S1 web: port the approved onboarding and sign-in into apps/tracer #32

Closed
opened 2026-10-06 15:42:34 +00:00 by kayg · 53 comments
Owner

Port the approved onboarding and sign-in mockup (branch mock-core, mockups/core/onboarding.*) into apps/tracer, after the auth review lands.

Includes:

  • the wizard stepper, the faulty-terminal backdrop, the glitch welcome (centred), count-up loading, and the tick primitive (D99);
  • action feedback: icon → tick swap for copy and download (D103), with icon-only buttons;
  • warm tooltips on every icon-only control;
  • the font rules (preload, no swap) and the monospace UI;
  • username and display name on the server and client (DECISIONS 97–98).

UI primitives go into packages/ui (reuse rule). Review from screenshots of a production build on a phone.

Port the approved onboarding and sign-in mockup (branch mock-core, `mockups/core/onboarding.*`) into `apps/tracer`, after the auth review lands. Includes: - the wizard stepper, the faulty-terminal backdrop, the glitch welcome (centred), count-up loading, and the tick primitive (D99); - action feedback: icon → tick swap for copy and download (D103), with icon-only buttons; - warm tooltips on every icon-only control; - the font rules (preload, no swap) and the monospace UI; - username and display name on the server and client (DECISIONS 97–98). UI primitives go into `packages/ui` (reuse rule). Review from screenshots of a production build on a phone.
Author
Owner

Starting auth onboarding port on t-web at 639ac27684 and t-server at 7948ff00fb. Both worktrees are clean; #32 has no comments. Read parent #21 history and the approved mockup modules; current DECISIONS.md ends at 86, so D96–103 are carried by those owner-approved module comments and #21 history. No edits yet.

Starting auth onboarding port on t-web at 639ac276841832b76828ca09abedf98cca663296 and t-server at 7948ff00fb6370e0a257a14b9fb00f3d6fc3cf7a. Both worktrees are clean; #32 has no comments. Read parent #21 history and the approved mockup modules; current DECISIONS.md ends at 86, so D96–103 are carried by those owner-approved module comments and #21 history. No edits yet.
Author
Owner

Contract proposal before edits: keep setup-start legacy accepted; add with username unique, display name ≤64 Unicode characters defaulting to username, reject ambiguous mixed forms. Store both account fields, use username/display_name for WebAuthn user.name/displayName, and add both to authenticated session (public status remains only). This preserves existing setup/status callers while making the new enrollment path explicit. This follows the job's stricter username rule over #21's older regex note.

Contract proposal before edits: keep setup-start legacy accepted; add with username unique, display name ≤64 Unicode characters defaulting to username, reject ambiguous mixed forms. Store both account fields, use username/display_name for WebAuthn user.name/displayName, and add both to authenticated session (public status remains only). This preserves existing setup/status callers while making the new enrollment path explicit. This follows the job's stricter username rule over #21's older regex note.
Author
Owner

Proposal clarification: preserve legacy setup-start with token and name. Add setup-start with token, username, and optional display_name. Require username to match [a-z0-9_-]{2,32} and be unique; cap display_name at 64 Unicode characters and default it to username. Reject requests mixing legacy and new fields. Store both account fields, pass username/display_name as WebAuthn user.name/displayName, and add both fields to the authenticated session; public status stays setup_required only. This preserves existing callers and follows the job rule over #21 older regex note.

Proposal clarification: preserve legacy setup-start with token and name. Add setup-start with token, username, and optional display_name. Require username to match [a-z0-9_-]{2,32} and be unique; cap display_name at 64 Unicode characters and default it to username. Reject requests mixing legacy and new fields. Store both account fields, pass username/display_name as WebAuthn user.name/displayName, and add both fields to the authenticated session; public status stays setup_required only. This preserves existing callers and follows the job rule over #21 older regex note.
Author
Owner

Findings before implementation: packages/ui is absent (ls packages shows only protocol), so I will create it for shared auth primitives. docs/PROCESS.md is absent from t-web (sed returns no such file), so I will follow the supplied Definition of Done text directly. D108 is still conditional/blocked on #30; the UI will use the current two-stage PRF unwrap contract.

Findings before implementation: packages/ui is absent (ls packages shows only protocol), so I will create it for shared auth primitives. docs/PROCESS.md is absent from t-web (sed returns no such file), so I will follow the supplied Definition of Done text directly. D108 is still conditional/blocked on #30; the UI will use the current two-stage PRF unwrap contract.
Author
Owner

Server regression test added first and reproduced the gap: account_identity_schema_has_unique_usernames_and_display_names fails because users has no username column (crates/infraternal-server/src/auth/tests.rs:1896). The failure is isolated to schema coverage; I am adding the migration next.

Server regression test added first and reproduced the gap: account_identity_schema_has_unique_usernames_and_display_names fails because users has no username column (crates/infraternal-server/src/auth/tests.rs:1896). The failure is isolated to schema coverage; I am adding the migration next.
Author
Owner

Server schema step committed as 95fd00a. Migration 0006 adds username/display_name, backfills four legacy-name cases, and creates a unique username index. Evidence: infraternal-server library tests 24 passed, 0 failed, 1 ignored; focused migration backfill test passed. Existing name/setup callers remain accepted pending the next auth API step.

Server schema step committed as 95fd00a. Migration 0006 adds username/display_name, backfills four legacy-name cases, and creates a unique username index. Evidence: infraternal-server library tests 24 passed, 0 failed, 1 ignored; focused migration backfill test passed. Existing name/setup callers remain accepted pending the next auth API step.
Author
Owner

Auth API regression test is red before implementation: setup_start_accepts_username_and_display_name gets HTTP 400 for the proposed additive request (crates/infraternal-server/src/auth/tests.rs:1893). The current Setup deserializer only accepts token/name; I am extending it while retaining that form.

Auth API regression test is red before implementation: setup_start_accepts_username_and_display_name gets HTTP 400 for the proposed additive request (crates/infraternal-server/src/auth/tests.rs:1893). The current Setup deserializer only accepts token/name; I am extending it while retaining that form.
Author
Owner

Server auth step committed as 2884249. setup-start now accepts both the additive username/display_name body and legacy name; registration persists the fields, WebAuthn labels use each field, duplicate usernames map to 409, and authenticated session returns both names. Evidence: infraternal-server library tests 26 passed, 0 failed, 1 ignored; production HTTP setup and session-response test passed.

Server auth step committed as 2884249. setup-start now accepts both the additive username/display_name body and legacy name; registration persists the fields, WebAuthn labels use each field, duplicate usernames map to 409, and authenticated session returns both names. Evidence: infraternal-server library tests 26 passed, 0 failed, 1 ignored; production HTTP setup and session-response test passed.
Author
Owner

Server auth contract committed as a74eae5. contracts/keywrap.md now documents 2–32 ASCII usernames, 64 Unicode-scalar display labels, the legacy setup body, authenticated session fields, migration backfill, and 409 duplicate handling. Public status remains setup_required only.

Server auth contract committed as a74eae5. contracts/keywrap.md now documents 2–32 ASCII usernames, 64 Unicode-scalar display labels, the legacy setup body, authenticated session fields, migration backfill, and 409 duplicate handling. Public status remains setup_required only.
Author
Owner

Finding: the first browser run recorded non-input layout shifts during step 1 → 2 (5 entries; largest 0.0818). The approved height animation changed the centered card footprint in Layout Instability. Reserving the maximum measured step pane before first paint so the card and footer stay fixed while step contents animate.

Finding: the first browser run recorded non-input layout shifts during step 1 → 2 (5 entries; largest 0.0818). The approved height animation changed the centered card footprint in Layout Instability. Reserving the maximum measured step pane before first paint so the card and footer stay fixed while step contents animate.
Author
Owner

Client shared-primitives step: added packages/ui at 4f140e4b37 (seven approved browser modules plus shared styles, AGPL-3.0-only). Tracer production build: built main.js, chunk-sxzajhbf.js, chunk-k5nvdjvh.js.

Client shared-primitives step: added packages/ui at 4f140e4b370aa25c72dfe49acafc320dd7cf98b0 (seven approved browser modules plus shared styles, AGPL-3.0-only). Tracer production build: built main.js, chunk-sxzajhbf.js, chunk-k5nvdjvh.js.
Author
Owner

Performance finding: on the same Linux 6.12 x86_64, 8-vCPU host, 30 fresh production pages measured the supplied client at setup p50/p99 105.4/148 ms and sign-in 101.9/135.3 ms; the new flow measured 247.2/504.1 ms and 185.7/223.6 ms. The screenshot upper bound included deferred WebGL shader compilation (load averages were about 1.7–1.8 / 3.1–3.4 / 3.0–3.1). I am moving that decorative work behind a short post-paint delay and will rerun the comparison.

Performance finding: on the same Linux 6.12 x86_64, 8-vCPU host, 30 fresh production pages measured the supplied client at setup p50/p99 105.4/148 ms and sign-in 101.9/135.3 ms; the new flow measured 247.2/504.1 ms and 185.7/223.6 ms. The screenshot upper bound included deferred WebGL shader compilation (load averages were about 1.7–1.8 / 3.1–3.4 / 3.0–3.1). I am moving that decorative work behind a short post-paint delay and will rerun the comparison.
Author
Owner

Finding: the new desktop anchor check failed after the production Playwright run: #setup-card center moved 127 px between setup steps (apps/tracer/test/onboarding.spec.ts:115). Non-input LayoutShift entries remained empty, so CLS alone missed the card movement. Investigating per-step bounds before changing the reservation.

Finding: the new desktop anchor check failed after the production Playwright run: #setup-card center moved 127 px between setup steps (apps/tracer/test/onboarding.spec.ts:115). Non-input LayoutShift entries remained empty, so CLS alone missed the card movement. Investigating per-step bounds before changing the reservation.
Author
Owner

Finding: clipboard rejection still displayed the copied tick. The production browser regression test failed against the previous helper at apps/tracer/test/onboarding.spec.ts:81 (data-done was present immediately after rejected writeText). Updating the shared primitive to confirm only after writeText resolves.

Finding: clipboard rejection still displayed the copied tick. The production browser regression test failed against the previous helper at apps/tracer/test/onboarding.spec.ts:81 (data-done was present immediately after rejected writeText). Updating the shared primitive to confirm only after writeText resolves.
Author
Owner

Client core step: 4844069 ports setup identity, passkey, recovery acknowledgement, welcome, discoverable sign-in, shared motion modules and delayed canvas lifecycle. TypeScript check and production build pass; the current dist also passes the real Rust-router setup/login/PRF check. Final browser suite then reported first-paint p50/p99 66.8/279.5 ms setup and 66.2/93.7 ms sign-in (30 fresh surfaces, Linux 6.12.111 x86_64, 8 vCPUs; post-run load 3.56/1.94/1.99).

Client core step: 4844069 ports setup identity, passkey, recovery acknowledgement, welcome, discoverable sign-in, shared motion modules and delayed canvas lifecycle. TypeScript check and production build pass; the current dist also passes the real Rust-router setup/login/PRF check. Final browser suite then reported first-paint p50/p99 66.8/279.5 ms setup and 66.2/93.7 ms sign-in (30 fresh surfaces, Linux 6.12.111 x86_64, 8 vCPUs; post-run load 3.56/1.94/1.99).
Author
Owner

Client browser evidence step: 828c592 passes all four Playwright tests. The onboarding test covers the 400 ms count-up threshold, copy rejection/no tick, successful copy/download tick, welcome center ≤1 px, stable reserved card height, no non-input layout shifts, terminal handoff and canvas removal. Terminal probe: 500/500 keys, p50/p99 15.1/16.7 ms; resume 79.5 ms. Twelve production-build dark screenshots (393×852 and 1440×900) are in /home/kayg/Developer/infraternal/.worktrees/t-web/apps/tracer/test-results/onboarding-setup-recovery--906cd--match-the-browser-contract/: step-1-token-iphone-15.png, step-1-token-desktop.png, step-2-identity-iphone-15.png, step-2-identity-desktop.png, step-3-passkey-iphone-15.png, step-3-passkey-desktop.png, step-4-recovery-iphone-15.png, step-4-recovery-desktop.png, welcome-iphone-15.png, welcome-desktop.png, returning-sign-in-iphone-15.png, returning-sign-in-desktop.png.

Client browser evidence step: 828c592 passes all four Playwright tests. The onboarding test covers the 400 ms count-up threshold, copy rejection/no tick, successful copy/download tick, welcome center ≤1 px, stable reserved card height, no non-input layout shifts, terminal handoff and canvas removal. Terminal probe: 500/500 keys, p50/p99 15.1/16.7 ms; resume 79.5 ms. Twelve production-build dark screenshots (393×852 and 1440×900) are in /home/kayg/Developer/infraternal/.worktrees/t-web/apps/tracer/test-results/onboarding-setup-recovery--906cd--match-the-browser-contract/: step-1-token-iphone-15.png, step-1-token-desktop.png, step-2-identity-iphone-15.png, step-2-identity-desktop.png, step-3-passkey-iphone-15.png, step-3-passkey-desktop.png, step-4-recovery-iphone-15.png, step-4-recovery-desktop.png, welcome-iphone-15.png, welcome-desktop.png, returning-sign-in-iphone-15.png, returning-sign-in-desktop.png.
Author
Owner

Server browser-integration step: cab5c55 updates the paired production browser check for username/display-name enrollment, recovery confirmation and authenticated session labels. Re-ran against t-web/apps/tracer/dist with the scratch-built Rust fixture: PASS (resident PRF setup, empty-storage discoverable login, DEK unwrap; Chromium 149.0.7827.55). The browser test verifies username=production_integration and display_name=Production integration in /auth/session.

Server browser-integration step: cab5c55 updates the paired production browser check for username/display-name enrollment, recovery confirmation and authenticated session labels. Re-ran against t-web/apps/tracer/dist with the scratch-built Rust fixture: PASS (resident PRF setup, empty-storage discoverable login, DEK unwrap; Chromium 149.0.7827.55). The browser test verifies username=production_integration and display_name=Production integration in /auth/session.
Author
Owner

Finding: the required t-server fast gate stopped clippy on crates/infraternal-server/src/auth/http.rs::execute (109/100 lines) and src/auth/webauthn.rs::registration_finish (104/100 lines). The auth additions pushed both over the configured pedantic limit; splitting the added branches/helpers before retrying the gate.

Finding: the required t-server fast gate stopped clippy on crates/infraternal-server/src/auth/http.rs::execute (109/100 lines) and src/auth/webauthn.rs::registration_finish (104/100 lines). The auth additions pushed both over the configured pedantic limit; splitting the added branches/helpers before retrying the gate.
Author
Owner

Server lint follow-up: 6e5d956 extracts the legacy/additive setup wire dispatch and proof/envelope validation before the existing writer-locked transaction. cargo fmt --all --check and cargo clippy --workspace --all-targets --locked -- -D warnings both pass. The disabled lint approach was rejected; uniqueness remains checked under the SQLite writer lock.

Server lint follow-up: 6e5d956 extracts the legacy/additive setup wire dispatch and proof/envelope validation before the existing writer-locked transaction. cargo fmt --all --check and cargo clippy --workspace --all-targets --locked -- -D warnings both pass. The disabled lint approach was rejected; uniqueness remains checked under the SQLite writer lock.
Author
Owner

Final t-server scripts/verify fast at 6e5d956 passed. Rust summary output:\n+---------------------------+--------------+\n| Rust gate | Result |\n+---------------------------+--------------+\n| fmt | PASS |\n| clippy | PASS |\n| test | PASS |\n| cargo-deny | PASS |\n| static builds | PASS |\n+---------------------------+--------------+\n\nThe TypeScript protocol compiler passed, the Rust stream oracle built, and bun test reported 7 pass, 0 fail, 34,390 expectations across 2 files.

Final t-server scripts/verify fast at 6e5d956 passed. Rust summary output:\n+---------------------------+--------------+\n| Rust gate | Result |\n+---------------------------+--------------+\n| fmt | PASS |\n| clippy | PASS |\n| test | PASS |\n| cargo-deny | PASS |\n| static builds | PASS |\n+---------------------------+--------------+\n\nThe TypeScript protocol compiler passed, the Rust stream oracle built, and bun test reported 7 pass, 0 fail, 34,390 expectations across 2 files.
Author
Owner

Correction to the preceding gate note: its line breaks were escaped. Here is the exact t-server Rust summary from scripts/verify fast at 6e5d956:

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

The TypeScript protocol compiler passed, the Rust stream oracle built, and bun test reported 7 pass, 0 fail, 34,390 expectations across 2 files.

Correction to the preceding gate note: its line breaks were escaped. Here is the exact t-server Rust summary from scripts/verify fast at 6e5d956: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ The TypeScript protocol compiler passed, the Rust stream oracle built, and bun test reported 7 pass, 0 fail, 34,390 expectations across 2 files.
Author
Owner

Finished issue #32. Final heads:

The required scripts/verify fast command passed in both worktrees. Verbatim t-web Rust summary:
+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

Verbatim t-server Rust summary:
+---------------------------+--------------+
| Rust gate | Result |
+---------------------------+--------------+
| fmt | PASS |
| clippy | PASS |
| test | PASS |
| cargo-deny | PASS |
| static builds | PASS |
+---------------------------+--------------+

Both TypeScript fast subgates passed: 7 pass, 0 fail, 34,390 expectations across 2 files; differential coverage was 134 streams / 11,439 schedules with zero divergences. Tracer Playwright: 4 passed. Final real Rust-router browser check: PASS (resident PRF enrollment, recovery confirmation, discoverable login and DEK unwrap; Chromium 149.0.7827.55). The 12 dark viewport screenshots and their paths are recorded in the prior evidence comment.

Performance follow-up on Linux 6.12.111 x86_64, 8 vCPUs, post-run load 3.56/1.94/1.99: 30 fresh surfaces measured setup FCP p50/p99 66.8/279.5 ms and sign-in 66.2/93.7 ms. Against the 639ac27 baseline, FCP medians rose 9.7/10.2 ms; setup p99 rose 191.7 ms. Screenshot upper bounds were 151.1/367.2 ms setup and 123/162 ms sign-in, below the existing 1,000 ms p99 gate. These are local desktop measurements, not phone, network or concurrent-user results.

Finished issue #32. Final heads: - t-web: 75ffa9376e902b819d40d81a3c09a290da24d44f - t-server: 6e5d9563f349f7e399d0b4e4618a20a2cfae3333 The required scripts/verify fast command passed in both worktrees. Verbatim t-web Rust summary: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ Verbatim t-server Rust summary: +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ Both TypeScript fast subgates passed: 7 pass, 0 fail, 34,390 expectations across 2 files; differential coverage was 134 streams / 11,439 schedules with zero divergences. Tracer Playwright: 4 passed. Final real Rust-router browser check: PASS (resident PRF enrollment, recovery confirmation, discoverable login and DEK unwrap; Chromium 149.0.7827.55). The 12 dark viewport screenshots and their paths are recorded in the prior evidence comment. Performance follow-up on Linux 6.12.111 x86_64, 8 vCPUs, post-run load 3.56/1.94/1.99: 30 fresh surfaces measured setup FCP p50/p99 66.8/279.5 ms and sign-in 66.2/93.7 ms. Against the 639ac27 baseline, FCP medians rose 9.7/10.2 ms; setup p99 rose 191.7 ms. Screenshot upper bounds were 151.1/367.2 ms setup and 123/162 ms sign-in, below the existing 1,000 ms p99 gate. These are local desktop measurements, not phone, network or concurrent-user results.
Author
Owner

Starting the two setup-screen bug fixes after reading #32 and parent #21 history. Fast-forwarded t-web (795acbba) and t-server (3721631) to deployed local dev base 0c47cd42e6. Both worktrees are clean; no fix edits yet.

Starting the two setup-screen bug fixes after reading #32 and parent #21 history. Fast-forwarded t-web (795acbba) and t-server (3721631) to deployed local dev base 0c47cd42e6afeec7ca9d6d430bc11b319eb677c8. Both worktrees are clean; no fix edits yet.
Author
Owner

Finding: the server exposes setup-start but no setup-check action. The new real HTTP route test is red on t-server at crates/infraternal-server/src/auth/tests.rs:2041: POST /auth/setup-check with a wrong token returns 400 (unknown action), expected 401. The test also covers valid 204 without consuming enrollment and 404 after owner creation.

Finding: the server exposes setup-start but no setup-check action. The new real HTTP route test is red on t-server at crates/infraternal-server/src/auth/tests.rs:2041: POST /auth/setup-check with a wrong token returns 400 (unknown action), expected 401. The test also covers valid 204 without consuming enrollment and 404 after owner creation.
Author
Owner

Server endpoint step committed as 87f11da. The new real-router test passes (1 passed, 0 failed): wrong token → 401; valid token → 204 with no Set-Cookie, then real software-authenticator enrollment succeeds; after owner creation → 404. The initial red run returned 400 for the missing action.

Server endpoint step committed as 87f11da. The new real-router test passes (1 passed, 0 failed): wrong token → 401; valid token → 204 with no Set-Cookie, then real software-authenticator enrollment succeeds; after owner creation → 404. The initial red run returned 400 for the missing action.
Author
Owner

Auth contract step committed as 36b3c92. contracts/keywrap.md now documents additive POST /auth/setup-check, constant-time validation, 204/401/404, inherited admissions and no state/cookie/ceremony changes; existing setup-start and one-shot enrollment remain unchanged. Diff check passed.

Auth contract step committed as 36b3c92. contracts/keywrap.md now documents additive POST /auth/setup-check, constant-time validation, 204/401/404, inherited admissions and no state/cookie/ceremony changes; existing setup-start and one-shot enrollment remain unchanged. Diff check passed.
Author
Owner

Finding: the new production-browser regression fails on deployed-base t-web. apps/tracer/test/onboarding.spec.ts:25 expects the step-1 heading after Continue with an invalid token; it is gone because the UI has advanced. The run used the production build and test double at TRACER_TEST_PORT=4432.

Finding: the new production-browser regression fails on deployed-base t-web. apps/tracer/test/onboarding.spec.ts:25 expects the step-1 heading after Continue with an invalid token; it is gone because the UI has advanced. The run used the production build and test double at TRACER_TEST_PORT=4432.
Author
Owner

Finding: the production-browser alignment assertion is red at apps/tracer/test/onboarding.spec.ts:70. On the 393×852 viewport the primary button is 167 px from the card content edge (required ≤1 px); the test saved test-results/onboarding-setup-primary-b-f436d-t-edge-on-phone-and-desktop/attachments/step-1-button-iphone-15-51007fd88e55c51e42681be886cd31d380f68dd0.png. Cause: auth.css declares margin-left:auto before its later margin:0 shorthand, and the hidden back button is removed by the global hidden rule, leaving the footer’s sole item at flex start.

Finding: the production-browser alignment assertion is red at apps/tracer/test/onboarding.spec.ts:70. On the 393×852 viewport the primary button is 167 px from the card content edge (required ≤1 px); the test saved test-results/onboarding-setup-primary-b-f436d-t-edge-on-phone-and-desktop/attachments/step-1-button-iphone-15-51007fd88e55c51e42681be886cd31d380f68dd0.png. Cause: auth.css declares margin-left:auto before its later margin:0 shorthand, and the hidden back button is removed by the global hidden rule, leaving the footer’s sole item at flex start.
Author
Owner

The red alignment regression now lives in its own module, apps/tracer/test/setup-layout.spec.ts:26. Re-running the deployed-base production build at 393×852 again measured a 167 px content-edge gap (limit 1 px); failure artifact: test-results/setup-layout-setup-primary-a4143-t-edge-on-phone-and-desktop/attachments/step-1-button-iphone-15-579aac54113a17d2486c32038382b3616ce2e632.png.

The red alignment regression now lives in its own module, apps/tracer/test/setup-layout.spec.ts:26. Re-running the deployed-base production build at 393×852 again measured a 167 px content-edge gap (limit 1 px); failure artifact: test-results/setup-layout-setup-primary-a4143-t-edge-on-phone-and-desktop/attachments/step-1-button-iphone-15-579aac54113a17d2486c32038382b3616ce2e632.png.
Author
Owner

Client token-validation step committed as 7fe608f. TypeScript check passed; production-build Playwright passed (1/1) at port 4434. Wrong token stayed on step 1, count-up was hidden at 180 ms and visible after 480 ms, inline guidance appeared and focus returned to the token field; a correct token advanced to step 2. Screenshots: apps/tracer/test-results/onboarding-invalid-setup-t-ae0f0-tep-1-with-focused-guidance/step-1-wrong-token-iphone-15.png and step-1-wrong-token-desktop.png.

Client token-validation step committed as 7fe608f. TypeScript check passed; production-build Playwright passed (1/1) at port 4434. Wrong token stayed on step 1, count-up was hidden at 180 ms and visible after 480 ms, inline guidance appeared and focus returned to the token field; a correct token advanced to step 2. Screenshots: apps/tracer/test-results/onboarding-invalid-setup-t-ae0f0-tep-1-with-focused-guidance/step-1-wrong-token-iphone-15.png and step-1-wrong-token-desktop.png.
Author
Owner

Button alignment step committed as 1b737d7. The production-browser assertion passes at 393×852 and 1440×900: measured right inset is 0.000 px at each viewport (limit ≤1 px), 1 test passed. Screenshots: apps/tracer/test-results/setup-layout-setup-primary-a4143-t-edge-on-phone-and-desktop/step-1-button-iphone-15.png and step-1-button-desktop.png.

Button alignment step committed as 1b737d7. The production-browser assertion passes at 393×852 and 1440×900: measured right inset is 0.000 px at each viewport (limit ≤1 px), 1 test passed. Screenshots: apps/tracer/test-results/setup-layout-setup-primary-a4143-t-edge-on-phone-and-desktop/step-1-button-iphone-15.png and step-1-button-desktop.png.
Author
Owner

Gate finding: first t-server scripts/verify fast stopped clippy on infraternal-server/src/auth/setup.rs:44 (pedantic items-after-statements) because the dummy digest const followed executable statements. The focused route test passed; I am moving the constant to module scope before rerunning the gate.

Gate finding: first t-server scripts/verify fast stopped clippy on infraternal-server/src/auth/setup.rs:44 (pedantic items-after-statements) because the dummy digest const followed executable statements. The focused route test passed; I am moving the constant to module scope before rerunning the gate.
Author
Owner

Clippy follow-up committed as 7918537. The dummy digest is at module scope; cargo fmt --all --check and cargo clippy --locked -p infraternal-server --all-targets -- -D warnings both pass. The full t-server gate will be rerun before completion.

Clippy follow-up committed as 7918537. The dummy digest is at module scope; cargo fmt --all --check and cargo clippy --locked -p infraternal-server --all-targets -- -D warnings both pass. The full t-server gate will be rerun before completion.
Author
Owner

Server gate finding at 7918537: the second full scripts/verify fast run reached 60 passed, 1 failed, 4 ignored. Unrelated timing-sensitive test review3_backend_reply_has_a_deadline failed at crates/infraternal-server/src/auth/tests.rs:1503 with 'server closed websocket unexpectedly'; it passed in the earlier full run. Clippy and fmt passed; investigating the isolated test while static builds finish.

Server gate finding at 7918537: the second full scripts/verify fast run reached 60 passed, 1 failed, 4 ignored. Unrelated timing-sensitive test review3_backend_reply_has_a_deadline failed at crates/infraternal-server/src/auth/tests.rs:1503 with 'server closed websocket unexpectedly'; it passed in the earlier full run. Clippy and fmt passed; investigating the isolated test while static builds finish.
Author
Owner

Follow-up: the deadline case passed isolated (1/1, 60.45s): failure response at 60003ms, late backend success ignored, next Ping succeeded, terminal rows=0. This supports a parallel-suite timing interaction; rerunning the full server gate.

Follow-up: the deadline case passed isolated (1/1, 60.45s): failure response at 60003ms, late backend success ignored, next Ping succeeded, terminal rows=0. This supports a parallel-suite timing interaction; rerunning the full server gate.
Author
Owner

t-server gate passed at 7918537. Verbatim summary:\n| Rust gate | Result |\n|---------------------------|--------------|\n| fmt | PASS |\n| clippy | PASS |\n| test | PASS |\n| cargo-deny | PASS |\n| static builds | PASS |\nThe server library suite reported: test result: 61 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 151.81s. Static artifacts: x86_64 15071488 bytes, aarch64 12636216 bytes, PT_INTERP absent for both. TypeScript protocol tests: 7 pass, 0 fail; differential run: 134 streams, 11439 schedules, zero divergences.

t-server gate passed at 7918537. Verbatim summary:\n| Rust gate | Result |\n|---------------------------|--------------|\n| fmt | PASS |\n| clippy | PASS |\n| test | PASS |\n| cargo-deny | PASS |\n| static builds | PASS |\nThe server library suite reported: test result: 61 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 151.81s. Static artifacts: x86_64 15071488 bytes, aarch64 12636216 bytes, PT_INTERP absent for both. TypeScript protocol tests: 7 pass, 0 fail; differential run: 134 streams, 11439 schedules, zero divergences.
Author
Owner

t-web gate passed at 1b737d7. Verbatim summary:\n| Rust gate | Result |\n|---------------------------|--------------|\n| fmt | PASS |\n| clippy | PASS |\n| test | PASS |\n| cargo-deny | PASS |\n| static builds | PASS |\nServer library: test result: 60 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 163.75s. Static artifacts: x86_64 15068128 bytes, aarch64 12632736 bytes, PT_INTERP absent for both. TypeScript protocol tests: 7 pass, 0 fail; differential run: 134 streams, 11439 schedules, zero divergences.

t-web gate passed at 1b737d7. Verbatim summary:\n| Rust gate | Result |\n|---------------------------|--------------|\n| fmt | PASS |\n| clippy | PASS |\n| test | PASS |\n| cargo-deny | PASS |\n| static builds | PASS |\nServer library: test result: 60 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 163.75s. Static artifacts: x86_64 15068128 bytes, aarch64 12632736 bytes, PT_INTERP absent for both. TypeScript protocol tests: 7 pass, 0 fail; differential run: 134 streams, 11439 schedules, zero divergences.
Author
Owner

Playwright finding at 1b737d7: required port-4421 full run completed 17/18 tests; terminal.spec.ts:264 still expected invalid setup token to advance to #setup-username and timed out at line 274 (30s). The new step-1 validation correctly prevented advancement. Updating the stale assertion to check the inline error and distinct passkey-cancel guidance, then rerunning the full suite on a fresh port.

Playwright finding at 1b737d7: required port-4421 full run completed 17/18 tests; terminal.spec.ts:264 still expected invalid setup token to advance to #setup-username and timed out at line 274 (30s). The new step-1 validation correctly prevented advancement. Updating the stale assertion to check the inline error and distinct passkey-cancel guidance, then rerunning the full suite on a fresh port.
Author
Owner

Client production-browser evidence at d51337b:

  • Full Playwright suite: 18 passed in 2.1 minutes on TRACER_TEST_PORT=4422.
  • Primary button right-edge inset: iPhone 15 0.000 px; desktop 0.000 px.
  • Terminal probe: 500/500 echoes painted; p50/p99 15.40/16.10 ms; resume 78.10 ms.

Wrong-token error state:

Wrong setup token on iPhone 15
Wrong setup token on desktop

Step-1 primary button alignment:

Setup button on iPhone 15
Setup button on desktop

Client production-browser evidence at d51337b: - Full Playwright suite: 18 passed in 2.1 minutes on TRACER_TEST_PORT=4422. - Primary button right-edge inset: iPhone 15 0.000 px; desktop 0.000 px. - Terminal probe: 500/500 echoes painted; p50/p99 15.40/16.10 ms; resume 78.10 ms. Wrong-token error state: ![Wrong setup token on iPhone 15](https://git.kayg.org/attachments/aa0b84da-9692-4c9f-ada0-11cd17067b65) ![Wrong setup token on desktop](https://git.kayg.org/attachments/a9964a2c-c6f5-41e8-88cc-577d6e0ab384) Step-1 primary button alignment: ![Setup button on iPhone 15](https://git.kayg.org/attachments/9e4eb989-f32f-42ea-b8fd-0d7cc9827408) ![Setup button on desktop](https://git.kayg.org/attachments/6756986b-025e-4e2a-98f7-c026ba027823)
Author
Owner

Commit body formatting correction: the test-only change is now 56c3162 (full SHA available in branch history); code is unchanged from the 18/18 Playwright run. This replaces the temporary d51337b hash in the screenshot evidence comment.

Commit body formatting correction: the test-only change is now 56c3162 (full SHA available in branch history); code is unchanged from the 18/18 Playwright run. This replaces the temporary d51337b hash in the screenshot evidence comment.
Author
Owner

Finished #32. No push or deploy.

Branch heads:

nix shell nixpkgs#binutils nixpkgs#file nixpkgs#gnumake nixpkgs#perl -c scripts/verify fast passed in both worktrees. Verbatim Rust gate summary:

Rust gate Result
fmt PASS
clippy PASS
test PASS
cargo-deny PASS
static builds PASS
  • t-server: test result: ok. 61 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 151.81s; static x86_64 15071488 bytes and aarch64 12636216 bytes, PT_INTERP absent.
  • t-web: test result: ok. 60 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 163.75s; static x86_64 15068128 bytes and aarch64 12632736 bytes, PT_INTERP absent.
  • Both protocol checks: Ghostty checksums OK; stream differential: 134 streams, 11439 schedules, zero divergences; 7 TypeScript protocol tests passed, 0 failed.

The full Playwright run at port 4421 found one stale assertion expecting rejection at the passkey step. After updating that assertion, the complete suite passed at fresh port 4422: 18 passed (2.1m). Step-1 button inset was 0.000 px at both iPhone 15 (393×852) and desktop (1440×900). The wrong-token error and button screenshots are attached in the preceding evidence comment.

Production-browser measurements: Linux 6.12.111 amd64, 8 vCPUs; load average at final review 1.76/4.58/8.68. The 500-key probe painted 500/500 echoes at p50/p99 15.40/16.10 ms (under the 20 ms gate), with 78.10 ms resume. Access render upper-bound p99 was 355.8 ms for setup and 144.2 ms for sign-in (under the 1,000 ms gate). The count-up test kept the indicator hidden at 180 ms and visible after 480 ms.

Reviewer checklist checked: status routing, owner identity and recovery acknowledgement, resident passkey/PRF flow, accessible icon controls, secret handling, and the attached phone/desktop screenshots.

Finished #32. No push or deploy. Branch heads: - t-web: 56c3162b40c41911848c579561e85b2ec2cbab04 - t-server: 79185377b954b441bd095b74dd6eacd29573306e `nix shell nixpkgs#binutils nixpkgs#file nixpkgs#gnumake nixpkgs#perl -c scripts/verify fast` passed in both worktrees. Verbatim Rust gate summary: | Rust gate | Result | |---------------------------|--------------| | fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | - t-server: `test result: ok. 61 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 151.81s`; static x86_64 15071488 bytes and aarch64 12636216 bytes, PT_INTERP absent. - t-web: `test result: ok. 60 passed; 0 failed; 4 ignored; 0 measured; 0 filtered out; finished in 163.75s`; static x86_64 15068128 bytes and aarch64 12632736 bytes, PT_INTERP absent. - Both protocol checks: Ghostty checksums OK; `stream differential: 134 streams, 11439 schedules, zero divergences`; 7 TypeScript protocol tests passed, 0 failed. The full Playwright run at port 4421 found one stale assertion expecting rejection at the passkey step. After updating that assertion, the complete suite passed at fresh port 4422: `18 passed (2.1m)`. Step-1 button inset was 0.000 px at both iPhone 15 (393×852) and desktop (1440×900). The wrong-token error and button screenshots are attached in the preceding evidence comment. Production-browser measurements: Linux 6.12.111 amd64, 8 vCPUs; load average at final review 1.76/4.58/8.68. The 500-key probe painted 500/500 echoes at p50/p99 15.40/16.10 ms (under the 20 ms gate), with 78.10 ms resume. Access render upper-bound p99 was 355.8 ms for setup and 144.2 ms for sign-in (under the 1,000 ms gate). The count-up test kept the indicator hidden at 180 ms and visible after 480 ms. Reviewer checklist checked: status routing, owner identity and recovery acknowledgement, resident passkey/PRF flow, accessible icon controls, secret handling, and the attached phone/desktop screenshots.
Author
Owner

Starting the sign-in cleanup and UI font switch on branch ui-tweaks at base 4daf8c0. The worktree is clean; changes stay in this worktree. I read #32 and parent #21 history.

Starting the sign-in cleanup and UI font switch on branch ui-tweaks at base 4daf8c0. The worktree is clean; changes stay in this worktree. I read #32 and parent #21 history.
Author
Owner

Test setup finding: the focused Playwright command could not start because apps/tracer has no installed playwright package (ERR_MODULE_NOT_FOUND at playwright.config.ts). Installing the checked-in frozen Bun dependencies before rerunning.

Test setup finding: the focused Playwright command could not start because apps/tracer has no installed playwright package (ERR_MODULE_NOT_FOUND at playwright.config.ts). Installing the checked-in frozen Bun dependencies before rerunning.
Author
Owner

Baseline regression reproduced in the production browser: apps/tracer/test/onboarding.spec.ts fails because .auth-signin contains 2 buttons (expected 1); the second is #login-other, “use a different passkey…”. The failing check is at line 182 on port 4531. Removing that secondary path and rebalance styling next.

Baseline regression reproduced in the production browser: apps/tracer/test/onboarding.spec.ts fails because .auth-signin contains 2 buttons (expected 1); the second is #login-other, “use a different passkey…”. The failing check is at line 182 on port 4531. Removing that secondary path and rebalance styling next.
Author
Owner

Fast-gate finding: scripts/verify fast reached the Rust mouse integration test and faithful_snapshot_preserves_public_mouse_encoder_reports failed at crates/infraternal-shim/tests/mouse.rs:18 (public mouse encoder probe failed; see t2-gate-public). The failure is outside the three UI files changed here; static builds are still running.

Fast-gate finding: `scripts/verify fast` reached the Rust mouse integration test and `faithful_snapshot_preserves_public_mouse_encoder_reports` failed at crates/infraternal-shim/tests/mouse.rs:18 (`public mouse encoder probe failed; see t2-gate-public`). The failure is outside the three UI files changed here; static builds are still running.
Author
Owner

Follow-up on the fast-gate failure: the native mouse test searches the workspace path target/debug/build for libghostty-vt.a, while this run correctly put Cargo output in the job cache. Its diagnostic is ValueError: no matching native archive. I will expose the cached target at the expected ignored target path for the retry; no source change is implicated.

Follow-up on the fast-gate failure: the native mouse test searches the workspace path target/debug/build for libghostty-vt.a, while this run correctly put Cargo output in the job cache. Its diagnostic is ValueError: no matching native archive. I will expose the cached target at the expected ignored target path for the retry; no source change is implicated.
Author
Owner

Decision record proposal per owner request: update D98 so sign-in has one usernameless “Continue with passkey” button; its discoverable platform chooser lists every passkey for this site, so the separate “use a different passkey…” link is unnecessary. This supersedes D98’s earlier secondary-action sentence. I will record the owner change with today’s date.

Decision record proposal per owner request: update D98 so sign-in has one usernameless “Continue with passkey” button; its discoverable platform chooser lists every passkey for this site, so the separate “use a different passkey…” link is unnecessary. This supersedes D98’s earlier secondary-action sentence. I will record the owner change with today’s date.
Author
Owner

Decision documentation committed: d2ce17a updates D98 per the owner request. Discoverable platform chooser lists every site passkey; the secondary link is removed from the decision. Fast gate has passed for the current code/test tree.

Decision documentation committed: d2ce17a updates D98 per the owner request. Discoverable platform chooser lists every site passkey; the secondary link is removed from the decision. Fast gate has passed for the current code/test tree.
Author
Owner

Completed the sign-in source step: 84a3733 removes the second passkey action and its duplicate ceremony state in apps/tracer/src/onboarding.ts. Fast gate exited 0; Rust server 66 passed, shim 61 passed, protocol 7 passed, static builds and cargo-deny passed.

Completed the sign-in source step: 84a3733 removes the second passkey action and its duplicate ceremony state in apps/tracer/src/onboarding.ts. Fast gate exited 0; Rust server 66 passed, shim 61 passed, protocol 7 passed, static builds and cargo-deny passed.
Author
Owner

Completed the sign-in presentation step: 575eaec removes the redundant link styling and rebalances the card at desktop and phone breakpoints. The fast gate above ran against this tree and exited 0.

Completed the sign-in presentation step: 575eaec removes the redundant link styling and rebalances the card at desktop and phone breakpoints. The fast gate above ran against this tree and exited 0.
Author
Owner

Completed the sign-in test step: 4892902 asserts the four required welcome-card elements, exactly one button, and no #login-other. The focused production test passed earlier (1 passed, 15.4 s); the fast gate also passed on this tree.

Completed the sign-in test step: 4892902 asserts the four required welcome-card elements, exactly one button, and no #login-other. The focused production test passed earlier (1 passed, 15.4 s); the fast gate also passed on this tree.
Author
Owner

Finished on branch ui-tweaks at ff505cb. D98 now matches the discoverable passkey flow; sign-in shows the mark, welcome, lede, and one passkey button. The full production Playwright suite passed (23 passed, 3.2m). No merge or deploy was performed.

Final scripts/verify fast output (verbatim summary):

test result: ok. 66 passed; 0 failed; 5 ignored; 0 measured; 0 filtered out; finished in 152.55s
test result: ok. 61 passed; 0 failed; 12 ignored; 0 measured; 0 filtered out; finished in 35.26s
advisories ok, bans ok, licenses ok, sources ok
PT_INTERP: absent; size: 16096816 bytes
PT_INTERP: absent; size: 13515080 bytes
server subcommand: exit 2, not implemented
agent subcommand: exit 2, not implemented

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

dist/ghostty-web.js: OK
dist/ghostty-vt.wasm: OK
dist/index.d.ts: OK

 7 pass
 0 fail
 34390 expect() calls
Ran 7 tests across 2 files. [2.31s]
Finished on branch ui-tweaks at ff505cb. D98 now matches the discoverable passkey flow; sign-in shows the mark, welcome, lede, and one passkey button. The full production Playwright suite passed (23 passed, 3.2m). No merge or deploy was performed. Final `scripts/verify fast` output (verbatim summary): ```text test result: ok. 66 passed; 0 failed; 5 ignored; 0 measured; 0 filtered out; finished in 152.55s test result: ok. 61 passed; 0 failed; 12 ignored; 0 measured; 0 filtered out; finished in 35.26s advisories ok, bans ok, licenses ok, sources ok PT_INTERP: absent; size: 16096816 bytes PT_INTERP: absent; size: 13515080 bytes server subcommand: exit 2, not implemented agent subcommand: exit 2, not implemented +---------------------------+--------------+ | Rust gate | Result | +---------------------------+--------------+ | fmt | PASS | | VT native fmt | PASS | | clippy | PASS | | test | PASS | | cargo-deny | PASS | | static builds | PASS | +---------------------------+--------------+ dist/ghostty-web.js: OK dist/ghostty-vt.wasm: OK dist/index.d.ts: OK 7 pass 0 fail 34390 expect() calls Ran 7 tests across 2 files. [2.31s] ```
Author
Owner

Closed by the orchestrator during the 2026-10-28 handover (#254), on the owner's request to close finished issues. Evidence: Sign-in/onboarding port merged as 56bb6f51d (in dev) with D98 updated in d2ce17a; tests and screenshots posted. Reopen if anything is still missing.

Closed by the orchestrator during the 2026-10-28 handover (#254), on the owner's request to close finished issues. Evidence: Sign-in/onboarding port merged as 56bb6f51d (in dev) with D98 updated in d2ce17a; tests and screenshots posted. Reopen if anything is still missing.
kayg closed this issue 2026-10-28 17:52:22 +00:00
Sign in to join this conversation.
No labels
launch-qa
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
kayg/infraternal#32
No description provided.