Plan: merge order and conflicts for merge round 7b #867

Capture the merged API scopes in the owned performance ledger (#867)

kayg referenced this issue from a commit

Exercise Daily Log collaboration through stable Note identities (#606, #867)

kayg referenced this issue from a commit

Install Tags in the background migration fixture (#618, #867)

kayg referenced this issue from a commit

Keep Mail 626 SQL checks pinned to their reviewed migration (#867)

kayg referenced this issue from a commit

Return canonical stable Note links from Search (#867, DESIGN 33)

kayg referenced this issue from a commit

Record the complete fixed merge set and current verification evidence (#867)

kayg referenced this issue from a commit

Format the deployed migration identity fixture (#867)

kayg referenced this issue from a commit

2026-10-03 16:25:09 +00:00

Run the composed grant fixture on the production worker stack (#824, #867)

kayg referenced this issue

MCP session DELETE differs from its retained status contract #1023

kayg referenced this issue from a commit

Capture public Auth themes without account writes and use the current SPA (#867)

kayg referenced this issue from a commit

Remove the duplicate tablet viewport from the Notes browser fixture (#867)

kayg referenced this issue from a commit

Verify Quick Look closes before targeting the restored deck (#611, #867)

kayg referenced this issue from a commit

Keep Money captures signed in and inspect the actual feed and modal tray (#867)

kayg referenced this issue from a commit

Decode finite MCP replies and use production stacks in live fixtures (#867)

kayg referenced this issue from a commit

Read Money cancellation results from the shared API fixture field (#462, #867)

kayg referenced this issue from a commit

Record the remaining release blockers and standalone passing gates (#867)

kayg referenced this issue from a commit

2026-10-03 16:34:52 +00:00

Keep one doc attribute style in the Connected Account tests (#867)

kayg commented

Author

Owner

Merge round 7b, round 2 — not ready for production

Issue: #867. Branch: job/merge-round-7b2. Start: 6aaacdb80a42a3cb3f1c0fadf95640230c1e9c5e. Head: 79a0c14c9b914dde1bbbb1d01ede64cf8d7ddf96.

7b READY FOR PRODUCTION: no. Required Rust and browser gates remain red or incomplete. No staging replacement, push, production action or promotion was done. The four-hour job limit applies.

Built

Task metadata and Batch G are complete. The fixed set has 72 input merges, with one merge commit per input. No heads approved after round 1 began were added. The audit is docs/audits/merge-round-7b2.md. The one origin/dev fetch and merge check before final gates returned Already up to date. Origin/dev was d4e7188810a89fb0e8e6b162279917f7e23989f9. Migration identities were checked against that head.

The complete API now publishes 373 operations and 354 generated tools. OpenAPI, the action registry and the API client are regenerated. Explicit policies cover the new routes. The merged Money opt-in now reveals its Tab through the shared saved visibility rule. Notes identity hydration runs independently of cached body loading. Search returns canonical stable Note links. PDF.js 6 selection transforms and PDF page-shell coverage are reconciled. Shared focus tokens replace nine local rules. Settings and Mail retain one benchmark sampler. CLI contracts and the authorization inventory include the merged routes.

Files: contracts/, packages/api-client/src/generated.ts, apps/web/src/, apps/web/e2e/, packages/ui/, crates/calternal-server/, crates/calternal-collab/, Files/Mail/Notes plugin files, bench/, tests/adversarial/, Cargo.lock, and the audit. Complete round-2 path inventory: artifacts/round2-changed-files.txt; own non-merge files: artifacts/round2-owned-files.txt. Review artifacts are not committed.

Verbatim gates

Web check (bun run --cwd apps/web check):

perf-lint: PASS; 0 violations; 19194 scoped exceptions
svelte-check found 0 errors and 2 warnings in 2 files

The two warnings are empty CSS rules in AttachmentDeck and AgendaList. The scoped bootstrap has 19,195 ledger entries, including the bundle budget; 19,194 exceptions have owners and expiry. This is captured debt, not a performance measurement or a claim that debt is fixed.

Full web tests (bun run --cwd apps/web test --maxWorkers=2):

 Test Files  215 passed (215)
      Tests  1453 passed (1453)

The later Money opt-in change passed 16 focused tests in two files, including three new shared-rule cases. The full web run above preceded that small change; it is not described as a new full-suite run on final HEAD.

Production web build completed. Its final output includes:

Compressed 532 static variants; saved 9729681 bytes.

All Cargo commands use line-tables-only, no incremental builds, jobs=3 and the worktree TMPDIR. Tests use four threads. cargo fmt --check final exit is recorded in artifacts/round2-fmt-final.log (empty output on success). The per-crate runner output is:

2026-10-03T13:25:16Z clippy async-imap
2026-10-03T13:52:46Z test async-imap
2026-10-03T14:06:40Z PASS async-imap
2026-10-03T14:06:40Z clippy calternal-api
2026-10-03T14:07:15Z test calternal-api
2026-10-03T14:07:42Z PASS calternal-api
2026-10-03T14:07:42Z clippy calternal-auth
2026-10-03T14:17:18Z test calternal-auth
2026-10-03T14:27:44Z PASS calternal-auth
2026-10-03T14:27:44Z clippy calternal-cli
2026-10-03T14:30:26Z test calternal-cli
2026-10-03T14:30:59Z PASS calternal-cli
2026-10-03T14:30:59Z clippy calternal-collab
2026-10-03T14:43:25Z test calternal-collab
2026-10-03T15:41:37Z clippy calternal-server
2026-10-03T15:44:21Z test calternal-server
2026-10-03T15:48:02Z RESULT calternal-server clippy=0 test=101
2026-10-03T15:48:02Z clippy calternal-plugin-mail
2026-10-03T15:48:49Z test calternal-plugin-mail
2026-10-03T15:49:02Z RESULT calternal-plugin-mail clippy=0 test=101
2026-10-03T15:49:02Z clippy calternal-plugin-notes
2026-10-03T15:49:47Z test calternal-plugin-notes
2026-10-03T15:55:29Z RESULT calternal-plugin-notes clippy=0 test=101
2026-10-03T15:55:29Z clippy calternal-collab
2026-10-03T15:59:21Z test calternal-collab
2026-10-03T16:08:24Z RESULT calternal-collab clippy=0 test=0
2026-10-03T16:08:24Z clippy calternal-plugin-files
2026-10-03T16:14:12Z test calternal-plugin-files
2026-10-03T16:28:33Z RESULT calternal-plugin-files clippy=0 test=0
2026-10-03T16:28:33Z clippy calternal-dav
2026-10-03T16:29:00Z test calternal-dav
2026-10-03T16:29:28Z RESULT calternal-dav clippy=0 test=0
2026-10-03T16:29:28Z clippy calternal-db
2026-10-03T16:29:40Z test calternal-db
2026-10-03T16:30:09Z RESULT calternal-db clippy=101 test=0
2026-10-03T16:30:09Z clippy calternal-embed
2026-10-03T16:31:09Z test calternal-embed

The first Collab failure was repaired; its full Clippy and test rerun passed. Server, Mail and Notes full-run failures are retained below. Focused fixes do not constitute a passing full rerun.

cargo test -p async-imap -- --test-threads=4:

test result: ok. 70 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.07s
test result: ok. 1 passed; 0 failed; 6 ignored; 0 measured; 0 filtered out; finished in 0.10s

cargo test -p calternal-api -- --test-threads=4:

test result: ok. 17 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.07s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

cargo test -p calternal-auth -- --test-threads=4:

test result: ok. 109 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 167.15s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

cargo test -p calternal-cli -- --test-threads=4:

test result: ok. 51 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.77s
test result: ok. 17 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 5.34s

cargo test -p calternal-collab -- --test-threads=4:

test result: ok. 38 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.05s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.66s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.10s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 195.16s
test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 10.22s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.63s
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.70s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 10.90s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.13s
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 48.72s
test result: ok. 15 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.13s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.58s
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 41.84s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

cargo test -p calternal-server -- --test-threads=4:

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 213 filtered out; finished in 14.92s
test result: FAILED. 204 passed; 2 failed; 8 ignored; 0 measured; 0 filtered out; finished in 30.29s

cargo test -p calternal-plugin-mail -- --test-threads=4:

test result: FAILED. 64 passed; 1 failed; 3 ignored; 0 measured; 0 filtered out; finished in 11.57s

cargo test -p calternal-plugin-notes -- --test-threads=4:

test result: FAILED. 251 passed; 5 failed; 1 ignored; 0 measured; 0 filtered out; finished in 251.27s

cargo test -p calternal-plugin-files -- --test-threads=4:

test result: ok. 227 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 395.85s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

Focused regressions and helper checks (verbatim summaries):

artifacts/round2-notes-migration-focused.log:

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 0.81s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.00s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 0.86s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.00s

artifacts/round2-notes-contract-focused.log:

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 0.14s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.00s

artifacts/round2-mcp-schema-focused.log:

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 213 filtered out; finished in 0.41s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 1 filtered out; finished in 0.00s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 1 filtered out; finished in 0.00s

artifacts/round2-files-grant-final.log:

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 229 filtered out; finished in 0.97s

artifacts/round2-mcp-direct-worker-stack.log:

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 213 filtered out; finished in 17.42s

artifacts/round2-mail-delta-alone.log:

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 67 filtered out; finished in 6.72s

artifacts/round2-notes-rebuild-alone.log:

test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 256 filtered out; finished in 178.37s

artifacts/round2-notes-unicode-alone.log:

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 81.41s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.02s

artifacts/round2-files-storm-alone.log:

test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 229 filtered out; finished in 275.02s

artifacts/round2-harness-current.log:

# pass 10
# fail 0
# skipped 1

Additional checks passed: 24 docs tests; 25 action-registry Python tests; 26 Cross-User classifier tests; 15 Admin classifier tests; two benchmark sampler tests; one proxy cleanup regression. Offline inventory output:

Cross-User classification gate: 375 operations classified
Generated entry point classification: 1062 tools classified

These classifier checks are not a live isolation matrix.

Browser verification and smoke

All runs use a real local server and the production SPA. macOS platform signals are enabled. Mail layouts, Admin denial, Calendar resize, theme capture and Auth workflows passed. The Admin run includes 951 controlled authorization requests and all 390/820/1440 px light/dark checks. Money's real User flow writes the expected Markdown and its derived views show the resulting data; its complete workflow disposition is in the fresh log. Agenda has 59 screenshots across the required widths and themes, plus nine finite media contract checks. The complete Agenda workflow remains red after Escape leaves Quick Look over the next pointer target. Owner visual review is still required. Screenshots stay under artifacts/ and are not committed. The fj CLI has no attachment command; screenshots are not claimed as uploaded.

The ordinary workflow run stopped to prioritize headline repairs. It is incomplete, not a passing full e2e suite. Exact completed dispositions from that run:

test:e2e: exit 1.
test:e2e:settings-50: exit 1.
test:e2e:money: exit 1.
test:e2e:notes: exit 1.
test:e2e:tasks: exit 1.
test:e2e:ai: exit 1.
test:e2e:analytics: exit 1.
test:e2e:ask: exit 1.
test:e2e:app-passwords: exit 1.
test:e2e:mail-sync-613: exit 1.
test:e2e:calendar: exit 1.
test:e2e:gaps-827-828: exit 1.
test:e2e:weekstate-609: exit 1.
test:e2e:calendar-crossday: exit 1.
test:e2e:calendar-doc-stack: exit 1.
test:e2e:preview-attach: exit 1.
test:e2e:hidden-activity: exit 1.
test:e2e:calendar-view-switcher: exit 1.
test:e2e:calendar-task-overflow: exit 1.
test:e2e:taskday-655-657: exit 1.
test:e2e:calendar-resize: exit 0.
test:e2e:composer: exit 1.
test:e2e:pill-feedback: exit 1.
test:e2e:mobile-focus: exit 1.
test:e2e:overflow-511: exit 1.
test:e2e:theme: exit 0.
test:e2e:theme-variants-506: exit 1.
test:e2e:settings-shortcut: exit 1.
test:e2e:settings-open-642: exit 1.
test:e2e:settings-blaze-641: exit 1.
test:e2e:midnight: exit 1.
test:e2e:maintenance: exit 1.
test:e2e:overlay-title: exit 1.
test:e2e:auth: exit 0.
test:e2e:files: exit 1.
test:e2e:bg-stability-535: exit 1.
test:e2e:files-paste: exit 1.
test:e2e:chrome-surfaces: exit 1.

Fresh artifacts/round2-money-modal-final.log tail:

PASS the User flow wrote exact Markdown and the screens show the derived numbers
SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-light-390.png
SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-end-light-390.png
SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-import-form-light-390.png
SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-import-review-light-390.png
518 | 					await review.getByRole('button', { name: 'Cancel' }).click();
519 | 					assert.equal((await cancelResponse).status(), 204, 'the web cancel action discards its preview');
520 | 					await review.waitFor({ state: 'hidden' });
521 | 					const budgets = await api(shotPage, '/api/v1/money/budgets');
522 | 					assert.equal(budgets.status, 200);
523 | 					assert.ok(!budgets.body.some((budget) => budget.name === title), 'cancelling leaves no Budget behind');
                              ^
TypeError: undefined is not an object (evaluating 'budgets.body.some')
      at /home/kayg/Developer/calternal-wt/merge-round-7b2/apps/web/e2e/money.mjs:523:25

Bun v1.4.2 (Linux x64)

Fresh artifacts/round2-agenda-final.log tail:

  "    - locator resolved to <div data-direction=\"horizontal\" class=\"attachment-deck svelte-5ho3lu\">…</div>",
  "  - attempting hover action", "    2 × waiting for element to be visible and stable",
  "      - element is visible and stable", "      - scrolling into view if needed", "      - done scrolling",
  "      - <img alt=\"\" decoding=\"async\" draggable=\"false\" class=\"svelte-1e76gna animate\" src=\"/api/v1/files/download?path=Attachments%2FDeck+photo+0.jpg&inline=true\"/> from <div data-calternal-overlay-layer=\"\">…</div> subtree intercepts pointer events",
  "    - retrying hover action", "    - waiting 20ms", "    2 × waiting for element to be visible and stable",
  "      - element is visible and stable", "      - scrolling into view if needed", "      - done scrolling",
  "      - <img alt=\"\" decoding=\"async\" draggable=\"false\" class=\"svelte-1e76gna animate\" src=\"/api/v1/files/download?path=Attachments%2FDeck+photo+0.jpg&inline=true\"/> from <div data-calternal-overlay-layer=\"\">…</div> subtree intercepts pointer events",
  "    - retrying hover action", "      - waiting 100ms", "    160 × waiting for element to be visible and stable",
  "        - element is visible and stable", "        - scrolling into view if needed",
  "        - done scrolling", "        - <img alt=\"\" decoding=\"async\" draggable=\"false\" class=\"svelte-1e76gna animate\" src=\"/api/v1/files/download?path=Attachments%2FDeck+photo+0.jpg&inline=true\"/> from <div data-calternal-overlay-layer=\"\">…</div> subtree intercepts pointer events",
  "      - retrying hover action", "        - waiting 500ms"
],

      at /home/kayg/Developer/calternal-wt/merge-round-7b2/tests/adversarial/node_modules/playwright-core/lib/coreBundle.js:57694:13

Bun v1.4.2 (Linux x64)

Fresh artifacts/round2-shell-final.log tail:

HTTP 401 /api/v1/appearance on /setup
HTTP 401 /api/v1/appearance on /setup
SCREENSHOT setup-recovery-390-tokyo-night.png
PASS first-owner setup, passkey registration, recovery-code handoff
SCREENSHOT notes-all-1440-paper.png
SCREENSHOT notes-all-390-paper.png
SCREENSHOT notes-all-1440-tokyo-night.png
SCREENSHOT notes-all-390-tokyo-night.png
CSP REPORTS shell: 0 across 2 pages
waitFor: Timeout 30000ms exceeded.
Call log:
[2m  - waiting for getByRole('group', { name: 'New log entry' }) to be visible[22m

    at createRealLogEntry (/home/kayg/Developer/calternal-wt/merge-round-7b2/apps/web/e2e/shell.mjs:446:59)
    at async runE2E (/home/kayg/Developer/calternal-wt/merge-round-7b2/apps/web/e2e/shell.mjs:1757:26)
    at processTicksAndRejections (native:7:39)

Settings redesign: partial phone captures; canonical search navigation fails. Agenda decks: all-width/theme captures and finite media checks; Escape workflow red. Mail layouts: passed locally. Money generated-fixture import: preview/review and cancel checks run in the local workflow; full disposition above. Voice memos/transcription: incomplete; real model preparation did not finish in the time-boxed run. Undo and Notes HHMM: focused web tests pass, but full headline smoke is incomplete. These are local checks, not staging smoke.

Remaining defects and known gaps

#1020 records browser integration blockers: Settings canonical search destination, Files-origin Daily Log editor after reload, platform shortcut fixtures, Agenda Quick Look after Escape, and the shell log-entry flow. Complete individual logs are in artifacts/round2-e2e/; fresh headline logs are above. Existing assertions remain.
#1021: Mail delta catch-up fails alone with Transport; classified logging identifies the cache/storage step. No origin/dev baseline result is claimed.
#1022: the 650-entry Notes projection rebuild fails alone with Index busy. Unicode Task creation receives 503 in the full run; its focused disposition is above. Its original 201 expectation and fixture size stay unchanged.
MCP status mismatch is tracked in #1023. Disabled-Plugin assertions now survive finite SSE decoding, but the owner session DELETE check returns 202 where the original expectation requires 200. This is filed separately. The 200 expectation is retained. The normal full server gate still needs a passing rerun. A direct default-stack invocation overflowed; the full live-app harness and focused production-stack run use 4 MiB.
#1005 nested Daily Log preservation passes in the Notes full run. It is not labelled a baseline failure. Files #1000/#942 storm verification is recorded only when a focused result exists; no unrun origin/dev baseline is claimed.
The remaining per-crate gates, full e2e suite, live cross-user isolation and broad adversarial probe are incomplete. Broad autonomous attack matrices were not executed; bounded negative authorization fixtures and offline classification checks ran. No release binary/image was built.
Performance profiles are present, but no perf VM measurements were run: this is not a performance issue under the current merge-round policy. No local numbers are substituted for perf VM evidence.

Staging

The staging helper was read. The existing staging health check returned:

staging health HTTP 200

No replacement tag was deployed. New staging tag: none. Existing staging remains unchanged. No authenticated headline staging smoke or Apple-client interop ran. Production was not touched.

UX gaps closed

Money opt-in now shows the acknowledged Tab through the same saved visibility rule as Settings. Note identity hydration no longer depends on a body-cache miss. Search uses the stable canonical Note route. Duplicate PhotoViewer callbacks and imports are removed. Shared focus paint remains consistent. Session-end cleanup discards late Admin payloads. Public Auth captures do not attempt account writes.

UX gaps left

The browser blockers above, incomplete headline Undo/HHMM/Voice smoke, incomplete device/theme evidence for Settings and other failed screens, and owner visual review. The feature set is not called UX-complete.

Decisions and expectation review

No OPEN design choice was built. New action declarations use explicit authority and continuation policies. Changes needs Account and Data authority. Writes never replay uncertain results. Voice status never replays because it can enqueue a job. Money multipart preview stays direct-HTTP until adapters can encode its boundary. Public Auth theme capture uses a read-only seam; signed-in captures keep the real Appearance writer.

Fixture expectation changes are limited to merged behavior: #746 untrusted tool data envelope (data) and publication metadata; #836 tool-result failures; #775 ninth Location route for Undo restore; #921 Features display label; #407 Calendar's moved card; #606 stable Daily Note IDs; #665 HTTP Note response; Notes migrations 1–31 and deployed epoch identity 26; Mail's reviewed SQL remains pinned to migration 9. The owner must review these changes. MCP DELETE 200, Unicode 201, all body-preservation assertions, cursor limits, and geometry assertions were not weakened.

Cleanup and final stopped-work dispositions are recorded in the final issue comment. 7b READY FOR PRODUCTION: no.

Final gate disposition and commands left

Eight crates pass both full Clippy and tests: async-imap, calternal-api, calternal-auth, calternal-cli, calternal-collab, calternal-plugin-files, calternal-dav, calternal-db. The Db Clippy error was a mixed inner/outer doc attribute on its test module. The final fix keeps both comments inside the module, without changing behavior. Final Db Clippy output:

    Checking calternal-db v0.1.0 (/home/kayg/Developer/calternal-wt/merge-round-7b2/crates/calternal-db)
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 3.05s

Final Db test summaries:

test result: ok. 28 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.33s
test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 4.33s
test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.56s
test result: ok. 17 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 1.68s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.12s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.21s

The Unicode Notes case passes alone. The full Notes run still needs a green result. Files' full suite and its storm pass, so #1000/#942 are not marked as known red baseline failures. The final Money fixture field correction uses json instead of body, as the shared harness declares. Syntax checks pass; a complete rerun after that correction remains. The preceding real preview cancel route returned 204. No claim is made that every final cancellation assertion passed.

The runner stopped at the incomplete Embed Clippy gate to stay within the job limit. Nineteen crate gates remain: calternal-embed, calternal-fs, calternal-imap, calternal-location, calternal-media, calternal-money, calternal-notes-core, calternal-path, calternal-plugin, calternal-plugin-ai, calternal-plugin-analytics, calternal-plugin-calendar, calternal-plugin-money, calternal-plugin-notifications, calternal-plugin-photos, calternal-plugin-video, calternal-search, calternal-sync, calternal-tags. For each, run cargo clippy -p <crate> --all-targets -- -D warnings and cargo test -p <crate> -- --test-threads=4, sequentially with jobs=3. Resolve #1021, #1022 and #1023, then rerun the three red full crate gates.

Resolve #1020 and run the remaining ordinary e2e commands in apps/web/package.json against the current production build, with macOS platform signals and one browser at a time. Full device/theme review is still needed. The broad probe and live isolation matrix remain unrun. After all required gates pass, build the release image, use the read staging helper with a SHA tag, and smoke the requested headline flows on staging. Do not promote this head to production.

Cleanup

Final worktree status is clean. No owned operational processes remain. cargo clean output, verbatim:

     Removed 29618 files, 27.7GiB total

Own web build output (apps/web/build and apps/web/.svelte-kit/output) is deleted. Review screenshots and logs remain under artifacts/. Cargo fmt final output is empty and the check passed. No pushes, production deploys or promotions were performed. New staging tag: none. 7b READY FOR PRODUCTION: no.

# Merge round 7b, round 2 — not ready for production Issue: #867. Branch: `job/merge-round-7b2`. Start: `6aaacdb80a42a3cb3f1c0fadf95640230c1e9c5e`. Head: `79a0c14c9b914dde1bbbb1d01ede64cf8d7ddf96`. **7b READY FOR PRODUCTION: no.** Required Rust and browser gates remain red or incomplete. No staging replacement, push, production action or promotion was done. The four-hour job limit applies. ## Built Task metadata and Batch G are complete. The fixed set has 72 input merges, with one merge commit per input. No heads approved after round 1 began were added. The audit is `docs/audits/merge-round-7b2.md`. The one origin/dev fetch and merge check before final gates returned `Already up to date.` Origin/dev was `d4e7188810a89fb0e8e6b162279917f7e23989f9`. Migration identities were checked against that head. The complete API now publishes 373 operations and 354 generated tools. OpenAPI, the action registry and the API client are regenerated. Explicit policies cover the new routes. The merged Money opt-in now reveals its Tab through the shared saved visibility rule. Notes identity hydration runs independently of cached body loading. Search returns canonical stable Note links. PDF.js 6 selection transforms and PDF page-shell coverage are reconciled. Shared focus tokens replace nine local rules. Settings and Mail retain one benchmark sampler. CLI contracts and the authorization inventory include the merged routes. Files: `contracts/`, `packages/api-client/src/generated.ts`, `apps/web/src/`, `apps/web/e2e/`, `packages/ui/`, `crates/calternal-server/`, `crates/calternal-collab/`, Files/Mail/Notes plugin files, `bench/`, `tests/adversarial/`, `Cargo.lock`, and the audit. Complete round-2 path inventory: `artifacts/round2-changed-files.txt`; own non-merge files: `artifacts/round2-owned-files.txt`. Review artifacts are not committed. ## Verbatim gates Web check (`bun run --cwd apps/web check`): ``` perf-lint: PASS; 0 violations; 19194 scoped exceptions svelte-check found 0 errors and 2 warnings in 2 files ``` The two warnings are empty CSS rules in AttachmentDeck and AgendaList. The scoped bootstrap has 19,195 ledger entries, including the bundle budget; 19,194 exceptions have owners and expiry. This is captured debt, not a performance measurement or a claim that debt is fixed. Full web tests (`bun run --cwd apps/web test --maxWorkers=2`): ``` Test Files 215 passed (215) Tests 1453 passed (1453) ``` The later Money opt-in change passed 16 focused tests in two files, including three new shared-rule cases. The full web run above preceded that small change; it is not described as a new full-suite run on final HEAD. Production web build completed. Its final output includes: ``` Compressed 532 static variants; saved 9729681 bytes. ``` All Cargo commands use line-tables-only, no incremental builds, jobs=3 and the worktree TMPDIR. Tests use four threads. `cargo fmt --check` final exit is recorded in `artifacts/round2-fmt-final.log` (empty output on success). The per-crate runner output is: ``` 2026-10-03T13:25:16Z clippy async-imap 2026-10-03T13:52:46Z test async-imap 2026-10-03T14:06:40Z PASS async-imap 2026-10-03T14:06:40Z clippy calternal-api 2026-10-03T14:07:15Z test calternal-api 2026-10-03T14:07:42Z PASS calternal-api 2026-10-03T14:07:42Z clippy calternal-auth 2026-10-03T14:17:18Z test calternal-auth 2026-10-03T14:27:44Z PASS calternal-auth 2026-10-03T14:27:44Z clippy calternal-cli 2026-10-03T14:30:26Z test calternal-cli 2026-10-03T14:30:59Z PASS calternal-cli 2026-10-03T14:30:59Z clippy calternal-collab 2026-10-03T14:43:25Z test calternal-collab 2026-10-03T15:41:37Z clippy calternal-server 2026-10-03T15:44:21Z test calternal-server 2026-10-03T15:48:02Z RESULT calternal-server clippy=0 test=101 2026-10-03T15:48:02Z clippy calternal-plugin-mail 2026-10-03T15:48:49Z test calternal-plugin-mail 2026-10-03T15:49:02Z RESULT calternal-plugin-mail clippy=0 test=101 2026-10-03T15:49:02Z clippy calternal-plugin-notes 2026-10-03T15:49:47Z test calternal-plugin-notes 2026-10-03T15:55:29Z RESULT calternal-plugin-notes clippy=0 test=101 2026-10-03T15:55:29Z clippy calternal-collab 2026-10-03T15:59:21Z test calternal-collab 2026-10-03T16:08:24Z RESULT calternal-collab clippy=0 test=0 2026-10-03T16:08:24Z clippy calternal-plugin-files 2026-10-03T16:14:12Z test calternal-plugin-files 2026-10-03T16:28:33Z RESULT calternal-plugin-files clippy=0 test=0 2026-10-03T16:28:33Z clippy calternal-dav 2026-10-03T16:29:00Z test calternal-dav 2026-10-03T16:29:28Z RESULT calternal-dav clippy=0 test=0 2026-10-03T16:29:28Z clippy calternal-db 2026-10-03T16:29:40Z test calternal-db 2026-10-03T16:30:09Z RESULT calternal-db clippy=101 test=0 2026-10-03T16:30:09Z clippy calternal-embed 2026-10-03T16:31:09Z test calternal-embed ``` The first Collab failure was repaired; its full Clippy and test rerun passed. Server, Mail and Notes full-run failures are retained below. Focused fixes do not constitute a passing full rerun. `cargo test -p async-imap -- --test-threads=4`: ``` test result: ok. 70 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.07s test result: ok. 1 passed; 0 failed; 6 ignored; 0 measured; 0 filtered out; finished in 0.10s ``` `cargo test -p calternal-api -- --test-threads=4`: ``` test result: ok. 17 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.07s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s ``` `cargo test -p calternal-auth -- --test-threads=4`: ``` test result: ok. 109 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 167.15s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s ``` `cargo test -p calternal-cli -- --test-threads=4`: ``` test result: ok. 51 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.77s test result: ok. 17 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 5.34s ``` `cargo test -p calternal-collab -- --test-threads=4`: ``` test result: ok. 38 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.05s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.66s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.10s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 195.16s test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 10.22s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.63s test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.70s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 10.90s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.13s test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 48.72s test result: ok. 15 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.13s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.58s test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 41.84s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s ``` `cargo test -p calternal-server -- --test-threads=4`: ``` test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 213 filtered out; finished in 14.92s test result: FAILED. 204 passed; 2 failed; 8 ignored; 0 measured; 0 filtered out; finished in 30.29s ``` `cargo test -p calternal-plugin-mail -- --test-threads=4`: ``` test result: FAILED. 64 passed; 1 failed; 3 ignored; 0 measured; 0 filtered out; finished in 11.57s ``` `cargo test -p calternal-plugin-notes -- --test-threads=4`: ``` test result: FAILED. 251 passed; 5 failed; 1 ignored; 0 measured; 0 filtered out; finished in 251.27s ``` `cargo test -p calternal-plugin-files -- --test-threads=4`: ``` test result: ok. 227 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 395.85s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s ``` Focused regressions and helper checks (verbatim summaries): `artifacts/round2-notes-migration-focused.log`: ``` test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 0.81s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.00s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 0.86s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.00s ``` `artifacts/round2-notes-contract-focused.log`: ``` test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 0.14s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.00s ``` `artifacts/round2-mcp-schema-focused.log`: ``` test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 213 filtered out; finished in 0.41s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 1 filtered out; finished in 0.00s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 1 filtered out; finished in 0.00s ``` `artifacts/round2-files-grant-final.log`: ``` test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 229 filtered out; finished in 0.97s ``` `artifacts/round2-mcp-direct-worker-stack.log`: ``` test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 213 filtered out; finished in 17.42s ``` `artifacts/round2-mail-delta-alone.log`: ``` test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 67 filtered out; finished in 6.72s ``` `artifacts/round2-notes-rebuild-alone.log`: ``` test result: FAILED. 0 passed; 1 failed; 0 ignored; 0 measured; 256 filtered out; finished in 178.37s ``` `artifacts/round2-notes-unicode-alone.log`: ``` test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 256 filtered out; finished in 81.41s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 2 filtered out; finished in 0.02s ``` `artifacts/round2-files-storm-alone.log`: ``` test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 229 filtered out; finished in 275.02s ``` `artifacts/round2-harness-current.log`: ``` # pass 10 # fail 0 # skipped 1 ``` Additional checks passed: 24 docs tests; 25 action-registry Python tests; 26 Cross-User classifier tests; 15 Admin classifier tests; two benchmark sampler tests; one proxy cleanup regression. Offline inventory output: ``` Cross-User classification gate: 375 operations classified Generated entry point classification: 1062 tools classified ``` These classifier checks are not a live isolation matrix. ## Browser verification and smoke All runs use a real local server and the production SPA. macOS platform signals are enabled. Mail layouts, Admin denial, Calendar resize, theme capture and Auth workflows passed. The Admin run includes 951 controlled authorization requests and all 390/820/1440 px light/dark checks. Money's real User flow writes the expected Markdown and its derived views show the resulting data; its complete workflow disposition is in the fresh log. Agenda has 59 screenshots across the required widths and themes, plus nine finite media contract checks. The complete Agenda workflow remains red after Escape leaves Quick Look over the next pointer target. Owner visual review is still required. Screenshots stay under `artifacts/` and are not committed. The fj CLI has no attachment command; screenshots are not claimed as uploaded. The ordinary workflow run stopped to prioritize headline repairs. It is incomplete, not a passing full e2e suite. Exact completed dispositions from that run: - `test:e2e`: exit 1. - `test:e2e:settings-50`: exit 1. - `test:e2e:money`: exit 1. - `test:e2e:notes`: exit 1. - `test:e2e:tasks`: exit 1. - `test:e2e:ai`: exit 1. - `test:e2e:analytics`: exit 1. - `test:e2e:ask`: exit 1. - `test:e2e:app-passwords`: exit 1. - `test:e2e:mail-sync-613`: exit 1. - `test:e2e:calendar`: exit 1. - `test:e2e:gaps-827-828`: exit 1. - `test:e2e:weekstate-609`: exit 1. - `test:e2e:calendar-crossday`: exit 1. - `test:e2e:calendar-doc-stack`: exit 1. - `test:e2e:preview-attach`: exit 1. - `test:e2e:hidden-activity`: exit 1. - `test:e2e:calendar-view-switcher`: exit 1. - `test:e2e:calendar-task-overflow`: exit 1. - `test:e2e:taskday-655-657`: exit 1. - `test:e2e:calendar-resize`: exit 0. - `test:e2e:composer`: exit 1. - `test:e2e:pill-feedback`: exit 1. - `test:e2e:mobile-focus`: exit 1. - `test:e2e:overflow-511`: exit 1. - `test:e2e:theme`: exit 0. - `test:e2e:theme-variants-506`: exit 1. - `test:e2e:settings-shortcut`: exit 1. - `test:e2e:settings-open-642`: exit 1. - `test:e2e:settings-blaze-641`: exit 1. - `test:e2e:midnight`: exit 1. - `test:e2e:maintenance`: exit 1. - `test:e2e:overlay-title`: exit 1. - `test:e2e:auth`: exit 0. - `test:e2e:files`: exit 1. - `test:e2e:bg-stability-535`: exit 1. - `test:e2e:files-paste`: exit 1. - `test:e2e:chrome-surfaces`: exit 1. Fresh `artifacts/round2-money-modal-final.log` tail: ``` PASS the User flow wrote exact Markdown and the screens show the derived numbers SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-light-390.png SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-end-light-390.png SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-import-form-light-390.png SCREENSHOT /home/kayg/Developer/calternal-wt/merge-round-7b2/artifacts/money/money-budget-import-review-light-390.png 518 | await review.getByRole('button', { name: 'Cancel' }).click(); 519 | assert.equal((await cancelResponse).status(), 204, 'the web cancel action discards its preview'); 520 | await review.waitFor({ state: 'hidden' }); 521 | const budgets = await api(shotPage, '/api/v1/money/budgets'); 522 | assert.equal(budgets.status, 200); 523 | assert.ok(!budgets.body.some((budget) => budget.name === title), 'cancelling leaves no Budget behind'); ^ TypeError: undefined is not an object (evaluating 'budgets.body.some') at /home/kayg/Developer/calternal-wt/merge-round-7b2/apps/web/e2e/money.mjs:523:25 Bun v1.4.2 (Linux x64) ``` Fresh `artifacts/round2-agenda-final.log` tail: ``` " - locator resolved to <div data-direction=\"horizontal\" class=\"attachment-deck svelte-5ho3lu\">…</div>", " - attempting hover action", " 2 × waiting for element to be visible and stable", " - element is visible and stable", " - scrolling into view if needed", " - done scrolling", " - <img alt=\"\" decoding=\"async\" draggable=\"false\" class=\"svelte-1e76gna animate\" src=\"/api/v1/files/download?path=Attachments%2FDeck+photo+0.jpg&inline=true\"/> from <div data-calternal-overlay-layer=\"\">…</div> subtree intercepts pointer events", " - retrying hover action", " - waiting 20ms", " 2 × waiting for element to be visible and stable", " - element is visible and stable", " - scrolling into view if needed", " - done scrolling", " - <img alt=\"\" decoding=\"async\" draggable=\"false\" class=\"svelte-1e76gna animate\" src=\"/api/v1/files/download?path=Attachments%2FDeck+photo+0.jpg&inline=true\"/> from <div data-calternal-overlay-layer=\"\">…</div> subtree intercepts pointer events", " - retrying hover action", " - waiting 100ms", " 160 × waiting for element to be visible and stable", " - element is visible and stable", " - scrolling into view if needed", " - done scrolling", " - <img alt=\"\" decoding=\"async\" draggable=\"false\" class=\"svelte-1e76gna animate\" src=\"/api/v1/files/download?path=Attachments%2FDeck+photo+0.jpg&inline=true\"/> from <div data-calternal-overlay-layer=\"\">…</div> subtree intercepts pointer events", " - retrying hover action", " - waiting 500ms" ], at /home/kayg/Developer/calternal-wt/merge-round-7b2/tests/adversarial/node_modules/playwright-core/lib/coreBundle.js:57694:13 Bun v1.4.2 (Linux x64) ``` Fresh `artifacts/round2-shell-final.log` tail: ``` HTTP 401 /api/v1/appearance on /setup HTTP 401 /api/v1/appearance on /setup SCREENSHOT setup-recovery-390-tokyo-night.png PASS first-owner setup, passkey registration, recovery-code handoff SCREENSHOT notes-all-1440-paper.png SCREENSHOT notes-all-390-paper.png SCREENSHOT notes-all-1440-tokyo-night.png SCREENSHOT notes-all-390-tokyo-night.png CSP REPORTS shell: 0 across 2 pages waitFor: Timeout 30000ms exceeded. Call log: [2m - waiting for getByRole('group', { name: 'New log entry' }) to be visible[22m at createRealLogEntry (/home/kayg/Developer/calternal-wt/merge-round-7b2/apps/web/e2e/shell.mjs:446:59) at async runE2E (/home/kayg/Developer/calternal-wt/merge-round-7b2/apps/web/e2e/shell.mjs:1757:26) at processTicksAndRejections (native:7:39) ``` Settings redesign: partial phone captures; canonical search navigation fails. Agenda decks: all-width/theme captures and finite media checks; Escape workflow red. Mail layouts: passed locally. Money generated-fixture import: preview/review and cancel checks run in the local workflow; full disposition above. Voice memos/transcription: incomplete; real model preparation did not finish in the time-boxed run. Undo and Notes HHMM: focused web tests pass, but full headline smoke is incomplete. These are local checks, not staging smoke. ## Remaining defects and known gaps - #1020 records browser integration blockers: Settings canonical search destination, Files-origin Daily Log editor after reload, platform shortcut fixtures, Agenda Quick Look after Escape, and the shell log-entry flow. Complete individual logs are in `artifacts/round2-e2e/`; fresh headline logs are above. Existing assertions remain. - #1021: Mail delta catch-up fails alone with Transport; classified logging identifies the cache/storage step. No origin/dev baseline result is claimed. - #1022: the 650-entry Notes projection rebuild fails alone with Index busy. Unicode Task creation receives 503 in the full run; its focused disposition is above. Its original 201 expectation and fixture size stay unchanged. - MCP status mismatch is tracked in #1023. Disabled-Plugin assertions now survive finite SSE decoding, but the owner session DELETE check returns 202 where the original expectation requires 200. This is filed separately. The 200 expectation is retained. The normal full server gate still needs a passing rerun. A direct default-stack invocation overflowed; the full live-app harness and focused production-stack run use 4 MiB. - #1005 nested Daily Log preservation passes in the Notes full run. It is not labelled a baseline failure. Files #1000/#942 storm verification is recorded only when a focused result exists; no unrun origin/dev baseline is claimed. - The remaining per-crate gates, full e2e suite, live cross-user isolation and broad adversarial probe are incomplete. Broad autonomous attack matrices were not executed; bounded negative authorization fixtures and offline classification checks ran. No release binary/image was built. - Performance profiles are present, but no perf VM measurements were run: this is not a performance issue under the current merge-round policy. No local numbers are substituted for perf VM evidence. ## Staging The staging helper was read. The existing staging health check returned: ``` staging health HTTP 200 ``` No replacement tag was deployed. New staging tag: **none**. Existing staging remains unchanged. No authenticated headline staging smoke or Apple-client interop ran. Production was not touched. ## UX gaps closed Money opt-in now shows the acknowledged Tab through the same saved visibility rule as Settings. Note identity hydration no longer depends on a body-cache miss. Search uses the stable canonical Note route. Duplicate PhotoViewer callbacks and imports are removed. Shared focus paint remains consistent. Session-end cleanup discards late Admin payloads. Public Auth captures do not attempt account writes. ## UX gaps left The browser blockers above, incomplete headline Undo/HHMM/Voice smoke, incomplete device/theme evidence for Settings and other failed screens, and owner visual review. The feature set is not called UX-complete. ## Decisions and expectation review No OPEN design choice was built. New action declarations use explicit authority and continuation policies. Changes needs Account and Data authority. Writes never replay uncertain results. Voice status never replays because it can enqueue a job. Money multipart preview stays direct-HTTP until adapters can encode its boundary. Public Auth theme capture uses a read-only seam; signed-in captures keep the real Appearance writer. Fixture expectation changes are limited to merged behavior: #746 untrusted tool data envelope (`data`) and publication metadata; #836 tool-result failures; #775 ninth Location route for Undo restore; #921 Features display label; #407 Calendar's moved card; #606 stable Daily Note IDs; #665 HTTP Note response; Notes migrations 1–31 and deployed epoch identity 26; Mail's reviewed SQL remains pinned to migration 9. The owner must review these changes. MCP DELETE 200, Unicode 201, all body-preservation assertions, cursor limits, and geometry assertions were not weakened. Cleanup and final stopped-work dispositions are recorded in the final issue comment. **7b READY FOR PRODUCTION: no.** ## Final gate disposition and commands left Eight crates pass both full Clippy and tests: async-imap, calternal-api, calternal-auth, calternal-cli, calternal-collab, calternal-plugin-files, calternal-dav, calternal-db. The Db Clippy error was a mixed inner/outer doc attribute on its test module. The final fix keeps both comments inside the module, without changing behavior. Final Db Clippy output: ``` Checking calternal-db v0.1.0 (/home/kayg/Developer/calternal-wt/merge-round-7b2/crates/calternal-db) Finished `dev` profile [unoptimized + debuginfo] target(s) in 3.05s ``` Final Db test summaries: ``` test result: ok. 28 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.33s test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 4.33s test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.56s test result: ok. 17 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 1.68s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.12s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.21s ``` The Unicode Notes case passes alone. The full Notes run still needs a green result. Files' full suite and its storm pass, so #1000/#942 are not marked as known red baseline failures. The final Money fixture field correction uses `json` instead of `body`, as the shared harness declares. Syntax checks pass; a complete rerun after that correction remains. The preceding real preview cancel route returned 204. No claim is made that every final cancellation assertion passed. The runner stopped at the incomplete Embed Clippy gate to stay within the job limit. Nineteen crate gates remain: calternal-embed, calternal-fs, calternal-imap, calternal-location, calternal-media, calternal-money, calternal-notes-core, calternal-path, calternal-plugin, calternal-plugin-ai, calternal-plugin-analytics, calternal-plugin-calendar, calternal-plugin-money, calternal-plugin-notifications, calternal-plugin-photos, calternal-plugin-video, calternal-search, calternal-sync, calternal-tags. For each, run `cargo clippy -p <crate> --all-targets -- -D warnings` and `cargo test -p <crate> -- --test-threads=4`, sequentially with jobs=3. Resolve #1021, #1022 and #1023, then rerun the three red full crate gates. Resolve #1020 and run the remaining ordinary e2e commands in `apps/web/package.json` against the current production build, with macOS platform signals and one browser at a time. Full device/theme review is still needed. The broad probe and live isolation matrix remain unrun. After all required gates pass, build the release image, use the read staging helper with a SHA tag, and smoke the requested headline flows on staging. Do not promote this head to production. ## Cleanup Final worktree status is clean. No owned operational processes remain. `cargo clean` output, verbatim: ``` Removed 29618 files, 27.7GiB total ``` Own web build output (`apps/web/build` and `apps/web/.svelte-kit/output`) is deleted. Review screenshots and logs remain under `artifacts/`. Cargo fmt final output is empty and the check passed. No pushes, production deploys or promotions were performed. New staging tag: none. **7b READY FOR PRODUCTION: no.**

kayg commented

2026-10-03 16:38:06 +00:00

Author

Owner

Round 3 starts on job/merge-round-7b2 at 79a0c14c9b. No new inputs. Fix #1020–#1023, complete sequential crate and browser gates, then staging only. Broad autonomous adversarial workflows are outside the available verification scope; ordinary regression and isolation tests remain in scope. Production readiness requires all required evidence.

Round 3 starts on job/merge-round-7b2 at 79a0c14c9b914dde1bbbb1d01ede64cf8d7ddf96. No new inputs. Fix #1020–#1023, complete sequential crate and browser gates, then staging only. Broad autonomous adversarial workflows are outside the available verification scope; ordinary regression and isolation tests remain in scope. Production readiness requires all required evidence.

kayg commented

2026-10-03 16:44:53 +00:00

Author

Owner

Starting round-3 continuation on job/merge-round-7b2. Requested continuation base: 79a0c14c9b; checkout already contains follow-up commit 527c7f74cd. I am preserving the existing uncommitted Mail diagnostic diff while investigating #1020–#1023.

Starting round-3 continuation on job/merge-round-7b2. Requested continuation base: 79a0c14c9b914dde1bbbb1d01ede64cf8d7ddf96; checkout already contains follow-up commit 527c7f74cd3ecae79c7dec8dd6fe6badc74729e2. I am preserving the existing uncommitted Mail diagnostic diff while investigating #1020–#1023.

kayg commented

2026-10-03 18:42:45 +00:00

Author

Owner

Round 3 finding: cargo test -p calternal-server -- --test-threads=4 passed 206 unit tests (8 ignored) and the performance guard, then failed private_index_permissions at PRAGMA journal_mode=DELETE with SQLite code 5. The same test failed alone (0 passed, 1 failed), so this was a test fixture assumption, not load: closing only readers left the authority and checkpoint connections open. I changed the fixture to snapshot, close the full Db, and then run the unchanged 0600 rollback-journal assertion through one bounded connection. The focused integration test now passes alone (1 passed, 0 failed; 0.45s). Server clippy and full test rerun are pending.

kayg commented

2026-10-03 18:55:36 +00:00

Author

Owner

Server private-index follow-up: full clippy and test gates now pass after fixing the test fixture, not the production implementation. The fixture first snapshots the private index, closes all Db pools and connections, then opens one bounded SQLx connection to switch journal mode and retain the original 0600 file-mode assertion. The isolated regression passes unchanged: 1 passed, 0 failed. Full Server output: 206 passed, 0 failed, 8 ignored (74.17s); perf guard: 1 passed (19.37s); private_index_permissions: 1 passed (0.17s). The fixture change is committed in 0f509d198 before the latest UI commits. Server clippy passed with -D warnings.

Server private-index follow-up: full clippy and test gates now pass after fixing the test fixture, not the production implementation. The fixture first snapshots the private index, closes all Db pools and connections, then opens one bounded SQLx connection to switch journal mode and retain the original 0600 file-mode assertion. The isolated regression passes unchanged: 1 passed, 0 failed. Full Server output: 206 passed, 0 failed, 8 ignored (74.17s); perf guard: 1 passed (19.37s); private_index_permissions: 1 passed (0.17s). The fixture change is committed in 0f509d198 before the latest UI commits. Server clippy passed with -D warnings.

kayg commented

2026-10-03 20:10:11 +00:00

Author

Owner

Round 3 finding #1 (commit 538c3ee0d): both Calendar plugin failures reproduced when each test ran alone. cargo test -p calternal-plugin-calendar --lib feeds::publication::tests::feed_capability_changes_commit_on_full -- --exact --test-threads=1 returned 403 because the fixture omitted the account and data scopes required by the feed-management route. The fixture now supplies both scopes; the isolated test passes.

Round 3 finding #2: routes::tests::omitted_event_calendar_uses_synced_default returned 500 because its test database applied Calendar migrations but not the built-in migration that creates integration_accounts, which list_accounts reads. The fixture now applies both migration sets; the isolated test passes.

No expected status or production authorization rule changed. The full Calendar crate test gate now passes: test result: ok. 98 passed; 0 failed; 1 ignored; finished in 10.33s.

Round 3 finding #1 (commit 538c3ee0d): both Calendar plugin failures reproduced when each test ran alone. `cargo test -p calternal-plugin-calendar --lib feeds::publication::tests::feed_capability_changes_commit_on_full -- --exact --test-threads=1` returned 403 because the fixture omitted the `account` and `data` scopes required by the feed-management route. The fixture now supplies both scopes; the isolated test passes. Round 3 finding #2: `routes::tests::omitted_event_calendar_uses_synced_default` returned 500 because its test database applied Calendar migrations but not the built-in migration that creates `integration_accounts`, which `list_accounts` reads. The fixture now applies both migration sets; the isolated test passes. No expected status or production authorization rule changed. The full Calendar crate test gate now passes: `test result: ok. 98 passed; 0 failed; 1 ignored; finished in 10.33s`.

kayg commented

2026-10-03 20:15:41 +00:00

Author

Owner

Round 3 Video gate finding (commit 9c7256d2d): cargo clippy -p calternal-plugin-video --all-targets -- -D warnings failed to compile the test target because serve_cache_file now requires request headers, while cache_file_response_does_not_wait_for_access_time_write still used the old call signature. The production playlist route passes an empty HeaderMap; the test now does the same. Video clippy and tests pass. The full test gate reports test result: ok. 16 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.10s.

Round 3 Video gate finding (commit 9c7256d2d): `cargo clippy -p calternal-plugin-video --all-targets -- -D warnings` failed to compile the test target because `serve_cache_file` now requires request headers, while `cache_file_response_does_not_wait_for_access_time_write` still used the old call signature. The production playlist route passes an empty `HeaderMap`; the test now does the same. Video clippy and tests pass. The full test gate reports `test result: ok. 16 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.10s`.

kayg commented

2026-10-03 20:23:35 +00:00

Author

Owner

Round 3 Tags gate found a test-schema defect. cargo test -p calternal-tags -- --test-threads=4 failed five corruption-recovery tests with no such table: note_items. index::items_page reads canonical Note titles from the Notes-owned note_items table; the new recovery fixtures created the Tags and Files tables but omitted the Notes table. Existing title-row tests define that table explicitly. I will add the same minimal Notes projection schema to the shared Tags test fixtures and rerun the package gate. This is a fixture correction; it does not alter production behavior or test expectations.

Round 3 Tags gate found a test-schema defect. `cargo test -p calternal-tags -- --test-threads=4` failed five corruption-recovery tests with `no such table: note_items`. `index::items_page` reads canonical Note titles from the Notes-owned `note_items` table; the new recovery fixtures created the Tags and Files tables but omitted the Notes table. Existing title-row tests define that table explicitly. I will add the same minimal Notes projection schema to the shared Tags test fixtures and rerun the package gate. This is a fixture correction; it does not alter production behavior or test expectations.

kayg commented

2026-10-03 20:25:08 +00:00

Author

Owner

Fixed the Tags fixture defect in d03f1bfb6 (test(tags): provide Notes schema to recovery fixtures). A shared test helper now installs the Notes-owned note_items projection schema in Tag page fixtures. No production code or test expectations changed. Verification: cargo clippy -p calternal-tags --all-targets -- -D warnings passed; cargo test -p calternal-tags -- --test-threads=4 passed with 18 passed; 0 failed; 0 ignored.

Fixed the Tags fixture defect in `d03f1bfb6` (`test(tags): provide Notes schema to recovery fixtures`). A shared test helper now installs the Notes-owned `note_items` projection schema in Tag page fixtures. No production code or test expectations changed. Verification: `cargo clippy -p calternal-tags --all-targets -- -D warnings` passed; `cargo test -p calternal-tags -- --test-threads=4` passed with `18 passed; 0 failed; 0 ignored`.

kayg commented

2026-10-03 20:34:41 +00:00

Author

Owner

Round 3 Rust gates are green for all 22 required packages, after resolving the Calendar authority fixture, Video cache helper call, and Tags Notes-schema fixture. No test expectation changed. #1021 Mail catch-up and #1022 Notes projection rebuild pass in their full package runs; #1023 MCP session ownership tests pass in the full Server run.

Verbatim package results: Server 206 passed; 0 failed; 8 ignored, perf guard 1 passed, private-index permission 1 passed; Mail 65 passed; 0 failed; 3 ignored; Notes 256 passed; 0 failed; 2 ignored, Apple replay 2 passed; Tags 18 passed; 0 failed. cargo fmt --check and every required per-package Clippy command passed. The Tags fixture fix is committed as d03f1bfb6. Full Web, browser, adversarial, cross-user, and staging checks are next; this does not claim production readiness.

Round 3 Rust gates are green for all 22 required packages, after resolving the Calendar authority fixture, Video cache helper call, and Tags Notes-schema fixture. No test expectation changed. #1021 Mail catch-up and #1022 Notes projection rebuild pass in their full package runs; #1023 MCP session ownership tests pass in the full Server run. Verbatim package results: Server `206 passed; 0 failed; 8 ignored`, perf guard `1 passed`, private-index permission `1 passed`; Mail `65 passed; 0 failed; 3 ignored`; Notes `256 passed; 0 failed; 2 ignored`, Apple replay `2 passed`; Tags `18 passed; 0 failed`. `cargo fmt --check` and every required per-package Clippy command passed. The Tags fixture fix is committed as `d03f1bfb6`. Full Web, browser, adversarial, cross-user, and staging checks are next; this does not claim production readiness.

kayg commented

2026-10-03 20:41:39 +00:00

Author

Owner

Round 3 E2E finding: bun e2e/ai.mjs passes server validation, Settings, ask flows, and turn-history rendering, then fails the unchanged assertion at apps/web/e2e/ai.mjs:511: the finished turn's changed-file links do not match the stable /f/<item_id> values read from the Files change feed. The sequential E2E run continues; I am tracing the file mutation identity projection before changing code.

Round 3 E2E finding: `bun e2e/ai.mjs` passes server validation, Settings, ask flows, and turn-history rendering, then fails the unchanged assertion at `apps/web/e2e/ai.mjs:511`: the finished turn's changed-file links do not match the stable `/f/<item_id>` values read from the Files change feed. The sequential E2E run continues; I am tracing the file mutation identity projection before changing code.

kayg commented

2026-10-03 20:48:53 +00:00

Author

Owner

E2E harness repair: test:e2e:gaps-827-828 stopped before launching a browser because gaps-827-828.mjs imported resolve from node:fs. The named export is from node:path. Commit 4012c07da corrects the import. node --check apps/web/e2e/gaps-827-828.mjs passes. I will rerun the real browser workflow after the current sequential pass; no product code or test expectation changed.

E2E harness repair: `test:e2e:gaps-827-828` stopped before launching a browser because `gaps-827-828.mjs` imported `resolve` from `node:fs`. The named export is from `node:path`. Commit `4012c07da` corrects the import. `node --check apps/web/e2e/gaps-827-828.mjs` passes. I will rerun the real browser workflow after the current sequential pass; no product code or test expectation changed.

kayg commented

2026-10-03 20:59:23 +00:00

Author

Owner

E2E disposition: test:e2e:voice-619 reached the runner's 600-second per-script bound (exit 124) while the local Voice worker was still processing the generated long backfill. The script log has only the Playwright version and no assertion failure; process evidence showed the server's Voice worker consuming CPU. I classify this as SLOW-only load, not a correctness failure. Voice memo behavior remains covered by the Notes workflow and the required staging smoke.

E2E disposition: `test:e2e:voice-619` reached the runner's 600-second per-script bound (exit 124) while the local Voice worker was still processing the generated long backfill. The script log has only the Playwright version and no assertion failure; process evidence showed the server's Voice worker consuming CPU. I classify this as SLOW-only load, not a correctness failure. Voice memo behavior remains covered by the Notes workflow and the required staging smoke.

kayg commented

2026-10-03 21:01:18 +00:00

Author

Owner

E2E finding: test:e2e:app-passwords times out at its route assertion after opening the legacy /settings/account/app-passwords path. The current Settings registry canonicalizes this group to /settings/apps-devices/app-passwords (DESIGN §33; sections.ts stable section ID apps-devices), while App Passwords Copy link headings and the E2E assertion still emit /settings/apps/app-passwords. I will use settingsHref('apps-devices', 'app-passwords') for those heading links and update the E2E to assert the canonical destination while retaining the legacy entry URL.

E2E finding: `test:e2e:app-passwords` times out at its route assertion after opening the legacy `/settings/account/app-passwords` path. The current Settings registry canonicalizes this group to `/settings/apps-devices/app-passwords` (DESIGN §33; `sections.ts` stable section ID `apps-devices`), while App Passwords Copy link headings and the E2E assertion still emit `/settings/apps/app-passwords`. I will use `settingsHref('apps-devices', 'app-passwords')` for those heading links and update the E2E to assert the canonical destination while retaining the legacy entry URL.

kayg commented

2026-10-03 23:15:12 +00:00

Author

Owner

The Notes standalone run reproduced the 77.6875 px inset failure in assertNoteCardStack. The rendered section Card ends 18 px after the actual last editor block, but the e2e measurement collected only blocks carrying cal-card-member; the #632 performance change decorates only section boundaries, so ordinary middle paragraphs were omitted. The test traversal now mirrors NoteCardsLayer and still expects the same 18 px inset. The Notes workflow is rerunning to reach its later persistence checks.

The Notes standalone run reproduced the 77.6875 px inset failure in `assertNoteCardStack`. The rendered section Card ends 18 px after the actual last editor block, but the e2e measurement collected only blocks carrying `cal-card-member`; the #632 performance change decorates only section boundaries, so ordinary middle paragraphs were omitted. The test traversal now mirrors `NoteCardsLayer` and still expects the same 18 px inset. The Notes workflow is rerunning to reach its later persistence checks.

kayg commented

2026-10-03 23:17:26 +00:00

Author

Owner

The Notes rerun now passes the Card geometry check and continues through the file-origin Daily Log editor save/reload steps before stopping at the heading Copy link check. That check still expected mutable heading slugs, but #881 and DESIGN §33 changed copied heading links to stable block IDs. I updated the e2e assertion to verify the stable #^<id> format and to require the same identity across read, edit and touch Copy link actions. No product code or behavior expectation changed outside #881.

The Notes rerun now passes the Card geometry check and continues through the file-origin Daily Log editor save/reload steps before stopping at the heading Copy link check. That check still expected mutable heading slugs, but #881 and DESIGN §33 changed copied heading links to stable block IDs. I updated the e2e assertion to verify the stable `#^<id>` format and to require the same identity across read, edit and touch Copy link actions. No product code or behavior expectation changed outside #881.

kayg commented

2026-10-03 23:36:01 +00:00

Author

Owner

The repeated Note heading Copy action exposed a real editor state write: ensureAnchorAt called setBlockAnchor on an already valid unique ID, dispatching docChanged and allowing the heading widget to redraw during focus transfer. A new focused anchors.svelte.test.ts test failed with 1 document change on the second copy (expected 0); after the guard, it passes. The Notes browser run later hit an unrelated 5-second Format bubble timeout before reaching this interaction on its most recent attempt, so the browser-level confirmation remains pending.

The repeated Note heading Copy action exposed a real editor state write: `ensureAnchorAt` called `setBlockAnchor` on an already valid unique ID, dispatching `docChanged` and allowing the heading widget to redraw during focus transfer. A new focused `anchors.svelte.test.ts` test failed with 1 document change on the second copy (expected 0); after the guard, it passes. The Notes browser run later hit an unrelated 5-second Format bubble timeout before reaching this interaction on its most recent attempt, so the browser-level confirmation remains pending.

kayg commented

2026-10-03 23:41:47 +00:00

Author

Owner

The App Password e2e run reached its setup flow, then failed because it compared the clipboard's full credential with .copyable-value-text, which is deliberately shortened for display. The test also printed the credential in Node's assertion diff. I removed the failed raw log and changed the test to compare against the in-memory creation response with a value-free assertion message, as CopyableValue requires full-copy behavior (#723). The rerun will use redacted output capture.

The App Password e2e run reached its setup flow, then failed because it compared the clipboard's full credential with `.copyable-value-text`, which is deliberately shortened for display. The test also printed the credential in Node's assertion diff. I removed the failed raw log and changed the test to compare against the in-memory creation response with a value-free assertion message, as CopyableValue requires full-copy behavior (#723). The rerun will use redacted output capture.

kayg commented

2026-10-03 23:45:01 +00:00

Author

Owner

After fixing the App Password test's full-value CopyableValue assertion, the MCP setup flow passed password creation, list, and the User-level switch. Its tools/call then returned HTTP 400 with missing required Mcp-Name header for tools/call. The test client sent mcp-method but omitted the required tool identity header; it now sends mcp-name from the requested tool name. The failure log was redacted, and no credential value was written to the issue.

After fixing the App Password test's full-value CopyableValue assertion, the MCP setup flow passed password creation, list, and the User-level switch. Its `tools/call` then returned HTTP 400 with `missing required Mcp-Name header for tools/call`. The test client sent `mcp-method` but omitted the required tool identity header; it now sends `mcp-name` from the requested tool name. The failure log was redacted, and no credential value was written to the issue.

kayg commented

2026-10-03 23:47:32 +00:00

Author

Owner

The App Password flow now gets through MCP write after the test client sends mcp-name. It then stopped in the photo-backup setup because the e2e treated a shortened CopyableValue display string as the full server URL, user name and password. CopyableValue intentionally separates display from clipboard value (#723). The test now uses the one-time creation response in memory for full-value copy and WebDAV requests, checks the visible URL's shortened form without exposing its contents, and keeps assertions value-free.

The App Password flow now gets through MCP write after the test client sends `mcp-name`. It then stopped in the photo-backup setup because the e2e treated a shortened CopyableValue display string as the full server URL, user name and password. CopyableValue intentionally separates display from clipboard value (#723). The test now uses the one-time creation response in memory for full-value copy and WebDAV requests, checks the visible URL's shortened form without exposing its contents, and keeps assertions value-free.

kayg commented

2026-10-04 00:02:02 +00:00

Author

Owner

Round 3 E2E evidence: the App Password photo backup duplicate PUT returned HTTP 204, but the response included a Location for the collision-renamed APP_PASSWORD_E2E_849 2.HEIC. The test now checks that the User's original HEIC bytes are unchanged and that the renamed file contains the second upload. This matches the existing Files DAV unit test upload_preconditions_and_upload_only_name_collisions_are_enforced and DESIGN §26's no-replace rule.

The same E2E then showed /api/v1/photos/timeline?days=365&tiles_per_day=200 returns HTTP 400 bad_request; Photos limits a request to 90 days. I changed the test query to the supported 90-day window because the fixture has no capture date and uses its upload day. The focused workflow is rerunning now. No server behavior or existing product expectation changed.

Round 3 E2E evidence: the App Password photo backup duplicate PUT returned HTTP 204, but the response included a Location for the collision-renamed `APP_PASSWORD_E2E_849 2.HEIC`. The test now checks that the User's original HEIC bytes are unchanged and that the renamed file contains the second upload. This matches the existing Files DAV unit test `upload_preconditions_and_upload_only_name_collisions_are_enforced` and DESIGN §26's no-replace rule. The same E2E then showed `/api/v1/photos/timeline?days=365&tiles_per_day=200` returns HTTP 400 `bad_request`; Photos limits a request to 90 days. I changed the test query to the supported 90-day window because the fixture has no capture date and uses its upload day. The focused workflow is rerunning now. No server behavior or existing product expectation changed.

kayg commented

2026-10-04 00:16:42 +00:00

Author

Owner

App Passwords focused workflow now passes against the real production web build and local server, with macOS platform emulation. Verbatim result:

app password e2e: grants, MCP setup and tools, device setup, copy actions, and QR contrast passed

The run produced 114 real screenshots under artifacts/app-passwords-review/, covering the chooser and setup at 390, 820 and 1440 px in light and dark themes, plus the zoomed row captures. The sharp runtime works with a worktree-local C++ library shim; no dependency or system library changed.

The regression now verifies App Password copy actions against the in-memory create response without printing credentials. It checks the photo upload-only grant cannot read, escape its Home folder, or overwrite an existing photo: a duplicate PUT resolves to a new filename and leaves the original bytes unchanged. It also uses the supported 90-day Photos timeline request. Commits: 891f62756 and de2a448aa.

The installed fj issue client has no attachment subcommand. The screenshots remain in the shared worktree under the path above; they are not committed or claimed as uploaded.

App Passwords focused workflow now passes against the real production web build and local server, with macOS platform emulation. Verbatim result: ``` app password e2e: grants, MCP setup and tools, device setup, copy actions, and QR contrast passed ``` The run produced 114 real screenshots under `artifacts/app-passwords-review/`, covering the chooser and setup at 390, 820 and 1440 px in light and dark themes, plus the zoomed row captures. The `sharp` runtime works with a worktree-local C++ library shim; no dependency or system library changed. The regression now verifies App Password copy actions against the in-memory create response without printing credentials. It checks the photo upload-only grant cannot read, escape its Home folder, or overwrite an existing photo: a duplicate PUT resolves to a new filename and leaves the original bytes unchanged. It also uses the supported 90-day Photos timeline request. Commits: `891f62756` and `de2a448aa`. The installed `fj issue` client has no attachment subcommand. The screenshots remain in the shared worktree under the path above; they are not committed or claimed as uploaded.

kayg commented

2026-10-04 01:32:39 +00:00

Author

Owner

Round 3 finding for #1020: Composer E2E with macOS platform emulation closed after Event Send, but no /api/v1/calendar/events response occurred and the CalDAV fixture only held its two setup objects. The frozen Event snapshot intentionally carried calendarId: null for the synced default (#827), while commitEvent rejected null before calling the existing default-aware endpoint. Removed that guard and made the Note-linked Event endpoint accept an omitted destination through the same resolve_event_calendar helper. Added web and server regressions and regenerated OpenAPI/types. Commit: 273b01f4a. Calendar clippy/test and server clippy/test passed; web check and production build passed. A subsequent Composer run reached the CalDAV provider assertion, then showed two Send all test shortcuts still used Control on the macOS-emulated page; changed those to Meta and am rerunning.

Round 3 finding for #1020: Composer E2E with macOS platform emulation closed after Event Send, but no `/api/v1/calendar/events` response occurred and the CalDAV fixture only held its two setup objects. The frozen Event snapshot intentionally carried `calendarId: null` for the synced default (#827), while `commitEvent` rejected null before calling the existing default-aware endpoint. Removed that guard and made the Note-linked Event endpoint accept an omitted destination through the same `resolve_event_calendar` helper. Added web and server regressions and regenerated OpenAPI/types. Commit: `273b01f4a`. Calendar clippy/test and server clippy/test passed; web check and production build passed. A subsequent Composer run reached the CalDAV provider assertion, then showed two Send all test shortcuts still used Control on the macOS-emulated page; changed those to Meta and am rerunning.

kayg commented

2026-10-04 02:22:11 +00:00

Author

Owner

Round 3 finding — deferred overlay focus:

The Composer E2E showed the Composer dialog open while #route-content remained active in a fresh responsive browser context. The focus trap treated the unchanged opener as a new outside focus claim and skipped its deferred focus move.

Fix committed as a134334b43c7f1f4a878843d0f1af99ad6316d4e. Deferred focus now distinguishes the unchanged opener from a different outside control. The regression test covers both cases.

Evidence:

bunx vitest run src/lib/a11y/focusTrap.test.ts --maxWorkers=2: 6 tests passed.
bun run check: 0 errors, 2 existing CSS empty-ruleset warnings.
The refreshed Composer E2E passed the responsive focus checks and continued to the keyboard Undo scenario.

The Composer E2E currently stops at Undo restores the keyboard-discarded draft. I am collecting the draft and card state after Undo before deciding whether that is a product defect or a test setup issue.

Round 3 finding — deferred overlay focus: The Composer E2E showed the `Composer` dialog open while `#route-content` remained active in a fresh responsive browser context. The focus trap treated the unchanged opener as a new outside focus claim and skipped its deferred focus move. Fix committed as `a134334b43c7f1f4a878843d0f1af99ad6316d4e`. Deferred focus now distinguishes the unchanged opener from a different outside control. The regression test covers both cases. Evidence: - `bunx vitest run src/lib/a11y/focusTrap.test.ts --maxWorkers=2`: 6 tests passed. - `bun run check`: 0 errors, 2 existing CSS empty-ruleset warnings. - The refreshed Composer E2E passed the responsive focus checks and continued to the keyboard Undo scenario. The Composer E2E currently stops at `Undo restores the keyboard-discarded draft`. I am collecting the draft and card state after Undo before deciding whether that is a product defect or a test setup issue.

kayg commented

2026-10-04 04:07:38 +00:00

Author

Owner

Round 3 progress from head f93b49120 (commits 85b17c5c3, b455e2b96, f93b49120).

The real Composer E2E exposed two integration defects. Reopening Composer while its earlier surface was still closing let the old focus restore steal the new text field. Undo also lost mouse activation because pointer-down scrolled the composer capsule; the Undo control now prevents mouse focus scroll and uses the shared touch target. The production flow now passes pointer, keyboard, and touch Undo, including a tap in the transparent area outside the painted label.

The cold-parse E2E timer also included the 150 ms polling helper and parser request startup, rather than measuring the pending row's visible paint. The test now measures the row directly, then confirms it remains visible while the NLP response is held. The 50 ms budget is unchanged.

Evidence: bun run check passed with 0 errors and 2 existing empty CSS ruleset warnings; focused Vitest passed 38 tests; test:e2e:composer passed with macOS shortcut emulation, 390/820/1440 px screenshots in light/dark, held-NLP coverage, and 0 CSP reports across 24 pages. The Composer perf-ledger fingerprints were refreshed with the exception count unchanged.

Round 3 progress from head `f93b49120` (commits `85b17c5c3`, `b455e2b96`, `f93b49120`). The real Composer E2E exposed two integration defects. Reopening Composer while its earlier surface was still closing let the old focus restore steal the new text field. Undo also lost mouse activation because pointer-down scrolled the composer capsule; the Undo control now prevents mouse focus scroll and uses the shared touch target. The production flow now passes pointer, keyboard, and touch Undo, including a tap in the transparent area outside the painted label. The cold-parse E2E timer also included the 150 ms polling helper and parser request startup, rather than measuring the pending row's visible paint. The test now measures the row directly, then confirms it remains visible while the NLP response is held. The 50 ms budget is unchanged. Evidence: `bun run check` passed with 0 errors and 2 existing empty CSS ruleset warnings; focused Vitest passed 38 tests; `test:e2e:composer` passed with macOS shortcut emulation, 390/820/1440 px screenshots in light/dark, held-NLP coverage, and 0 CSP reports across 24 pages. The Composer perf-ledger fingerprints were refreshed with the exception count unchanged.

kayg commented

2026-10-04 05:57:23 +00:00

Author

Owner

The full production shell E2E exposed two stale screenshot readiness checks. Calendar views are now live Calendar navigation links per DESIGN §34, so the capture now opens the sidebar and verifies the selected Week link. The tag endpoint returned both log_entry and note; the Tag page renders the Log entry group as “Journal”, which matches CONTEXT.md, so both screenshot paths now wait for that documented label. I am rerunning the full shell flow with macOS emulation and the production build.

The full production shell E2E exposed two stale screenshot readiness checks. Calendar views are now live Calendar navigation links per DESIGN §34, so the capture now opens the sidebar and verifies the selected Week link. The tag endpoint returned both `log_entry` and `note`; the Tag page renders the Log entry group as “Journal”, which matches CONTEXT.md, so both screenshot paths now wait for that documented label. I am rerunning the full shell flow with macOS emulation and the production build.

kayg commented

2026-10-04 06:02:58 +00:00

Author

Owner

The next shell capture failed because it still expected the retired “Week and Day” / #opt-grid setting. CalendarsSection.svelte and settings/sections.ts show that issue #624 retired that control and the compatibility route now maps to the live “Dragging Items” group. I changed the capture to use the canonical Calendar dragging route, check the live “Snap to other items” control, and name the screenshot for the current group.

The next shell capture failed because it still expected the retired “Week and Day” / `#opt-grid` setting. `CalendarsSection.svelte` and `settings/sections.ts` show that issue #624 retired that control and the compatibility route now maps to the live “Dragging Items” group. I changed the capture to use the canonical Calendar dragging route, check the live “Snap to other items” control, and name the screenshot for the current group.

kayg commented

2026-10-04 06:15:23 +00:00

Author

Owner

Mode reordering reached the focused Calendar Tab with the expected macOS Meta+Shift+ArrowRight key sequence. The test still read calternal.settings from raw localStorage, but #555 moved that preference into per-User storage; the supported test seam is window.__userStorageTest. I changed the reorder checks to read and write through that seam, and compare the visible order while preserving hidden Tab slots as DESIGN §34 requires.

Mode reordering reached the focused Calendar Tab with the expected macOS `Meta+Shift+ArrowRight` key sequence. The test still read `calternal.settings` from raw localStorage, but #555 moved that preference into per-User storage; the supported test seam is `window.__userStorageTest`. I changed the reorder checks to read and write through that seam, and compare the visible order while preserving hidden Tab slots as DESIGN §34 requires.

kayg commented

2026-10-04 06:26:33 +00:00

Author

Owner

The phone reorder probe reached the Tab Bar but left the saved order unchanged. The recorded geometry showed its drag began on Photos while the 390 px Tab Bar showed only earlier items; the touch event was outside the visible scroll window. The probe now checks Notes and Tasks, which are visible in the first phone window, and still verifies the User-scoped full order and route after the gesture.

kayg commented

2026-10-04 07:30:38 +00:00

Author

Owner

Round 3 update: the focused production Tab Bar width matrix passes for 4, 5, 6 and 7 live modes at 390, 820 and 1440 px in both themes. A Settings switch returned 204 but did not reach the expected aria state once during the combined shell run; a fresh width-only run then passed all four mode counts and all real Settings switches, so this did not reproduce as an API/UI defect.

The phone flow exposed a test timing issue: after returning to Calendar, the second touch used the old Files coordinates while the selected-label scroll reveal was still moving. The recorded click target was Calendar, with Files at x=303–353 and the tray ending at x=309. The focused phone test now waits for the selection morph and scroll position to settle before it taps Files again.

That phone run then measured selected-label transitions at 203.4 ms and 202.3 ms against the existing 200 ms threshold. These are the only remaining findings from this pass; the threshold stays unchanged. The values are recorded as load-sensitive timing evidence under the owner rule that performance is not a merge gate.

Round 3 update: the focused production Tab Bar width matrix passes for 4, 5, 6 and 7 live modes at 390, 820 and 1440 px in both themes. A Settings switch returned 204 but did not reach the expected aria state once during the combined shell run; a fresh width-only run then passed all four mode counts and all real Settings switches, so this did not reproduce as an API/UI defect. The phone flow exposed a test timing issue: after returning to Calendar, the second touch used the old Files coordinates while the selected-label scroll reveal was still moving. The recorded click target was Calendar, with Files at x=303–353 and the tray ending at x=309. The focused phone test now waits for the selection morph and scroll position to settle before it taps Files again. That phone run then measured selected-label transitions at 203.4 ms and 202.3 ms against the existing 200 ms threshold. These are the only remaining findings from this pass; the threshold stays unchanged. The values are recorded as load-sensitive timing evidence under the owner rule that performance is not a merge gate.

kayg commented

2026-10-04 07:39:24 +00:00

Author

Owner

Round 3 finding: the production shell E2E run reached Settings account, then failed at apps/web/e2e/shell.mjs:2066 waiting for an H1 named “Server configuration”. The current Settings tree gives admin/configuration the section label “Configuration” (apps/web/src/routes/settings/sections.ts → sectionLabel), while the “Server configuration” text is the group card title in ConfigGroup.svelte. The existing expectation is unchanged per the owner rule. Earlier tray reorder and passkey flows passed; the shell suite did not complete. The full log is artifacts/round3-full-e2e-shell-final.log.

Round 3 finding: the production shell E2E run reached Settings account, then failed at `apps/web/e2e/shell.mjs:2066` waiting for an H1 named “Server configuration”. The current Settings tree gives `admin/configuration` the section label “Configuration” (`apps/web/src/routes/settings/sections.ts` → `sectionLabel`), while the “Server configuration” text is the group card title in `ConfigGroup.svelte`. The existing expectation is unchanged per the owner rule. Earlier tray reorder and passkey flows passed; the shell suite did not complete. The full log is `artifacts/round3-full-e2e-shell-final.log`.

kayg commented

2026-10-04 07:43:37 +00:00

Author

Owner

Round 3 auth E2E finding and fix: page.waitForURL() returned after the SPA changed to the calendar URL, but before the calendar view mounted. Starting the next full-page login navigation at that point intermittently aborted (ERR_ABORTED). Waiting for .calendar before signing out keeps the original route and auth assertions. CALTERNAL_E2E_MAC=1 bun run test:e2e:auth now passes end-to-end: setup, passkey sign-in, Settings, invitations, config validation, role change and recovery; CSP reports: 0 across 3 pages. Detailed output: artifacts/round3-auth-e2e-final.log.

Round 3 auth E2E finding and fix: `page.waitForURL()` returned after the SPA changed to the calendar URL, but before the calendar view mounted. Starting the next full-page login navigation at that point intermittently aborted (`ERR_ABORTED`). Waiting for `.calendar` before signing out keeps the original route and auth assertions. `CALTERNAL_E2E_MAC=1 bun run test:e2e:auth` now passes end-to-end: setup, passkey sign-in, Settings, invitations, config validation, role change and recovery; CSP reports: 0 across 3 pages. Detailed output: `artifacts/round3-auth-e2e-final.log`.

kayg commented

2026-10-04 08:22:07 +00:00

Author

Owner

Merge round 7b — round 3 final report

Issue: #867
Branch: job/merge-round-7b2
Head: 0c4fd513e79d624ef30d04f0e85d2769cd4ebecb

7b READY FOR PRODUCTION: no. The round reached its four-hour work limit before staging. No new inputs were added. There was no push, merge, production action, or staging deployment.

Built

Addressed #1020 browser integration repros across Settings deep links, Calendar authority, overlay focus, Quick Look dismissal, and Notes/composer route readiness. Added the E2E wait for the Calendar view to mount before the Auth flow starts its next navigation; this removed an observed intermittent ERR_ABORTED while keeping every assertion intact.
#1021 Mail delta catch-up now retries through checkpoint contention.
#1022 Notes projection rebuild retries transient SQLite contention.
#1023 preserved the MCP session DELETE status contract.
The round changed 65 files (git diff --stat 79a0c14c9..HEAD: 3,715 insertions, 2,228 deletions). Main areas: crates/plugins/mail/, crates/plugins/notes/, crates/calternal-server/, crates/calternal-db/, Settings/Calendar/Mail/Notes/Composer web code, production E2E workflows, and adversarial fixtures. The final round-three Auth-only test change is apps/web/e2e/auth.mjs.

Gates

cargo fmt --check exited 0 with empty output (artifacts/round3-rust-gates/fmt.log). Final per-crate cargo clippy -p <crate> --all-targets -- -D warnings and cargo test -p <crate> -- --test-threads=4 logs are under artifacts/round3-rust-gates/. All 22 final per-crate Clippy logs ended successfully, and all 22 test logs contain only passing summaries. This includes Server, Mail, and Notes.

Verbatim final Rust test summaries:

calternal-embed:
test result: ok. 36 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 21.57s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-fs:
test result: ok. 85 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 15.38s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.08s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.50s
test result: ok. 47 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 13.08s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.05s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-imap:
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 30 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.15s
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-location:
test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.19s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-media:
test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-money:
test result: ok. 16 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.06s
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s
test result: ok. 12 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.30s
test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.45s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.37s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-notes-core:
test result: ok. 544 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.15s
test result: ok. 19 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.26s
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.04s
test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.36s
test result: ok. 12 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-path:
test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-ai:
test result: ok. 13 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.56s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-analytics:
test result: ok. 34 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.63s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-calendar:
test result: ok. 99 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 7.00s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.24s
test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.14s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-mail:
test result: ok. 65 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 8.50s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-money:
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 60 filtered out; finished in 26.11s
test result: ok. 60 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 26.14s
test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 4.90s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-notes:
test result: ok. 256 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 137.02s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.91s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-notifications:
test result: ok. 28 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.49s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-photos:
test result: ok. 52 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 3.29s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin:
test result: ok. 39 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 4.21s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-plugin-video:
test result: ok. 16 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.11s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-search:
test result: ok. 51 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 7.41s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.64s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.05s
test result: ok. 24 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 228.46s
test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.02s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.02s
test result: ok. 1 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 3.87s
test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-server:
test result: ok. 206 passed; 0 failed; 8 ignored; 0 measured; 0 filtered out; finished in 42.35s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 17.25s
test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.13s
calternal-sync:
test result: ok. 60 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.27s
test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s
calternal-tags:
test result: ok. 18 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.27s
test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s

The final web check output was:

perf-lint: PASS; 0 violations; 19189 scoped exceptions
svelte-check found 0 errors and 2 warnings in 2 files

Full web Vitest output:

 Test Files  216 passed (216)
      Tests  1464 passed (1464)

The production web build completed with:

Compressed 532 static variants; saved 9719833 bytes.

The packaged E2E sweep ran 67 workflows: 8 passed, 56 exited non-zero, and 3 timed out. After the sweep, the focused Auth workflow passed end-to-end (setup, passkey sign-in, Settings, invitations, config validation, role change, recovery; CSP reports: 0 across 3 pages; artifacts/round3-auth-e2e-final.log). The last full test:e2e shell run reached account screenshots and then failed at apps/web/e2e/shell.mjs:2066: the existing test expects the H1 “Server configuration”, while the current section heading is “Configuration” and “Server configuration” is the card title. The test expectation was not changed. The 390 px phone workflow passed its interactions; five measured label morphs included 203.4 ms and 202.3 ms against the unchanged 200 ms assertion.

The full tests/adversarial/run.sh round exited 1. Route classification passed for 375 operations and 1,062 generated tools; Admin classification covered 39 operations. The live Admin authorization matrix passed 2,348 requests across four identities and 18 App Password scope classes. The Cross-User matrix stopped at a shared Photo missing from its Calendar range. Other recorded findings include a Journal attachment append concurrency check reporting lost child links, a daily Log GET returning 404, two Notes IMAP FETCH/replay failures, and a Photos Undo request receiving 502 while the runner reported its local server unavailable. Existing reminder status assertions (DELETE: expected 204, got 200) and malformed Calendar cursor assertion (expected 200, got 400) were left unchanged. Several reminder, semantic recall, zip, and Calendar duplicate operations were SLOW under load. Two CLI attack phases could not start because the runner was given a shared server binary but the matching calternal CLI binary was absent. Full evidence is in artifacts/round3-adversarial.log.

Cleanup completed. Verbatim cargo clean output:

     Removed 33366 files, 33.0GiB total

apps/web/build and apps/web/.svelte-kit/output were removed. The worktree is clean.

Staging and smoke

Staging was not built or deployed. No staging tag was created. The staging Notes IMAP edge variables were not re-enabled and no environment backup was made. The requested Settings redesign, Agenda decks, Mail layouts, generated Money import fixture, Voice Memos, Undo flows, Notes HHMM, and Notes-over-IMAP edge smokes were not run.

UX gaps

Closed: the Auth test now waits for real Calendar content before navigating away; the route/focus fixes listed above are covered by the focused tests recorded in this round.

Left: the complete browser workflow sweep remains red; not every touched screen has the required 390/820/1440 light/dark evidence in this round; staging smoke and owner visual review remain pending. The Calendar/Notes shared Photo projection and Journal child-link concurrency findings need code fixes and focused regression tests before readiness.

Decisions

No product behavior was chosen outside docs/DESIGN.md. The test’s current H1 expectation and existing response-status expectations were preserved under the owner rule; the orchestrator should decide whether the H1 expectation or the UI title is out of date.

# Merge round 7b — round 3 final report Issue: #867 Branch: `job/merge-round-7b2` Head: `0c4fd513e79d624ef30d04f0e85d2769cd4ebecb` **7b READY FOR PRODUCTION: no.** The round reached its four-hour work limit before staging. No new inputs were added. There was no push, merge, production action, or staging deployment. ## Built - Addressed #1020 browser integration repros across Settings deep links, Calendar authority, overlay focus, Quick Look dismissal, and Notes/composer route readiness. Added the E2E wait for the Calendar view to mount before the Auth flow starts its next navigation; this removed an observed intermittent `ERR_ABORTED` while keeping every assertion intact. - #1021 Mail delta catch-up now retries through checkpoint contention. - #1022 Notes projection rebuild retries transient SQLite contention. - #1023 preserved the MCP session DELETE status contract. - The round changed 65 files (`git diff --stat 79a0c14c9..HEAD`: 3,715 insertions, 2,228 deletions). Main areas: `crates/plugins/mail/`, `crates/plugins/notes/`, `crates/calternal-server/`, `crates/calternal-db/`, Settings/Calendar/Mail/Notes/Composer web code, production E2E workflows, and adversarial fixtures. The final round-three Auth-only test change is `apps/web/e2e/auth.mjs`. ## Gates `cargo fmt --check` exited 0 with empty output (`artifacts/round3-rust-gates/fmt.log`). Final per-crate `cargo clippy -p <crate> --all-targets -- -D warnings` and `cargo test -p <crate> -- --test-threads=4` logs are under `artifacts/round3-rust-gates/`. All 22 final per-crate Clippy logs ended successfully, and all 22 test logs contain only passing summaries. This includes Server, Mail, and Notes. Verbatim final Rust test summaries: ```text calternal-embed: test result: ok. 36 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 21.57s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-fs: test result: ok. 85 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 15.38s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.08s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.50s test result: ok. 47 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 13.08s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.05s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-imap: test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 30 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.15s test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-location: test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.19s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-media: test result: ok. 6 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-money: test result: ok. 16 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.06s test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.03s test result: ok. 12 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 6.30s test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.45s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.37s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-notes-core: test result: ok. 544 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.15s test result: ok. 19 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.26s test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.04s test result: ok. 7 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.36s test result: ok. 12 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-path: test result: ok. 8 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.01s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-ai: test result: ok. 13 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.56s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-analytics: test result: ok. 34 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.63s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-calendar: test result: ok. 99 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 7.00s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.24s test result: ok. 3 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.14s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-mail: test result: ok. 65 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 8.50s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-money: test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 60 filtered out; finished in 26.11s test result: ok. 60 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 26.14s test result: ok. 27 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 4.90s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-notes: test result: ok. 256 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 137.02s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.91s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-notifications: test result: ok. 28 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.49s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-photos: test result: ok. 52 passed; 0 failed; 3 ignored; 0 measured; 0 filtered out; finished in 3.29s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin: test result: ok. 39 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 4.21s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-plugin-video: test result: ok. 16 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.11s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-search: test result: ok. 51 passed; 0 failed; 1 ignored; 0 measured; 0 filtered out; finished in 7.41s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.64s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.05s test result: ok. 24 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 228.46s test result: ok. 5 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.02s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.02s test result: ok. 1 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 3.87s test result: ok. 4 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-server: test result: ok. 206 passed; 0 failed; 8 ignored; 0 measured; 0 filtered out; finished in 42.35s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 17.25s test result: ok. 1 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.13s calternal-sync: test result: ok. 60 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 1.27s test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s calternal-tags: test result: ok. 18 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.27s test result: ok. 0 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.00s ``` The final web check output was: ```text perf-lint: PASS; 0 violations; 19189 scoped exceptions svelte-check found 0 errors and 2 warnings in 2 files ``` Full web Vitest output: ```text Test Files 216 passed (216) Tests 1464 passed (1464) ``` The production web build completed with: ```text Compressed 532 static variants; saved 9719833 bytes. ``` The packaged E2E sweep ran 67 workflows: 8 passed, 56 exited non-zero, and 3 timed out. After the sweep, the focused Auth workflow passed end-to-end (setup, passkey sign-in, Settings, invitations, config validation, role change, recovery; CSP reports: 0 across 3 pages; `artifacts/round3-auth-e2e-final.log`). The last full `test:e2e` shell run reached account screenshots and then failed at `apps/web/e2e/shell.mjs:2066`: the existing test expects the H1 “Server configuration”, while the current section heading is “Configuration” and “Server configuration” is the card title. The test expectation was not changed. The 390 px phone workflow passed its interactions; five measured label morphs included 203.4 ms and 202.3 ms against the unchanged 200 ms assertion. The full `tests/adversarial/run.sh` round exited 1. Route classification passed for 375 operations and 1,062 generated tools; Admin classification covered 39 operations. The live Admin authorization matrix passed 2,348 requests across four identities and 18 App Password scope classes. The Cross-User matrix stopped at a shared Photo missing from its Calendar range. Other recorded findings include a Journal attachment append concurrency check reporting lost child links, a daily Log GET returning 404, two Notes IMAP FETCH/replay failures, and a Photos Undo request receiving 502 while the runner reported its local server unavailable. Existing reminder status assertions (DELETE: expected 204, got 200) and malformed Calendar cursor assertion (expected 200, got 400) were left unchanged. Several reminder, semantic recall, zip, and Calendar duplicate operations were SLOW under load. Two CLI attack phases could not start because the runner was given a shared server binary but the matching `calternal` CLI binary was absent. Full evidence is in `artifacts/round3-adversarial.log`. Cleanup completed. Verbatim `cargo clean` output: ```text Removed 33366 files, 33.0GiB total ``` `apps/web/build` and `apps/web/.svelte-kit/output` were removed. The worktree is clean. ## Staging and smoke Staging was not built or deployed. No staging tag was created. The staging Notes IMAP edge variables were not re-enabled and no environment backup was made. The requested Settings redesign, Agenda decks, Mail layouts, generated Money import fixture, Voice Memos, Undo flows, Notes HHMM, and Notes-over-IMAP edge smokes were not run. ## UX gaps Closed: the Auth test now waits for real Calendar content before navigating away; the route/focus fixes listed above are covered by the focused tests recorded in this round. Left: the complete browser workflow sweep remains red; not every touched screen has the required 390/820/1440 light/dark evidence in this round; staging smoke and owner visual review remain pending. The Calendar/Notes shared Photo projection and Journal child-link concurrency findings need code fixes and focused regression tests before readiness. ## Decisions No product behavior was chosen outside `docs/DESIGN.md`. The test’s current H1 expectation and existing response-status expectations were preserved under the owner rule; the orchestrator should decide whether the H1 expectation or the UI title is out of date.

kayg commented

2026-10-04 08:26:17 +00:00

Author

Owner

7bfix-photos started on branch job/7bfix-photos, base 0c4fd513e. Scope: Calendar Photo projection, thumbnail worker, Search normalization and the listed focused probe findings. I will compare each finding with production dev 6074f71d1 and commit fixes separately. No push or deploy.

7bfix-photos started on branch job/7bfix-photos, base 0c4fd513e. Scope: Calendar Photo projection, thumbnail worker, Search normalization and the listed focused probe findings. I will compare each finding with production dev 6074f71d1 and commit fixes separately. No push or deploy.

kayg commented

2026-10-04 08:26:25 +00:00

Author

Owner

Started 7bfix-data on job/7bfix-data, base 0c4fd513e79d624ef30d04f0e85d2769cd4ebecb. Scope: Journal attachment persistence, oversized-body response handling, block reminders, Daily GET, and bounded voice failure. I will compare with production revision 6074f71d1, make one commit per finding, and run focused defensive regression tests. No push or deploy.

Started 7bfix-data on `job/7bfix-data`, base `0c4fd513e79d624ef30d04f0e85d2769cd4ebecb`. Scope: Journal attachment persistence, oversized-body response handling, block reminders, Daily GET, and bounded voice failure. I will compare with production revision `6074f71d1`, make one commit per finding, and run focused defensive regression tests. No push or deploy.

kayg commented

2026-10-04 08:26:39 +00:00

Author

Owner

Starting the 7bfix-e2e sweep triage. Branch: job/7bfix-e2e. Base SHA: 0c4fd513e7. The current issue description says read-only merge-order audit, while the explicit job prompt assigns the packaged e2e sweep. I am following the e2e job prompt and will report the discrepancy with the final findings.

Starting the 7bfix-e2e sweep triage. Branch: job/7bfix-e2e. Base SHA: 0c4fd513e79d624ef30d04f0e85d2769cd4ebecb. The current issue description says read-only merge-order audit, while the explicit job prompt assigns the packaged e2e sweep. I am following the e2e job prompt and will report the discrepancy with the final findings.

kayg commented

2026-10-04 08:34:26 +00:00

Author

Owner

Findings on job/7bfix-data:

Journal: production appends repeated targets; round 7b deduplicates them. The round-3 check sent one target with 12 labels, then required all labels. The route already has a distinct-file concurrency regression (journal_append_attachments_is_checked_idempotent_and_concurrent). Commit e7a281140 checks target persistence and retry idempotency, with four helper tests. This is a changed-contract probe mismatch, not established data loss.
Daily: POST extracts DailyQuery, not JSON. The probe created today then read 2099-01-02. Commit ba4b7931f uses the declared query. Production GET creates a missing date, masking this fixture error; round 7b correctly makes GET read-only (#752).
Reminder DELETE: #768 adds a 200 body with the Undo revision. Commit ddfb93e52 checks that response and revision. Production returned 204.
Transport: the recorded Photos 502 body is emitted only by editor-proxy.mjs. The supplied evidence does not contain its socket error code or backend lifecycle log. Synthetic early-rejection tests pass at 200 KB, 5 MiB, and 8 MiB+1, before and after Continue. No server crash is established; no live oversized-body finding is cleared.
Reminder create 126: the route caps all records at 128 and indexes the whole reminder projection on each mutation. Production uses the same cap and projection shape. No isolated timing result was obtained. The timeout remains unclassified as load versus a defect.
Voice: production has no voice.rs route. Current decode timeout is 15 minutes; slot wait is 5 minutes; model download can wait 30 minutes. The supplied 30-second processing result does not identify which stage was active. This remains unresolved.

The required fetch and merge was attempted once. origin/dev is 6074f71d18. The merge has 18 conflicts. The server startup architectures conflict (#549 versus #1011), and Notes migration 0028 conflicts (Task instant versus deployed headingless Log rebuild). I aborted it to preserve scoped commits. This needs the merge-round owner.

Only defensive contract and synthetic transport tests ran in this job. Live hostile-input/DoS reproduction did not run. READY remains no.

Findings on `job/7bfix-data`: - Journal: production appends repeated targets; round 7b deduplicates them. The round-3 check sent one target with 12 labels, then required all labels. The route already has a distinct-file concurrency regression (`journal_append_attachments_is_checked_idempotent_and_concurrent`). Commit e7a281140 checks target persistence and retry idempotency, with four helper tests. This is a changed-contract probe mismatch, not established data loss. - Daily: POST extracts `DailyQuery`, not JSON. The probe created today then read 2099-01-02. Commit ba4b7931f uses the declared query. Production GET creates a missing date, masking this fixture error; round 7b correctly makes GET read-only (#752). - Reminder DELETE: #768 adds a 200 body with the Undo revision. Commit ddfb93e52 checks that response and revision. Production returned 204. - Transport: the recorded Photos 502 body is emitted only by `editor-proxy.mjs`. The supplied evidence does not contain its socket error code or backend lifecycle log. Synthetic early-rejection tests pass at 200 KB, 5 MiB, and 8 MiB+1, before and after Continue. No server crash is established; no live oversized-body finding is cleared. - Reminder create 126: the route caps all records at 128 and indexes the whole reminder projection on each mutation. Production uses the same cap and projection shape. No isolated timing result was obtained. The timeout remains unclassified as load versus a defect. - Voice: production has no `voice.rs` route. Current decode timeout is 15 minutes; slot wait is 5 minutes; model download can wait 30 minutes. The supplied 30-second processing result does not identify which stage was active. This remains unresolved. The required fetch and merge was attempted once. `origin/dev` is 6074f71d18abe73b2b4255acb564f275a9acc851. The merge has 18 conflicts. The server startup architectures conflict (#549 versus #1011), and Notes migration 0028 conflicts (Task instant versus deployed headingless Log rebuild). I aborted it to preserve scoped commits. This needs the merge-round owner. Only defensive contract and synthetic transport tests ran in this job. Live hostile-input/DoS reproduction did not run. READY remains no.

kayg referenced this issue

2026-10-04 08:34:27 +00:00

VOICE: revive and merge local Parakeet transcription (job/voice, #373) + backfill-transcribe existing voice memos #619

kayg referenced this issue

2026-10-04 08:34:28 +00:00

Investigate API transport timeouts during mixed read/write stress #368

kayg commented

2026-10-04 08:36:21 +00:00

Author

Owner

READY: no

Built five atomic test commits on job/7bfix-data. Head: 44a49ceb85bb9e8ec4d56fc6def4a7396db2f561. No push or deploy. No product Rust or web code changed.

Files:

tests/adversarial/attack.py: use the Daily query contract; require reminder Remove's Undo revision; check retry identity by file target.
tests/adversarial/dav_probe_contracts.py: shared attachment identity check.
tests/adversarial/test_dav_probe.py: four attachment identity regressions, including direct test invocation.
tests/adversarial/proxy-early-response.test.mjs: six synthetic early-413 cases across the 4 MiB buffering boundary, before and after Continue.

Finding	Regression/pre-existing	Fixed/filed
Journal missing retry labels	Regression in probe expectation after target deduplication; no data loss established	Fixed probe, `e7a281140`; current distinct-file route test was inspected, not rerun
Photos oversized Undo 502	Pre-existing proxy code; live cause unverified	Evidence filed on #368; NOT cleared
Finder oversized AppleDouble Broken pipe	Pre-existing early-close transport handling; live cause unverified	Evidence filed on #368; NOT cleared
Reminder create 126 timeout	Same cap and projection on dev; regression status unverified	Evidence filed on #368; NOT cleared
Daily GET 404	Regression in probe compatibility with read-only GET; POST ignored JSON on both revisions	Fixed probe, `ba4b7931f`
Reminder DELETE 200	Regression in probe compatibility with #768 Undo response	Fixed probe, `ddfb93e52`
Voice processing after 30 seconds	Regression/new feature; production has no Voice route	Evidence filed on #619; NOT cleared

Classification limits: production comparison used git show 6074f71d1:<file>, not live reproduction. The transport and timeout classifications are source comparisons, not proven behavior comparisons. No real-server before/after run was obtained. Hostile-input and DoS reproduction did not run; tests were defensive contract checks and synthetic protocol fixtures.

Known gaps: server lifecycle and socket-error evidence is missing for 502/Broken pipe; no isolated reminder timing exists; Voice's stage at timeout is unknown. Its configured decode/slot/model deadlines exceed the probe's 30 seconds. These findings must stay open.

Merge: ran git fetch origin and attempted git merge origin/dev once. origin/dev was 6074f71d18abe73b2b4255acb564f275a9acc851. There were 18 conflicts, including incompatible startup lifecycles (#549/#1011) and Notes migration 0028 with two meanings (Task instant versus deployed headingless Log rebuild). Aborted the merge, preserving the scoped commits. The merge-round owner must resolve it; this branch is not integrated with dev.

Gates (verbatim):

cargo fmt --check: exit 0, no output.

python3 tests/adversarial/test_dav_probe.py:

.....................
----------------------------------------------------------------------
Ran 21 tests in 0.085s

OK
KNOWN litmus owner_modify: 403 Forbidden
KNOWN litmus complex_cond_put: 400 Bad Request

node --test tests/adversarial/proxy-early-response.test.mjs:

TAP version 13
# Subtest: proxy preserves 413 for 200000 bytes (Continue first: false)
ok 1 - proxy preserves 413 for 200000 bytes (Continue first: false)
  ---
  duration_ms: 418.679023
  type: 'test'
  ...
# Subtest: proxy preserves 413 for 200000 bytes (Continue first: true)
ok 2 - proxy preserves 413 for 200000 bytes (Continue first: true)
  ---
  duration_ms: 650.222228
  type: 'test'
  ...
# Subtest: proxy preserves 413 for 5242880 bytes (Continue first: false)
ok 3 - proxy preserves 413 for 5242880 bytes (Continue first: false)
  ---
  duration_ms: 628.854439
  type: 'test'
  ...
# Subtest: proxy preserves 413 for 5242880 bytes (Continue first: true)
ok 4 - proxy preserves 413 for 5242880 bytes (Continue first: true)
  ---
  duration_ms: 518.43946
  type: 'test'
  ...
# Subtest: proxy preserves 413 for 8388609 bytes (Continue first: false)
ok 5 - proxy preserves 413 for 8388609 bytes (Continue first: false)
  ---
  duration_ms: 719.498268
  type: 'test'
  ...
# Subtest: proxy preserves 413 for 8388609 bytes (Continue first: true)
ok 6 - proxy preserves 413 for 8388609 bytes (Continue first: true)
  ---
  duration_ms: 638.807761
  type: 'test'
  ...
1..6
# tests 6
# suites 0
# pass 6
# fail 0
# cancelled 0
# skipped 0
# todo 0
# duration_ms 3812.729504

python3 -m py_compile tests/adversarial/attack.py tests/adversarial/dav_probe_contracts.py tests/adversarial/test_dav_probe.py: exit 0, no output.

git diff --check: exit 0, no output.

cargo clean:

     Removed 1 file, 356B total

Rust clippy/test and web check/Vitest were not run: no Rust crate or web source changed. No web build output was produced. No dependencies changed; existing test dependencies were installed with bun install --frozen-lockfile. Comments in all four changed files were re-read.

Decisions: no product decisions. Retain documented target deduplication (#421), read-only Daily GET with explicit query-based POST (#752), and revision-bearing reminder Remove (#768). Keep the original server findings open; passing synthetic transport checks do not prove them fixed.

UX gaps closed/left: not applicable; no UI changes.

For the merge round: after reconciling dev and migration numbers, run tests/adversarial/run.sh once with the matching server and CLI builds. It must prove distinct-file attachment persistence, Daily POST/GET parity, reminder Undo revision handling, backend 413 with process survival, bounded reminder mutations, and terminal invalid-audio failure without model work. Retain backend lifecycle logs and content-free write-stage/socket diagnostics. Full Rust/web gates belong to that combined round. No performance measurement ran: this job did not change a user-facing hot path and is not a performance issue.

READY: no Built five atomic test commits on `job/7bfix-data`. Head: `44a49ceb85bb9e8ec4d56fc6def4a7396db2f561`. No push or deploy. No product Rust or web code changed. Files: - `tests/adversarial/attack.py`: use the Daily query contract; require reminder Remove's Undo revision; check retry identity by file target. - `tests/adversarial/dav_probe_contracts.py`: shared attachment identity check. - `tests/adversarial/test_dav_probe.py`: four attachment identity regressions, including direct test invocation. - `tests/adversarial/proxy-early-response.test.mjs`: six synthetic early-413 cases across the 4 MiB buffering boundary, before and after Continue. | Finding | Regression/pre-existing | Fixed/filed | | --- | --- | --- | | Journal missing retry labels | Regression in probe expectation after target deduplication; no data loss established | Fixed probe, e7a281140; current distinct-file route test was inspected, not rerun | | Photos oversized Undo 502 | Pre-existing proxy code; live cause unverified | Evidence filed on #368; NOT cleared | | Finder oversized AppleDouble Broken pipe | Pre-existing early-close transport handling; live cause unverified | Evidence filed on #368; NOT cleared | | Reminder create 126 timeout | Same cap and projection on dev; regression status unverified | Evidence filed on #368; NOT cleared | | Daily GET 404 | Regression in probe compatibility with read-only GET; POST ignored JSON on both revisions | Fixed probe, ba4b7931f | | Reminder DELETE 200 | Regression in probe compatibility with #768 Undo response | Fixed probe, ddfb93e52 | | Voice processing after 30 seconds | Regression/new feature; production has no Voice route | Evidence filed on #619; NOT cleared | Classification limits: production comparison used `git show 6074f71d1:<file>`, not live reproduction. The transport and timeout classifications are source comparisons, not proven behavior comparisons. No real-server before/after run was obtained. Hostile-input and DoS reproduction did not run; tests were defensive contract checks and synthetic protocol fixtures. Known gaps: server lifecycle and socket-error evidence is missing for 502/Broken pipe; no isolated reminder timing exists; Voice's stage at timeout is unknown. Its configured decode/slot/model deadlines exceed the probe's 30 seconds. These findings must stay open. Merge: ran `git fetch origin` and attempted `git merge origin/dev` once. `origin/dev` was `6074f71d18abe73b2b4255acb564f275a9acc851`. There were 18 conflicts, including incompatible startup lifecycles (#549/#1011) and Notes migration 0028 with two meanings (Task instant versus deployed headingless Log rebuild). Aborted the merge, preserving the scoped commits. The merge-round owner must resolve it; this branch is not integrated with dev. Gates (verbatim): `cargo fmt --check`: exit 0, no output. `python3 tests/adversarial/test_dav_probe.py`: ```text ..................... ---------------------------------------------------------------------- Ran 21 tests in 0.085s OK KNOWN litmus owner_modify: 403 Forbidden KNOWN litmus complex_cond_put: 400 Bad Request ``` `node --test tests/adversarial/proxy-early-response.test.mjs`: ```text TAP version 13 # Subtest: proxy preserves 413 for 200000 bytes (Continue first: false) ok 1 - proxy preserves 413 for 200000 bytes (Continue first: false) --- duration_ms: 418.679023 type: 'test' ... # Subtest: proxy preserves 413 for 200000 bytes (Continue first: true) ok 2 - proxy preserves 413 for 200000 bytes (Continue first: true) --- duration_ms: 650.222228 type: 'test' ... # Subtest: proxy preserves 413 for 5242880 bytes (Continue first: false) ok 3 - proxy preserves 413 for 5242880 bytes (Continue first: false) --- duration_ms: 628.854439 type: 'test' ... # Subtest: proxy preserves 413 for 5242880 bytes (Continue first: true) ok 4 - proxy preserves 413 for 5242880 bytes (Continue first: true) --- duration_ms: 518.43946 type: 'test' ... # Subtest: proxy preserves 413 for 8388609 bytes (Continue first: false) ok 5 - proxy preserves 413 for 8388609 bytes (Continue first: false) --- duration_ms: 719.498268 type: 'test' ... # Subtest: proxy preserves 413 for 8388609 bytes (Continue first: true) ok 6 - proxy preserves 413 for 8388609 bytes (Continue first: true) --- duration_ms: 638.807761 type: 'test' ... 1..6 # tests 6 # suites 0 # pass 6 # fail 0 # cancelled 0 # skipped 0 # todo 0 # duration_ms 3812.729504 ``` `python3 -m py_compile tests/adversarial/attack.py tests/adversarial/dav_probe_contracts.py tests/adversarial/test_dav_probe.py`: exit 0, no output. `git diff --check`: exit 0, no output. `cargo clean`: ```text Removed 1 file, 356B total ``` Rust clippy/test and web check/Vitest were not run: no Rust crate or web source changed. No web build output was produced. No dependencies changed; existing test dependencies were installed with `bun install --frozen-lockfile`. Comments in all four changed files were re-read. Decisions: no product decisions. Retain documented target deduplication (#421), read-only Daily GET with explicit query-based POST (#752), and revision-bearing reminder Remove (#768). Keep the original server findings open; passing synthetic transport checks do not prove them fixed. UX gaps closed/left: not applicable; no UI changes. For the merge round: after reconciling dev and migration numbers, run `tests/adversarial/run.sh` once with the matching server and CLI builds. It must prove distinct-file attachment persistence, Daily POST/GET parity, reminder Undo revision handling, backend 413 with process survival, bounded reminder mutations, and terminal invalid-audio failure without model work. Retain backend lifecycle logs and content-free write-stage/socket diagnostics. Full Rust/web gates belong to that combined round. No performance measurement ran: this job did not change a user-facing hot path and is not a performance issue.

kayg commented

2026-10-04 08:37:42 +00:00

Author

Owner

Commit 13a646590 fixes a probe error: round-3 queried the JPEG on 1904-01-01 because xuser_matrix.py used tiles[0], the paired video, to choose its day. The probe now uses the requested JPEG's stable Files item ID. Offline regression: Ran 8 tests in 0.091s; OK. The positive Share/revoke contract is running against the local merge-round binary.

Fresh-Instance live evidence at 0c4fd513e: burst 120 pages 18; Search found unicodenfcsentinel after 2 requests and unicodenfdsentinel after 1 request. The broad Calendar probe stops after 200 pages of 7 (1,400 items), but Search setup first adds 20,000 files. A capped scan cannot prove the burst is absent. The far-future cursor has the retired three-field shape; current cursors include the source field. DESIGN §39 requires thumbnail kind parameters; the reported text-card URL is for a document, not evidence of a Photo renderer mismatch. The CSP self-test reports an expected blocked inline script and the log explicitly reports HOSTILE BYTES FINDINGS 0.

The isolated valid PDF publishes a preview. An equal-byte text card does not publish in the same live run. I am tracing that worker result. Production-dev comparison is still in progress; no final classification is claimed here.

Commit 13a646590 fixes a probe error: round-3 queried the JPEG on 1904-01-01 because xuser_matrix.py used tiles[0], the paired video, to choose its day. The probe now uses the requested JPEG's stable Files item ID. Offline regression: Ran 8 tests in 0.091s; OK. The positive Share/revoke contract is running against the local merge-round binary. Fresh-Instance live evidence at 0c4fd513e: burst 120 pages 18; Search found unicodenfcsentinel after 2 requests and unicodenfdsentinel after 1 request. The broad Calendar probe stops after 200 pages of 7 (1,400 items), but Search setup first adds 20,000 files. A capped scan cannot prove the burst is absent. The far-future cursor has the retired three-field shape; current cursors include the source field. DESIGN §39 requires thumbnail kind parameters; the reported text-card URL is for a document, not evidence of a Photo renderer mismatch. The CSP self-test reports an expected blocked inline script and the log explicitly reports HOSTILE BYTES FINDINGS 0. The isolated valid PDF publishes a preview. An equal-byte text card does not publish in the same live run. I am tracing that worker result. Production-dev comparison is still in progress; no final classification is claimed here.

kayg commented

2026-10-04 08:38:01 +00:00

Author

Owner

Finding: the latest shell E2E log (artifacts/round3-full-e2e-shell-final.log, 2026-10-04 09:36 CEST) reaches the Admin configuration page and then times out at apps/web/e2e/shell.mjs:2066 because it expects the H1 “Server configuration”. DESIGN §50 S8 names the Settings page “Configuration”; “Server configuration” remains the card title. This is a stale E2E assertion caused by the Settings redesign, not a product failure. I will update the H1 assertion and keep the card-title assertion.

kayg referenced this issue from a commit

test(journal): check file identity across attachment retries (#867)

kayg referenced this issue from a commit

test(notes): create the requested Daily date through its query contract (#867)

kayg referenced this issue from a commit

test(reminders): require the revision returned for safe Undo (#867)

kayg referenced this issue from a commit

test(proxy): cover streamed early rejections across buffering boundary (#867)

kayg referenced this issue from a commit