.63s]
kayg commented 2026-10-09 05:23:19 +00:00
Author
Owner
Copy link

Completed step ab96a5913b: Extend the existing real-QUIC fixture to retain a prefix and once-only greeting, retire the channel, and require all cached read/write/authority operations to refuse. All three branches failed locally before the fix. Reuse the already admitted socket fixture rather than introducing fake channels or another greeting builder. D64, D128 and D130. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prefixed-retirement-tests-fast-1791521867214618713-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16119184 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [6.09s]

Completed step ab96a5913bb958183d3847d57f3c811a376bdeef: Extend the existing real-QUIC fixture to retain a prefix and once-only greeting, retire the channel, and require all cached read/write/authority operations to refuse. All three branches failed locally before the fix. Reuse the already admitted socket fixture rather than introducing fake channels or another greeting builder. D64, D128 and D130. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-prefixed-retirement-tests-fast-1791521867214618713-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18082576 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16119184 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [6.09s]
kayg commented 2026-10-09 05:28:55 +00:00
Author
Owner
Copy link

The rejected unmounted paired-ledger prototype reproduced its authority crossing on real SQLite: relay-control record passed ordinary EnrollmentStore::admit instead of being refused (0 passed/1 failed,0.14s). Raw run3-role-isolation-negative-1791523469890638415.log. The runner restored both committed relay_store source/test files; tracked tree is clean atab96a591. Never commit that remote_certificates insertion. Add isolated relay_control_certificates tied to the durable role generation, then prove rollback, separate endpoint/control/host admission, and purpose-scoped real CA leaves. D131/D132/D168; additive ledger proposal already recorded on103.

The rejected unmounted paired-ledger prototype reproduced its authority crossing on real SQLite: relay-control record passed ordinary EnrollmentStore::admit instead of being refused (0 passed/1 failed,0.14s). Raw run3-role-isolation-negative-1791523469890638415.log. The runner restored both committed relay_store source/test files; tracked tree is clean atab96a591. Never commit that remote_certificates insertion. Add isolated relay_control_certificates tied to the durable role generation, then prove rollback, separate endpoint/control/host admission, and purpose-scoped real CA leaves. D131/D132/D168; additive ledger proposal already recorded on103.
kayg commented 2026-10-09 05:47:44 +00:00
Author
Owner
Copy link

Completed step a88df230af: The rejected paired-ledger prototype inserted its control record into remote_certificates and passed ordinary host admission; the real SQLite assertion failed0/1 in0.14s before this change. Create a separate public control-certificate ledger bound to relay-role host and generation, with fixed key/fingerprint sizes and seven-day validity. Reject granting relay control ordinary host execution authority. The schema alone mounts no listener or signer. D131, D132 and D168. Evidence run3-role-isolation-negative-1791523469890638415.log under HOME cache/relay-direct. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-ledger-schema-fast-1791523736006286940-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.24s]

Completed step a88df230af624647a07146b983dc1bf8063a2f8f: The rejected paired-ledger prototype inserted its control record into remote_certificates and passed ordinary host admission; the real SQLite assertion failed0/1 in0.14s before this change. Create a separate public control-certificate ledger bound to relay-role host and generation, with fixed key/fingerprint sizes and seven-day validity. Reject granting relay control ordinary host execution authority. The schema alone mounts no listener or signer. D131, D132 and D168. Evidence run3-role-isolation-negative-1791523469890638415.log under HOME cache/relay-direct. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-ledger-schema-fast-1791523736006286940-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.24s]
kayg commented 2026-10-09 05:49:24 +00:00
Author
Owner
Copy link

Additive relay enrollment contract for D131/D132/D168/D176: owner/session-protected role invitation and disable endpoints; bounded canonical-CSR join endpoint over pinned HTTPS; paired server-only endpoint and client-only relay-control leaves from one locally proven key. The control URI is urn:infraternal:relay-control:, with authority isolated from ordinary host execution and bound to the current enabled role generation. Both certificate records and one-use capability consumption commit together. No host signer reuse, private key persistence, public token query parameters or new foundational engine. Runtime role disable must retire active forwarding before acknowledgement; ledger-only disable is not production completion. Existing transport Call/Result framing will be reused for the host-checkbox backend, with bounded action payloads and separate parser fuzz coverage. Production wiring, real browser/300 ms path and scale measurements remain pending.

Additive relay enrollment contract for D131/D132/D168/D176: owner/session-protected role invitation and disable endpoints; bounded canonical-CSR join endpoint over pinned HTTPS; paired server-only endpoint and client-only relay-control leaves from one locally proven key. The control URI is urn:infraternal:relay-control:<host UUID>, with authority isolated from ordinary host execution and bound to the current enabled role generation. Both certificate records and one-use capability consumption commit together. No host signer reuse, private key persistence, public token query parameters or new foundational engine. Runtime role disable must retire active forwarding before acknowledgement; ledger-only disable is not production completion. Existing transport Call/Result framing will be reused for the host-checkbox backend, with bounded action payloads and separate parser fuzz coverage. Production wiring, real browser/300 ms path and scale measurements remain pending.
kayg commented 2026-10-09 06:06:33 +00:00
Author
Owner
Copy link

Completed step 1b7fd42c70: The rejected draft passed a relay-control record to ordinary host execution admission; its real SQLite regression failed locally before this fix. Publish endpoint and control records together with capability consumption, using the isolated control ledger and role generation. Require different fingerprints with the same proven key, issuer and validity. Reuse the existing redemption transaction rather than duplicate authorization or consume the token early. D131, D132 and D168. Negative run3-role-isolation-negative-1791523469890638415.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-pair-source-fast-1791524914765767316-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.63s]

Completed step 1b7fd42c70779d4b9201ccc0e4bd0dd6bb681d17: The rejected draft passed a relay-control record to ordinary host execution admission; its real SQLite regression failed locally before this fix. Publish endpoint and control records together with capability consumption, using the isolated control ledger and role generation. Require different fingerprints with the same proven key, issuer and validity. Reuse the existing redemption transaction rather than duplicate authorization or consume the token early. D131, D132 and D168. Negative run3-role-isolation-negative-1791523469890638415.log. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-pair-source-fast-1791524914765767316-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.63s]
kayg commented 2026-10-09 06:07:36 +00:00
Author
Owner
Copy link

Paired-publication source committed 1b7fd42c after full fast exit0 (verbatim gate table already posted). Targeted real SQLite: 7 passed, 0 failed, 0.27 s; raw run3-control-pair-positive-1791524885130333260.log. Includes wrong key/issuer/lifetime/fingerprint refusal, purpose isolation, and second-insert failure rolling back endpoint publication and capability consumption. X86 static 18086768 B, ARM 16123360 B; raw run3-control-pair-source-fast-1791524914765767316-attempt1.log. Retained test-file commit is gating separately. Control admission/signer/broker and production runtime remain next. No latency/connected-RSS acceptance claimed.

Paired-publication source committed 1b7fd42c after full fast exit0 (verbatim gate table already posted). Targeted real SQLite: 7 passed, 0 failed, 0.27 s; raw run3-control-pair-positive-1791524885130333260.log. Includes wrong key/issuer/lifetime/fingerprint refusal, purpose isolation, and second-insert failure rolling back endpoint publication and capability consumption. X86 static 18086768 B, ARM 16123360 B; raw run3-control-pair-source-fast-1791524914765767316-attempt1.log. Retained test-file commit is gating separately. Control admission/signer/broker and production runtime remain next. No latency/connected-RSS acceptance claimed.
kayg commented 2026-10-09 06:14:11 +00:00
Author
Owner
Copy link

Completed step ce03def006: Retain the locally failing control/host authority regression and add transactional rollback evidence. Ordinary host admission refuses both paired records; endpoint admission refuses control. Wrong key, issuer, validity or reused fingerprint leaves the invitation valid. Failure on the control insert rolls back endpoint publication and token consumption. Reject mocked storage outcomes. D131, D132 and D168. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-pair-tests-fast-1791524914765767316-attempt1.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.57s]

Completed step ce03def0062a9dbbd3458941e3fefdff7b679a27: Retain the locally failing control/host authority regression and add transactional rollback evidence. Ordinary host admission refuses both paired records; endpoint admission refuses control. Wrong key, issuer, validity or reused fingerprint leaves the invitation valid. Failure on the control insert rolls back endpoint publication and token consumption. Reject mocked storage outcomes. D131, D132 and D168. Real component counts and static sizes appear in the raw fast log. No production listener/CLI/control wiring or full-route benchmark acceptance is claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-pair-tests-fast-1791524914765767316-attempt1.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.57s]
kayg commented 2026-10-09 06:14:49 +00:00
Author
Owner
Copy link

Existing real HTTP router fails the new relay enrollment regression: /relay/invitations returns 404 instead of the required private unauthenticated 401 (0 passed/1 failed). Existing SQLite, Auth, signer, CSR and request helpers reused. Implement additive pinned-HTTPS enrollment routes with session/Origin ownership and canonical bounded CSR; never mount ledger-only disable without active forwarding retirement. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-negative-1791526463970024452.log. D131/D132/D168/D176.

Existing real HTTP router fails the new relay enrollment regression: /relay/invitations returns 404 instead of the required private unauthenticated 401 (0 passed/1 failed). Existing SQLite, Auth, signer, CSR and request helpers reused. Implement additive pinned-HTTPS enrollment routes with session/Origin ownership and canonical bounded CSR; never mount ledger-only disable without active forwarding retirement. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-negative-1791526463970024452.log. D131/D132/D168/D176.
kayg commented 2026-10-09 06:15:20 +00:00
Author
Owner
Copy link

Actual SQLite test rejects sharing endpoint admission with relay-control admission: valid paired control is denied, 0 passed/1 failed. Implement a separate ledger lookup with identical current-role/generation/time checks; neither identity grants host execution. D131/D132/D168. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-negative-1791526489559866421.log.

Actual SQLite test rejects sharing endpoint admission with relay-control admission: valid paired control is denied, 0 passed/1 failed. Implement a separate ledger lookup with identical current-role/generation/time checks; neither identity grants host execution. D131/D132/D168. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-negative-1791526489559866421.log.
kayg commented 2026-10-09 06:30:35 +00:00
Author
Owner
Copy link

Completed step 5e8e6f49a5: The real SQLite regression refuses an endpoint-ledger alias for control. Expose distinct control admission with shared current-role validation, including generation, host removal and leaf revocation/expiry. Fixed SQL statements avoid dynamic per-admission query allocation; reuse role decoding rather than duplicate policy. D131, D132, D168. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-negative-1791526489559866421.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-positive-1791526489559866421.log. Runtime retirement wiring remains pending; no performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-source-fast-1791526489559866421.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.83s]

Completed step 5e8e6f49a5f5c44e0ca1fdd48ce01d59a856e0e0: The real SQLite regression refuses an endpoint-ledger alias for control. Expose distinct control admission with shared current-role validation, including generation, host removal and leaf revocation/expiry. Fixed SQL statements avoid dynamic per-admission query allocation; reuse role decoding rather than duplicate policy. D131, D132, D168. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-negative-1791526489559866421.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-positive-1791526489559866421.log. Runtime retirement wiring remains pending; no performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-source-fast-1791526489559866421.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.83s]
kayg commented 2026-10-09 06:31:21 +00:00
Author
Owner
Copy link

Control admission source committed5e8e6f49; full fast exit0 and verbatim rows alreadyposted. Actual SQLite regression: rejected endpoint/control alias0/1 in0.12s; isolated admission8/0 in0.36s. Raw run3-control-admission-{negative,positive}-1791526489559866421.log. Checks unknown and wrong-purpose fingerprint, certificate time/revocation, role replacement/disable and host removal. Lookup chooses only fixed SQL and shares role decoding. The separate regression-test commit is gating. CA control signer, paired enrollment broker and production relay control remain next.

Control admission source committed5e8e6f49; full fast exit0 and verbatim rows alreadyposted. Actual SQLite regression: rejected endpoint/control alias0/1 in0.12s; isolated admission8/0 in0.36s. Raw run3-control-admission-{negative,positive}-1791526489559866421.log. Checks unknown and wrong-purpose fingerprint, certificate time/revocation, role replacement/disable and host removal. Lookup chooses only fixed SQL and shares role decoding. The separate regression-test commit is gating. CA control signer, paired enrollment broker and production relay control remain next.
kayg commented 2026-10-09 06:38:08 +00:00
Author
Owner
Copy link

Completed step 2561d21035: Keep failing-first real SQLite checks for purpose separation, unknown fingerprint, certificate boundaries/revocation, role replacement/disable and host removal. Verify control-leaf revocation leaves endpoint admission intact. Reject mocked policy outcomes. D131, D132, D168. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-negative-1791526489559866421.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-positive-1791526489559866421.log. Runtime retirement wiring remains pending; no performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-tests-fast-1791526489559866421.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.21s]

Completed step 2561d21035e66987a778c8c7d8054f758bf700a4: Keep failing-first real SQLite checks for purpose separation, unknown fingerprint, certificate boundaries/revocation, role replacement/disable and host removal. Verify control-leaf revocation leaves endpoint admission intact. Reject mocked policy outcomes. D131, D132, D168. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-negative-1791526489559866421.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-positive-1791526489559866421.log. Runtime retirement wiring remains pending; no performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-control-admission-tests-fast-1791526489559866421.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.21s]
kayg commented 2026-10-09 06:38:40 +00:00
Author
Owner
Copy link

Real signed-X509 regression rejects the previous host/control alias: relay-control namespace assertion failed (0 passed, 1 failed). No credential or certificate DER logged. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-negative-1791527888908563002.log. Implement a separate client-only signer purpose; never publish control identity as a host. D131/D132/D168.

Real signed-X509 regression rejects the previous host/control alias: relay-control namespace assertion failed (0 passed, 1 failed). No credential or certificate DER logged. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-negative-1791527888908563002.log. Implement a separate client-only signer purpose; never publish control identity as a host. D131/D132/D168.
kayg commented 2026-10-09 06:55:25 +00:00
Author
Owner
Copy link

Completed step 8efb498777: The rejected host-signer alias failed the real X509 namespace regression. Add a closed certificate-purpose enum and dedicated relay-control URI, preserving server-only endpoint and client-only host purposes. Reuse chain verification and issuer lifetime validation rather than duplicate signing or grant host execution authority. D131, D132, D168. Raw negative /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-negative-1791527888908563002.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-positive-1791527888908563002.log. No production relay wiring or path performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-source-fast-1791527888908563002.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.34s]

Completed step 8efb49877745017ef1c5af1fa52f8c0e655c174f: The rejected host-signer alias failed the real X509 namespace regression. Add a closed certificate-purpose enum and dedicated relay-control URI, preserving server-only endpoint and client-only host purposes. Reuse chain verification and issuer lifetime validation rather than duplicate signing or grant host execution authority. D131, D132, D168. Raw negative /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-negative-1791527888908563002.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-positive-1791527888908563002.log. No production relay wiring or path performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-source-fast-1791527888908563002.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.34s]
kayg commented 2026-10-09 06:56:24 +00:00
Author
Owner
Copy link

Dedicated control signer committed8efb4987; full fast exit0 and verbatim gate rows alreadyposted. Real signed-X509 namespace regression rejected the host-signer alias0/1 in0.02s; dedicated purpose positive4/0/1diagnosticignored in0.02s. Raw run3-ca-control-{negative,positive}-1791527888908563002.log. OpenSSL verifies issuer signature and accepts control only for SSL_CLIENT, refusing SSL_SERVER; pair shares proven key/issuer/lifetime with distinct fingerprint. Static sizes18086768/16123360B. Retained certificate-test commit is gating separately; no production relay listener or path-latency acceptance yet.

Dedicated control signer committed8efb4987; full fast exit0 and verbatim gate rows alreadyposted. Real signed-X509 namespace regression rejected the host-signer alias0/1 in0.02s; dedicated purpose positive4/0/1diagnosticignored in0.02s. Raw run3-ca-control-{negative,positive}-1791527888908563002.log. OpenSSL verifies issuer signature and accepts control only for SSL_CLIENT, refusing SSL_SERVER; pair shares proven key/issuer/lifetime with distinct fingerprint. Static sizes18086768/16123360B. Retained certificate-test commit is gating separately; no production relay listener or path-latency acceptance yet.
kayg commented 2026-10-09 07:00:36 +00:00
Author
Owner
Copy link

Reuse finding: crates/infraternal/src/remote.rs129-201 already implements pinned TLS1.3 HTTPS POST, strict join URL validation, an8192-byte reply limit and AbortConnection cleanup. Relay CLI enrollment should extend that transport through a shared non-generic byte-body helper, preserving existing host joining, rather than add a second HTTP/TLS engine or duplicate connector code. CSR/key proof stays local and capabilities stay in private stdin/control bodies, never argv. Existing run_remote_agent bounded stdin reader is main.rs239-246. New CLI regression draft requires relay --udp-port N, optional TCP port and refusal of zero/overflow/config/token arguments; not run yet. D131/D132/D176. Production client/control wiring remains pending after signed broker gates.

Reuse finding: crates/infraternal/src/remote.rs129-201 already implements pinned TLS1.3 HTTPS POST, strict join URL validation, an8192-byte reply limit and AbortConnection cleanup. Relay CLI enrollment should extend that transport through a shared non-generic byte-body helper, preserving existing host joining, rather than add a second HTTP/TLS engine or duplicate connector code. CSR/key proof stays local and capabilities stay in private stdin/control bodies, never argv. Existing run_remote_agent bounded stdin reader is main.rs239-246. New CLI regression draft requires relay --udp-port N, optional TCP port and refusal of zero/overflow/config/token arguments; not run yet. D131/D132/D176. Production client/control wiring remains pending after signed broker gates.
kayg commented 2026-10-09 07:13:44 +00:00
Author
Owner
Copy link

Completed step 4dd4dacea1: Retain failing-first signed-X509 checks for dedicated control URI, real issuer signature, matching paired key/issuer/lifetime, distinct fingerprint and SSL client/server purpose verification. Reject certificate text-only purpose assertions. D131, D132, D168. Raw negative /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-negative-1791527888908563002.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-positive-1791527888908563002.log. No production relay wiring or path performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-tests-fast-1791527888908563002.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.76s]

Completed step 4dd4dacea12b6f6f6086df6172f0b252ebee269b: Retain failing-first signed-X509 checks for dedicated control URI, real issuer signature, matching paired key/issuer/lifetime, distinct fingerprint and SSL client/server purpose verification. Reject certificate text-only purpose assertions. D131, D132, D168. Raw negative /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-negative-1791527888908563002.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-positive-1791527888908563002.log. No production relay wiring or path performance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-ca-control-tests-fast-1791527888908563002.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.76s]
kayg commented 2026-10-09 07:14:26 +00:00
Author
Owner
Copy link

Real SQLite/CSR/CA broker regression fails when publication reuses the preflight second at the invitation boundary (0 passed/1 failed). Recompute a conservative local publication instant after blocking signing, and recheck role/capability atomically. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-negative-1791530024500556126.log. D131/D132/D168. No signed identity or capability logged.

Real SQLite/CSR/CA broker regression fails when publication reuses the preflight second at the invitation boundary (0 passed/1 failed). Recompute a conservative local publication instant after blocking signing, and recheck role/capability atomically. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-negative-1791530024500556126.log. D131/D132/D168. No signed identity or capability logged.
kayg commented 2026-10-09 07:31:18 +00:00
Author
Owner
Copy link

Broker source full fast failed only fmt: removing temporary test registration left one extra trailing blank line in relay_enrollment.rs85. Native fmt/clippy/test/cargo-deny/static rows all PASS; x86/ARM18086768/16123360B. Raw run3-relay-enrollment-source-fast-1791530024500556126.log. No gate exception claimed or source committed. cargo fmt --all corrected the blank line; rerun the whole gate before source/test commits. Actual broker red0/1 .05s and positive4/0 .25s already recorded, raw run3-relay-enrollment-{negative,positive}-1791530024500556126.log.

Broker source full fast failed only fmt: removing temporary test registration left one extra trailing blank line in relay_enrollment.rs85. Native fmt/clippy/test/cargo-deny/static rows all PASS; x86/ARM18086768/16123360B. Raw run3-relay-enrollment-source-fast-1791530024500556126.log. No gate exception claimed or source committed. cargo fmt --all corrected the blank line; rerun the whole gate before source/test commits. Actual broker red0/1 .05s and positive4/0 .25s already recorded, raw run3-relay-enrollment-{negative,positive}-1791530024500556126.log.
kayg commented 2026-10-09 07:44:46 +00:00
Author
Owner
Copy link

Next control-admission contract (D128/D131/D132): a forwarding grant jointly requires current enabled relay-control/endpoint authority and an ordinary live enrolled peer certificate. Its deadline is min(now+60s, control certificate expiry, peer certificate expiry); never use the sixty-second ceiling to extend certificate authority. Endpoint revocation must refuse new forwarding even when independent control admission remains valid. Use current durable role generation and purpose-separated ledgers; no new crypto/library selection. Prepared actual-SQLite regression draft .scratch/relay-direct/run3-tunnel-grant-regression.rs has ten-second certificate-boundary cases, wrong-purpose/unknown fingerprints, endpoint revocation and role disable. Draft not run; authorize_tunnel API not implemented. Run failing-first after current sole broker writer finishes, then wire authenticated control.

Next control-admission contract (D128/D131/D132): a forwarding grant jointly requires current enabled relay-control/endpoint authority and an ordinary live enrolled peer certificate. Its deadline is min(now+60s, control certificate expiry, peer certificate expiry); never use the sixty-second ceiling to extend certificate authority. Endpoint revocation must refuse new forwarding even when independent control admission remains valid. Use current durable role generation and purpose-separated ledgers; no new crypto/library selection. Prepared actual-SQLite regression draft .scratch/relay-direct/run3-tunnel-grant-regression.rs has ten-second certificate-boundary cases, wrong-purpose/unknown fingerprints, endpoint revocation and role disable. Draft not run; authorize_tunnel API not implemented. Run failing-first after current sole broker writer finishes, then wire authenticated control.
kayg commented 2026-10-09 07:45:39 +00:00
Author
Owner
Copy link

Completed step 398321a714: Bind a bounded locally proven CSR to the invited server-owned relay role and separate endpoint/control signer purposes. Blocking signing cannot freeze token validity: publication conservatively includes elapsed local time and rechecks generation, owner role and capability in the atomic paired store transaction. Refuse signing/commit failures without consuming an invitation. Reject ordinary host certificates as relay-control identities. D131, D132, D168, D176. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-negative-1791530024500556126.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-positive-1791530024500556126.log. CLI/HTTP/active listener wiring and full-route benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-source-fast-1791531094423240051.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.95s]

Completed step 398321a714d617dae5cde2227d07682cbb228c07: Bind a bounded locally proven CSR to the invited server-owned relay role and separate endpoint/control signer purposes. Blocking signing cannot freeze token validity: publication conservatively includes elapsed local time and rechecks generation, owner role and capability in the atomic paired store transaction. Refuse signing/commit failures without consuming an invitation. Reject ordinary host certificates as relay-control identities. D131, D132, D168, D176. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-negative-1791530024500556126.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-positive-1791530024500556126.log. CLI/HTTP/active listener wiring and full-route benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-source-fast-1791531094423240051.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.95s]
kayg commented 2026-10-09 07:47:12 +00:00
Author
Owner
Copy link

Broker source398321a7 committedaftercorrectedfullfast exit0 (verbatimrowsalreadyon103), rawrun3-relay-enrollment-source-fast-1791531094423240051.log. Real CA/CSR/SQLite boundary negative0/1 .05s beforeelapsed-time publication fix; positive4/0 .25s. Rawrun3-relay-enrollment-{negative,positive}-1791530024500556126.log. Covers invalid invitation before signer access, cold-locked signer/malformed proof preserving capability, paired current identity, role disable racing paused signing and last-valid-second expiry. Source uses dedicated client-only control signer; never publishes control as ordinary host. Static18086768/16123360B. Testcommitgatingnow; HTTP/CLI/active control and real300ms/scalebench stillpending.

Broker source398321a7 committedaftercorrectedfullfast exit0 (verbatimrowsalreadyon103), rawrun3-relay-enrollment-source-fast-1791531094423240051.log. Real CA/CSR/SQLite boundary negative0/1 .05s beforeelapsed-time publication fix; positive4/0 .25s. Rawrun3-relay-enrollment-{negative,positive}-1791530024500556126.log. Covers invalid invitation before signer access, cold-locked signer/malformed proof preserving capability, paired current identity, role disable racing paused signing and last-valid-second expiry. Source uses dedicated client-only control signer; never publishes control as ordinary host. Static18086768/16123360B. Testcommitgatingnow; HTTP/CLI/active control and real300ms/scalebench stillpending.
kayg commented 2026-10-09 08:02:24 +00:00
Author
Owner
Copy link

Completed step 70a75bed19: Retain failing-first invitation-boundary evidence with real CA/CSR/SQLite, invalid invitation before signer access, locked signer and malformed CSR preserving capability, committed paired identity admission, and a disable racing paused signing. No mock certificate ledger or real private key files. D131, D132, D168. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-negative-1791530024500556126.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-positive-1791530024500556126.log. CLI/HTTP/active listener wiring and full-route benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-tests-fast-1791531094423240051.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.03s]

Completed step 70a75bed191a97530f9f6fbcbd5e08c244904d69: Retain failing-first invitation-boundary evidence with real CA/CSR/SQLite, invalid invitation before signer access, locked signer and malformed CSR preserving capability, committed paired identity admission, and a disable racing paused signing. No mock certificate ledger or real private key files. D131, D132, D168. Negative /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-negative-1791530024500556126.log, positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-positive-1791530024500556126.log. CLI/HTTP/active listener wiring and full-route benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-enrollment-tests-fast-1791531094423240051.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18086768 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16123360 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.03s]
kayg commented 2026-10-09 08:03:56 +00:00
Author
Owner
Copy link

Current real clap CLI fails the relay --udp-port N regression before implementation (0 passed/1 failed). Retain explicit nonzero UDP/optional TCP settings, no config argument and no token argv. Reuse bounded private-stdin enrollment and the existing pinned HTTPS transport. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-cli-negative-1791532990284447929.log. D131/D132/D176. Main source restored after this negative run.

Current real clap CLI fails the relay --udp-port N regression before implementation (0 passed/1 failed). Retain explicit nonzero UDP/optional TCP settings, no config argument and no token argv. Reuse bounded private-stdin enrollment and the existing pinned HTTPS transport. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-cli-negative-1791532990284447929.log. D131/D132/D176. Main source restored after this negative run.
kayg commented 2026-10-09 08:22:07 +00:00
Author
Owner
Copy link

HTTPsource fullfast failed Clippy needless_pass_by_value in relay_http.rs128 and ARMstatic size16216560B >16196376B by20184B. X86 18204272B passes; fullnative/auth/test/deny/fmt rowsPASS. Rawrun3-relay-http-source-fast-1791533038326236304.log. Noexception/budgetchange/commitclaimed. Earlierprogress incorrectlyassumedClippypassed; actualerrorconfirmed. Fixhelperborrow and reuseexistingremote_http Api/body/admission middleware acrosshost+relayroutes, eliminating duplicate Axum state/middleware instantiations and independent idle state. Recheck realHTTPtests and both static sizes beforewholegate/commit.

HTTPsource fullfast failed Clippy needless_pass_by_value in relay_http.rs128 and ARMstatic size16216560B >16196376B by20184B. X86 18204272B passes; fullnative/auth/test/deny/fmt rowsPASS. Rawrun3-relay-http-source-fast-1791533038326236304.log. Noexception/budgetchange/commitclaimed. Earlierprogress incorrectlyassumedClippypassed; actualerrorconfirmed. Fixhelperborrow and reuseexistingremote_http Api/body/admission middleware acrosshost+relayroutes, eliminating duplicate Axum state/middleware instantiations and independent idle state. Recheck realHTTPtests and both static sizes beforewholegate/commit.
kayg commented 2026-10-09 08:34:09 +00:00
Author
Owner
Copy link

Size correction after the rejected HTTP source gate: relay routes now share remote_http::Api, the existing eight-slot admission and 8192-byte upload middleware; removed the duplicate Axum state/router/middleware instantiations. Error mapper now borrows EnrollmentError. Actual combined HTTP tests7pass0fail1diagnosticignored .27s, raw run3-relay-http-shared-positive.log; all-target server Clippy exit0 in39.21s, raw run3-relay-http-shared-clippy.log. Full source gate still running: run3-relay-http-source-fast-1791534451149628818.log. No HTTP commit yet and no size success claimed. Budgets unchanged. CLI/control/browser/scale work remains pending.

Size correction after the rejected HTTP source gate: relay routes now share remote_http::Api, the existing eight-slot admission and 8192-byte upload middleware; removed the duplicate Axum state/router/middleware instantiations. Error mapper now borrows EnrollmentError. Actual combined HTTP tests7pass0fail1diagnosticignored .27s, raw run3-relay-http-shared-positive.log; all-target server Clippy exit0 in39.21s, raw run3-relay-http-shared-clippy.log. Full source gate still running: run3-relay-http-source-fast-1791534451149628818.log. No HTTP commit yet and no size success claimed. Budgets unchanged. CLI/control/browser/scale work remains pending.
kayg commented 2026-10-09 08:43:41 +00:00
Author
Owner
Copy link

Completed step 9612f01425: The existing HTTP router failed the new invitation route assertion (404 instead of private 401). Add owner/session/Origin-authorized role invitations and canonical bounded CSR joining, with server-selected ports, names and trust metadata. Reuse canonical CSR decoding and migrated signer/store primitives; share the existing eight-slot admission and upload middleware to avoid duplicate Axum state instantiations and ARM size overhead. Return paired certificates only after durable publication and current role recheck. Do not expose disable before active runtime retirement. Reject public token URLs and host/control identity reuse. D131, D132, D168, D176. Negative run3-relay-http-negative-1791526463970024452.log; positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-shared-positive.log. CLI, active role control, full browser path and scale benchmarks still pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-source-fast-1791534451149628818.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.18s]

Completed step 9612f01425f5128c2ecd5406b405b480259a3c56: The existing HTTP router failed the new invitation route assertion (404 instead of private 401). Add owner/session/Origin-authorized role invitations and canonical bounded CSR joining, with server-selected ports, names and trust metadata. Reuse canonical CSR decoding and migrated signer/store primitives; share the existing eight-slot admission and upload middleware to avoid duplicate Axum state instantiations and ARM size overhead. Return paired certificates only after durable publication and current role recheck. Do not expose disable before active runtime retirement. Reject public token URLs and host/control identity reuse. D131, D132, D168, D176. Negative run3-relay-http-negative-1791526463970024452.log; positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-shared-positive.log. CLI, active role control, full browser path and scale benchmarks still pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-source-fast-1791534451149628818.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.18s]
kayg commented 2026-10-09 08:45:25 +00:00
Author
Owner
Copy link

HTTP source 9612f014 passed every fast-gate row: x86 18,158,928B, ARM16,180,144B, TS15pass/0fail/34,537assertions; exact output recorded in prior completion comment. Separate tests gate run3-relay-http-tests-fast-1791534451149628818.log found Clippy too_many_lines113/100 at remote_http_tests.rs325. Stopped that owned gate scope after the deterministic error; no test commit or stability exception. Extracted refused CSR checks into a borrowed helper, preserving same-capability successful redemption. Running all-target Clippy and real HTTP tests before a fresh full test gate. No production CLI/control/path acceptance claimed.

HTTP source 9612f014 passed every fast-gate row: x86 18,158,928B, ARM16,180,144B, TS15pass/0fail/34,537assertions; exact output recorded in prior completion comment. Separate tests gate run3-relay-http-tests-fast-1791534451149628818.log found Clippy too_many_lines113/100 at remote_http_tests.rs325. Stopped that owned gate scope after the deterministic error; no test commit or stability exception. Extracted refused CSR checks into a borrowed helper, preserving same-capability successful redemption. Running all-target Clippy and real HTTP tests before a fresh full test gate. No production CLI/control/path acceptance claimed.
kayg commented 2026-10-09 08:47:03 +00:00
Author
Owner
Copy link

Owner decision D198 (main eecf6ace): every role (relay, mesh, SSH, backup, agent, server) is a plugin behind infraternal-plugin-api — a cargo feature per plugin (removable at build time), switchable per node at run time with nothing left running when off, actions/settings/defaults/role dependencies declared, not hand-wired. Build new work in that form.

Owner decision D198 (main eecf6ace): every role (relay, mesh, SSH, backup, agent, server) is a plugin behind infraternal-plugin-api — a cargo feature per plugin (removable at build time), switchable per node at run time with nothing left running when off, actions/settings/defaults/role dependencies declared, not hand-wired. Build new work in that form.
kayg commented 2026-10-09 08:52:06 +00:00
Author
Owner
Copy link

Completed step 748ba1ec77: Retain the locally failing real-router regression, using existing request/Auth/SQLite/CSR helpers. Verify unauthenticated and hostile-Origin refusal, private server-selected invitation metadata, proven-key and separate-namespace certificates, and one-use redemption. Reject mocked signing or invented endpoint success. D131, D132, D168, D176. Negative run3-relay-http-negative-1791526463970024452.log; positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-tests-positive-1791535509901275659.log. CLI, active role control, full browser path and scale benchmarks still pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-tests-fast-1791535509901275659.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [5.68s]

Completed step 748ba1ec77edc1a193f9ffb604aee7026dda8905: Retain the locally failing real-router regression, using existing request/Auth/SQLite/CSR helpers. Verify unauthenticated and hostile-Origin refusal, private server-selected invitation metadata, proven-key and separate-namespace certificates, and one-use redemption. Reject mocked signing or invented endpoint success. D131, D132, D168, D176. Negative run3-relay-http-negative-1791526463970024452.log; positive /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-tests-positive-1791535509901275659.log. CLI, active role control, full browser path and scale benchmarks still pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-http-tests-fast-1791535509901275659.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [5.68s]
kayg commented 2026-10-09 08:54:02 +00:00
Author
Owner
Copy link

Actual SQLite regression rejects a fixed sixty-second grant when either certificate has only ten seconds left. Minimal existing-admission implementation failed0/1 before deadline fix; no production grant API installed. Implement joint current relay-control/endpoint and ordinary-peer admission with earliest certificate expiry and role generation. Source/test files restored. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-negative-1791535978227163518.log. D128/D131/D132.

Actual SQLite regression rejects a fixed sixty-second grant when either certificate has only ten seconds left. Minimal existing-admission implementation failed0/1 before deadline fix; no production grant API installed. Implement joint current relay-control/endpoint and ordinary-peer admission with earliest certificate expiry and role generation. Source/test files restored. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-negative-1791535978227163518.log. D128/D131/D132.
kayg commented 2026-10-09 09:11:00 +00:00
Author
Owner
Copy link

Completed step 08b2354683: Read relay-control identity, its same-key same-issuer current-generation endpoint and the ordinary host peer in one SQLite snapshot before forwarding allocation. Cap the sixty-second lease by all three certificate deadlines; reject disabled roles and any revoked identity. Separate-purpose ledgers prevent host or endpoint certificates from obtaining relay-control authority. Reject independent reads that can observe different generations and a fixed sixty-second cap that outlives a certificate. This public-record query does not replace active runtime retirement subscriptions. D128, D131, D132, D176. Real targeted tests: /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-positive-1791536046962700727.log. Listener/control/CLI wiring and full-path/scale benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-source-fast-1791536046962700727.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.71s]

Completed step 08b235468386c52586f6c99750f28699738453cf: Read relay-control identity, its same-key same-issuer current-generation endpoint and the ordinary host peer in one SQLite snapshot before forwarding allocation. Cap the sixty-second lease by all three certificate deadlines; reject disabled roles and any revoked identity. Separate-purpose ledgers prevent host or endpoint certificates from obtaining relay-control authority. Reject independent reads that can observe different generations and a fixed sixty-second cap that outlives a certificate. This public-record query does not replace active runtime retirement subscriptions. D128, D131, D132, D176. Real targeted tests: /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-positive-1791536046962700727.log. Listener/control/CLI wiring and full-path/scale benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-source-fast-1791536046962700727.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.71s]
kayg commented 2026-10-09 09:17:12 +00:00
Author
Owner
Copy link

Read new owner D198 from main:DECISIONS.md1500-1517 and existing contracts/plugin.md. Local infraternal-plugin-api/src/lib.rs and plugins/ssh/src/lib.rs are descriptor/interface placeholders; there is no crates/infraternal/src/plugins.rs registry yet. Production relay work will follow the existing static/versioned PluginDescriptor+ActionDef contract: own optional plugin crate/Cargo feature, declared permissions/settings/defaults/dependencies, generated adapters, and per-node up/down/status with full listener/task retirement. Preserve relay --udp-port compatibility as an adapter of the same declared up action, alongside D199 role commands; no independent hand-wired relay dispatcher. Current joint-authority test gate can finish independently. No new foundational library or engine chosen. HTTP/ledger/admission prerequisites are committed; production listener/control/UI and a-c/d acceptance remain pending.

Read new owner D198 from main:DECISIONS.md1500-1517 and existing contracts/plugin.md. Local infraternal-plugin-api/src/lib.rs and plugins/ssh/src/lib.rs are descriptor/interface placeholders; there is no crates/infraternal/src/plugins.rs registry yet. Production relay work will follow the existing static/versioned PluginDescriptor+ActionDef contract: own optional plugin crate/Cargo feature, declared permissions/settings/defaults/dependencies, generated adapters, and per-node up/down/status with full listener/task retirement. Preserve relay --udp-port compatibility as an adapter of the same declared up action, alongside D199 role commands; no independent hand-wired relay dispatcher. Current joint-authority test gate can finish independently. No new foundational library or engine chosen. HTTP/ledger/admission prerequisites are committed; production listener/control/UI and a-c/d acceptance remain pending.
kayg commented 2026-10-09 09:19:05 +00:00
Author
Owner
Copy link

Completed step d1f8609a5a: Preserve the failing-first real SQLite deadline regression: either control or peer certificate with ten seconds remaining must cap the forwarding grant at ten seconds. Verify exact-boundary refusal, wrong-purpose and unknown identities, endpoint revocation despite a still-valid control identity, and disabled role rejection. Reject a mocked authority ledger or a lease that independently prolongs revocation. D128, D131, D132, D176. Real targeted tests: /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-positive-1791536046962700727.log. Listener/control/CLI wiring and full-path/scale benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-tests-fast-1791536046962700727.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.92s]

Completed step d1f8609a5ab3160f9906d54abdd4cc7d6637c923: Preserve the failing-first real SQLite deadline regression: either control or peer certificate with ten seconds remaining must cap the forwarding grant at ten seconds. Verify exact-boundary refusal, wrong-purpose and unknown identities, endpoint revocation despite a still-valid control identity, and disabled role rejection. Reject a mocked authority ledger or a lease that independently prolongs revocation. D128, D131, D132, D176. Real targeted tests: /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-positive-1791536046962700727.log. Listener/control/CLI wiring and full-path/scale benchmarks remain pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-tunnel-grant-tests-fast-1791536046962700727.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.92s]
kayg commented 2026-10-09 09:22:52 +00:00
Author
Owner
Copy link

Additive D198/plugin-contract implementation proposal before production migration: implement the already-described static/versioned PluginDescriptor and ActionDef metadata in infraternal-plugin-api (contracts/plugin.md sections2-3), preserving action/schema/permission/effect/dry-run definitions once. Add role metadata for default settings, allowed role dependencies and references to the same up/down/status action IDs; core registration validates unique plugin/action identifiers, API version and destructive dry-run requirements. The relay implementation becomes an optional own crate with agent/server parts and one descriptor, discovered by the generic compiled registry rather than a relay-specific core dispatcher. Build without the relay feature omits registration/listeners/UI; per-node off retires its tasks and transports before acknowledgement. No WASM host, foundational schema engine, crypto or transport library is selected here; schema references remain part of the existing static contract. Existing frozen terminal framing/writers and persisted public authority remain compatible. Independent grant tests committedd1f8609a and exact gate posted; enrollment-parser fuzz preparation is now gating.

Additive D198/plugin-contract implementation proposal before production migration: implement the already-described static/versioned PluginDescriptor and ActionDef metadata in infraternal-plugin-api (contracts/plugin.md sections2-3), preserving action/schema/permission/effect/dry-run definitions once. Add role metadata for default settings, allowed role dependencies and references to the same up/down/status action IDs; core registration validates unique plugin/action identifiers, API version and destructive dry-run requirements. The relay implementation becomes an optional own crate with agent/server parts and one descriptor, discovered by the generic compiled registry rather than a relay-specific core dispatcher. Build without the relay feature omits registration/listeners/UI; per-node off retires its tasks and transports before acknowledgement. No WASM host, foundational schema engine, crypto or transport library is selected here; schema references remain part of the existing static contract. Existing frozen terminal framing/writers and persisted public authority remain compatible. Independent grant tests committedd1f8609a and exact gate posted; enrollment-parser fuzz preparation is now gating.
kayg commented 2026-10-09 09:28:39 +00:00
Author
Owner
Copy link

Completed step bb87655e8a: Expose the already-selected workspace serde_json dependency directly to the existing fuzz harness. No new codec/foundation/version is selected; root lock changes only the harness dependency list. Keep dependency registration separate from target code. D40, D131, D132. No production relay/path benchmark acceptance is implied. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-json-fuzz-dependency-fast-1791537640219997836.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.84s]

Completed step bb87655e8af91865f233c6825c21942c2feacff8: Expose the already-selected workspace serde_json dependency directly to the existing fuzz harness. No new codec/foundation/version is selected; root lock changes only the harness dependency list. Keep dependency registration separate from target code. D40, D131, D132. No production relay/path benchmark acceptance is implied. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-json-fuzz-dependency-fast-1791537640219997836.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.84s]
kayg commented 2026-10-09 09:31:13 +00:00
Author
Owner
Copy link

Fuzz codec dependency committedbb87655e after fullfastPASS; exactoutputalreadyposted. First fuzz startup stopped beforecompilation/iterations: Rust version string does not start with rustc; rawrun3-relay-json-fuzz-1791537640219997836.log. rustup which --toolchain nightly rustc reports compiler not installed despite its component manifest. Existing dated nightly-2026-10-01 rustc1.101.0-nightly21b707e3f2026-09-30 works. Resumed explicitly with that already-installed toolchain, same8KiB/600sec/2048MiB fuzz bounds, rawrun3-relay-json-fuzz-1791538238195314457.log; targetnotcommittedyet. No toolchain install, budget change or fuzz-success claim.

Fuzz codec dependency committedbb87655e after fullfastPASS; exactoutputalreadyposted. First fuzz startup stopped beforecompilation/iterations: Rust version string does not start with rustc; rawrun3-relay-json-fuzz-1791537640219997836.log. rustup which --toolchain nightly rustc reports compiler not installed despite its component manifest. Existing dated nightly-2026-10-01 rustc1.101.0-nightly21b707e3f2026-09-30 works. Resumed explicitly with that already-installed toolchain, same8KiB/600sec/2048MiB fuzz bounds, rawrun3-relay-json-fuzz-1791538238195314457.log; targetnotcommittedyet. No toolchain install, budget change or fuzz-success claim.
kayg commented 2026-10-09 09:36:50 +00:00
Author
Owner
Copy link

Datednightly fuzz build progressed but failed beforeiterations because vendoredOpenSSL could notfind make. Rawrun3-relay-json-fuzz-1791538238195314457.log. Retrying under repository-standard Nix buildtools gnumake/perl/zig0.15 with existingnightly2026-10-01 and identical fuzzbounds; active rawrun3-relay-json-fuzz-1791538532992336744.log. Parser target remainsuncommitted; no fuzz-success/coverage claim.

Datednightly fuzz build progressed but failed beforeiterations because vendoredOpenSSL could notfind make. Rawrun3-relay-json-fuzz-1791538238195314457.log. Retrying under repository-standard Nix buildtools gnumake/perl/zig0.15 with existingnightly2026-10-01 and identical fuzzbounds; active rawrun3-relay-json-fuzz-1791538532992336744.log. Parser target remainsuncommitted; no fuzz-success/coverage claim.
kayg commented 2026-10-09 09:48:32 +00:00
Author
Owner
Copy link

D198 interface preview only (uninstalled): existing static descriptor contract plus additive per-node lifecycle/default/dependency metadata; API major0 follows current plugin-api package0.1.0, independent of wirev1. An unchecked constructor failed the destructive-action mandatory-dry-run regression0/1 .01s; actual validator passed6/0 .01s, strict standaloneClippy pedantic/warningsPASS. Rawrun3-plugin-api-negative.log/run3-plugin-api-positive.log/run3-plugin-api-preview-clippy.log. Covers namespace/duplicate IDs/API mismatch, host-scoped distinct lifecycle/effects and self/repeated role dependencies. Optional dependency declarations do not load absent roles; immutable registry starts no IO/tasks/timers. Source/tests onlyin.scratch/run3-plugin-{descriptor,registry,registry-tests}.rs; no interface/runtime installation, action-handler generation or production relay completion implied. Enrollment fuzz currentlyactive separately; no benchmark run alongsidebuild.

D198 interface preview only (uninstalled): existing static descriptor contract plus additive per-node lifecycle/default/dependency metadata; API major0 follows current plugin-api package0.1.0, independent of wirev1. An unchecked constructor failed the destructive-action mandatory-dry-run regression0/1 .01s; actual validator passed6/0 .01s, strict standaloneClippy pedantic/warningsPASS. Rawrun3-plugin-api-negative.log/run3-plugin-api-positive.log/run3-plugin-api-preview-clippy.log. Covers namespace/duplicate IDs/API mismatch, host-scoped distinct lifecycle/effects and self/repeated role dependencies. Optional dependency declarations do not load absent roles; immutable registry starts no IO/tasks/timers. Source/tests onlyin.scratch/run3-plugin-{descriptor,registry,registry-tests}.rs; no interface/runtime installation, action-handler generation or production relay completion implied. Enrollment fuzz currentlyactive separately; no benchmark run alongsidebuild.
kayg commented 2026-10-09 09:53:03 +00:00
Author
Owner
Copy link

Relay JSON fuzz run completed locally on infraternal-dev; 8 KiB input ceiling; 600-second requested fuzz window; synthetic corpus, no signing/network/authority publication. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-json-fuzz-1791538532992336744.log. Current load 15.50 19.11 19.52 6/3844 4160390. Verbatim completion:
#19495932 DONE cov: 2089 ft: 7454 corp: 2703/220Kb lim: 8192 exec/s: 32439 rss: 565Mb
Done 19495932 runs in 601 second(s)

Relay JSON fuzz run completed locally on infraternal-dev; 8 KiB input ceiling; 600-second requested fuzz window; synthetic corpus, no signing/network/authority publication. Raw /home/kayg/.cache/codex-scratch/relay-direct/run3-relay-json-fuzz-1791538532992336744.log. Current load 15.50 19.11 19.52 6/3844 4160390. Verbatim completion: #19495932 DONE cov: 2089 ft: 7454 corp: 2703/220Kb lim: 8192 exec/s: 32439 rss: 565Mb Done 19495932 runs in 601 second(s)
kayg commented 2026-10-09 10:00:37 +00:00
Author
Owner
Copy link

Run 4 started, sole writer on relay-direct at base/head 860297d59bfb80505b7234c2cf69d12019d47aa6. Read #103 history (140 comments), parent #5, repository contract, architecture/process and D192/D198 brief. Resume WIP fuzz target: previous instrumented run completed 19,495,932 executions in 601 seconds (recorded run3 log). Fold target and registration into an atomic gated commit, then implement the existing static plugin contract with additive lifecycle metadata and optional relay crate before production wiring. Benchmarks local CPUWeight=10000; browser port4803; no push/deploy or other hosts. Local DECISIONS ends at D112; later decisions are supplied by the binding brief/issue, not edited by this job.

Run 4 started, sole writer on relay-direct at base/head 860297d59bfb80505b7234c2cf69d12019d47aa6. Read #103 history (140 comments), parent #5, repository contract, architecture/process and D192/D198 brief. Resume WIP fuzz target: previous instrumented run completed 19,495,932 executions in 601 seconds (recorded run3 log). Fold target and registration into an atomic gated commit, then implement the existing static plugin contract with additive lifecycle metadata and optional relay crate before production wiring. Benchmarks local CPUWeight=10000; browser port4803; no push/deploy or other hosts. Local DECISIONS ends at D112; later decisions are supplied by the binding brief/issue, not edited by this job.
kayg commented 2026-10-09 10:02:11 +00:00
Author
Owner
Copy link

WIP target fast gate found 15 Clippy manual_assert_eq errors (fuzz/fuzz_targets/relay_enrollment_json.rs). The assertions deliberately suppress capability/body contents on panic; retain boolean assertions with a documented target-local lint allowance rather than assert_eq diagnostics. Stopped only owned gate scope run-p34440-i21002632.scope after this deterministic error; no gate success claimed. Raw run4-fuzz-fast.log. Rerun full gate after correction.

WIP target fast gate found 15 Clippy manual_assert_eq errors (fuzz/fuzz_targets/relay_enrollment_json.rs). The assertions deliberately suppress capability/body contents on panic; retain boolean assertions with a documented target-local lint allowance rather than assert_eq diagnostics. Stopped only owned gate scope run-p34440-i21002632.scope after this deterministic error; no gate success claimed. Raw run4-fuzz-fast.log. Rerun full gate after correction.
kayg commented 2026-10-09 10:09:24 +00:00
Author
Owner
Copy link

Completed fuzz at 7b21ed452c. Previous dated-nightly instrumented run completed 19,495,932 executions in 601 seconds, cov2089/ft7454, maximum input8192 bytes; raw run3-relay-json-fuzz-1791538532992336744.log. Self-review: bounded parser only, no logging, signing or authority publication. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-fuzz-fast-corrected.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [167.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.86s]

Completed fuzz at 7b21ed452ced878bf6c3d3c68118fa3783ae6f38. Previous dated-nightly instrumented run completed 19,495,932 executions in 601 seconds, cov2089/ft7454, maximum input8192 bytes; raw run3-relay-json-fuzz-1791538532992336744.log. Self-review: bounded parser only, no logging, signing or authority publication. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-fuzz-fast-corrected.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [167.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.86s]
kayg commented 2026-10-09 10:15:57 +00:00
Author
Owner
Copy link

Completed plugin-descriptor at cdb197047e. Descriptor metadata only; no listeners, handlers or generated adapters are installed by this slice. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-plugin-descriptor-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [225.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.01s]

Completed plugin-descriptor at cdb197047e57c7933d8a60f48d7e0c9f38ef4801. Descriptor metadata only; no listeners, handlers or generated adapters are installed by this slice. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-plugin-descriptor-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [225.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.01s]
kayg commented 2026-10-09 10:22:49 +00:00
Author
Owner
Copy link

Completed plugin-registry at 5c962dd7ee. Prior standalone regression logs run3-plugin-api-negative.log and run3-plugin-api-positive.log; immutable lookup starts no tasks or timers. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-plugin-registry-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [194.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.89s]

Completed plugin-registry at 5c962dd7ee6a74c8a8e643b58b63f305f24bb0e4. Prior standalone regression logs run3-plugin-api-negative.log and run3-plugin-api-positive.log; immutable lookup starts no tasks or timers. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-plugin-registry-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [194.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.89s]
kayg commented 2026-10-09 10:30:36 +00:00
Author
Owner
Copy link

Completed plugin-registry-tests at 91a74f27c0. Six registry tests use the production validator; no runtime activation or production relay acceptance claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-plugin-registry-tests-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [481.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [6.19s]

Completed plugin-registry-tests at 91a74f27c0b1503195a6d34055b605289d528b07. Six registry tests use the production validator; no runtime activation or production relay acceptance claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-plugin-registry-tests-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [481.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [6.19s]
kayg commented 2026-10-09 10:36:37 +00:00
Author
Owner
Copy link

Shared-fixture gate with two Rust test threads failed the existing remote_client::tests::winning_quic_drops_the_unfinished_tls_dial_before_control_dispatch: AddrInUse at remote_client_tests.rs179, 56pass/1fail/8ignored in9.57s. No fixture commit or baseline exception accepted. Raw run4-shared-transport-fixture-fast.log; shared load38.68/41.26/34.79. Stopped only owned scope run-p296942-i21265050.scope, returning future gates to one test thread and repeating full gate unchanged. No production network behavior or timeout changes.

Shared-fixture gate with two Rust test threads failed the existing remote_client::tests::winning_quic_drops_the_unfinished_tls_dial_before_control_dispatch: AddrInUse at remote_client_tests.rs179, 56pass/1fail/8ignored in9.57s. No fixture commit or baseline exception accepted. Raw run4-shared-transport-fixture-fast.log; shared load38.68/41.26/34.79. Stopped only owned scope run-p296942-i21265050.scope, returning future gates to one test thread and repeating full gate unchanged. No production network behavior or timeout changes.
kayg commented 2026-10-09 10:51:25 +00:00
Author
Owner
Copy link

Completed shared-transport-fixture at 2c8184fbd4. Same generated certificate fixtures and real transport assertions; no production transport or signing behavior change. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-shared-transport-fixture-serial-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [186.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.41s]

Completed shared-transport-fixture at 2c8184fbd483310fed7dcd98d8de06f54767ddc6. Same generated certificate fixtures and real transport assertions; no production transport or signing behavior change. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-shared-transport-fixture-serial-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [186.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.41s]
kayg commented 2026-10-09 11:00:25 +00:00
Author
Owner
Copy link

Completed relay-plugin-package at 3672268c42. Package-only slice; default and --no-default-features checks pass. Raw run4-plugin-package-check.log. Runtime registration and activation remain subsequent work. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-plugin-package-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [165.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.54s]

Completed relay-plugin-package at 3672268c426eb3955d8afacbf6f429e742ccac5b. Package-only slice; default and --no-default-features checks pass. Raw run4-plugin-package-check.log. Runtime registration and activation remain subsequent work. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-plugin-package-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [165.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.54s]
kayg commented 2026-10-09 11:14:49 +00:00
Author
Owner
Copy link

Completed relay-forward-plugin at 7e06ca14f3. Nine real socket/deadline tests retain exact 1MiB forwarding and independent inner TLS identity evidence. Relocation only; no production runtime activation claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-forward-plugin-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [229.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.98s]

Completed relay-forward-plugin at 7e06ca14f3ccda7ae252319fa8e4804178f370da. Nine real socket/deadline tests retain exact 1MiB forwarding and independent inner TLS identity evidence. Relocation only; no production runtime activation claimed. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-forward-plugin-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [229.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.98s]
kayg commented 2026-10-09 11:16:07 +00:00
Author
Owner
Copy link

Relay descriptor choice before implementation [D198]: host-scoped relay.role.up/down/status share permissions/settings metadata in the own crate. Extra relays default off. UDP default7445 avoids native server UDP/TCP7444; actual server-selected enrollment settings override the default. Optional existing infraternal agent management is declared without enabling ordinary execution on relay-only nodes. TCP is a declared future setting; runtime must refuse it until implemented. This is metadata preparation, not startup/CLI/route completion.

Relay descriptor choice before implementation [D198]: host-scoped relay.role.up/down/status share permissions/settings metadata in the own crate. Extra relays default off. UDP default7445 avoids native server UDP/TCP7444; actual server-selected enrollment settings override the default. Optional existing infraternal agent management is declared without enabling ordinary execution on relay-only nodes. TCP is a declared future setting; runtime must refuse it until implemented. This is metadata preparation, not startup/CLI/route completion.
kayg commented 2026-10-09 11:31:09 +00:00
Author
Owner
Copy link

Completed relay-tunnel-plugin at eb95e0860e. --no-default-features package check passes, raw run4-relay-plugin-no-feature.log. No host-core relay_forward or relay_tunnel module remains. Server ledger/control/CLI and route acceptance still pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-tunnel-plugin-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [267.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.55s]

Completed relay-tunnel-plugin at eb95e0860e946bfd5b734d703deb8158246d75bf. --no-default-features package check passes, raw run4-relay-plugin-no-feature.log. No host-core relay_forward or relay_tunnel module remains. Server ledger/control/CLI and route acceptance still pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-tunnel-plugin-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [267.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.55s]
kayg commented 2026-10-09 11:39:41 +00:00
Author
Owner
Copy link

Completed relay-descriptor at de3834071e. Descriptor only: explicit-off default, nonzero port schema and optional infraternal agent management dependency. No runtime action handler or CLI success is implied; production control/route acceptance remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-descriptor-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [208.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.91s]

Completed relay-descriptor at de3834071e31b3f07ac784b5a6b2ca20a4f5d72e. Descriptor only: explicit-off default, nonzero port schema and optional infraternal agent management dependency. No runtime action handler or CLI success is implied; production control/route acceptance remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-descriptor-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [208.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.91s]
kayg commented 2026-10-09 11:44:22 +00:00
Author
Owner
Copy link

Relay descriptor source de383407 passed full fast. Actual no-feature registration test initially failed to compile (E0716 at descriptor_tests.rs7): temporary array cannot satisfy Registry::new static lifetime. Raw run4-relay-descriptor-no-feature-tests.log. Use an explicit static compiled list, matching the production registry ownership model; no API relaxation, transport or authority change. Test slice remains uncommitted until its no-feature test and full gate pass.

Relay descriptor source de383407 passed full fast. Actual no-feature registration test initially failed to compile (E0716 at descriptor_tests.rs7): temporary array cannot satisfy Registry::new static lifetime. Raw run4-relay-descriptor-no-feature-tests.log. Use an explicit static compiled list, matching the production registry ownership model; no API relaxation, transport or authority change. Test slice remains uncommitted until its no-feature test and full gate pass.
kayg commented 2026-10-09 11:51:20 +00:00
Author
Owner
Copy link

Completed relay-descriptor-tests at be7ac5ee28. Actual descriptor registration passes with --no-default-features (one test), raw run4-relay-descriptor-no-feature-tests.log. This does not claim runtime quiescence or end-to-end relay startup. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-descriptor-tests-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [140.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.09s]

Completed relay-descriptor-tests at be7ac5ee28a933f8b5191ffbb8ddcebf5bd89d2d. Actual descriptor registration passes with --no-default-features (one test), raw run4-relay-descriptor-no-feature-tests.log. This does not claim runtime quiescence or end-to-end relay startup. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-descriptor-tests-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [140.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.09s]
kayg commented 2026-10-09 11:57:40 +00:00
Author
Owner
Copy link

Run 4 checkpoint, not completion of #103. Branch relay-direct, head be7ac5ee28; ten atomic commits since bb87655e (WIP860297d5 soft-reset and folded into parser7b21ed45). Every committed slice passed full fast. No push/deploy or other benchmark hosts.

Landed: bounded enrollment parser target; static plugin descriptors/registry and six regressions; shared synthetic certificate support; optional crates/plugins/relay; forwarding and ordered tunnel modules removed from host core; actual relay descriptor and registration test. Relay crate tests10pass/0fail in1.85s (nine socket/deadline cases plus descriptor); --no-default-features registration1pass/0fail. Normal feature-off dependency tree contains only infraternal-plugin-api (run4-feature-off-tree.log). Previous instrumented parser evidence remains19,495,932 executions/601s/max8192B, raw run3-relay-json-fuzz-1791538532992336744.log; no new fuzz iteration count claimed. Explicit-off metadata, optional ordinary host management and no transport registration side effects were self-reviewed. Forwarding preserves fixed destination, current authority, earliest deadlines, bounds and independent inner TLS identity.

Unfinished: server relay ledger/enrollment/router still in core; generic server/plugin registration and authenticated control; runtime up/down/status/host management; production inventory/probes; per-writer route handover (do not replace RemoteHosts primary source); str0m adapter off by default; browser-server-relay-infraternal-agent RTT300 and D170 CPU/RSS/overhead/switch acceptance. CLI relay remains unimplemented. A metadata test is not runtime-off proof. Continue with an additive generic server service/router extension to move server parts into the plugin while retaining the shared enrollment admission/body middleware (prior ARM duplication regression is on this issue). Reuse current pinned HTTPS client and dedicated control/endpoint ledgers; never grant control ordinary host execution.

Two rejected attempts recorded above: parser manual_assert_eq lint; test-thread2 TCP/UDP port collision. Serial complete retry passed, with no timeout/production change or baseline exception. Actual descriptor test temporary-list lifetime corrected to explicit static registrations. All owned gate writers finished. Job caches and source drafts are retained for this unfinished task; no new temporary checkout was created. Source tree has only pre-existing untracked .scratch/, scripts/pycache/, target symlink.

Final completed slice scripts/verify fast exit0, raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-descriptor-tests-fast.log. Verbatim output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)

15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.09s]

Run 4 checkpoint, not completion of #103. Branch relay-direct, head be7ac5ee28a933f8b5191ffbb8ddcebf5bd89d2d; ten atomic commits since bb87655e (WIP860297d5 soft-reset and folded into parser7b21ed45). Every committed slice passed full fast. No push/deploy or other benchmark hosts. Landed: bounded enrollment parser target; static plugin descriptors/registry and six regressions; shared synthetic certificate support; optional crates/plugins/relay; forwarding and ordered tunnel modules removed from host core; actual relay descriptor and registration test. Relay crate tests10pass/0fail in1.85s (nine socket/deadline cases plus descriptor); --no-default-features registration1pass/0fail. Normal feature-off dependency tree contains only infraternal-plugin-api (run4-feature-off-tree.log). Previous instrumented parser evidence remains19,495,932 executions/601s/max8192B, raw run3-relay-json-fuzz-1791538532992336744.log; no new fuzz iteration count claimed. Explicit-off metadata, optional ordinary host management and no transport registration side effects were self-reviewed. Forwarding preserves fixed destination, current authority, earliest deadlines, bounds and independent inner TLS identity. Unfinished: server relay ledger/enrollment/router still in core; generic server/plugin registration and authenticated control; runtime up/down/status/host management; production inventory/probes; per-writer route handover (do not replace RemoteHosts primary source); str0m adapter off by default; browser-server-relay-infraternal-agent RTT300 and D170 CPU/RSS/overhead/switch acceptance. CLI relay remains unimplemented. A metadata test is not runtime-off proof. Continue with an additive generic server service/router extension to move server parts into the plugin while retaining the shared enrollment admission/body middleware (prior ARM duplication regression is on this issue). Reuse current pinned HTTPS client and dedicated control/endpoint ledgers; never grant control ordinary host execution. Two rejected attempts recorded above: parser manual_assert_eq lint; test-thread2 TCP/UDP port collision. Serial complete retry passed, with no timeout/production change or baseline exception. Actual descriptor test temporary-list lifetime corrected to explicit static registrations. All owned gate writers finished. Job caches and source drafts are retained for this unfinished task; no new temporary checkout was created. Source tree has only pre-existing untracked .scratch/, scripts/__pycache__/, target symlink. Final completed slice scripts/verify fast exit0, raw /home/kayg/.cache/codex-scratch/relay-direct/run4-relay-descriptor-tests-fast.log. Verbatim output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.09s]
kayg commented 2026-10-09 13:07:39 +00:00
Author
Owner
Copy link

Run 5 started as sole writer on relay-direct at base/head be7ac5ee28. Read #103 and parent #5 history, CLAUDE.md, DECISIONS.md, architecture/process and run-4 checkpoint. Continue D198 server-side plugin migration using an additive generic server service interface, retaining shared eight-slot/8192-byte/5-second enrollment middleware and separate control/host authority. Binding D192/D196/D197/D201/D202 supplied by brief; local decisions end at112. No push/deploy or other hosts. Local benchmarks CPUWeight10000; browser port4803; HOME-cache scratch. Production startup/control, routing, switching and str0m/acceptance remain the requested subsequent steps.

Run 5 started as sole writer on relay-direct at base/head be7ac5ee28a933f8b5191ffbb8ddcebf5bd89d2d. Read #103 and parent #5 history, CLAUDE.md, DECISIONS.md, architecture/process and run-4 checkpoint. Continue D198 server-side plugin migration using an additive generic server service interface, retaining shared eight-slot/8192-byte/5-second enrollment middleware and separate control/host authority. Binding D192/D196/D197/D201/D202 supplied by brief; local decisions end at112. No push/deploy or other hosts. Local benchmarks CPUWeight10000; browser port4803; HOME-cache scratch. Production startup/control, routing, switching and str0m/acceptance remain the requested subsequent steps.
kayg commented 2026-10-09 13:08:13 +00:00
Author
Owner
Copy link

Additive D198 server-service seam before migration: a dependency-free ServerService trait in infraternal-plugin-api uses adapter-owned context/request/response types and references the existing descriptor. The core retains bounded upload/admission/deadline middleware; plugins declare their compatibility enrollment paths and own dispatch. Runtime validates namespace/duplicate paths against the compiled descriptor before serving. No new web/schema/database/transport engine, signer unlock, host authority or frozen wire change. Relay store/broker/HTTP move to the server feature of crates/plugins/relay; binary composition selects compiled services, so core does not depend on a relay implementation.

Additive D198 server-service seam before migration: a dependency-free ServerService trait in infraternal-plugin-api uses adapter-owned context/request/response types and references the existing descriptor. The core retains bounded upload/admission/deadline middleware; plugins declare their compatibility enrollment paths and own dispatch. Runtime validates namespace/duplicate paths against the compiled descriptor before serving. No new web/schema/database/transport engine, signer unlock, host authority or frozen wire change. Relay store/broker/HTTP move to the server feature of crates/plugins/relay; binary composition selects compiled services, so core does not depend on a relay implementation.
kayg commented 2026-10-09 13:14:47 +00:00
Author
Owner
Copy link

Completed run5-server-api at 5a77725613. Dependency-free interface only, no activated handlers/listeners. Author review: object-safe lifetime/cancellation contract; private request types need no Debug. Six existing registry tests remain green. Production relay acceptance remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-api-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [137.00ms]
$ bun run build.ts
built main.js, chunk-0gjma2ke.js, chunk-566673p4.js, backdrop.worker.js, faulty-terminal.worker.js, chunk-vfjy0kzh.js, chunk-5hf4zvs5.js, chunk-cdfbtn3m.js, chunk-axbkkqp1.js, chunk-c43rw71e.js
chunk-566673p4.js: identity=6464 gzip=3111 brotli=2835 bytes
chunk-5hf4zvs5.js: identity=185 gzip=150 brotli=130 bytes
faulty-terminal.worker.js: identity=1021 gzip=534 brotli=459 bytes
backdrop.worker.js: identity=241 gzip=212 brotli=160 bytes
chunk-vfjy0kzh.js: identity=9549 gzip=3889 brotli=3341 bytes
chunk-axbkkqp1.js: identity=90 gzip=90 brotli=78 bytes
chunk-c43rw71e.js: identity=711965 gzip=206534 brotli=167586 bytes
chunk-0gjma2ke.js: identity=7799 gzip=3331 brotli=2978 bytes
main.js: identity=79936 gzip=29892 brotli=26043 bytes
chunk-cdfbtn3m.js: identity=29244 gzip=10353 brotli=9093 bytes
JavaScript transfer: identity=846494 gzip=258096/307200 brotli=212703/256000 bytes
fonts/MapleMonoNF-Italic.latin.woff2: identity=37148 gzip=37148 brotli=37148 bytes
fonts/MapleMonoNF-Medium.latin.woff2: identity=33904 gzip=33904 brotli=33904 bytes
fonts/MapleMonoNF-Regular.latin.woff2: identity=32852 gzip=32852 brotli=32852 bytes
fonts/MapleMonoNF-SemiBold.latin.woff2: identity=34160 gzip=34160 brotli=34160 bytes
fonts/MapleMonoNF-SemiBoldItalic.latin.woff2: identity=38228 gzip=38228 brotli=38228 bytes
fonts/MapleMonoNF-symbols.woff2: identity=223748 gzip=223748 brotli=223748 bytes
Font transfer: critical=32852/36864 latin=176292/184320 symbols=223748/235520 bytes

==> Rust stream oracle (for the differential test)

==> TypeScript protocol tests
bun test v1.4.2 (744846f84)

src/codec.test.ts:
(pass) wire protocol golden vectors > encodes and decodes each frame exactly as the shared manifest describes [5.99ms]
(pass) wire protocol golden vectors > returns borrowed payload views and writes into reusable buffers [0.20ms]
(pass) wire protocol golden vectors > decodes muxed messages split across WebSocket messages [0.72ms]
(pass) wire protocol golden vectors > rejects oversized, non-canonical, unknown, and malformed frame bytes [0.16ms]
(pass) split mux header preserves trailing Close [0.14ms]

src/stream.test.ts:
stream differential: 134 streams, 11439 schedules, zero divergences
(pass) Rust and TypeScript preserve every frame at all boundaries and random multi-splits [2909.26ms]
(pass) maximum frame size and fragmented oversized prefix obey the allocation boundary [7.95ms]

src/routes.test.ts:
(pass) server relay defaults, then complete-path medians choose the fastest measured route [0.99ms]
(pass) RTT outliers and exact ties preserve a stable attached route [0.24ms]
(pass) network changes reject old replies even if a fresh probe is pending [0.13ms]
(pass) wrong route/generation does not consume the genuine ticket; exact expiry rejects [0.09ms]
(pass) override survives route failure, but audit policy always wins [0.48ms]
(pass) changing route kind cancels pending probes and does not reuse its measurements [0.07ms]
(pass) inventory is atomic and bounds both path count and outstanding probes [1.70ms]
(pass) preferences isolate server/host, clear explicitly, and tolerate denied storage [0.38ms]

15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.94s]

Completed run5-server-api at 5a777256139edf11f6628e2a63d0bd7475550bf4. Dependency-free interface only, no activated handlers/listeners. Author review: object-safe lifetime/cancellation contract; private request types need no Debug. Six existing registry tests remain green. Production relay acceptance remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-api-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [137.00ms] $ bun run build.ts built main.js, chunk-0gjma2ke.js, chunk-566673p4.js, backdrop.worker.js, faulty-terminal.worker.js, chunk-vfjy0kzh.js, chunk-5hf4zvs5.js, chunk-cdfbtn3m.js, chunk-axbkkqp1.js, chunk-c43rw71e.js chunk-566673p4.js: identity=6464 gzip=3111 brotli=2835 bytes chunk-5hf4zvs5.js: identity=185 gzip=150 brotli=130 bytes faulty-terminal.worker.js: identity=1021 gzip=534 brotli=459 bytes backdrop.worker.js: identity=241 gzip=212 brotli=160 bytes chunk-vfjy0kzh.js: identity=9549 gzip=3889 brotli=3341 bytes chunk-axbkkqp1.js: identity=90 gzip=90 brotli=78 bytes chunk-c43rw71e.js: identity=711965 gzip=206534 brotli=167586 bytes chunk-0gjma2ke.js: identity=7799 gzip=3331 brotli=2978 bytes main.js: identity=79936 gzip=29892 brotli=26043 bytes chunk-cdfbtn3m.js: identity=29244 gzip=10353 brotli=9093 bytes JavaScript transfer: identity=846494 gzip=258096/307200 brotli=212703/256000 bytes fonts/MapleMonoNF-Italic.latin.woff2: identity=37148 gzip=37148 brotli=37148 bytes fonts/MapleMonoNF-Medium.latin.woff2: identity=33904 gzip=33904 brotli=33904 bytes fonts/MapleMonoNF-Regular.latin.woff2: identity=32852 gzip=32852 brotli=32852 bytes fonts/MapleMonoNF-SemiBold.latin.woff2: identity=34160 gzip=34160 brotli=34160 bytes fonts/MapleMonoNF-SemiBoldItalic.latin.woff2: identity=38228 gzip=38228 brotli=38228 bytes fonts/MapleMonoNF-symbols.woff2: identity=223748 gzip=223748 brotli=223748 bytes Font transfer: critical=32852/36864 latin=176292/184320 symbols=223748/235520 bytes ==> Rust stream oracle (for the differential test) ==> TypeScript protocol tests bun test v1.4.2 (744846f84) src/codec.test.ts: (pass) wire protocol golden vectors > encodes and decodes each frame exactly as the shared manifest describes [5.99ms] (pass) wire protocol golden vectors > returns borrowed payload views and writes into reusable buffers [0.20ms] (pass) wire protocol golden vectors > decodes muxed messages split across WebSocket messages [0.72ms] (pass) wire protocol golden vectors > rejects oversized, non-canonical, unknown, and malformed frame bytes [0.16ms] (pass) split mux header preserves trailing Close [0.14ms] src/stream.test.ts: stream differential: 134 streams, 11439 schedules, zero divergences (pass) Rust and TypeScript preserve every frame at all boundaries and random multi-splits [2909.26ms] (pass) maximum frame size and fragmented oversized prefix obey the allocation boundary [7.95ms] src/routes.test.ts: (pass) server relay defaults, then complete-path medians choose the fastest measured route [0.99ms] (pass) RTT outliers and exact ties preserve a stable attached route [0.24ms] (pass) network changes reject old replies even if a fresh probe is pending [0.13ms] (pass) wrong route/generation does not consume the genuine ticket; exact expiry rejects [0.09ms] (pass) override survives route failure, but audit policy always wins [0.48ms] (pass) changing route kind cancels pending probes and does not reuse its measurements [0.07ms] (pass) inventory is atomic and bounds both path count and outstanding probes [1.70ms] (pass) preferences isolate server/host, clear explicitly, and tolerate denied storage [0.38ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.94s]
kayg commented 2026-10-09 13:22:45 +00:00
Author
Owner
Copy link

Dependency-only fast stopped in existing infraternal-agent readiness regression, then next socket test stalled; stopped only owned scope run-p1442138-i22410144.scope. Exact unchanged readiness recheck 0pass/1fail in0.00s: expected WouldBlock, got NotFound (daemon_readiness_tests.rs68). Evidence run5-core-api-dep-readiness-recheck.log; old hard-coded HOME-cache remote-agents parent is absent. sessions::state uses nonrecursive secure leaf creation; readiness/daemon fixtures assume another test/job created that parent. First interface gate passed when it existed; no product defect or load exception claimed. Preserve pending dependency patch, fix only fixture setup in a separate prerequisite: honor existing INFRATERNAL_TEST_STATE_ROOT and create the synthetic parent; migrate seven inherited fixture roots away from another job. Deadlines, production filesystem security and payloads stay unchanged. Recheck exact failure and complete fast before commit.

Dependency-only fast stopped in existing infraternal-agent readiness regression, then next socket test stalled; stopped only owned scope run-p1442138-i22410144.scope. Exact unchanged readiness recheck 0pass/1fail in0.00s: expected WouldBlock, got NotFound (daemon_readiness_tests.rs68). Evidence run5-core-api-dep-readiness-recheck.log; old hard-coded HOME-cache remote-agents parent is absent. sessions::state uses nonrecursive secure leaf creation; readiness/daemon fixtures assume another test/job created that parent. First interface gate passed when it existed; no product defect or load exception claimed. Preserve pending dependency patch, fix only fixture setup in a separate prerequisite: honor existing INFRATERNAL_TEST_STATE_ROOT and create the synthetic parent; migrate seven inherited fixture roots away from another job. Deadlines, production filesystem security and payloads stay unchanged. Recheck exact failure and complete fast before commit.
kayg commented 2026-10-09 13:31:07 +00:00
Author
Owner
Copy link

Completed run5-fixture-root at 9489a4598a. Failing-first unchanged readiness test: expected WouldBlock, got NotFound, 0pass/1fail in0.00s. Isolated setup fix:1pass/0fail in0.01s; native socket path101B/107B. Seven fixture roots honor this job environment; no production deadlines or directory checks changed. Exact evidence run5-core-api-dep-readiness-recheck.log and run5-fixture-positive.log. Self-reviewed ownership, cleanup and path length. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-fixture-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [274.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.55s]

Completed run5-fixture-root at 9489a4598a1d01d1f19a8d038fcc56d693178483. Failing-first unchanged readiness test: expected WouldBlock, got NotFound, 0pass/1fail in0.00s. Isolated setup fix:1pass/0fail in0.01s; native socket path101B/107B. Seven fixture roots honor this job environment; no production deadlines or directory checks changed. Exact evidence run5-core-api-dep-readiness-recheck.log and run5-fixture-positive.log. Self-reviewed ownership, cleanup and path length. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-fixture-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [274.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.55s]
kayg commented 2026-10-09 13:49:13 +00:00
Author
Owner
Copy link

Completed run5-core-api-dependency at cc4765d106. Dependency-only slice: no new external crate, engine, feature or handler. Offline check passed in6.69s. Complete retry passes after separately committed fixture-root correction; no baseline exception or threshold change. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-core-api-dep-retry-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [218.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.00s]

Completed run5-core-api-dependency at cc4765d10688b92b06213b77592f09f582f84b70. Dependency-only slice: no new external crate, engine, feature or handler. Offline check passed in6.69s. Complete retry passes after separately committed fixture-root correction; no baseline exception or threshold change. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-core-api-dep-retry-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18158928 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16180144 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [218.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.00s]
kayg commented 2026-10-09 13:56:16 +00:00
Author
Owner
Copy link

Production inventory finding to resolve at routing step: RelayStore currently caps the GLOBAL enabled-role table at31 (relay_store.rs14,155; inventory_capacity_is_atomic_and_existing_role_replacement_uses_its_own_slot proves refusal of the32nd). The browser selector independently caps its working inventory at32 routes and96 outstanding probes. D196/D201 permit relay components across the D170 fleet (1000nodes,10000headroom); a global31 limit must not silently become a fleet capability limit. Separate node admission/inventory bounds from the per-client measured candidate window when wiring production routing; preserve bounded probes and stored overrides. No capacity or contract change made during the generic server migration.

Production inventory finding to resolve at routing step: RelayStore currently caps the GLOBAL enabled-role table at31 (relay_store.rs14,155; inventory_capacity_is_atomic_and_existing_role_replacement_uses_its_own_slot proves refusal of the32nd). The browser selector independently caps its working inventory at32 routes and96 outstanding probes. D196/D201 permit relay components across the D170 fleet (1000nodes,10000headroom); a global31 limit must not silently become a fleet capability limit. Separate node admission/inventory bounds from the per-client measured candidate window when wiring production routing; preserve bounded probes and stored overrides. No capacity or contract change made during the generic server migration.
kayg commented 2026-10-09 14:13:11 +00:00
Author
Owner
Copy link

Completed run5-server-adapter at 905d9ee920. All-target server Clippy passed in40.23s; complete fast green. Static sizes x86=18176656B (+17728), ARM=16194136B (+13992); ARM budget has2240B remaining. Reused the existing middleware/state; no new engine, transport task, timer or production registration. Author review: namespace/collision checks, bounded body before dispatch, transaction recheck requirement, no Debug on private requests. Dedicated dispatcher regressions are the next slice; runtime relay acceptance remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-adapter-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [210.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.55s]

Completed run5-server-adapter at 905d9ee920079478ae568a7bc100204f52406855. All-target server Clippy passed in40.23s; complete fast green. Static sizes x86=18176656B (+17728), ARM=16194136B (+13992); ARM budget has2240B remaining. Reused the existing middleware/state; no new engine, transport task, timer or production registration. Author review: namespace/collision checks, bounded body before dispatch, transaction recheck requirement, no Debug on private requests. Dedicated dispatcher regressions are the next slice; runtime relay acceptance remains pending. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-adapter-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [210.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.55s]
kayg commented 2026-10-09 14:16:15 +00:00
Author
Owner
Copy link

Shared server fixture extraction first all-target/test-support Clippy failed: unused CA-test imports/re-export and missing Panics docs on three newly public test fixture functions. Raw run5-server-fixtures-clippy.log. Remove obsolete imports and document fixture construction/readiness panics; preserve synthetic source, SQLite data and exact clock/watchdog behavior. No gate/commit success claimed yet; no production crypto or timeout changes.

Shared server fixture extraction first all-target/test-support Clippy failed: unused CA-test imports/re-export and missing Panics docs on three newly public test fixture functions. Raw run5-server-fixtures-clippy.log. Remove obsolete imports and document fixture construction/readiness panics; preserve synthetic source, SQLite data and exact clock/watchdog behavior. No gate/commit success claimed yet; no production crypto or timeout changes.
kayg commented 2026-10-09 14:35:50 +00:00
Author
Owner
Copy link

Completed run5-server-fixtures at 6f651f4d24. Standalone test-support cargo check passed16.45s; corrected all-target/test-support Clippy passed1m35s. Initial lint-only failure is recorded separately. Full fast passes; production static sizes unchanged18176656/16194136B. Existing signer, real SQLite and controlled socket tests retain their assertions. Self-review: synthetic-only keys, gated module, original constraints/data and exact90s watchdog preserved; no operator credential read or new engine. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-fixtures-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [292.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.99s]

Completed run5-server-fixtures at 6f651f4d24abd09a941147b4357ed37eddfc88c9. Standalone test-support cargo check passed16.45s; corrected all-target/test-support Clippy passed1m35s. Initial lint-only failure is recorded separately. Full fast passes; production static sizes unchanged18176656/16194136B. Existing signer, real SQLite and controlled socket tests retain their assertions. Self-review: synthetic-only keys, gated module, original constraints/data and exact90s watchdog preserved; no operator credential read or new engine. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-fixtures-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [292.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.99s]
kayg commented 2026-10-09 14:37:40 +00:00
Author
Owner
Copy link

Generic dispatcher regression first targeted compile failed E0658: Range Index is not a stable const trait in the invalid-path table (plugin_tests.rs). Raw run5-server-tests-targeted.log. Use direct array-to-slice coercion and explicit static-lifetime case tables, retaining the registration lifetime contract rather than relaxing it. Test slice uncommitted; no runtime change or success claim.

Generic dispatcher regression first targeted compile failed E0658: Range Index is not a stable const trait in the invalid-path table (plugin_tests.rs). Raw run5-server-tests-targeted.log. Use direct array-to-slice coercion and explicit static-lifetime case tables, retaining the registration lifetime contract rather than relaxing it. Test slice uncommitted; no runtime change or success claim.
kayg commented 2026-10-09 14:42:25 +00:00
Author
Owner
Copy link

Generic dispatcher targeted tests pass3/0 in0.03s, raw run5-server-tests-targeted-corrected.log. All-target Clippy then found items_after_statements on the two constant case tables, raw run5-server-tests-clippy.log. Move tables to module scope; no lint allowance or API lifetime relaxation. Add cross-path admission evidence: eight pending plugin calls must also refuse a host join with429. Clarify the trait context supplies trusted server facilities, not authenticated caller authority; endpoints retain private capability/session checks and transaction-time authorization. Test slice remains uncommitted until complete fast passes.

Generic dispatcher targeted tests pass3/0 in0.03s, raw run5-server-tests-targeted-corrected.log. All-target Clippy then found items_after_statements on the two constant case tables, raw run5-server-tests-clippy.log. Move tables to module scope; no lint allowance or API lifetime relaxation. Add cross-path admission evidence: eight pending plugin calls must also refuse a host join with429. Clarify the trait context supplies trusted server facilities, not authenticated caller authority; endpoints retain private capability/session checks and transaction-time authorization. Test slice remains uncommitted until complete fast passes.
kayg commented 2026-10-09 15:01:13 +00:00
Author
Owner
Copy link

Completed server-dispatch-tests at e34752416e. Three dispatcher tests pass in 0.03 seconds: 8193-byte body refused before invocation; eight calls share host admission; all cancel at five seconds and release permits. Full fast gate passes. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-tests-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [228.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.39s]

Completed server-dispatch-tests at e34752416e9dd553a223a0fa638d7a561a9102da. Three dispatcher tests pass in 0.03 seconds: 8193-byte body refused before invocation; eight calls share host admission; all cancel at five seconds and release permits. Full fast gate passes. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-server-tests-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [228.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.39s]
kayg commented 2026-10-09 15:04:05 +00:00
Author
Owner
Copy link

Dependency preparation: the first fast gate refused its locked clippy/test rows because Cargo.lock had not yet been regenerated. Stopped only this job's scope run-p2404108-i23371747.scope; regenerated the graph with offline cargo check and restarted the full gate. Existing workspace engines/versions only; relay server feature and binary default relay feature are separately removable. Evidence: run5-relay-server-deps-fast.log lines 9–14 and run5-relay-server-deps-check.log under /home/kayg/.cache/codex-scratch/relay-direct/. No production module moved yet.

Dependency preparation: the first fast gate refused its locked clippy/test rows because Cargo.lock had not yet been regenerated. Stopped only this job's scope run-p2404108-i23371747.scope; regenerated the graph with offline cargo check and restarted the full gate. Existing workspace engines/versions only; relay server feature and binary default relay feature are separately removable. Evidence: run5-relay-server-deps-fast.log lines 9–14 and run5-relay-server-deps-check.log under /home/kayg/.cache/codex-scratch/relay-direct/. No production module moved yet.
kayg commented 2026-10-09 15:11:29 +00:00
Author
Owner
Copy link

Next module step (D198/D201): move only relay enrolment HTTP ownership to the compiled relay service, preserving /relay/invitations and /relay/join, existing JSON extraction/status semantics, the shared 8-slot/8192-byte/5-second guard, and durable owner rechecks. Binary composition registers compiled metadata; a no-relay build mounts no relay HTTP endpoints. The broker and ledger remain temporary core dependencies until their individually gated moves. Historical migrations remain unchanged to preserve applied checksums. This does not yet provide relay up/down/status or production control admission.

Next module step (D198/D201): move only relay enrolment HTTP ownership to the compiled relay service, preserving /relay/invitations and /relay/join, existing JSON extraction/status semantics, the shared 8-slot/8192-byte/5-second guard, and durable owner rechecks. Binary composition registers compiled metadata; a no-relay build mounts no relay HTTP endpoints. The broker and ledger remain temporary core dependencies until their individually gated moves. Historical migrations remain unchanged to preserve applied checksums. This does not yet provide relay up/down/status or production control admission.
kayg commented 2026-10-09 15:14:51 +00:00
Author
Owner
Copy link

Completed relay-server-dependencies at 5dc0bb78f9. Offline dependency check passes; full fast retry passes after lock regeneration. Static sizes remain x86_64 18176656 and ARM 16194136 bytes. No server module moved in this step. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-relay-server-deps-retry-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [131.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.95s]

Completed relay-server-dependencies at 5dc0bb78f9ea813302e535c15005847e6815b9d3. Offline dependency check passes; full fast retry passes after lock regeneration. Static sizes remain x86_64 18176656 and ARM 16194136 bytes. No server module moved in this step. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-relay-server-deps-retry-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18176656 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16194136 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [131.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.95s]
kayg commented 2026-10-09 15:15:33 +00:00
Author
Owner
Copy link

Failing-first D198 endpoint-removal proof: absent_relay_service_has_no_enrollment_endpoints returns 422 instead of 404 from the old core router with no registered services. Real SQLite/router test: 0 passed, 1 failed in 0.06s; raw run5-http-absence-red.log under /home/kayg/.cache/codex-scratch/relay-direct/. This proves the compiled-service migration must remove the legacy core routes, rather than merely add a second plugin adapter.

Failing-first D198 endpoint-removal proof: absent_relay_service_has_no_enrollment_endpoints returns 422 instead of 404 from the old core router with no registered services. Real SQLite/router test: 0 passed, 1 failed in 0.06s; raw run5-http-absence-red.log under /home/kayg/.cache/codex-scratch/relay-direct/. This proves the compiled-service migration must remove the legacy core routes, rather than merely add a second plugin adapter.
kayg commented 2026-10-09 15:18:24 +00:00
Author
Owner
Copy link

HTTP relocation compile check found three old test-only private-core references: cookie hash and two CaMaterial::issuer calls. Corrected the moved fixture to populate its synthetic session digest directly and use the shared public_issuer fixture accessor; production authentication and key material APIs remain unchanged. Evidence: run5-http-targeted.log E0603/E0624; retry log run5-http-targeted-retry.log. No parser behavior changed.

HTTP relocation compile check found three old test-only private-core references: cookie hash and two CaMaterial::issuer calls. Corrected the moved fixture to populate its synthetic session digest directly and use the shared public_issuer fixture accessor; production authentication and key material APIs remain unchanged. Evidence: run5-http-targeted.log E0603/E0624; retry log run5-http-targeted-retry.log. No parser behavior changed.
kayg commented 2026-10-09 15:21:10 +00:00
Author
Owner
Copy link

HTTP migration preview: paired/single-use and absent-service tests 2 passed/0 failed in 0.12s (run5-http-targeted-retry.log); workspace all-target clippy passed13.10s; infraternal --no-default-features check passed22.57s and normal dependency tree has zero infraternal-plugin-relay entries. Existing JSON fuzz target now imports plugin types; uninstrumented smoke1000 executions/0s is startup proof only, no coverage claim (run5-http-fuzz-smoke.log). Prior instrumented19,495,932-run parser evidence remains applicable to unchanged JSON types/fields. Full gate run5-http-fast.log active. Generic core no longer constructs RelayStore/RelayEnrollment or mounts relay routes; binary composes the descriptor-backed service. Broker/ledger migration still pending.

HTTP migration preview: paired/single-use and absent-service tests 2 passed/0 failed in 0.12s (run5-http-targeted-retry.log); workspace all-target clippy passed13.10s; infraternal --no-default-features check passed22.57s and normal dependency tree has zero infraternal-plugin-relay entries. Existing JSON fuzz target now imports plugin types; uninstrumented smoke1000 executions/0s is startup proof only, no coverage claim (run5-http-fuzz-smoke.log). Prior instrumented19,495,932-run parser evidence remains applicable to unchanged JSON types/fields. Full gate run5-http-fast.log active. Generic core no longer constructs RelayStore/RelayEnrollment or mounts relay routes; binary composes the descriptor-backed service. Broker/ledger migration still pending.
kayg commented 2026-10-09 15:22:55 +00:00
Author
Owner
Copy link

D198 review finding: crates/infraternal-server/src/ca.rs still owns LeafRole::RelayEndpoint/RelayControl, relay SAN namespaces and sign_relay/sign_relay_control. Moving only HTTP/broker/ledger therefore leaves relay-specific core behavior. Proposed subsequent module step: preserve the existing validated OpenSSL signing/chain/lifetime implementation; expose a trusted compile-time leaf template (client/server purpose, static namespace, optional static DNS prefix), with unchanged core host template. Relay plugin owns its endpoint/control templates and real purpose/SAN tests. Templates are trusted compiled declarations, never CSR/request fields; caller still must authorize the role before signing. No crypto engine/dependency or wire/DB identity format change. Separate atomic gated change after broker/ledger moves.

D198 review finding: crates/infraternal-server/src/ca.rs still owns LeafRole::RelayEndpoint/RelayControl, relay SAN namespaces and sign_relay/sign_relay_control. Moving only HTTP/broker/ledger therefore leaves relay-specific core behavior. Proposed subsequent module step: preserve the existing validated OpenSSL signing/chain/lifetime implementation; expose a trusted compile-time leaf template (client/server purpose, static namespace, optional static DNS prefix), with unchanged core host template. Relay plugin owns its endpoint/control templates and real purpose/SAN tests. Templates are trusted compiled declarations, never CSR/request fields; caller still must authorize the role before signing. No crypto engine/dependency or wire/DB identity format change. Separate atomic gated change after broker/ledger moves.
kayg commented 2026-10-09 15:38:28 +00:00
Author
Owner
Copy link

HTTP migration full gate: fmt, VT native fmt, clippy, test and cargo-deny PASS; static builds FAIL. x86_64=18,312,464B (max19,244,272); ARM=16,326,008B (max16,196,376), exceeds by129,632B. Previous sizes18,176,656/16,194,136: delta135,808/131,872B. Both static, PT_INTERP absent. Raw run5-http-fast.log; local infraternal-dev load18.65/24.37/20.89 at result. This is a real size failure, not a host-load exception; no commit or budget alteration. Measuring removal of redundant body extraction and future layers while preserving shared admission/body/deadline and existing JSON rejection behavior.

HTTP migration full gate: fmt, VT native fmt, clippy, test and cargo-deny PASS; static builds FAIL. x86_64=18,312,464B (max19,244,272); ARM=16,326,008B (max16,196,376), exceeds by129,632B. Previous sizes18,176,656/16,194,136: delta135,808/131,872B. Both static, PT_INTERP absent. Raw run5-http-fast.log; local infraternal-dev load18.65/24.37/20.89 at result. This is a real size failure, not a host-load exception; no commit or budget alteration. Measuring removal of redundant body extraction and future layers while preserving shared admission/body/deadline and existing JSON rejection behavior.
kayg commented 2026-10-09 15:42:34 +00:00
Author
Owner
Copy link

Size candidate1: removing the extra boxed async wrapper gives ARM16,325,352B, saving656B but still128,976B over budget; raw run5-http-arm-future.log, release2m36s. Candidate2 now measures a shared ServerRequest::json decoder over already bounded bytes, using existing Axum Json::from_bytes and the already-transitive mime0.3.17 media-type parser. No new parser engine/version; direct dependency preparation will be its own gated commit if the candidate is viable. Prepared48 status/value equivalence cases against Axum's existing extractor. Shared8-slot/8192-byte/5-second guards unchanged. Migration remains uncommitted; raw candidate run5-http-arm-json.log.

Size candidate1: removing the extra boxed async wrapper gives ARM16,325,352B, saving656B but still128,976B over budget; raw run5-http-arm-future.log, release2m36s. Candidate2 now measures a shared ServerRequest::json decoder over already bounded bytes, using existing Axum Json::from_bytes and the already-transitive mime0.3.17 media-type parser. No new parser engine/version; direct dependency preparation will be its own gated commit if the candidate is viable. Prepared48 status/value equivalence cases against Axum's existing extractor. Shared8-slot/8192-byte/5-second guards unchanged. Migration remains uncommitted; raw candidate run5-http-arm-json.log.
kayg commented 2026-10-09 15:50:23 +00:00
Author
Owner
Copy link

Size candidate2 (shared JSON decoder) ARM16,318,440B, saving6,912B versus candidate1 but still122,064B over budget; raw run5-http-arm-json.log release4m51s. Set it aside: restored existing extractor and removed proposed direct mime dependency/helper; draft retained in scratch, no commit. Candidate3 addresses a larger suspected cause: CLI currently awaits separate default and compiled-plugin startup futures. Add a generic optional-remote compiled-service startup entry and have CLI prepare remote coordinates before one shared await, preserving startup behavior. This avoids retaining separate full server future chains. Raw run5-http-arm-single-startup.log. No size-success claim yet. The JSON equivalence invocation used an incorrect test-name filter and ran0tests; only prepared cases exist, no48-case pass claim.

Size candidate2 (shared JSON decoder) ARM16,318,440B, saving6,912B versus candidate1 but still122,064B over budget; raw run5-http-arm-json.log release4m51s. Set it aside: restored existing extractor and removed proposed direct mime dependency/helper; draft retained in scratch, no commit. Candidate3 addresses a larger suspected cause: CLI currently awaits separate default and compiled-plugin startup futures. Add a generic optional-remote compiled-service startup entry and have CLI prepare remote coordinates before one shared await, preserving startup behavior. This avoids retaining separate full server future chains. Raw run5-http-arm-single-startup.log. No size-success claim yet. The JSON equivalence invocation used an incorrect test-name filter and ran0tests; only prepared cases exist, no48-case pass claim.
kayg commented 2026-10-09 15:56:16 +00:00
Author
Owner
Copy link

Size candidate3 single startup future: ARM16,323,704B, saving1,648B versus candidate1; still127,328B over budget. Raw run5-http-arm-single-startup.log release3m06s. Paired and absent-service HTTP tests pass2/0 in0.06s after this candidate. The startup-duplication hypothesis does not explain the dominant growth; proceeding to symbol/link-map measurements instead of more speculative refactors. No budget change, commit or acceptance claim. CurrentHEAD remains5dc0bb78f9ea813302e535c15005847e6815b9d3.

Size candidate3 single startup future: ARM16,323,704B, saving1,648B versus candidate1; still127,328B over budget. Raw run5-http-arm-single-startup.log release3m06s. Paired and absent-service HTTP tests pass2/0 in0.06s after this candidate. The startup-duplication hypothesis does not explain the dominant growth; proceeding to symbol/link-map measurements instead of more speculative refactors. No budget change, commit or acceptance claim. CurrentHEAD remains5dc0bb78f9ea813302e535c15005847e6815b9d3.
kayg commented 2026-10-09 16:01:20 +00:00
Author
Owner
Copy link

Link-map measurement: current plugin symbols total37,452B (16 symbols); dispatch poll23,692B, join poll11,424B. Core remote_http symbols84,712B; signer5,848B. These totals alone do not explain127,328B overage. Artifacts run5-http-arm.map/run5-http-arm-symbols.txt/run5-http-arm-unstripped. Profiling exact committedHEAD5dc0bb78 on the same branch: saved10 owned tracked draft files under size-preview/ with files.json, temporarily restored only those paths; untracked draft modules remain inert. Baseline symbol build run5-head-arm-symbols.log active. Restore with python /home/kayg/.cache/codex-scratch/relay-direct/profile_baseline.py restore after measurement. No other checkout/host, budget change or source commit.

Link-map measurement: current plugin symbols total37,452B (16 symbols); dispatch poll23,692B, join poll11,424B. Core remote_http symbols84,712B; signer5,848B. These totals alone do not explain127,328B overage. Artifacts run5-http-arm.map/run5-http-arm-symbols.txt/run5-http-arm-unstripped. Profiling exact committedHEAD5dc0bb78 on the same branch: saved10 owned tracked draft files under size-preview/ with files.json, temporarily restored only those paths; untracked draft modules remain inert. Baseline symbol build run5-head-arm-symbols.log active. Restore with `python /home/kayg/.cache/codex-scratch/relay-direct/profile_baseline.py restore` after measurement. No other checkout/host, budget change or source commit.
kayg commented 2026-10-09 16:15:06 +00:00
Author
Owner
Copy link

ExactHEAD symbol comparison complete (run5-head-arm-symbols.log release4m42s); all saved draft paths restored. Named text grows102,868B. Relay-plugin text37,452B replaces old relay_http34,424B and core remote_http shrinks35,144B; startup shrinks648B. The dominant remaining changes include additional SQLx generic pool futures: acquire+14,340B, connect+4,840B, execution flattening+8,236B, return-to-pool+2,776B and associated drop/channel/maintenance code. Head already has two large acquire copies; draft has three, indicating crate-boundary monomorphization growth rather than certificate signing (+0B). Artifacts run5-http-arm-text-delta.txt and both section/map/symbol files. Measuring the already-selected bundled LLD's conservative --icf=safe folding (address-significant code remains distinct), without changing source, compiler profile, engine or budgets; this is a candidate measurement, not a shipped build-policy decision.

ExactHEAD symbol comparison complete (run5-head-arm-symbols.log release4m42s); all saved draft paths restored. Named text grows102,868B. Relay-plugin text37,452B replaces old relay_http34,424B and core remote_http shrinks35,144B; startup shrinks648B. The dominant remaining changes include additional SQLx generic pool futures: acquire+14,340B, connect+4,840B, execution flattening+8,236B, return-to-pool+2,776B and associated drop/channel/maintenance code. Head already has two large acquire copies; draft has three, indicating crate-boundary monomorphization growth rather than certificate signing (+0B). Artifacts run5-http-arm-text-delta.txt and both section/map/symbol files. Measuring the already-selected bundled LLD's conservative --icf=safe folding (address-significant code remains distinct), without changing source, compiler profile, engine or budgets; this is a candidate measurement, not a shipped build-policy decision.
kayg commented 2026-10-09 16:23:56 +00:00
Author
Owner
Copy link

LLD conservative ICF candidate: ARM16,313,560B (saves10,144B) remains117,184B over limit; raw run5-http-arm-icf-safe.log release2m28s. It correctly refused folding native SQLite/crypto objects with incomplete address-significance metadata; no linker flag adopted. Next source-level candidate: erase only the three cold cross-crate admission futures (ledger invite/admit and broker join) at their defining crate using Pin<Box<dyn Future+Send>>. This prevents downstream instantiation of their complete SQLx pool futures, preserves existing awaited results/cancellation/authorization, and adds bounded cold-path allocations under the shared8-request guard. Byte forwarding never calls these APIs. Candidate run5-http-arm-boxed-admission.log; will measure size and real SQLite timing before adoption, and split any accepted module edits into independently gated commits.

LLD conservative ICF candidate: ARM16,313,560B (saves10,144B) remains117,184B over limit; raw run5-http-arm-icf-safe.log release2m28s. It correctly refused folding native SQLite/crypto objects with incomplete address-significance metadata; no linker flag adopted. Next source-level candidate: erase only the three cold cross-crate admission futures (ledger invite/admit and broker join) at their defining crate using Pin<Box<dyn Future+Send>>. This prevents downstream instantiation of their complete SQLx pool futures, preserves existing awaited results/cancellation/authorization, and adds bounded cold-path allocations under the shared8-request guard. Byte forwarding never calls these APIs. Candidate run5-http-arm-boxed-admission.log; will measure size and real SQLite timing before adoption, and split any accepted module edits into independently gated commits.
kayg commented 2026-10-09 16:33:50 +00:00
Author
Owner
Copy link

Boxed cold admission candidate ARM16,219,960B: saves103,744B versus the unboxed single-startup draft; still23,584B over max16,196,376. Raw run5-http-arm-boxed-admission.log release4m44s. Paired enrolment and absent-service HTTP tests pass2/0 in0.17s (run5-http-box-targeted.log). No source/engine/linker/budget changes committed. Next local candidate shares immutable ServerContext through Arc in routing state, avoiding per-request issuer-DER/facility cloning and temporary contexts. Auth origin, owner transaction, signer source and shared request guards unchanged. Raw run5-http-arm-shared-context.log active. Timing/heap-bound evidence and separate module gates remain required before adopting boxed admission.

Boxed cold admission candidate ARM16,219,960B: saves103,744B versus the unboxed single-startup draft; still23,584B over max16,196,376. Raw run5-http-arm-boxed-admission.log release4m44s. Paired enrolment and absent-service HTTP tests pass2/0 in0.17s (run5-http-box-targeted.log). No source/engine/linker/budget changes committed. Next local candidate shares immutable ServerContext through Arc in routing state, avoiding per-request issuer-DER/facility cloning and temporary contexts. Auth origin, owner transaction, signer source and shared request guards unchanged. Raw run5-http-arm-shared-context.log active. Timing/heap-bound evidence and separate module gates remain required before adopting boxed admission.
kayg commented 2026-10-09 16:51:26 +00:00
Author
Owner
Copy link

Raw generic handler candidate ARM16,215,688B: saves1,072B versus shared-context draft, remains19,312B over budget. Shared body/path/admission/deadline dispatch tests pass3/0 in0.07s (run5-http-raw-handler-dispatch-tests.log). Headers now move from the request; original body-limit extension remains for explicit bounded Bytes extraction before callback. Measuring explicit non-inlining of boxed cold admission constructors so fat LTO cannot reconstruct their concrete database futures at plugin callers. No engine, dependency, linker flag or budget change; no source commit yet. Raw run5-http-arm-opaque-admission.log active.

Raw generic handler candidate ARM16,215,688B: saves1,072B versus shared-context draft, remains19,312B over budget. Shared body/path/admission/deadline dispatch tests pass3/0 in0.07s (run5-http-raw-handler-dispatch-tests.log). Headers now move from the request; original body-limit extension remains for explicit bounded Bytes extraction before callback. Measuring explicit non-inlining of boxed cold admission constructors so fat LTO cannot reconstruct their concrete database futures at plugin callers. No engine, dependency, linker flag or budget change; no source commit yet. Raw run5-http-arm-opaque-admission.log active.
kayg commented 2026-10-09 16:59:50 +00:00
Author
Owner
Copy link

Explicit non-inlining candidate ARM16,215,816B (128B larger), still19,440B over budget; raw run5-http-arm-opaque-admission.log release3m06s. Removed the ineffective inline attributes; must_use retained on cold future APIs. Next source candidate splits invitation/join future state machines at the service boundary and moves headers into the existing extractor instead of cloning. Same-origin cookie admission is computed without IO but its refusal is returned only after JSON validation, preserving rejection priority and current owner checks. Raw run5-http-arm-split-handlers.log active. All source remains uncommitted pending a green size gate and cold-path timing evidence.

Explicit non-inlining candidate ARM16,215,816B (128B larger), still19,440B over budget; raw run5-http-arm-opaque-admission.log release3m06s. Removed the ineffective inline attributes; must_use retained on cold future APIs. Next source candidate splits invitation/join future state machines at the service boundary and moves headers into the existing extractor instead of cloning. Same-origin cookie admission is computed without IO but its refusal is returned only after JSON validation, preserving rejection priority and current owner checks. Raw run5-http-arm-split-handlers.log active. All source remains uncommitted pending a green size gate and cold-path timing evidence.
kayg commented 2026-10-09 17:23:40 +00:00
Author
Owner
Copy link

HTTP migration size finding: sharing the existing bounded JSON decoder reduced ARM from 16,216,568 to 16,203,912 bytes. Four dispatcher tests pass, including 48 media-type/body comparisons against the real Axum extractor (0.05s). The selected bundled LLD with --icf=safe measures 16,193,768 bytes (limit 16,196,376; 2,608 bytes spare). LLVM documents conservative handling of objects lacking address-significance metadata; linker warnings confirm those native objects are not folded. Separating this build optimization before the remaining module commits. Evidence: run5-http-arm-reuse-decoder.log, run5-http-arm-decoder-icf-safe.log, run5-http-json-final-tests.log under /home/kayg/.cache/codex-scratch/relay-direct. No budget or engine change.

HTTP migration size finding: sharing the existing bounded JSON decoder reduced ARM from 16,216,568 to 16,203,912 bytes. Four dispatcher tests pass, including 48 media-type/body comparisons against the real Axum extractor (0.05s). The selected bundled LLD with --icf=safe measures 16,193,768 bytes (limit 16,196,376; 2,608 bytes spare). LLVM documents conservative handling of objects lacking address-significance metadata; linker warnings confirm those native objects are not folded. Separating this build optimization before the remaining module commits. Evidence: run5-http-arm-reuse-decoder.log, run5-http-arm-decoder-icf-safe.log, run5-http-json-final-tests.log under /home/kayg/.cache/codex-scratch/relay-direct. No budget or engine change.
kayg commented 2026-10-09 23:11:03 +00:00
Author
Owner
Copy link

Completed safe-fold at 2a098d53ed. Build-only safe folding passed full fast gate. Static CLI x86=18,174,448 bytes; ARM=16,184,072 bytes; both have no PT_INTERP and are below unchanged limits. Separate post-gate musl functional check could not execute: the scratch target directory had disappeared; existing real-binary debug tests in the full gate passed. No claim for that additional invocation. Evidence: run5-safe-fold-musl-functional.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-safe-fold-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18174448 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16184072 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [297.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.25s]

Completed safe-fold at 2a098d53ed8f36b31a2d1a1ab1d529decb688dcc. Build-only safe folding passed full fast gate. Static CLI x86=18,174,448 bytes; ARM=16,184,072 bytes; both have no PT_INTERP and are below unchanged limits. Separate post-gate musl functional check could not execute: the scratch target directory had disappeared; existing real-binary debug tests in the full gate passed. No claim for that additional invocation. Evidence: run5-safe-fold-musl-functional.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-safe-fold-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18174448 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16184072 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [297.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.25s]
kayg commented 2026-10-09 23:28:46 +00:00
Author
Owner
Copy link

Cold admission boundary measurements, infraternal-dev x86_64, 8 cores, CPUWeight=10000. Alternating inline/boxed operations in the same fixtures and scheduler window (load 10.74/9.39/7.64): invitations n=2000/mode, inline p50/p99=139659/399561 ns; boxed=139913/438486 ns. Paired real OpenSSL/SQLite enrollments n=200/mode, inline=1284189/2528765 ns; boxed=1293628/2103469 ns. Median deltas +254 ns (+0.18%) and +9439 ns (+0.74%); these are shared-host samples, not statistical significance claims. Separate rounds drifted widely (baseline invitation p50 118006 then 79975 ns), so they cannot isolate allocation overhead. Actual boxed payloads: invite1256/admit144/join1832 bytes versus inline1304/144/1896; max8 concurrent requests means at most14656 bytes (14.3KiB) of these transient boxes, excluding allocator headers and the existing request/signer buffers. No idle-session box allocation, and forwarding/lease authorization paths are untouched. Combined cold boundary size candidate saved103744 ARM bytes; budget need justifies this small measured cold median cost. Does not satisfy the pending 300ms path/switch/200-session acceptance benchmarks. Temporary benchmark methods/tests restored; working tree has no tracked edits. Raw evidence: /home/kayg/.cache/codex-scratch/relay-direct/run5-admission-interleaved.log, run5-admission-interleaved-summary.log, run5-admission-summary.log and run5-admission-0..3 logs.

Cold admission boundary measurements, infraternal-dev x86_64, 8 cores, CPUWeight=10000. Alternating inline/boxed operations in the same fixtures and scheduler window (load 10.74/9.39/7.64): invitations n=2000/mode, inline p50/p99=139659/399561 ns; boxed=139913/438486 ns. Paired real OpenSSL/SQLite enrollments n=200/mode, inline=1284189/2528765 ns; boxed=1293628/2103469 ns. Median deltas +254 ns (+0.18%) and +9439 ns (+0.74%); these are shared-host samples, not statistical significance claims. Separate rounds drifted widely (baseline invitation p50 118006 then 79975 ns), so they cannot isolate allocation overhead. Actual boxed payloads: invite1256/admit144/join1832 bytes versus inline1304/144/1896; max8 concurrent requests means at most14656 bytes (14.3KiB) of these transient boxes, excluding allocator headers and the existing request/signer buffers. No idle-session box allocation, and forwarding/lease authorization paths are untouched. Combined cold boundary size candidate saved103744 ARM bytes; budget need justifies this small measured cold median cost. Does not satisfy the pending 300ms path/switch/200-session acceptance benchmarks. Temporary benchmark methods/tests restored; working tree has no tracked edits. Raw evidence: /home/kayg/.cache/codex-scratch/relay-direct/run5-admission-interleaved.log, run5-admission-interleaved-summary.log, run5-admission-summary.log and run5-admission-0..3 logs.
kayg commented 2026-10-09 23:54:17 +00:00
Author
Owner
Copy link

Completed run5-mime-dep at ddf88e0a7f. Dependency-only declaration of the existing transitive mime 0.3.17 parser; no new foundation or parser version. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-mime-dep-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18166160 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16184072 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [329.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.71s]

Completed run5-mime-dep at ddf88e0a7fc87cad4c8040ae51cfccd33f008045. Dependency-only declaration of the existing transitive mime 0.3.17 parser; no new foundation or parser version. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-mime-dep-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18166160 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16184072 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [329.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.71s]
kayg commented 2026-10-09 23:55:52 +00:00
Author
Owner
Copy link

Ledger boundary first gate stopped after clippy rejected redundant #[must_use] attributes on invite/admit (returns Pin<Box>, already must-use); 2 lint errors. Removed only these added attributes, including the prepared broker join attribute; no transaction, authority or cancellation logic changed. Stopped own run-p2202055-i27358683.scope before remaining checks could consume resources. Restarting the ledger full gate, retaining raw first log at /home/kayg/.cache/codex-scratch/relay-direct/run5-ledger-box-first-fast.log. Dependency step remains committed at ddf88e0a7f.

Ledger boundary first gate stopped after clippy rejected redundant #[must_use] attributes on invite/admit (returns Pin<Box<dyn Future>>, already must-use); 2 lint errors. Removed only these added attributes, including the prepared broker join attribute; no transaction, authority or cancellation logic changed. Stopped own run-p2202055-i27358683.scope before remaining checks could consume resources. Restarting the ledger full gate, retaining raw first log at /home/kayg/.cache/codex-scratch/relay-direct/run5-ledger-box-first-fast.log. Dependency step remains committed at ddf88e0a7fc87cad4c8040ae51cfccd33f008045.
kayg commented 2026-10-10 00:11:31 +00:00
Author
Owner
Copy link

Completed run5-ledger-box at 2cd929ed45. Cold ledger boundary only; measured real invitation transactions and frame payloads before/after. Evidence: run5-admission-summary.log and run5-admission--.log; forwarding unchanged. Same-window inline/boxed: invite p50/p99139659/399561 versus139913/438486 ns (n2000); paired1284189/2528765 versus1293628/2103469 ns (n200). Payload1256/144/1832B; eight-request peak14656B excluding allocator headers. Machine infraternal-dev, x86_64,8cores; load10.74/9.39/7.64; CPUWeight10000. Raw run5-admission-interleaved.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-ledger-box-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18164816 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16183752 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [203.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.26s]

Completed run5-ledger-box at 2cd929ed45d83d8f9e10595b0b82b76370b08258. Cold ledger boundary only; measured real invitation transactions and frame payloads before/after. Evidence: run5-admission-summary.log and run5-admission-*-*.log; forwarding unchanged. Same-window inline/boxed: invite p50/p99139659/399561 versus139913/438486 ns (n2000); paired1284189/2528765 versus1293628/2103469 ns (n200). Payload1256/144/1832B; eight-request peak14656B excluding allocator headers. Machine infraternal-dev, x86_64,8cores; load10.74/9.39/7.64; CPUWeight10000. Raw run5-admission-interleaved.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-ledger-box-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18164816 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16183752 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [203.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.26s]
kayg commented 2026-10-10 00:23:40 +00:00
Author
Owner
Copy link

Completed run5-broker-box at c9595ddec4. Paired broker boundary only; real signed-leaf transaction timings and boxed payload sizes recorded in run5-admission-summary.log. Existing signer-race/capability tests run in the full gate. Same-window inline/boxed: invite p50/p99139659/399561 versus139913/438486 ns (n2000); paired1284189/2528765 versus1293628/2103469 ns (n200). Payload1256/144/1832B; eight-request peak14656B excluding allocator headers. Machine infraternal-dev, x86_64,8cores; load10.74/9.39/7.64; CPUWeight10000. Raw run5-admission-interleaved.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-broker-box-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [120.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.14s]

Completed run5-broker-box at c9595ddec4015dc45df6a012b7ab33163c2bf0a2. Paired broker boundary only; real signed-leaf transaction timings and boxed payload sizes recorded in run5-admission-summary.log. Existing signer-race/capability tests run in the full gate. Same-window inline/boxed: invite p50/p99139659/399561 versus139913/438486 ns (n2000); paired1284189/2528765 versus1293628/2103469 ns (n200). Payload1256/144/1832B; eight-request peak14656B excluding allocator headers. Machine infraternal-dev, x86_64,8cores; load10.74/9.39/7.64; CPUWeight10000. Raw run5-admission-interleaved.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-broker-box-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [120.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.14s]
kayg commented 2026-10-10 00:38:05 +00:00
Author
Owner
Copy link

Completed run5-bounded-json at 5ed24d0c0f. Shared decoder reuses existing MIME and Axum engines; compatibility comparison measured 48 media/body cases before staging, run5-http-json-final-tests.log (4 dispatcher tests, 0.05s). scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-bounded-json-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [316.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.17s]

Completed run5-bounded-json at 5ed24d0c0f603de133d4ebb2e796d99d0ca82cc2. Shared decoder reuses existing MIME and Axum engines; compatibility comparison measured 48 media/body cases before staging, run5-http-json-final-tests.log (4 dispatcher tests, 0.05s). scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-bounded-json-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [316.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.17s]
kayg commented 2026-10-10 00:39:38 +00:00
Author
Owner
Copy link

Shared decoder module committed5ed24d0c0f603de133d4ebb2e796d99d0ca82cc2; full fast gate green, x8618164464B/ARM16183400B unchanged. Separate 48-case equivalence test gate stopped after Clippy rejected one redundant status closure at plugin_tests.rs:276. Replaced it with JsonRejection::status; assertions/cases unchanged. Stopped only own run-p2825464-i27981959.scope and retained firstlog at /home/kayg/.cache/codex-scratch/relay-direct/run5-bounded-json-tests-first-fast.log. Restarting only the pending test step; earlier support commits remain gated.

Shared decoder module committed5ed24d0c0f603de133d4ebb2e796d99d0ca82cc2; full fast gate green, x8618164464B/ARM16183400B unchanged. Separate 48-case equivalence test gate stopped after Clippy rejected one redundant status closure at plugin_tests.rs:276. Replaced it with JsonRejection::status; assertions/cases unchanged. Stopped only own run-p2825464-i27981959.scope and retained firstlog at /home/kayg/.cache/codex-scratch/relay-direct/run5-bounded-json-tests-first-fast.log. Restarting only the pending test step; earlier support commits remain gated.
kayg commented 2026-10-10 00:45:58 +00:00
Author
Owner
Copy link

Completed run5-bounded-json-tests at 144a1eb983. Real-extractor equivalence test covers 48 combinations alongside existing registration/body-limit/deadline tests; full fast gate output follows. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-bounded-json-tests-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [136.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.59s]

Completed run5-bounded-json-tests at 144a1eb983428bcc0173d0db61344e8011428fb1. Real-extractor equivalence test covers 48 combinations alongside existing registration/body-limit/deadline tests; full fast gate output follows. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-bounded-json-tests-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [136.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.59s]
kayg commented 2026-10-10 00:47:15 +00:00
Author
Owner
Copy link

Support steps complete at144a1eb983428bcc0173d0db61344e8011428fb1 (full fast green for each). Clock/proof helper visibility preflight stopped at cargo fmt: existing now was pub(crate); the edit retained that prefix before its doc comment. Corrected only visibility/comment placement before any gate or HTTP module wiring. No behavior change. Raw preflight: /home/kayg/.cache/codex-scratch/relay-direct/run5-http-steps-summary.log. Restarting the API step, followed by separately gated relayHTTP module, production composition and absence regression.

Support steps complete at144a1eb983428bcc0173d0db61344e8011428fb1 (full fast green for each). Clock/proof helper visibility preflight stopped at cargo fmt: existing now was pub(crate); the edit retained that prefix before its doc comment. Corrected only visibility/comment placement before any gate or HTTP module wiring. No behavior change. Raw preflight: /home/kayg/.cache/codex-scratch/relay-direct/run5-http-steps-summary.log. Restarting the API step, followed by separately gated relayHTTP module, production composition and absence regression.
kayg commented 2026-10-10 00:58:07 +00:00
Author
Owner
Copy link

Public proof/clock helper gate has exactly one runtimefailure: auth::tests::socket_tests::expired_handshake_releases_user_permits atsocket_tests.rs:151 expected handshake refusal503 (161 server tests passed,1 failed,17ignored;97.17s). Fmt/nativefmt/clippy/cargo-deny/staticbuild rows PASS; x8618164464/ARM16183400. Sourcechange visibility/docs only. Earlier tail progress missed this testfailure; no commit or green claim made. Following #118 authsocket baseline rule on own detacheddev checkout base e704e5b893 under /home/kayg/.cache/codex-scratch/relay-direct/baseline, exacttest in same currentload window. Raw completed gate run5-public-proof-first-fast.log. No externalhost or credential investigation.

Public proof/clock helper gate has exactly one runtimefailure: auth::tests::socket_tests::expired_handshake_releases_user_permits atsocket_tests.rs:151 expected handshake refusal503 (161 server tests passed,1 failed,17ignored;97.17s). Fmt/nativefmt/clippy/cargo-deny/staticbuild rows PASS; x8618164464/ARM16183400. Sourcechange visibility/docs only. Earlier tail progress missed this testfailure; no commit or green claim made. Following #118 authsocket baseline rule on own detacheddev checkout base e704e5b8937ddf0d1f9e3a4fc9c2c2a9d7597455 under /home/kayg/.cache/codex-scratch/relay-direct/baseline, exacttest in same currentload window. Raw completed gate run5-public-proof-first-fast.log. No externalhost or credential investigation.
kayg commented 2026-10-10 01:07:34 +00:00
Author
Owner
Copy link

Named #118 authsocket comparison: detacheddev e704e5b8 initial exacttest PASS1/1 (0.85s), changed checkout exacttest PASS1/1 (1.06s). Built binaries then alternated12 rounds each in one localCPUWeight10000 scope; dev12/12 PASS and changed12/12 PASS, hostload about20 on8cores. Failure at handshake503 assertion did not reproduce on baseline, so #118 exception is NOT applied. Restarting full public-proof fast gate; no source behavior change and no commit from the failed gate. Rawinitial run5-public-proof-dev-baseline-test.log/run5-public-proof-changed-test.log;24 exact outputs in run5-authsocket-window-00..11-{dev,changed}.log; run5-authsocket-window-summary.log/results.json; rawfullfailure run5-public-proof-first-fast.log. Baseline remains own clean detachedcheckout until cleanup.

Named #118 authsocket comparison: detacheddev e704e5b8 initial exacttest PASS1/1 (0.85s), changed checkout exacttest PASS1/1 (1.06s). Built binaries then alternated12 rounds each in one localCPUWeight10000 scope; dev12/12 PASS and changed12/12 PASS, hostload about20 on8cores. Failure at handshake503 assertion did not reproduce on baseline, so #118 exception is NOT applied. Restarting full public-proof fast gate; no source behavior change and no commit from the failed gate. Rawinitial run5-public-proof-dev-baseline-test.log/run5-public-proof-changed-test.log;24 exact outputs in run5-authsocket-window-00..11-{dev,changed}.log; run5-authsocket-window-summary.log/results.json; rawfullfailure run5-public-proof-first-fast.log. Baseline remains own clean detachedcheckout until cleanup.
kayg commented 2026-10-10 01:17:04 +00:00
Author
Owner
Copy link

Completed run5-public-proof at e2529e7509. Visibility/docs only; existing host and relay enrollment proof tests run unchanged in the full gate. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-public-proof-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [571.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [6.95s]

Completed run5-public-proof at e2529e75099a7f26fc775d3086c232e89f966417. Visibility/docs only; existing host and relay enrollment proof tests run unchanged in the full gate. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-public-proof-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [571.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [6.95s]
kayg commented 2026-10-10 01:23:30 +00:00
Author
Owner
Copy link

HTTP module gate stopped before commit: saturated_terminal_queues_preserve_every_byte_and_daemon_liveness failed during initial terminal.open with terminal open refused (agent_queues.rs:14; scripts/verify-agent-queues.py initialization). Raw output: run5-http-module-first-fast.log. Existing sessions::launch bounds shim readiness at two seconds; the relay module is not wired into production yet. This row is outside the #118 baseline exception, so no exception is applied. Investigating with the exact test under CPUWeight=10000; all gates will be rerun before committing.

HTTP module gate stopped before commit: saturated_terminal_queues_preserve_every_byte_and_daemon_liveness failed during initial terminal.open with terminal open refused (agent_queues.rs:14; scripts/verify-agent-queues.py initialization). Raw output: run5-http-module-first-fast.log. Existing sessions::launch bounds shim readiness at two seconds; the relay module is not wired into production yet. This row is outside the #118 baseline exception, so no exception is applied. Investigating with the exact test under CPUWeight=10000; all gates will be rerun before committing.
kayg commented 2026-10-10 01:30:23 +00:00
Author
Owner
Copy link

Queue failure investigation: exact saturated_terminal_queues_preserve_every_byte_and_daemon_liveness rerun passed 1/1 in36.04s under CPUWeight10000, raw run5-http-queue-recheck.log. Full gate rerun also passed this row and the server authentication/socket tests so far. No product change, timeout increase or #118 exception applied. The initial terminal-open refusal cause is not established; launch has an unchanged two-second shim readiness bound. Preserved first full-gate failure in run5-http-module-first-fast.log. Full static/TypeScript gates still pending.

Queue failure investigation: exact saturated_terminal_queues_preserve_every_byte_and_daemon_liveness rerun passed 1/1 in36.04s under CPUWeight10000, raw run5-http-queue-recheck.log. Full gate rerun also passed this row and the server authentication/socket tests so far. No product change, timeout increase or #118 exception applied. The initial terminal-open refusal cause is not established; launch has an unchanged two-second shim readiness bound. Preserved first full-gate failure in run5-http-module-first-fast.log. Full static/TypeScript gates still pending.
kayg commented 2026-10-10 01:38:08 +00:00
Author
Owner
Copy link

Completed run5-http-module at 48b10a8825. Standalone relay-owned HTTP module added; existing core handler remains until the separate production wiring step. Shared parser and real core paired-enrollment tests gate this intermediate step. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-http-module-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [633.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [4.43s]

Completed run5-http-module at 48b10a8825ea7c94b5c4ff05b5fa4102310d33eb. Standalone relay-owned HTTP module added; existing core handler remains until the separate production wiring step. Shared parser and real core paired-enrollment tests gate this intermediate step. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-http-module-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18164464 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16183400 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [633.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [4.43s]
kayg commented 2026-10-10 01:39:28 +00:00
Author
Owner
Copy link

HTTP service module committed48b10a8825ea7c94b5c4ff05b5fa4102310d33eb with full fast green; x8618164464B, ARM16183400B. Production wiring intermediate gate caught unused router test import at crates/plugins/relay/src/http_tests.rs:9: the absence regression that uses it is retained in the next atomic commit. Stopped own failed gate, removed the unused import from this intermediate step, and restarted the full gate. Final regression restore retains its needed import. Raw first output run5-http-wiring-first-fast.log; no product behavior change for this correction.

HTTP service module committed48b10a8825ea7c94b5c4ff05b5fa4102310d33eb with full fast green; x8618164464B, ARM16183400B. Production wiring intermediate gate caught unused router test import at crates/plugins/relay/src/http_tests.rs:9: the absence regression that uses it is retained in the next atomic commit. Stopped own failed gate, removed the unused import from this intermediate step, and restarted the full gate. Final regression restore retains its needed import. Raw first output run5-http-wiring-first-fast.log; no product behavior change for this correction.
kayg commented 2026-10-10 01:52:57 +00:00
Author
Owner
Copy link

Completed run5-http-wiring at e49f06cae0. Failing-first absence evidence run5-http-absence-red.log (422 versus expected404). Existing real paired/single-use enrollment test moved to the relay crate. Initial fully composed size preview ARM16193768B with safe folding; actual static sizes are in this gate. No routing/switch/300ms acceptance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-http-wiring-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18176016 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16193768 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [277.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.93s]

Completed run5-http-wiring at e49f06cae012b7bafca99dc0da4d4504941148af. Failing-first absence evidence run5-http-absence-red.log (422 versus expected404). Existing real paired/single-use enrollment test moved to the relay crate. Initial fully composed size preview ARM16193768B with safe folding; actual static sizes are in this gate. No routing/switch/300ms acceptance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-http-wiring-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18176016 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16193768 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [277.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.93s]
kayg commented 2026-10-10 02:00:46 +00:00
Author
Owner
Copy link

Completed run5-http-absence-tests at b67d454776. Absence regression retained alongside real paired enrollment; red evidence run5-http-absence-red.log predates the wiring change (expected404, observed422). scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-http-absence-tests-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18176016 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16193768 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [206.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.64s]

Completed run5-http-absence-tests at b67d454776f7a57af4b0f6f91b6b8da9c4f88c11. Absence regression retained alongside real paired enrollment; red evidence run5-http-absence-red.log predates the wiring change (expected404, observed422). scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-http-absence-tests-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18176016 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16193768 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [206.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.64s]
kayg commented 2026-10-10 02:11:19 +00:00
Author
Owner
Copy link

Next additive signer interface for D198: retain the existing OpenSSL/validated-CSR/CA implementation and host profile in core; expose a trusted compiled LeafTemplate (namespace, optional DNS prefix, exclusive client/server purpose). The relay plugin owns its relay and relay-control constants and real certificate-purpose tests. No request/CSR name or extension becomes a template, and ledger authorization remains independent. Sequence: shared public synthetic CSR fixture; additive generic signer with compatibility wrappers; plugin identity module/test relocation; broker wiring and removal of core relay wrappers. Generated staging code parses in scratch preview only; no signer behavior changes applied yet. No foundational engine/library change or wire/enrollment schema change.

Next additive signer interface for D198: retain the existing OpenSSL/validated-CSR/CA implementation and host profile in core; expose a trusted compiled LeafTemplate (namespace, optional DNS prefix, exclusive client/server purpose). The relay plugin owns its relay and relay-control constants and real certificate-purpose tests. No request/CSR name or extension becomes a template, and ledger authorization remains independent. Sequence: shared public synthetic CSR fixture; additive generic signer with compatibility wrappers; plugin identity module/test relocation; broker wiring and removal of core relay wrappers. Generated staging code parses in scratch preview only; no signer behavior changes applied yet. No foundational engine/library change or wire/enrollment schema change.
kayg commented 2026-10-10 02:16:22 +00:00
Author
Owner
Copy link

Broker relocation gate refused: all Rust rows except static builds passed; ARM16297224B exceeds16196376B by100848B, versus composed HTTP16193768B; x8618288240B versus18176016B. Raw run5-plugin-broker-first-fast.log. No commit and no budget exception. Preserved draft and restored it; measuring cold preflight/redemption future erasure at the remaining cross-crate ledger boundary before retrying. SQL/authority ordering will remain unchanged; forwarding remains untouched.

Broker relocation gate refused: all Rust rows except static builds passed; ARM16297224B exceeds16196376B by100848B, versus composed HTTP16193768B; x8618288240B versus18176016B. Raw run5-plugin-broker-first-fast.log. No commit and no budget exception. Preserved draft and restored it; measuring cold preflight/redemption future erasure at the remaining cross-crate ledger boundary before retrying. SQL/authority ordering will remain unchanged; forwarding remains untouched.
kayg commented 2026-10-10 02:26:36 +00:00
Author
Owner
Copy link

Measured remaining cold ledger boundary: ARM broker-relocation candidate16297224 ->16194344B (saved102880B; limit16196376B,2032B spare). Same-window ABBA real SQLite/OpenSSL paired admission n200/mode: inline p50/p991315864/2368751ns; boxed1347045/2436977ns (+2.37%/+2.88% point estimates, no significance claim). Outer payload1832 ->424B, preflight168B, redemption1424B; sequential peak1848B/request versus1832, eight-request peak14784B excluding allocator headers. No idle/forwarding allocation. Machine infraternal-dev x86_64,8cores;CPUWeight10000; measurement load4.01/9.67/14.14. Evidence run5-remaining-boundary-bench.log, run5-remaining-boundary-arm-preview.log and summary. Temporary benchmark/preview sources restored. Full gate started for the boundary change before independently retrying module moves; no budget/profile change.

Measured remaining cold ledger boundary: ARM broker-relocation candidate16297224 ->16194344B (saved102880B; limit16196376B,2032B spare). Same-window ABBA real SQLite/OpenSSL paired admission n200/mode: inline p50/p991315864/2368751ns; boxed1347045/2436977ns (+2.37%/+2.88% point estimates, no significance claim). Outer payload1832 ->424B, preflight168B, redemption1424B; sequential peak1848B/request versus1832, eight-request peak14784B excluding allocator headers. No idle/forwarding allocation. Machine infraternal-dev x86_64,8cores;CPUWeight10000; measurement load4.01/9.67/14.14. Evidence run5-remaining-boundary-bench.log, run5-remaining-boundary-arm-preview.log and summary. Temporary benchmark/preview sources restored. Full gate started for the boundary change before independently retrying module moves; no budget/profile change.
kayg commented 2026-10-10 02:39:06 +00:00
Author
Owner
Copy link

Completed run5-ledger-remaining-box at 362ea595dd. Same-window ABBA real paired signing/SQLite n200 per mode: inline p50/p991315864/2368751ns; boxed1347045/2436977ns. Peak payload per request1832 versus1848B (outer424+redemption1424; preflight168); eight-request peak14784B excluding allocator overhead. No idle-session allocation. Machine infraternal-dev x86_64,8cores,CPUWeight10000; measurement load4.01/9.67/14.14. Raw run5-remaining-boundary-bench.log. ARM candidate=16194344B; unboxed boundary=16297224B; limit=16196376B; saved=102880B scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-ledger-remaining-box-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18175504 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16193960 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [220.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.15s]

Completed run5-ledger-remaining-box at 362ea595dddbc16f0da7dd847049ca45e7786c4a. Same-window ABBA real paired signing/SQLite n200 per mode: inline p50/p991315864/2368751ns; boxed1347045/2436977ns. Peak payload per request1832 versus1848B (outer424+redemption1424; preflight168); eight-request peak14784B excluding allocator overhead. No idle-session allocation. Machine infraternal-dev x86_64,8cores,CPUWeight10000; measurement load4.01/9.67/14.14. Raw run5-remaining-boundary-bench.log. ARM candidate=16194344B; unboxed boundary=16297224B; limit=16196376B; saved=102880B scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-ledger-remaining-box-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18175504 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16193960 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [220.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.15s]
kayg commented 2026-10-10 02:53:30 +00:00
Author
Owner
Copy link

Completed run5-plugin-broker at 68b909963f. Broker and existing real paired-enrollment tests relocated without changing schemas or signing/admission ordering. No production routing or acceptance benchmark claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-broker-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18175888 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16194344 bytes (baseline 15934232 + allowance 262144 = max 16196376)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [227.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.91s]

Completed run5-plugin-broker at 68b909963f14f4b628358bc0e93b4f19950fa996. Broker and existing real paired-enrollment tests relocated without changing schemas or signing/admission ordering. No production routing or acceptance benchmark claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-broker-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18175888 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16194344 bytes (baseline 15934232 + allowance 262144 = max 16196376) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [227.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.91s]
kayg commented 2026-10-10 03:09:43 +00:00
Author
Owner
Copy link

Ledger relocation gate refused: ARM16301896B exceeds16196376B by105520B (prior broker16194344B); x8618288304B versus18175888B. Other five Rust rows passed, including moved real SQLite authority tests. Raw run5-plugin-store-first-fast.log. No commit, budget exception or foundation change. Preserved draft; generating an ARM link map to identify the duplicate code before choosing the smallest shared boundary change. A --no-default-features normal dependency graph has zero relay-crate rows (run5-no-default-normal-deps.log); final no-relay build/run still pending.

Ledger relocation gate refused: ARM16301896B exceeds16196376B by105520B (prior broker16194344B); x8618288304B versus18175888B. Other five Rust rows passed, including moved real SQLite authority tests. Raw run5-plugin-store-first-fast.log. No commit, budget exception or foundation change. Preserved draft; generating an ARM link map to identify the duplicate code before choosing the smallest shared boundary change. A --no-default-features normal dependency graph has zero relay-crate rows (run5-no-default-normal-deps.log); final no-relay build/run still pending.
kayg commented 2026-10-10 03:17:51 +00:00
Author
Owner
Copy link

Link-map finding: moving the ledger creates an additional SQLx pool instantiation. Previous/new ARM maps show pool symbol rows362/437 and summed symbol bytes201924/257320; acquire copies2/3, PoolConnection drop copies2/3. Largest added acquire copy about14KB; other connection maintenance/return paths also duplicate. Evidence run5-head-arm.map and run5-store-arm.map (older map is diagnostic, not an exact adjacent baseline). Proposed additive shared database boundary: core owns opaque connection/transaction leases; plugins borrow SqliteConnection references and retain existing SQL/binds, while pooled ownership and its maintenance/drop futures remain in core. Existing SQLx/SQLite engine, cancellation and rollback stay unchanged. Measure real transactions and static candidate before committing; draft restored and preserved in scratch.

Link-map finding: moving the ledger creates an additional SQLx pool instantiation. Previous/new ARM maps show pool symbol rows362/437 and summed symbol bytes201924/257320; acquire copies2/3, PoolConnection drop copies2/3. Largest added acquire copy about14KB; other connection maintenance/return paths also duplicate. Evidence run5-head-arm.map and run5-store-arm.map (older map is diagnostic, not an exact adjacent baseline). Proposed additive shared database boundary: core owns opaque connection/transaction leases; plugins borrow SqliteConnection references and retain existing SQL/binds, while pooled ownership and its maintenance/drop futures remain in core. Existing SQLx/SQLite engine, cancellation and rollback stay unchanged. Measure real transactions and static candidate before committing; draft restored and preserved in scratch.
kayg commented 2026-10-10 03:55:39 +00:00
Author
Owner
Copy link

Database-boundary candidates remain uncommitted: connection/transaction ownership alone produced ARM16235096B (saved66800B; excess38720B). Opaque pool ownership produced16235912B, adding816B without removing residual duplication, so that prototype was removed. Its map restores pool acquire/drop copies3->2, but adds SQLite executor stream/flume command copies (largest added stream8236B). Keeping execution and opaque row ownership in core now produces16206424B: saved95472B versus direct-ledger16301896B; still10048B over16196376B. Raw current candidate run5-database-executor-arm-preview.log and summary; opaque-pool evidence run5-database-pool-first-preview.log/summary and run5-leased-store-arm.map. A new executor map is running. No commit, engine/library choice, SQL/bind change, profile or budget exception.

Database-boundary candidates remain uncommitted: connection/transaction ownership alone produced ARM16235096B (saved66800B; excess38720B). Opaque pool ownership produced16235912B, adding816B without removing residual duplication, so that prototype was removed. Its map restores pool acquire/drop copies3->2, but adds SQLite executor stream/flume command copies (largest added stream8236B). Keeping execution and opaque row ownership in core now produces16206424B: saved95472B versus direct-ledger16301896B; still10048B over16196376B. Raw current candidate run5-database-executor-arm-preview.log and summary; opaque-pool evidence run5-database-pool-first-preview.log/summary and run5-leased-store-arm.map. A new executor map is running. No commit, engine/library choice, SQL/bind change, profile or budget exception.
kayg commented 2026-10-10 04:18:02 +00:00
Author
Owner
Copy link

Exact adjacent baseline map complete at 68b909963f. Normalizing crate names and compiler suffixes identifies +5228B in the invitation poll function; opaque executor candidate symbol sum grows9396B overall. Maps: run5-broker-baseline-arm.map and run5-executed-store-arm.map; normalized delta run5-normalized-map-delta.txt under /home/kayg/.cache/codex-scratch/relay-direct. Local non-inlined token generation/hash candidate16205752B ARM (limit16196376B),256B worse than local-unboxed candidate; discarded. Raw run5-token-codegen-arm-preview.log. Shared existing enrollment token producer comparison running. No budget, engine or wire-format changes; ledger remains in core pending a passing candidate.

Exact adjacent baseline map complete at 68b909963f14f4b628358bc0e93b4f19950fa996. Normalizing crate names and compiler suffixes identifies +5228B in the invitation poll function; opaque executor candidate symbol sum grows9396B overall. Maps: run5-broker-baseline-arm.map and run5-executed-store-arm.map; normalized delta run5-normalized-map-delta.txt under /home/kayg/.cache/codex-scratch/relay-direct. Local non-inlined token generation/hash candidate16205752B ARM (limit16196376B),256B worse than local-unboxed candidate; discarded. Raw run5-token-codegen-arm-preview.log. Shared existing enrollment token producer comparison running. No budget, engine or wire-format changes; ledger remains in core pending a passing candidate.
kayg commented 2026-10-10 04:22:27 +00:00
Author
Owner
Copy link

Contract proposal under docs/PROCESS.md section7 (before any baseline edit): preserve the existing ledger implementation and relocate it into the removable relay plugin, then record its intentional target-specific static size. Measured pure relocation: x86_64-musl18288304B (+112416B vs68b90996; below19244272B ceiling), ARM-musl16301896B (+107552B vs68b90996;105520B beyond16196376B ceiling). Raw run5-plugin-store-first-fast.log: every other Rust gate row passed, existing ten real-SQLite ledger tests preserved. Proposed ARM measured baseline16301896B with the unchanged262144B allowance (maximum16564040B); x86 baseline remains unchanged. PROCESS explicitly requires target measurements and issue rationale before an intentional baseline change. No owner decision, release profile, crypto/database engine, parser or authority check changes. Rationale: D198 necessarily splits existing pool/transaction state machines across crate boundaries; opaque connection/executor/row facade avoids about100KiB but adds boxed operations, row dispatch and another database API solely for static size. Shared token candidate16200760B still fails by4384B (run5-shared-token-arm-preview.log). Discarding that facade and token experiments favors correctness and runtime latency. Same-window real-SQLite core/plugin invitation benchmark now running under CPUWeight10000 on infraternal-dev before the baseline update; no forwarding/routing acceptance claim.

Contract proposal under docs/PROCESS.md section7 (before any baseline edit): preserve the existing ledger implementation and relocate it into the removable relay plugin, then record its intentional target-specific static size. Measured pure relocation: x86_64-musl18288304B (+112416B vs68b90996; below19244272B ceiling), ARM-musl16301896B (+107552B vs68b90996;105520B beyond16196376B ceiling). Raw run5-plugin-store-first-fast.log: every other Rust gate row passed, existing ten real-SQLite ledger tests preserved. Proposed ARM measured baseline16301896B with the unchanged262144B allowance (maximum16564040B); x86 baseline remains unchanged. PROCESS explicitly requires target measurements and issue rationale before an intentional baseline change. No owner decision, release profile, crypto/database engine, parser or authority check changes. Rationale: D198 necessarily splits existing pool/transaction state machines across crate boundaries; opaque connection/executor/row facade avoids about100KiB but adds boxed operations, row dispatch and another database API solely for static size. Shared token candidate16200760B still fails by4384B (run5-shared-token-arm-preview.log). Discarding that facade and token experiments favors correctness and runtime latency. Same-window real-SQLite core/plugin invitation benchmark now running under CPUWeight10000 on infraternal-dev before the baseline update; no forwarding/routing acceptance claim.
kayg commented 2026-10-10 04:34:59 +00:00
Author
Owner
Copy link

Production control gap confirmed: crates/infraternal-server/src/remote_runtime.rs242-246 and270-281 route all mutually authenticated native peers through ordinary EnrollmentStore/AuthorityGuard admission. RelayStore.admit_control already validates the isolated relay_control_certificates ledger (relay_store.rs356); using ordinary host admission for it would violate existing purpose isolation. Contract proposal before extension: additive registered native transport service beside the existing HTTP ServerService, using frozen Hello/Call/Result frames, bounded admission and the shared node retirement facility. Preserve existing host admission as fallback, and dispatch relay-control traffic only through the compiled plugin. Context also needs generic node action dispatch for D132: current EmbeddedAgent.control_call is crate-private (embedded.rs190), and ServerContext exposes only auth/pool/source/endpoint (plugins.rs16). No new transport/crypto engine or wire tag proposed. Existing host retirement signals are tied to clones of the same EnrollmentStore (remote_guard.rs18,91); reconstructing a store from the pool would miss immediate revocation, so production plugin wiring must retain that shared store.

Production control gap confirmed: crates/infraternal-server/src/remote_runtime.rs242-246 and270-281 route all mutually authenticated native peers through ordinary EnrollmentStore/AuthorityGuard admission. RelayStore.admit_control already validates the isolated relay_control_certificates ledger (relay_store.rs356); using ordinary host admission for it would violate existing purpose isolation. Contract proposal before extension: additive registered native transport service beside the existing HTTP ServerService, using frozen Hello/Call/Result frames, bounded admission and the shared node retirement facility. Preserve existing host admission as fallback, and dispatch relay-control traffic only through the compiled plugin. Context also needs generic node action dispatch for D132: current EmbeddedAgent.control_call is crate-private (embedded.rs190), and ServerContext exposes only auth/pool/source/endpoint (plugins.rs16). No new transport/crypto engine or wire tag proposed. Existing host retirement signals are tied to clones of the same EnrollmentStore (remote_guard.rs18,91); reconstructing a store from the pool would miss immediate revocation, so production plugin wiring must retain that shared store.
kayg commented 2026-10-10 04:37:37 +00:00
Author
Owner
Copy link

Completed run5-relay-size-baseline at 755c24d411. Pre-relocation68b90996 x86/ARM18175888/16194344B; pure relocation18288304/16301896B. ARM new maximum16564040B. Existing ten real-SQLite ledger tests pass. Raw run5-pure-ledger-bench.log; module remains unmoved in this gate. Same-window n2000 ABBA core/plugin benchmark: test store::tests::admission_interleaved ... MEASURE same-window invitations arch=x86_64 cores=8 load=13.83 15.08 12.19 3/9680 1672390
MEASURE same-window invitations mode=0 n=2000 p50_ns=97690 p99_ns=281905
MEASURE same-window invitations mode=1 n=2000 p50_ns=97898 p99_ns=281203 scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-relay-size-baseline-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18175888 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16194344 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [175.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.88s]

Completed run5-relay-size-baseline at 755c24d4115f8547abc7aab968d05f05d33b2fd2. Pre-relocation68b90996 x86/ARM18175888/16194344B; pure relocation18288304/16301896B. ARM new maximum16564040B. Existing ten real-SQLite ledger tests pass. Raw run5-pure-ledger-bench.log; module remains unmoved in this gate. Same-window n2000 ABBA core/plugin benchmark: test store::tests::admission_interleaved ... MEASURE same-window invitations arch=x86_64 cores=8 load=13.83 15.08 12.19 3/9680 1672390 MEASURE same-window invitations mode=0 n=2000 p50_ns=97690 p99_ns=281905 MEASURE same-window invitations mode=1 n=2000 p50_ns=97898 p99_ns=281203 scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-relay-size-baseline-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18175888 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16194344 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [175.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.88s]
kayg commented 2026-10-10 04:38:48 +00:00
Author
Owner
Copy link

Documentation-only gate failed real_binary_preserves_input_and_output_across_24_abrupt_stops at agent_daemon.rs18 / scripts/verify-agent.py99: terminal command queue full; reattach and replay unacknowledged input. Test0pass/1fail in11.70s; other five Rust rows passed, unchanged static sizes18175888/16194344B. Raw run5-relay-size-docs-fast.log. The immediately preceding full gate for755c24d4 passed this same test; only docs/PROCESS.md is pending. This is outside the #118 exception list, so no baseline exception, timeout or production modification accepted. Pipeline stopped before ledger migration. Rechecking the exact test, then repeating the full documentation gate.

Documentation-only gate failed real_binary_preserves_input_and_output_across_24_abrupt_stops at agent_daemon.rs18 / scripts/verify-agent.py99: terminal command queue full; reattach and replay unacknowledged input. Test0pass/1fail in11.70s; other five Rust rows passed, unchanged static sizes18175888/16194344B. Raw run5-relay-size-docs-fast.log. The immediately preceding full gate for755c24d4 passed this same test; only docs/PROCESS.md is pending. This is outside the #118 exception list, so no baseline exception, timeout or production modification accepted. Pipeline stopped before ledger migration. Rechecking the exact test, then repeating the full documentation gate.
kayg commented 2026-10-10 04:41:32 +00:00
Author
Owner
Copy link

Exact daemon restart test recheck passed1/1 in41.20s, unchanged code; raw run5-docs-daemon-recheck.log. No baseline exception or production/test timeout change. Complete documentation gate retry running, then the unchanged ledger relocation if green.

Exact daemon restart test recheck passed1/1 in41.20s, unchanged code; raw run5-docs-daemon-recheck.log. No baseline exception or production/test timeout change. Complete documentation gate retry running, then the unchanged ledger relocation if green.
kayg commented 2026-10-10 04:47:41 +00:00
Author
Owner
Copy link

Completed run5-relay-size-docs-retry at e7e6ba49db. Published ARM baseline16301896B/maximum16564040B agrees with scripts/verify.d/rust.sh; target numbers and rationale recorded on #103 before the gate change. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-relay-size-docs-retry-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18175888 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16194344 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [201.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.77s]

Completed run5-relay-size-docs-retry at e7e6ba49dbde3984ff5487dbffee9b4289ee7221. Published ARM baseline16301896B/maximum16564040B agrees with scripts/verify.d/rust.sh; target numbers and rationale recorded on #103 before the gate change. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-relay-size-docs-retry-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18175888 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16194344 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [201.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.77s]
kayg commented 2026-10-10 04:59:41 +00:00
Author
Owner
Copy link

Ledger relocation gate: static builds PASS x86_64-musl18288304B, ARM-musl16301896B; the other four non-test rows also PASS. Rust test row failed only auth::tests::socket_tests::expired_handshake_releases_user_permits at socket_tests.rs151 (expected503 at the expiry boundary): server suite146pass/1fail/17ignored in118.18s. Raw run5-plugin-store-fast.log. Earlier progress incorrectly said all Rust tests passed while following the log tail; final gate summary supersedes it. No ledger commit. Applying the prescribed #118 same-window audit: dev advanced from the earlier e704e5b8 baseline to7e8545c4b9bf7a8602dc2c79b53bb92264ce6f95; refreshed only this job's detached cache baseline and building/running the exact test there before interleaving it with the changed binary. No baseline exception or timeout change accepted yet.

Ledger relocation gate: static builds PASS x86_64-musl18288304B, ARM-musl16301896B; the other four non-test rows also PASS. Rust test row failed only auth::tests::socket_tests::expired_handshake_releases_user_permits at socket_tests.rs151 (expected503 at the expiry boundary): server suite146pass/1fail/17ignored in118.18s. Raw run5-plugin-store-fast.log. Earlier progress incorrectly said all Rust tests passed while following the log tail; final gate summary supersedes it. No ledger commit. Applying the prescribed #118 same-window audit: dev advanced from the earlier e704e5b8 baseline to7e8545c4b9bf7a8602dc2c79b53bb92264ce6f95; refreshed only this job's detached cache baseline and building/running the exact test there before interleaving it with the changed binary. No baseline exception or timeout change accepted yet.
kayg commented 2026-10-10 05:01:53 +00:00
Author
Owner
Copy link

Fresh dev7e8545c4 exact auth socket test passed1/1 in0.83s (run5-store-dev-socket.log). Same-scope interleaved audit then passed12/12 dev and12/12 changed runs; CPUWeight10000, machine infraternal-dev,8cores, load17.97–22.08. Raw run5-store-authsocket-window-00..11-{dev,changed}.log and run5-store-authsocket-window-results.json; summary run5-store-authsocket-window-summary.log. No matching baseline failure reproduced, so no #118 exception accepted. Repeating the entire unchanged ledger gate, retaining original run5-plugin-store-fast.log failure.

Fresh dev7e8545c4 exact auth socket test passed1/1 in0.83s (run5-store-dev-socket.log). Same-scope interleaved audit then passed12/12 dev and12/12 changed runs; CPUWeight10000, machine infraternal-dev,8cores, load17.97–22.08. Raw run5-store-authsocket-window-00..11-{dev,changed}.log and run5-store-authsocket-window-results.json; summary run5-store-authsocket-window-summary.log. No matching baseline failure reproduced, so no #118 exception accepted. Repeating the entire unchanged ledger gate, retaining original run5-plugin-store-fast.log failure.
kayg commented 2026-10-10 05:08:36 +00:00
Author
Owner
Copy link

Completed run5-plugin-store-retry at 96a9f7b5ee. Ledger and existing SQLite tests relocated; frozen migration files untouched. Joint current authority and request bounds preserved. Scale and routing acceptance remain pending. Same-window real-SQLite ABBA invitations (n=2000 each): core p50/p99=97.690/281.905us, plugin97.898/281.203us. Machine infraternal-dev, x86_64,8cores, CPUWeight10000, load13.83/15.08/12.19. Point estimates do not establish a material latency difference. Raw run5-pure-ledger-bench.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-store-retry-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288304 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301896 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [278.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.16s]

Completed run5-plugin-store-retry at 96a9f7b5eedd78977ee1c72f96fb7343f71bf257. Ledger and existing SQLite tests relocated; frozen migration files untouched. Joint current authority and request bounds preserved. Scale and routing acceptance remain pending. Same-window real-SQLite ABBA invitations (n=2000 each): core p50/p99=97.690/281.905us, plugin97.898/281.203us. Machine infraternal-dev, x86_64,8cores, CPUWeight10000, load13.83/15.08/12.19. Point estimates do not establish a material latency difference. Raw run5-pure-ledger-bench.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-store-retry-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288304 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301896 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [278.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.16s]
kayg commented 2026-10-10 05:15:24 +00:00
Author
Owner
Copy link

Shared CSR fixture gate failed the unchanged abrupt-stop oracle after38.97s; stopped only owned scope run-p2299656-i31637579.scope. Raw run5-shared-csr-fixture-fast.log. At iteration22/before-forward, all468224 payload bytes arrived, input fence remained17bytes unacknowledged, and the terminal reported command queue full. Evidence: daemon.rs336 bounds commands to8; agent_fixture.py95 refunds each64KiB; verify-agent.py replays payload plus a separate fence while intentionally holding acknowledgements. Those unnecessary refunds can consume the queue before the fence. Fixing the oracle, not queue capacity/policy: fund the bounded payload window between completed iterations and consume within-iteration output without auto-refunds. Keep exact output equality, offsets, acknowledgement/fence conditions and timeouts. Original failing real-binary test is the failing-first proof. Saved/reverted the pending CSR fixture files so this is an isolated atomic test fix; exact real-binary recheck running before full gate. No #118 exception applies.

Shared CSR fixture gate failed the unchanged abrupt-stop oracle after38.97s; stopped only owned scope run-p2299656-i31637579.scope. Raw run5-shared-csr-fixture-fast.log. At iteration22/before-forward, all468224 payload bytes arrived, input fence remained17bytes unacknowledged, and the terminal reported command queue full. Evidence: daemon.rs336 bounds commands to8; agent_fixture.py95 refunds each64KiB; verify-agent.py replays payload plus a separate fence while intentionally holding acknowledgements. Those unnecessary refunds can consume the queue before the fence. Fixing the oracle, not queue capacity/policy: fund the bounded payload window between completed iterations and consume within-iteration output without auto-refunds. Keep exact output equality, offsets, acknowledgement/fence conditions and timeouts. Original failing real-binary test is the failing-first proof. Saved/reverted the pending CSR fixture files so this is an isolated atomic test fix; exact real-binary recheck running before full gate. No #118 exception applies.
kayg commented 2026-10-10 05:27:48 +00:00
Author
Owner
Copy link

Completed run5-restart-credit-oracle at 6ef92b2bc6. Failing-first run5-shared-csr-fixture-fast.log: iteration 22 received all 468224 payload bytes, then refused before the final 17-byte fence acknowledgement. Exact fixed real-binary recheck passed 1/1 in 39.78 s: 24 abrupt stops, four positions each six times, 4710396 payload bytes verified. Raw run5-restart-credit-fixed.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-restart-credit-oracle-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288304 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301896 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [190.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.75s]

Completed run5-restart-credit-oracle at 6ef92b2bc61ff05566fa6664b34a29b30aee0bec. Failing-first run5-shared-csr-fixture-fast.log: iteration 22 received all 468224 payload bytes, then refused before the final 17-byte fence acknowledgement. Exact fixed real-binary recheck passed 1/1 in 39.78 s: 24 abrupt stops, four positions each six times, 4710396 payload bytes verified. Raw run5-restart-credit-fixed.log. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-restart-credit-oracle-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288304 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301896 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [190.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.75s]
kayg commented 2026-10-10 05:40:50 +00:00
Author
Owner
Copy link

Shared CSR fixture retry: test and static rows PASS (x86/ARM18288304/16301896B), Clippy rejected public requested_extensions_csr at test_support.rs198 without must_use. Raw run5-shared-csr-fixture-retry-fast.log. Earlier progress incorrectly called Clippy green; final table supersedes it. Added must_use to the proof-returning helper; no production change or fixture assertion relaxed. Full gate retry follows before commit; original failed logs retained.

Shared CSR fixture retry: test and static rows PASS (x86/ARM18288304/16301896B), Clippy rejected public requested_extensions_csr at test_support.rs198 without must_use. Raw run5-shared-csr-fixture-retry-fast.log. Earlier progress incorrectly called Clippy green; final table supersedes it. Added must_use to the proof-returning helper; no production change or fixture assertion relaxed. Full gate retry follows before commit; original failed logs retained.
kayg commented 2026-10-10 05:47:24 +00:00
Author
Owner
Copy link

Completed run5-shared-csr-fixture-lint-retry at 4a4479c83c. Existing real OpenSSL host/relay purpose and CSR-extension exclusion tests preserved; production signer unchanged. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-shared-csr-fixture-lint-retry-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288304 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301896 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [149.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.61s]

Completed run5-shared-csr-fixture-lint-retry at 4a4479c83cb9bbd121189e693731f2bb70f21c5a. Existing real OpenSSL host/relay purpose and CSR-extension exclusion tests preserved; production signer unchanged. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-shared-csr-fixture-lint-retry-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288304 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301896 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [149.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.61s]
kayg commented 2026-10-10 06:02:29 +00:00
Author
Owner
Copy link

Completed run5-generic-signer at e82c97f62d. Additive signer interface proposed on #103 before changes. Existing OpenSSL chain/purpose/name exclusion and time/issuer tests gate the preserved profiles; no wire or enrollment schema change. Same-window real signing measurements: test ca::tests::trusted_template_interleaved ... MEASURE template machine=infraternal-dev arch=x86_64 cores=8 load=11.52 12.13 14.00 20/9852 2879207
MEASURE template profile=0 mode=0 n=1000 p50_ns=259182 p99_ns=572214 total_der_bytes=398000
MEASURE template profile=0 mode=1 n=1000 p50_ns=257677 p99_ns=535508 total_der_bytes=398000
MEASURE template profile=1 mode=0 n=1000 p50_ns=258836 p99_ns=814521 total_der_bytes=454000
MEASURE template profile=1 mode=1 n=1000 p50_ns=258821 p99_ns=691536 total_der_bytes=454000
MEASURE template profile=2 mode=0 n=1000 p50_ns=260588 p99_ns=485929 total_der_bytes=409000
MEASURE template profile=2 mode=1 n=1000 p50_ns=260218 p99_ns=520408 total_der_bytes=409000 scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-generic-signer-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288144 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301800 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [157.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.93s]

Completed run5-generic-signer at e82c97f62d020780fecfd2d9208345d99e0344cc. Additive signer interface proposed on #103 before changes. Existing OpenSSL chain/purpose/name exclusion and time/issuer tests gate the preserved profiles; no wire or enrollment schema change. Same-window real signing measurements: test ca::tests::trusted_template_interleaved ... MEASURE template machine=infraternal-dev arch=x86_64 cores=8 load=11.52 12.13 14.00 20/9852 2879207 MEASURE template profile=0 mode=0 n=1000 p50_ns=259182 p99_ns=572214 total_der_bytes=398000 MEASURE template profile=0 mode=1 n=1000 p50_ns=257677 p99_ns=535508 total_der_bytes=398000 MEASURE template profile=1 mode=0 n=1000 p50_ns=258836 p99_ns=814521 total_der_bytes=454000 MEASURE template profile=1 mode=1 n=1000 p50_ns=258821 p99_ns=691536 total_der_bytes=454000 MEASURE template profile=2 mode=0 n=1000 p50_ns=260588 p99_ns=485929 total_der_bytes=409000 MEASURE template profile=2 mode=1 n=1000 p50_ns=260218 p99_ns=520408 total_der_bytes=409000 scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-generic-signer-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288144 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301800 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [157.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.93s]
kayg commented 2026-10-10 06:19:02 +00:00
Author
Owner
Copy link

Completed run5-template-label-tests at d330b2d983. Two tests cover36 refused declarations and a real OpenSSL signed63-byte boundary certificate; no new identity or protocol format. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-template-label-tests-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288144 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301800 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [194.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.48s]

Completed run5-template-label-tests at d330b2d9837b1c4e80e84bdd5bd981e2f5dd5fda. Two tests cover36 refused declarations and a real OpenSSL signed63-byte boundary certificate; no new identity or protocol format. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-template-label-tests-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288144 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301800 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [194.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.48s]
kayg commented 2026-10-10 06:31:19 +00:00
Author
Owner
Copy link

Completed run5-plugin-identity at 55aad8a511. Two existing real chain/purpose tests relocated with declarations; endpoint/control namespaces and validity unchanged. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-identity-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288144 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301800 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [160.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.75s]

Completed run5-plugin-identity at 55aad8a511c5cfad2c4cff6ab6b22dfd0c357291. Two existing real chain/purpose tests relocated with declarations; endpoint/control namespaces and validity unchanged. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-identity-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288144 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301800 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [160.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.75s]
kayg commented 2026-10-10 06:49:15 +00:00
Author
Owner
Copy link

Completed run5-plugin-signer-wiring at 9c71690483. Paired enrollment, exclusive purpose and single-use/current-authority tests gate composition; no forwarding or routing acceptance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-signer-wiring-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288240 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301880 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [214.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [3.25s]

Completed run5-plugin-signer-wiring at 9c71690483ea8ed5db5847c2f7ce707a94fd2860. Paired enrollment, exclusive purpose and single-use/current-authority tests gate composition; no forwarding or routing acceptance claim. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-plugin-signer-wiring-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288240 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301880 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [214.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [3.25s]
kayg commented 2026-10-10 07:04:24 +00:00
Author
Owner
Copy link

Completed run5-native-service-api at f168608417. Inert interface only; no tasks, listeners, timers or new dependencies. Native control proposal recorded on #103 before this extension; production composition and real socket tests follow. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-native-service-api-fast.log. Verbatim gate output:

+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| VT native fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+
PT_INTERP: absent; size: 18288240 bytes (baseline 18982128 + allowance 262144 = max 19244272)
PT_INTERP: absent; size: 16301880 bytes (baseline 16301896 + allowance 262144 = max 16564040)
20 pass
0 fail
376 expect() calls
Ran 20 tests across 6 files. [317.00ms]
15 pass
0 fail
34537 expect() calls
Ran 15 tests across 3 files. [2.57s]

Completed run5-native-service-api at f1686084174383a06f9037012b6a1e19a3407afb. Inert interface only; no tasks, listeners, timers or new dependencies. Native control proposal recorded on #103 before this extension; production composition and real socket tests follow. scripts/verify fast exit0; raw /home/kayg/.cache/codex-scratch/relay-direct/run5-native-service-api-fast.log. Verbatim gate output: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ PT_INTERP: absent; size: 18288240 bytes (baseline 18982128 + allowance 262144 = max 19244272) PT_INTERP: absent; size: 16301880 bytes (baseline 16301896 + allowance 262144 = max 16564040) 20 pass 0 fail 376 expect() calls Ran 20 tests across 6 files. [317.00ms] 15 pass 0 fail 34537 expect() calls Ran 15 tests across 3 files. [2.57s]
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
j470b-server-mesh-rejoin
mesh-compat-382
j444b-d381-key-wins-flood
j461b-macos-utun-reader-thread
j475b-remove-host-motion
fix-reconnect-new-link
j481-d380
dev
j479-mesh-soak
fix/j476-host-facts
j475-motion-pill-split
j470-mac-node-join
root-upgrade-d376
j474-review-merge
latency-targets-120hz
nightly-suite-shards
gotatun-pin-j469
macos-utun-reader-thread
fix-flood-echo-stall
wip/j416
motion-fast-ease-out
mesh-addhost-j259
takeover-close-terminals-d368
kbd-instant-followups
fix-trust-key-double-click
nightly-suite-partial-report
appwin-icons-j207
rusqlite-slice1
swipe-content-first-d374
mesh-mtu-381
addhost-instant-open
glass-contrast-d365-mock
glass-contrast-d365
signout-followups-442
fix-startup-toast-red
research-gotatun-matrix-j317
themes-sync-j304
fix-load-flakes
fix-nix-zig-patch
fix-compute01-browser-baseline
ratchet-size-baseline-419
predict-auto-d130
revoke-coverage-d367
nightly-suite-queue-timeout
nightly-browser-suite
fix-rustsec-2026-0459
fonts-ioskeley-obel
log-retention-d375
fix-kbd-no-anim-427
fix-dev-red-specs-2
takeover-proof-305-j205
wip/j410-server-pump
ui-consistency-d372
fix-osc52-clipboard
libghostty-vt-1.0
forget-retired-host-d369
j404-merge-426
fix-setup-lock-race
fix-touch-select-426
revoke-access-report-ui
fix-predictive-echo-specs
fix-selinux-agent-exec
wip/j395-checkpoint2
wip/j395
j391-review-scratch
ui-consistency-j195-deferred-rebased
fix-rustc-paths-421
decisions-wording-341
fix-install-suite-red
fix-dev-rtt-specs
fix-pending-removal-align
fix-privileged-skip
fix-install-suite-404
review-j368
mesh-dns-j254
revoke-access-report-d356b
revoke-privilege-round5
fix-removal-confirm
review-j360
gotatun-rotation-367
deps-refresh-340
review-j354
review-J351
review-j350
addhost-takeover-d359
revoke-report-items-347
revoke-end-sessions-d362
review-j348
size-mesh-ssh-346
size-trim-337
revoke-access-diff-d356
j259-pre-rebase-b06c5a6
j305-ended-at-start-spec
addhost-sudo-consent
fix-layout-real-specs-411
fix-shell-open-411
fix-console-hosts-373
fix-shell-exit
fuzz-host-name-353
fix-host-details-410
macos-filter-j197
research-gotatun-tricks
revoke-p3-j216l
stale-state-pill
revoke-followups-j216k
revoke-progress-j216e
backup/macos01-20261106/lume/j197
backup/macos01-20261106/lume/appwin-icons-j207-stash0
backup/macos01-20261106/claude/stash-0
backup/macos01-20261106/claude/stash-1
backup/macos01-20261106/claude/J80-macos-s1
backup/macos01-20261106/claude/J80
backup/macos01-20261106/claude/J80-slice1-20261102
backup/macos01-20261106/claude/J16b-appwin-on-dev
backup/macos01-20261106/claude/appwin-launch-idle-146
backup/macos01-20261106/claude/J80-s1-run
wip/j259
trust-285-p3c
runtime-lock-j216f
mesh-ssh-p3-j175
tun-proof-p3b
trust-285-p3
terminal-polish-j271
tun-proof-p3
ssh-install-flakes-j216i
fuzz-fixes-j264
trust-285-a
trust-285-b
restart-oversized-fixture-j198
backup/macos01-20261106/lume/appwin-icons-j207
restart-bound-j216h
up-preflight-p3-j198
backup/macos01-20261106/lume/j251
up-preflight-j198
wip/up-preflight-j198-j245-p2
wip/up-preflight-j198-j245-p1
agent-fsync-offloop-j196
wip/up-preflight-j198-j245
settings-offscreen-j231
wip/up-preflight-j198-j226-inspection
wip/j103f
themes-j206
wip/up-preflight-j198-j226
flakes-j216c
backup/macos01-20261106/claude/j197
flakes-j216b
backup/macos01-20261106/claude/j197-red
wip/up-preflight-j198-review
flakes-j216
backup/macos01-20261106/claude/appwin-icons-j207
engineering-j208b
appwin-mock-j209
toast-stack-j165
engineering-j208
decisions-d351
decisions-d350
decisions-d349
appearance-presets-mock-j129
decisions-d348
docs-audit-spec
doc-coverage-baseline-raise
decisions-d347
decisions-d346b
decisions-d346
j88h-profile
ui-consistency-j195-deferred
ui-consistency-j195
engineering-practices-j194
preferred-login-ui-j166
signin-pools-263-j174
macos-node-265-s1
decisions-d345
decisions-d344
proxy-cgroup-j164
verify-quick-fuzz-329
proxy-start-io-j168
timing-flakes-j163
up-263-s1
mcp-events-j147
verify-quick-fuzz-j158
proxy-p3-j159
agent-resource-j157-detect
wip/j103d-final
remote-cancel-hang-j155
proxy-traefik-281-j62d
j149-verify-quick
proxy-traefik-281-j62b
j150-drop-unnamed
wip/j79d-red-r4
wip/j103d-d332
rootless-toast-j142
wip/j103d-head
wip/j103d
j125c-mcp
macos-build-j136
backup/macos01-20261106/claude/j136
wip/j79d-red-cli
wip/j79d-red-server
j121-link-liveness
j118c-tracer-failures
wip/j79c-head
release-desktop-j131
j124c-signed-macos
wip/j79c-red-p3
wip/j79c-red-p23
wip/j79c-red-p22
merge-j128-check
perf-deferred-j128
sidebar-toggle-260
j127-shortcut-changed-dot
j125-mcp
j118-tracer-failures
j126-account-summary
settings-sections-mock-j92
j120-pinch-lock
sidebar-292
backup/macos01-20261106/claude/j120
taste-2026-11-03
row-actions-242-215
backup/macos01-20261106/claude/stash-0-untracked
backup/macos01-20261106/claude/J80-265-s1-run
backup/macos01-20261106/claude/J80-macos-s1-20261102
j112-canvas-commands
ui-no-overshoot-scroll
fix-recovery-restart
backup/macos01-20261106/claude/stash-1-untracked
taste-2026-11-02
size-baseline-ratchet
machine-cards-structure-j108
j109-boot-copy-mockups
fix-dev-red
wip/up-263-s1
j95b-settings-rows
proxy-traefik-281
j95-option-rows
docs-human-j91
docs-keys-palette
fix-flood-1013-270-j38d
mesh-ssh-s1
mesh-ssh-s0
addhost-d291-tunnel
test-stability-load
mesh-relay-220
repro-296
shim-thp-off-24
mesh-pins-258-v2
mesh-dns
signin-d311-merged
docs-pages
ui-size-283
fix-ui-batch-specs
wip/j12b-edge-update
gate-on-arm
fix-rename-sync-293
trust-285-j60b
fix-focus-ring-clip
integ-ui-1101
edge-update-v2
signin-d311
taste-2026-11-01
layout-per-tab-250
ui-approved-splitopt-224
ui-approved-splitopt-224-on-dev
icon-text-centre-282
ui-approved-189-212-216
research-reverse-proxy
transfer-budget-320
review-mesh-pins-258-r2
sidebar-pill-259
review-mesh-relay-220
wip/j12-edge-update-v2
nix-package
status-crash-loop-237
fix-dev-test-failures
fix-fixture-socket-path
wip/mesh-relay-220-pre-dev-advance-rebase
fix-canvas-side-clip
research-head-to-head
frosted-real-256
review-mesh-pins-258
wip/mesh-relay-220-pre-final-rebase
review-layout-per-tab-250-r2
broad-suite-triage
nix-devshell
review-status-crash-loop-237-r3
wip/nix-devshell-pre-p3
taste-frosted-real
wip/code-serve-demo-2026-10-30
wip/code-serve-glass-2026-10-30
wip/code-serve-clusters-2026-10-30
wip/code-merge-release94-dev-2026-10-30
wip/rproxy-delta-scratch-2026-10-30
wip/stale-note-257-staged-2026-10-30
wip/d241-size-staged-2026-10-30
review-broad-suite-207
wip/mesh-pins-258-pre-split
wip/mesh-relay-220-pre-rebase
mesh-measure-220
plan-211-update
review-edge-autoupdate-r3
perf24-reconnect
review-stale-note-r2
release-prod-live
d241-shell-marker
review-mesh-pins-server-258
review-revocation-r3
review-setup-token-r2
review-sync-journal-r3
wip/review-mesh-pins-258-prune-review-1995fee46
frosted-scroll-256
sync-journal-r3
revocation-per-account-wip
review-tokio-asyncfd-0386
review-sync-p3s
wip/stash-8by64-stash0-2026-10-30
mesh-drag-r2
setup-token-data-wip
release-9-4
auth-busy-1013
mesh-view-overlap
sync-p3s
review-auth-busy-wake
review-readd-r8-wip
mesh-drag-bounds
audit-auth-93
app-windows-stream-start
review-authority-cap-r3
app-windows-r7-measure
readd-minimal-r6
readd-minimal-r5
app-windows-r7-soak
readd-minimal-r4
app-windows-r7
app-windows-r6
readd-minimal-r3
readd-minimal-r2
readd-minimal
merge-release8-dev
review-busy-page-liveness-r2
review-busy-page-liveness
mesh-view-trim
review-authority-read-retry-r2
review-remove-uninstall-r2
release-retry-reattach
release-retry
review-remove-uninstall
review-authority-read-retry
app-windows-ship
mesh-autoswitch-p2
mesh-ssh
mesh-relay
mesh-macos-node
screen-and-control
session-labels
review-real-logins-r5
review-real-logins-r4
review-real-logins-r3
shim-memory-r2
review-real-logins-r2
shim-memory
shadow-addhost-instant
shadow-real-logins
shadow-same-session-flood
shadow-prod-perf-fix
hosts-from-server
app-windows-real
app-windows-transport
release-prod-6w
mesh-policy-verify
quiet-close-toast
dev-merge-r4
review-mesh-agent-transport
review-app-windows-transport
settings-roundtrip
review-release-pipeline-r4
mesh-agent-transport
ended-strip-fix
review-audit-ca-update-d244
settings-rows
phone-width
signout-clear
fix-restart-tests
dev-merge-rp
signer-hardening
review-audit-ca-update
terminal-fidelity
review-audit-ca-update-wip
mesh-integrate
review-mesh-agent-transport-515d3712
review-mesh-dataplane-feec1988
mesh-dataplane
signout
research-canvas-node
wip/review-renderer-cost
wip/review-terminal-fidelity
padlock-glyph
review-release-pipeline-2e95b689
phone-polish
sidebar-polish
inbox-generic
review-mesh-dataplane-378d4055
prod-batch
ui-inbox
ui-ca-unlock
ui-machine-sections
review-mesh-dataplane-a5f744fb
tls-session-prod
session-bounds-prod
fix-dead-session
session-bounds
tls-session
palette-polish
shortcuts-d222
research-appmeta
phone-gutter
ui-centre-button
wip/ui-centre-button
wip/tls-session
review-addhost-breakit
ui-platform
renderer-cost
remote-agents
mesh-api
ssh-bootstrap
prod-hotfix2
prod-fixes
review-remote-agents
review-mesh-dataplane
session-life
research-browser
review-mesh-api
research-capture
review-mesh-api-round3
review-mesh-api-round1
gate-stability
review-gate-stability
review-relay-direct
review-mesh-api-findings
proto-sessions
font-picker
proto-addhost
proto-windows
mesh-core
latency-gate
research-bsd
research-discovery
keytray-mods
mock-core
ui-tweaks
fix-motion-sel
fix-motion
integrate2
fix-robust
t-web
t-server
vt-ghostty
fix-snap-echo
t-web-auth
review-vt2
t-agent
lane-protocol
lane-shim
lane-auth
s1-base
s1-contracts
s1-workspace
No results found.
Labels
Clear labels   
launch-qa
Findings from launch QA.

No labels
launch-qa
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/infraternal#103
No description provided.