MOTION: extend the tab-bar spring language app-wide (sheets, sidebar, segmented, popovers, toasts, lists)

kayg commented

2026-09-28 06:12:33 +00:00

Owner

Owner feedback (2026-09-28): "very good job on everything else btw. including the motion of the tab bar morphing. we need more springy motion like that which looks very good and feels very natural! amazing job!"

Goal: extend the motion language of #236 (capsule width morph, easeOutQuint, overdamped spring, rise with stagger; tokens in packages/ui/src/motion.ts) to the other places where things move, so the whole app feels like the tab-bar morph. Use ONLY the shared motion tokens and helpers. Add a token only when no existing one fits.

Candidates (audit the app and list what moves today and how):

Sheets on phone: present and dismiss with a spring, and an interactive drag-to-dismiss that follows the finger and springs back on cancel.
Sidebar collapse/expand and the resizable panels' snap-to-collapse and double-click reset.
Segmented controls and tab-like switches: the selected thumb slides with the spring.
Popovers and the inspector popover: scale and translate from their anchor.
Toast stack: newer toasts push older ones with the spring; swipe-to-dismiss springs back on cancel.
List reordering (the tab bar reorder from the tabbar job, pinned items in the sidebar) and items that insert or remove (FLIP with the spring).
The Search palette open and close.
Rules: at most 500 ms for each animation (owner cap), transform/opacity only (width only for small fixed elements, as with the capsule), interruptible (a new gesture retargets the spring from its current velocity, no jump), prefers-reduced-motion gives instant changes. Sub-view switches stay instant (#236 decision). Performance: no long tasks, no INP regression on a mid-range phone profile.
Evidence: recordings of each changed interaction, light and dark, plus reduced motion. The orchestrator reviews them.

Depends on: the tabbar job (context toolbar + reorder) and the glass job merging first, because they touch the same surfaces.

Owner feedback (2026-09-28): "very good job on everything else btw. including the motion of the tab bar morphing. we need more springy motion like that which looks very good and feels very natural! amazing job!" Goal: extend the motion language of #236 (capsule width morph, easeOutQuint, overdamped spring, rise with stagger; tokens in `packages/ui/src/motion.ts`) to the other places where things move, so the whole app feels like the tab-bar morph. Use ONLY the shared motion tokens and helpers. Add a token only when no existing one fits. Candidates (audit the app and list what moves today and how): - Sheets on phone: present and dismiss with a spring, and an interactive drag-to-dismiss that follows the finger and springs back on cancel. - Sidebar collapse/expand and the resizable panels' snap-to-collapse and double-click reset. - Segmented controls and tab-like switches: the selected thumb slides with the spring. - Popovers and the inspector popover: scale and translate from their anchor. - Toast stack: newer toasts push older ones with the spring; swipe-to-dismiss springs back on cancel. - List reordering (the tab bar reorder from the tabbar job, pinned items in the sidebar) and items that insert or remove (FLIP with the spring). - The Search palette open and close. Rules: at most 500 ms for each animation (owner cap), transform/opacity only (width only for small fixed elements, as with the capsule), interruptible (a new gesture retargets the spring from its current velocity, no jump), `prefers-reduced-motion` gives instant changes. Sub-view switches stay instant (#236 decision). Performance: no long tasks, no INP regression on a mid-range phone profile. Evidence: recordings of each changed interaction, light and dark, plus reduced motion. The orchestrator reviews them. Depends on: the tabbar job (context toolbar + reorder) and the glass job merging first, because they touch the same surfaces.

390-tokyo-night-transition-frames.png

892 KiB

1440-tokyo-night-transition-frames.png

1.3 MiB

390-paper.webm

1.2 MiB

390-paper-reduced-motion.png

66 KiB

390-tokyo-night.webm

1.2 MiB

1440-tokyo-night-reduced-motion.png

204 KiB

390-tokyo-night-reduced-motion.png

80 KiB

820-paper.webm

1.7 MiB

820-paper-reduced-motion.png

139 KiB

820-tokyo-night.webm

1.8 MiB

820-tokyo-night-reduced-motion.png

153 KiB

1440-paper.webm

2.2 MiB

1440-paper-reduced-motion.png

180 KiB

1440-tokyo-night.webm

3 MiB

390-paper.webm

1.3 MiB

390-tokyo-night.webm

1.3 MiB

820-paper.webm

1.5 MiB

820-tokyo-night.webm

1.7 MiB

1440-paper.webm

3.4 MiB

1440-tokyo-night.webm

3.1 MiB

390-paper-reduced-motion.png

66 KiB

1440-tokyo-night-reduced-motion.png

184 KiB

390-tokyo-night-reduced-motion.png

63 KiB

820-paper-reduced-motion.png

135 KiB

820-tokyo-night-reduced-motion.png

141 KiB

1440-paper-reduced-motion.png

179 KiB

kayg referenced this issue

2026-09-28 06:12:59 +00:00

Owner test tracker: every request and when it is testable on o2 #203

kayg commented

2026-09-28 07:34:34 +00:00

Author

Owner

The scan of dev at c9ec6aff84a9ec810124a2cfbdd07c189eea1b9c found 142 CSS transition/animation declarations with direct non-zero time values and 105 with direct easing values outside packages/ui/src/tokens.css. Repeated sources include Calendar GridColumn.svelte (6 of each), AgendaList.svelte (6 durations), and Notes NoteProperties.svelte (5 of each). Examples of local values are packages/ui/src/components/PrimaryPill.svelte:75 (160ms ease-out), PillGroup.svelte:338 (0.22s cubic-bezier(...)), and Toggle.svelte:72-88 (0.2s transitions).

I also found one explicit magnetic opt-out: PrimaryPill.svelte:39 passes magnetic={false} to PillGroup, whose default is true. The component comments explain why this is a separate single-action pill, but do not say why it opts out. Please review this with the shared interaction work. No new motion issue was created.

The scan of dev at `c9ec6aff84a9ec810124a2cfbdd07c189eea1b9c` found 142 CSS transition/animation declarations with direct non-zero time values and 105 with direct easing values outside `packages/ui/src/tokens.css`. Repeated sources include Calendar `GridColumn.svelte` (6 of each), `AgendaList.svelte` (6 durations), and Notes `NoteProperties.svelte` (5 of each). Examples of local values are `packages/ui/src/components/PrimaryPill.svelte:75` (`160ms ease-out`), `PillGroup.svelte:338` (`0.22s cubic-bezier(...)`), and `Toggle.svelte:72-88` (`0.2s` transitions). I also found one explicit magnetic opt-out: `PrimaryPill.svelte:39` passes `magnetic={false}` to `PillGroup`, whose default is true. The component comments explain why this is a separate single-action pill, but do not say why it opts out. Please review this with the shared interaction work. No new motion issue was created.

kayg referenced this issue

2026-09-28 08:31:30 +00:00

AUDIT: app-wide chrome consistency (headers, blur, breadcrumbs, pills, links, type scale) #237

kayg referenced this issue

2026-09-28 08:42:20 +00:00

RESEARCH: local voice transcription and rewrite (Superwhisper-style) — on-device/on-server, no cloud (NO implementation) #304

kayg commented

2026-09-28 13:01:20 +00:00

Author

Owner

Starting #291 on branch job/motion-spring, based on dev SHA 5431fd67c13a504b6aed0afb573de8675ef55692.

Audit: what moves today in the requested surfaces

Mode tray and segmented controls: the selected capsule slides and resizes over 240 ms with an overshooting cubic. Active labels grow through padding and grid-column transitions (140–160 ms). Tab reordering uses a 200 ms FLIP transition with the shared overdamped curve. Sub-view navigation stays instant under DESIGN §34 and §38.
Sheets: OverlaySurface fades the scrim in 180 ms and out 120 ms. Full sheets slide in 240 ms and out 180 ms with the drawer easing. Popover phone sheets use 240/160 ms. Sheet drag follows the pointer directly; releases over 96 px continue into the close transition, while shorter drags return with a 220 ms CSS transform transition. Reduced motion makes these transitions instant.
Anchored menus and popovers: FloatingSurface scales from its anchor and fades in over 180 ms, then fades out over 120 ms. Popover content enters in reading order with a 90 ms first delay, 25 ms spacing, and 310 ms rises, capped at 500 ms. Phone presentations use the sheet path. Reduced motion removes the movement.
Search palette: it uses the shared OverlaySurface pill transition: 180/120 ms opacity, 10 px vertical travel and a small scale change. Its result-driven size and position update without transitions. Sub-view changes remain immediate.
Sidebar: desktop collapse and width changes animate grid-template-columns over 240 ms, which moves layout. Resizing itself tracks the pointer without easing; snap-collapse and double-click reset then pass through the grid transition. The phone sidebar follows the finger and settles in 220 ms with drawer easing; recent pointer velocity chooses the destination but does not drive the settle. Reduced motion snaps it immediately.
Toasts: Sonner's stack uses fixed 180 ms transform/opacity transitions and 120 ms removal/swipe-out. A swipe follows the pointer directly. Under reduced motion the stack stays in place but still fades for 120 ms.
Lists and other local motion: the shared flip.ts helper uses transform-only FLIP transitions that can be called again during a transition. Agenda rows enter with a 240 ms fade/rise and an 18 ms per-row delay; calendar counters, image hovers and file drop feedback use separate local CSS durations and curves.

The current pieces already respect the 500 ms ceiling in these target interactions, but easing and reduced-motion behavior differ by component. I’ll converge the requested surfaces on the shared spring tokens/helpers, keep direct manipulation immediate, and preserve the DESIGN rule that sub-view switches do not animate.

Starting #291 on branch `job/motion-spring`, based on dev SHA `5431fd67c13a504b6aed0afb573de8675ef55692`. Audit: what moves today in the requested surfaces - Mode tray and segmented controls: the selected capsule slides and resizes over 240 ms with an overshooting cubic. Active labels grow through padding and grid-column transitions (140–160 ms). Tab reordering uses a 200 ms FLIP transition with the shared overdamped curve. Sub-view navigation stays instant under DESIGN §34 and §38. - Sheets: `OverlaySurface` fades the scrim in 180 ms and out 120 ms. Full sheets slide in 240 ms and out 180 ms with the drawer easing. Popover phone sheets use 240/160 ms. Sheet drag follows the pointer directly; releases over 96 px continue into the close transition, while shorter drags return with a 220 ms CSS transform transition. Reduced motion makes these transitions instant. - Anchored menus and popovers: `FloatingSurface` scales from its anchor and fades in over 180 ms, then fades out over 120 ms. Popover content enters in reading order with a 90 ms first delay, 25 ms spacing, and 310 ms rises, capped at 500 ms. Phone presentations use the sheet path. Reduced motion removes the movement. - Search palette: it uses the shared `OverlaySurface` pill transition: 180/120 ms opacity, 10 px vertical travel and a small scale change. Its result-driven size and position update without transitions. Sub-view changes remain immediate. - Sidebar: desktop collapse and width changes animate `grid-template-columns` over 240 ms, which moves layout. Resizing itself tracks the pointer without easing; snap-collapse and double-click reset then pass through the grid transition. The phone sidebar follows the finger and settles in 220 ms with drawer easing; recent pointer velocity chooses the destination but does not drive the settle. Reduced motion snaps it immediately. - Toasts: Sonner's stack uses fixed 180 ms transform/opacity transitions and 120 ms removal/swipe-out. A swipe follows the pointer directly. Under reduced motion the stack stays in place but still fades for 120 ms. - Lists and other local motion: the shared `flip.ts` helper uses transform-only FLIP transitions that can be called again during a transition. Agenda rows enter with a 240 ms fade/rise and an 18 ms per-row delay; calendar counters, image hovers and file drop feedback use separate local CSS durations and curves. The current pieces already respect the 500 ms ceiling in these target interactions, but easing and reduced-motion behavior differ by component. I’ll converge the requested surfaces on the shared spring tokens/helpers, keep direct manipulation immediate, and preserve the DESIGN rule that sub-view switches do not animate.

kayg commented

2026-09-28 13:25:20 +00:00

Author

Owner

Finding: a touch pointercancel after dragging a sheet more than 96 px called the close path and navigated away after the 180 ms close delay. The new regression test reproduced this before the fix and passes after it. Cancel and short release now retarget the shared overdamped spring to rest while preserving its sampled velocity; the sheet entrance and dismiss also use the shared 310/200 ms roles.

Fix committed as d82fdc11 (fix(ui): spring sheet drag cancellation back to rest).

Finding: a touch `pointercancel` after dragging a sheet more than 96 px called the close path and navigated away after the 180 ms close delay. The new regression test reproduced this before the fix and passes after it. Cancel and short release now retarget the shared overdamped spring to rest while preserving its sampled velocity; the sheet entrance and dismiss also use the shared 310/200 ms roles. Fix committed as `d82fdc11` (`fix(ui): spring sheet drag cancellation back to rest`).

kayg commented

2026-09-28 14:20:50 +00:00

Author

Owner

Finding: The normal phone sidebar toggle and Escape path wrote mobileOpen = false directly. The {#if mobileOpen} branch then destroyed the dialog before the sheet could animate; only drag settle used the fixed 220 ms timer. The sidebar now keeps its Bits UI dialog and focus trap mounted through the shared retargetable spring, so a rapid close or reopen retargets the current position. The shared spring helper also avoids layout reads when visible-only staggering is not requested.

Focused evidence: bun run test src/lib/components/OverlaySurface.svelte.test.ts src/lib/capsule-motion.test.ts passed 2 files / 13 tests. After adding the layout-read guard, bun run test src/lib/capsule-motion.test.ts passed 1 file / 8 tests.

Finding: The normal phone sidebar toggle and Escape path wrote `mobileOpen = false` directly. The `{#if mobileOpen}` branch then destroyed the dialog before the sheet could animate; only drag settle used the fixed 220 ms timer. The sidebar now keeps its Bits UI dialog and focus trap mounted through the shared retargetable spring, so a rapid close or reopen retargets the current position. The shared spring helper also avoids layout reads when visible-only staggering is not requested. Focused evidence: `bun run test src/lib/components/OverlaySurface.svelte.test.ts src/lib/capsule-motion.test.ts` passed 2 files / 13 tests. After adding the layout-read guard, `bun run test src/lib/capsule-motion.test.ts` passed 1 file / 8 tests.

kayg commented

2026-09-28 14:59:23 +00:00

Author

Owner

Finding during the sheet review: the scrim click handler ignored a deliberate tap while settling was true, so a tap during the opening spring could not retarget the sheet closed. Gesture-generated clicks already pass through the separate swallowClick capture handler. I am removing the settling guard and adding a production-build capture step that taps the scrim during sheet entry.

Finding during the sheet review: the scrim click handler ignored a deliberate tap while `settling` was true, so a tap during the opening spring could not retarget the sheet closed. Gesture-generated clicks already pass through the separate `swallowClick` capture handler. I am removing the settling guard and adding a production-build capture step that taps the scrim during sheet entry.

kayg commented

2026-09-28 17:51:39 +00:00

Author

Owner

Web gate finding after the one-time dev merge: bun run check reported sheetMotion.target may be undefined in apps/web/src/lib/components/app-sidebar.svelte; I am adding an explicit state guard. It also reported that module qr is missing from the current node_modules, while apps/web/package.json and bun.lock already contain the locked qr@0.7.0 dependency. I will restore the frozen install and rerun the web gates.

Web gate finding after the one-time dev merge: `bun run check` reported `sheetMotion.target` may be undefined in `apps/web/src/lib/components/app-sidebar.svelte`; I am adding an explicit state guard. It also reported that module `qr` is missing from the current `node_modules`, while `apps/web/package.json` and `bun.lock` already contain the locked `qr@0.7.0` dependency. I will restore the frozen install and rerun the web gates.

kayg commented

2026-09-28 18:30:51 +00:00

Author

Owner

Evidence review: the production capture opened the phone Info sheet and closed it with Escape, but did not record the sheet's drag path. The capture now drags past the 96 px close threshold, checks that the sheet follows the touch, cancels and checks return to rest, then repeats and dismisses with a completed drag. I will verify this in the 390 px production recording.

kayg commented

2026-09-28 19:36:00 +00:00

Author

Owner

Build finding: the production-server capture build remained in the vendored openssl-sys CMake/make step for over an hour on the shared host. The process tree showed that native build waiting without CPU progress. The host provides system OpenSSL 3.5.7 through pkg-config, and openssl-sys supports OPENSSL_NO_VENDOR=1. I stopped the stalled build and will use that environment variable for local Cargo commands only; it does not change repository files or the dependency graph.

Build finding: the production-server capture build remained in the vendored `openssl-sys` CMake/make step for over an hour on the shared host. The process tree showed that native build waiting without CPU progress. The host provides system OpenSSL 3.5.7 through `pkg-config`, and `openssl-sys` supports `OPENSSL_NO_VENDOR=1`. I stopped the stalled build and will use that environment variable for local Cargo commands only; it does not change repository files or the dependency graph.

kayg commented

2026-09-28 19:45:15 +00:00

Author

Owner

Capture finding: at 390 px, the new touch-cancel probe passed both checks: the Info sheet followed a 128 px pull and returned to rest after cancellation. The next capture step tried to click Show Info while the sheet was correctly still open, so its scrim intercepted the click (aria-expanded="true"). I will close the retained sheet with Escape before reopening it for the completed-drag dismissal recording. This is a capture-script sequencing issue, not a product failure.

Capture finding: at 390 px, the new touch-cancel probe passed both checks: the Info sheet followed a 128 px pull and returned to rest after cancellation. The next capture step tried to click Show Info while the sheet was correctly still open, so its scrim intercepted the click (`aria-expanded="true"`). I will close the retained sheet with Escape before reopening it for the completed-drag dismissal recording. This is a capture-script sequencing issue, not a product failure.

kayg commented

2026-09-28 19:51:15 +00:00

Author

Owner

Evidence progression: the updated capture passed the 390 px phone run in light and dark, including sheet cancel/spring-back and completed-drag dismissal. Both 820 px tablet runs also passed. At 1440 px light, the toast swipe probe timed out waiting for Sonner's data-swiping state immediately after a mouse press. I am checking the hit target and pointer event path before classifying this as an app regression.

Evidence progression: the updated capture passed the 390 px phone run in light and dark, including sheet cancel/spring-back and completed-drag dismissal. Both 820 px tablet runs also passed. At 1440 px light, the toast swipe probe timed out waiting for Sonner's `data-swiping` state immediately after a mouse press. I am checking the hit target and pointer event path before classifying this as an app regression.

kayg commented

2026-09-28 19:56:59 +00:00

Author

Owner

Capture finding: a full production run sampled the phone sheet 360 ms after touch-cancel and observed 1.28463 px of residual translation. The shared interactive spring is capped at 500 ms, so this was an early assertion, not a missed return. I am changing the capture to wait, up to 650 ms for frame-scheduling slack, until the sheet is within 1 px of rest; the pass condition stays the same.

kayg referenced this issue

2026-09-28 20:14:22 +00:00

SEARCH step D: 1M-document query returns 503; measure every target on a QUIET host #338

kayg commented

2026-09-28 20:17:05 +00:00

Author

Owner

Production evidence from the real app is attached. Recordings cover the spring interactions at 390, 820, and 1440 px in both themes. Reduced-motion screenshots cover all six viewport/theme combinations.

Recordings:

390 px — paper, Tokyo Night
820 px — paper, Tokyo Night
1440 px — paper, Tokyo Night

Reduced-motion screenshots:

390 px — paper, Tokyo Night
820 px — paper, Tokyo Night
1440 px — paper, Tokyo Night

The local captures remain under artifacts/motion-spring/; no review artifact is committed.

Production evidence from the real app is attached. Recordings cover the spring interactions at 390, 820, and 1440 px in both themes. Reduced-motion screenshots cover all six viewport/theme combinations. Recordings: - 390 px — [paper](https://git.kayg.org/attachments/8ac12bf1-7022-4a0f-b063-f8436f9f03dd), [Tokyo Night](https://git.kayg.org/attachments/d91eaca3-b160-43cc-85b4-b63ecdc8e12e) - 820 px — [paper](https://git.kayg.org/attachments/21c2dd4f-94e9-41c9-89fb-43ef8836f9df), [Tokyo Night](https://git.kayg.org/attachments/ddeefd59-4805-4dc9-9bd0-32ae3fd0dd03) - 1440 px — [paper](https://git.kayg.org/attachments/95b4c443-369c-40cd-a304-37702d17b752), [Tokyo Night](https://git.kayg.org/attachments/815ebf77-ad20-4a4e-8bcb-37e9c40be949) Reduced-motion screenshots: - 390 px — [paper](https://git.kayg.org/attachments/623f3442-b31f-4901-95d3-545f63e83ec0), [Tokyo Night](https://git.kayg.org/attachments/7630f19a-fe09-446d-bf99-b78f9b99cd12) - 820 px — [paper](https://git.kayg.org/attachments/d82769da-5ac4-4892-ac22-49f2ed5c92c1), [Tokyo Night](https://git.kayg.org/attachments/cbc3ce88-5c73-4281-a398-b843a357406e) - 1440 px — [paper](https://git.kayg.org/attachments/ddaccf3b-2d3c-48d9-9651-1e453353fe0a), [Tokyo Night](https://git.kayg.org/attachments/2c815681-9693-418a-ab81-2e1515ad1d8d) The local captures remain under `artifacts/motion-spring/`; no review artifact is committed.

kayg referenced this issue

2026-09-28 20:22:47 +00:00

SECURITY: Preserve cross-User search availability during staged rebuild #345

kayg referenced this issue

2026-09-28 20:22:47 +00:00

Search: hybrid keyword + semantic retrieval #59

kayg referenced this issue

2026-09-28 20:22:47 +00:00

Search: sources — PDF text extraction, events, bookmarks, tasks, tags in the index #58

kayg referenced this issue

2026-09-28 20:23:45 +00:00

Calendar: calternal as a CalDAV server — an adapter over log entries plus connected calendars #41

kayg referenced this issue

2026-09-28 20:36:13 +00:00

Search rebuild temporarily hides committed results #362

kayg referenced this issue

2026-09-28 20:38:26 +00:00

App password profile issuance returns 403 for the Owner #374

kayg referenced this issue

2026-09-28 20:44:30 +00:00

SECURITY: concurrent dedup can corrupt files across Users #375

kayg commented

2026-09-28 23:08:15 +00:00

Author

Owner

Finished — motion spring (#291)

Implemented the shared, interruptible spring motion across the application and UI surfaces. The common motion driver caps motion at 500 ms, uses transform/opacity, settles interrupted gestures from their current position and velocity, and makes reduced motion instant. Sub-view switches remain instant. Added production-browser evidence for the phone sheet drag/cancel/dismiss gestures and the related shell, overlay, toast, and list interactions.

Files changed for this job:

packages/ui/src/motion.ts, packages/ui/src/tokens.css, packages/ui/src/index.ts
packages/ui/src/components/FloatingSidebar.svelte, OverlaySurface.svelte, SegmentedControl.svelte, menu/FloatingSurface.svelte, calendar/AgendaList.svelte
apps/web/src/routes/layout.css
apps/web/src/lib/capsule-motion.test.ts
apps/web/src/lib/components/AppToaster.svelte, OverlaySurface.svelte.test.ts, SidebarLinks.svelte, app-sidebar.svelte
apps/web/src/lib/files/FilesSidebar.svelte, apps/web/src/lib/notifications/InboxPanel.svelte
apps/web/e2e/motion-spring-evidence.mjs

Six production recordings and six reduced-motion screenshots are attached in the preceding comment. No screenshot or video is committed.

Branch: job/motion-spring
Head: b0887f627e13f06bde4e4bb8ed06f4bef09f67f0 (matches origin/job/motion-spring). The one dev merge is e94f991569d2c063202d899338396cf1d9f74b34. dev advanced after that merge; I did not merge it a second time.

Gates

cargo fmt --check — exit 0; stdout and stderr were empty.
cargo clippy --all-targets -- -D warnings — passed. Final output:

    Finished `dev` profile [unoptimized + debuginfo] target(s) in 82m 59s

cargo test — exit 101. The API, auth, and CLI suites passed (1/1, 53/53, 10/10). The collaboration suite had 14 passes and one existing 2-second performance assertion exceeded under shared-host load:

10,000-block collaboration phases: parse=2.087111477s, Yrs=924.271549ms, block-index=51.993165ms, first-sync=107.364299ms (616204 bytes), snapshot=86.895113ms (616199 bytes), total=3.257635603s
thread 'session::public_edit_limit_tests::ten_thousand_markdown_blocks_open_and_sync_within_two_seconds' panicked at crates/calternal-collab/src/session.rs:370:9:
10,000-block open, first sync and snapshot took 3.257635603s
test result: FAILED. 14 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.50s
error: test failed, to rerun pass `-p calternal-collab --lib`

bun run check — passed:

$ svelte-kit sync && svelte-check --tsconfig ./tsconfig.json
Loading svelte-check in workspace: /home/kayg/Developer/calternal-wt/motion-spring/apps/web
Getting Svelte diagnostics...

svelte-check found 0 errors and 0 warnings

bun run test — exit 1; 749/750 passed. The font-weight-token source scan exceeded Vitest's 5-second timeout (reported duration 15.5 seconds):

 ❯ |unit| src/lib/styles/font-weight-token.test.ts (3 tests | 1 failed) 15510ms
Error: Test timed out in 5000ms.
Test Files  1 failed | 112 passed (113)
Tests  1 failed | 749 passed (750)
error: script "test" exited with code 1

No test expectation was changed and neither failed gate was rerun. cargo clean removed 18,977 files (12.8 GiB); apps/web/build and apps/web/.svelte-kit were deleted.

Adversarial round and known gaps

The one time-boxed real-server round exited 1. It reported concurrent dedup upload/copy byte mismatches, cross-User overwrite changes, and a missing restored path. I filed this merge-blocking data-integrity/security finding as #375 and linked it from #331. The round also produced reports on #374 (Owner app-profile issue returned 403), #362/#345/#338 (Search results during rebuild), #368 (transport timeouts), #347 (Calendar photo and recurring Log mismatches), #41 (accepted malformed DAV alarm), #58 (PDF text search), and #59 (semantic recall). The CLI probe did not run because it hard-codes <repo>/target/debug/calternal while this job uses the required external CARGO_TARGET_DIR; that harness gap is recorded on #350. Calendar setup also requested its own server build because the harness did not receive CALTERNAL_SERVER_BIN; later User-matrix fixtures returned HTTP -1. No second adversarial pass was run.

In the 820 px Tokyo Night reduced-motion capture, the palette count says five navigation entries while three rows are visible. This screenshot is attached for review; the other width/theme captures show the full list.

Decisions not specified in DESIGN

I used an overdamped ratio of 1.08, a 310 ms item rise, a 90 ms first-item delay, 25 ms staggering capped at four indices, and a 200 ms surface morph. These choices keep the final staggered item within the 500 ms cap and avoid bounce. Reduced-motion transitions settle immediately.

## Finished — motion spring (#291) Implemented the shared, interruptible spring motion across the application and UI surfaces. The common motion driver caps motion at 500 ms, uses transform/opacity, settles interrupted gestures from their current position and velocity, and makes reduced motion instant. Sub-view switches remain instant. Added production-browser evidence for the phone sheet drag/cancel/dismiss gestures and the related shell, overlay, toast, and list interactions. Files changed for this job: - `packages/ui/src/motion.ts`, `packages/ui/src/tokens.css`, `packages/ui/src/index.ts` - `packages/ui/src/components/FloatingSidebar.svelte`, `OverlaySurface.svelte`, `SegmentedControl.svelte`, `menu/FloatingSurface.svelte`, `calendar/AgendaList.svelte` - `apps/web/src/routes/layout.css` - `apps/web/src/lib/capsule-motion.test.ts` - `apps/web/src/lib/components/AppToaster.svelte`, `OverlaySurface.svelte.test.ts`, `SidebarLinks.svelte`, `app-sidebar.svelte` - `apps/web/src/lib/files/FilesSidebar.svelte`, `apps/web/src/lib/notifications/InboxPanel.svelte` - `apps/web/e2e/motion-spring-evidence.mjs` Six production recordings and six reduced-motion screenshots are attached in the preceding comment. No screenshot or video is committed. Branch: `job/motion-spring` Head: `b0887f627e13f06bde4e4bb8ed06f4bef09f67f0` (matches `origin/job/motion-spring`). The one `dev` merge is `e94f991569d2c063202d899338396cf1d9f74b34`. `dev` advanced after that merge; I did not merge it a second time. ## Gates - `cargo fmt --check` — exit 0; stdout and stderr were empty. - `cargo clippy --all-targets -- -D warnings` — passed. Final output: ```text Finished `dev` profile [unoptimized + debuginfo] target(s) in 82m 59s ``` - `cargo test` — exit 101. The API, auth, and CLI suites passed (1/1, 53/53, 10/10). The collaboration suite had 14 passes and one existing 2-second performance assertion exceeded under shared-host load: ```text 10,000-block collaboration phases: parse=2.087111477s, Yrs=924.271549ms, block-index=51.993165ms, first-sync=107.364299ms (616204 bytes), snapshot=86.895113ms (616199 bytes), total=3.257635603s thread 'session::public_edit_limit_tests::ten_thousand_markdown_blocks_open_and_sync_within_two_seconds' panicked at crates/calternal-collab/src/session.rs:370:9: 10,000-block open, first sync and snapshot took 3.257635603s test result: FAILED. 14 passed; 1 failed; 0 ignored; 0 measured; 0 filtered out; finished in 3.50s error: test failed, to rerun pass `-p calternal-collab --lib` ``` - `bun run check` — passed: ```text $ svelte-kit sync && svelte-check --tsconfig ./tsconfig.json Loading svelte-check in workspace: /home/kayg/Developer/calternal-wt/motion-spring/apps/web Getting Svelte diagnostics... svelte-check found 0 errors and 0 warnings ``` - `bun run test` — exit 1; 749/750 passed. The font-weight-token source scan exceeded Vitest's 5-second timeout (reported duration 15.5 seconds): ```text ❯ |unit| src/lib/styles/font-weight-token.test.ts (3 tests | 1 failed) 15510ms Error: Test timed out in 5000ms. Test Files 1 failed | 112 passed (113) Tests 1 failed | 749 passed (750) error: script "test" exited with code 1 ``` No test expectation was changed and neither failed gate was rerun. `cargo clean` removed 18,977 files (12.8 GiB); `apps/web/build` and `apps/web/.svelte-kit` were deleted. ## Adversarial round and known gaps The one time-boxed real-server round exited 1. It reported concurrent dedup upload/copy byte mismatches, cross-User overwrite changes, and a missing restored path. I filed this merge-blocking data-integrity/security finding as #375 and linked it from #331. The round also produced reports on #374 (Owner app-profile issue returned 403), #362/#345/#338 (Search results during rebuild), #368 (transport timeouts), #347 (Calendar photo and recurring Log mismatches), #41 (accepted malformed DAV alarm), #58 (PDF text search), and #59 (semantic recall). The CLI probe did not run because it hard-codes `<repo>/target/debug/calternal` while this job uses the required external `CARGO_TARGET_DIR`; that harness gap is recorded on #350. Calendar setup also requested its own server build because the harness did not receive `CALTERNAL_SERVER_BIN`; later User-matrix fixtures returned HTTP -1. No second adversarial pass was run. In the 820 px Tokyo Night reduced-motion capture, the palette count says five navigation entries while three rows are visible. This screenshot is attached for review; the other width/theme captures show the full list. ## Decisions not specified in DESIGN I used an overdamped ratio of 1.08, a 310 ms item rise, a 90 ms first-item delay, 25 ms staggering capped at four indices, and a 200 ms surface morph. These choices keep the final staggered item within the 500 ms cap and avoid bounce. Reduced-motion transitions settle immediately.

kayg commented

2026-09-28 23:10:58 +00:00

Author

Owner

Review of job/motion-spring `b0887f62` (Claude): one round

Reduced motion leaves the Search palette clipped. 820-tokyo-night-reduced-motion.png: the palette shows 3 of 5 Navigate rows, and the footer action bar overlaps the list. The height morph never settles. With prefers-reduced-motion, every spring must jump to its final value synchronously (no rAF, no transition), and the measured size must come from the final layout. Add an e2e assertion: with reduced motion on, the palette's final height equals its content height within 1 px, 50 ms after opening.
The "tokyo-night" captures render the light theme. Set and assert the theme before capture (as in #326).
The recordings show only page load, not the motion. Record each surface doing its transition (open and close a sheet, collapse and expand the sidebar, switch a segmented control, open a popover and a menu, show and dismiss a toast, open search), 2–3 s each, at 1440 and 390, one theme. Also export a frame strip (6 frames per transition) so it can be reviewed as images.
Merge dev (it moved a lot tonight: selection pills, cal-grid, cal-switcher, mobile-focus later) and resolve with both sides in mind. Role tokens only.

Web gates verbatim (re-run the two timeouts: they came from host load; do not raise the timeouts).

## Review of job/motion-spring b0887f62 (Claude): one round 1. **Reduced motion leaves the Search palette clipped.** `820-tokyo-night-reduced-motion.png`: the palette shows 3 of 5 Navigate rows, and the footer action bar overlaps the list. The height morph never settles. With `prefers-reduced-motion`, every spring must jump to its final value synchronously (no rAF, no transition), and the measured size must come from the final layout. Add an e2e assertion: with reduced motion on, the palette's final height equals its content height within 1 px, 50 ms after opening. 2. **The "tokyo-night" captures render the light theme.** Set and assert the theme before capture (as in #326). 3. **The recordings show only page load**, not the motion. Record each surface doing its transition (open and close a sheet, collapse and expand the sidebar, switch a segmented control, open a popover and a menu, show and dismiss a toast, open search), 2–3 s each, at 1440 and 390, one theme. Also export a frame strip (6 frames per transition) so it can be reviewed as images. 4. Merge dev (it moved a lot tonight: selection pills, cal-grid, cal-switcher, mobile-focus later) and resolve with both sides in mind. Role tokens only. Web gates verbatim (re-run the two timeouts: they came from host load; do not raise the timeouts).

kayg commented

2026-09-28 23:12:34 +00:00

Author

Owner

Continuing #291 on job/motion-spring at b0887f627e13f06bde4e4bb8ed06f4bef09f67f0. The branch was based on dev 5431fd67c13a504b6aed0afb573de8675ef55692; its prior dev merge is e94f991569d2c063202d899338396cf1d9f74b34. I am addressing the four findings in the latest Claude review: reduced-motion Search palette sizing, explicit theme setup for Tokyo Night captures, recordings/frame strips that show each transition at 390 and 1440 px, and one current dev merge before final gates.

Continuing #291 on `job/motion-spring` at `b0887f627e13f06bde4e4bb8ed06f4bef09f67f0`. The branch was based on dev `5431fd67c13a504b6aed0afb573de8675ef55692`; its prior dev merge is `e94f991569d2c063202d899338396cf1d9f74b34`. I am addressing the four findings in the latest Claude review: reduced-motion Search palette sizing, explicit theme setup for Tokyo Night captures, recordings/frame strips that show each transition at 390 and 1440 px, and one current dev merge before final gates.

kayg commented

2026-09-28 23:16:43 +00:00

Author

Owner

Finding: 820-tokyo-night-reduced-motion.png shows the Search palette count as 5 while only 3 Navigate rows fit; the action bar overlaps the remaining list. search-dialog.svelte sends every compact-palette ResizeObserver update through requestAnimationFrame, even when reduced motion is active, so the measured height can lag the final result layout. I am adding a production e2e assertion for the final content height 50 ms after opening.

Finding: the capture's setTheme spreads the registered User's settings and replaces only theme and palette. A saved v2 family remains authoritative in normalizeSettings, so the resulting data-theme can differ from the requested Tokyo Night palette. The existing capture only checks .is-dark, which does not identify the selected palette. I will assert data-theme and data-mode before recording.

Finding: `820-tokyo-night-reduced-motion.png` shows the Search palette count as 5 while only 3 Navigate rows fit; the action bar overlaps the remaining list. `search-dialog.svelte` sends every compact-palette ResizeObserver update through `requestAnimationFrame`, even when reduced motion is active, so the measured height can lag the final result layout. I am adding a production e2e assertion for the final content height 50 ms after opening. Finding: the capture's `setTheme` spreads the registered User's settings and replaces only `theme` and `palette`. A saved v2 `family` remains authoritative in `normalizeSettings`, so the resulting `data-theme` can differ from the requested Tokyo Night palette. The existing capture only checks `.is-dark`, which does not identify the selected palette. I will assert `data-theme` and `data-mode` before recording.

kayg commented

2026-09-28 23:40:49 +00:00

Author

Owner

Finding reproduced on the real production app at 820 × 900 with reduced motion. The new e2e assertion fails 50 ms after opening: surface height 233 px, content height 365 px, Navigate count 5, and all 5 rows present. This confirms stale compact-palette measurement rather than missing Search results. I am making the reduced-motion ResizeObserver path measure synchronously; normal-motion measurements remain coalesced to one animation frame.

Finding reproduced on the real production app at 820 × 900 with reduced motion. The new e2e assertion fails 50 ms after opening: surface height `233` px, content height `365` px, Navigate count `5`, and all `5` rows present. This confirms stale compact-palette measurement rather than missing Search results. I am making the reduced-motion ResizeObserver path measure synchronously; normal-motion measurements remain coalesced to one animation frame.

kayg commented

2026-09-28 23:45:08 +00:00

Author

Owner

Capture setup finding: the first post-fix e2e run still returned 233 px because the local server binary embeds apps/web/build and had been built before the updated SPA bundle. The browser was still exercising the old JavaScript. I am rebuilding the server after the SPA bundle, then will rerun the same assertion against that binary before deciding whether the product fix needs more work.

Capture setup finding: the first post-fix e2e run still returned `233` px because the local server binary embeds `apps/web/build` and had been built before the updated SPA bundle. The browser was still exercising the old JavaScript. I am rebuilding the server after the SPA bundle, then will rerun the same assertion against that binary before deciding whether the product fix needs more work.

kayg commented

2026-09-28 23:56:26 +00:00

Author

Owner

Finding reproduced: the immediate post-setTheme assertion passes, but the capture-time assertion after navigation to /today fails. The requested state is tokyo-night / dark; the actual document state is tokyo-night-day / light. The capture was asserting before a later settings hydration, so the existing Tokyo Night recordings were not dark captures. I am tracing the authoritative appearance state and will set it before recording, then assert at capture time.

Finding reproduced: the immediate post-`setTheme` assertion passes, but the capture-time assertion after navigation to `/today` fails. The requested state is `tokyo-night` / `dark`; the actual document state is `tokyo-night-day` / `light`. The capture was asserting before a later settings hydration, so the existing Tokyo Night recordings were not dark captures. I am tracing the authoritative appearance state and will set it before recording, then assert at capture time.

kayg commented

2026-09-29 00:20:58 +00:00

Author

Owner

Capture finding and fix: the first exact 390 px theme assertion observed the document at noir with local family mono while the requested local family was tokyo-night and the server scheme was dark. The capture setup now persists the server-backed scheme through /api/v1/appearance, writes the versioned family value, and checks the active palette after reload and navigation. A repeat production run passed Paper and Tokyo Night at 390 px and emitted the six-frame interaction strip.

Capture finding and fix: the first exact 390 px theme assertion observed the document at `noir` with local family `mono` while the requested local family was `tokyo-night` and the server scheme was `dark`. The capture setup now persists the server-backed scheme through `/api/v1/appearance`, writes the versioned family value, and checks the active palette after reload and navigation. A repeat production run passed Paper and Tokyo Night at 390 px and emitted the six-frame interaction strip.

kayg commented

2026-09-29 00:43:17 +00:00

Author

Owner

Finding: after the one-time dev merge and a fresh production capture, 820 px Tokyo Night with reduced motion still rendered five search rows inside a 233 px palette although the content height was 365 px, 50 ms after opening. This reproduces the clipped palette in the merged build; the scroller's fixed border box does not trigger ResizeObserver when its scrollHeight grows. I added a reduced-motion effect that measures after Svelte renders the result count, with no animation-frame delay. I am validating it against a rebuilt production server now.

kayg commented

2026-09-29 00:45:56 +00:00

Author

Owner

The 820 px reduced-motion sizing fix passed against a freshly rebuilt production server. MOTION_SPRING_PROFILE=820 bun e2e/motion-spring-evidence.mjs exited 0 for Paper and Tokyo Night; the 50 ms check confirms the compact Search palette matches the final content height within 1 px. Tokyo Night also passed the exact data-theme/mode assertion after navigation and was dark at capture time. Fix committed as 091068ab (fix(web): measure reduced-motion search results after render).

The 820 px reduced-motion sizing fix passed against a freshly rebuilt production server. `MOTION_SPRING_PROFILE=820 bun e2e/motion-spring-evidence.mjs` exited 0 for Paper and Tokyo Night; the 50 ms check confirms the compact Search palette matches the final content height within 1 px. Tokyo Night also passed the exact `data-theme`/mode assertion after navigation and was dark at capture time. Fix committed as `091068ab` (`fix(web): measure reduced-motion search results after render`).

kayg commented

2026-09-29 00:53:26 +00:00

Author

Owner

Capture finding: a fresh 390 px run failed before the first interaction. The setup requested Paper/light, but after navigating to the app the document and local cache were mono/light, while the server scheme was light. The authenticated app had already hydrated its default family before the capture helper wrote the requested family; hydration could then persist Mono over the capture setting. I am moving both the server mode and local family setup before the app's first route load, following the ordering used in #326, and retaining the exact palette/mode assertion before capture.

Capture finding: a fresh 390 px run failed before the first interaction. The setup requested Paper/light, but after navigating to the app the document and local cache were `mono`/light, while the server scheme was light. The authenticated app had already hydrated its default family before the capture helper wrote the requested family; hydration could then persist Mono over the capture setting. I am moving both the server mode and local family setup before the app's first route load, following the ordering used in #326, and retaining the exact palette/mode assertion before capture.

kayg commented

2026-09-29 00:55:36 +00:00

Author

Owner

The phone rerun reached the Inspector cancel check after the requested Paper theme had rendered correctly. Its unchanged 650 ms wait timed out, but the diagnostic read immediately afterward measured the sheet at 0px. This points to delayed browser frame polling under shared-host load, not a sheet left translated. I did not raise the timeout. The previously passing 390 px production recordings and Tokyo Night frame strip remain available; I am refreshing the desktop captures separately.

The phone rerun reached the Inspector cancel check after the requested Paper theme had rendered correctly. Its unchanged 650 ms wait timed out, but the diagnostic read immediately afterward measured the sheet at `0px`. This points to delayed browser frame polling under shared-host load, not a sheet left translated. I did not raise the timeout. The previously passing 390 px production recordings and Tokyo Night frame strip remain available; I am refreshing the desktop captures separately.

kayg commented

2026-09-29 00:58:28 +00:00

Author

Owner

Capture finding: the 1440 px run timed out waiting for the Inspector after Control+Shift+.. In the Files route, that chord is reserved for hidden files; the Inspector opens from the real Show info toolbar action. I am updating this capture to click that action, then keeping the role=dialog assertion and recording the open/close transition.

Capture finding: the 1440 px run timed out waiting for the Inspector after `Control+Shift+.`. In the Files route, that chord is reserved for hidden files; the Inspector opens from the real `Show info` toolbar action. I am updating this capture to click that action, then keeping the role=dialog assertion and recording the open/close transition.

kayg commented

2026-09-29 01:04:20 +00:00

Author

Owner

Capture finding: the 1440 Tokyo Night run reached the toast probe, but the plain confirmation expired before the F6 focus assertion. Plain confirmations use the real 4-second duration. The capture waited 2.2 seconds for the show frames before focusing the toast, leaving too little margin under host scheduling. I am moving F6 focus into the toast creation action so AppToaster pauses the real toast timer during the recorded show; its duration and dismissal behavior remain unchanged.

kayg referenced this issue

2026-09-29 01:20:36 +00:00

E2E: one shared theme/capture helper — dark screenshots render light across suites #380

kayg commented

2026-09-29 01:22:52 +00:00

Author

Owner

Adversarial-round harness finding: Search chaos passed with 32 concurrent renames. The nested calendar_event_tags.mjs and mail_api.mjs browser probes could not start because I passed ADVERSARIAL_SERVER_BIN but omitted CALTERNAL_SERVER_BIN; their helper then looked for target/debug/calternal-server, which is absent because this job uses the required external CARGO_TARGET_DIR. The one round is still running and will not be restarted.

Adversarial-round harness finding: Search chaos passed with 32 concurrent renames. The nested `calendar_event_tags.mjs` and `mail_api.mjs` browser probes could not start because I passed `ADVERSARIAL_SERVER_BIN` but omitted `CALTERNAL_SERVER_BIN`; their helper then looked for `target/debug/calternal-server`, which is absent because this job uses the required external `CARGO_TARGET_DIR`. The one round is still running and will not be restarted.

kayg referenced this issue

2026-09-29 01:25:56 +00:00

Editor undo and redo changes block separation after paste #363

kayg commented

2026-09-29 01:41:35 +00:00

Author

Owner

Web gate finding: the final bun run test completed with 768/769 tests passing; the existing src/lib/styles/font-weight-token.test.ts source scan timed out at its 5,000 ms limit and reported 15,449 ms. I left the timeout unchanged, as requested. The exact summary is:

 FAIL  |unit| src/lib/styles/font-weight-token.test.ts > font weight token guard > rejects numeric font-weight declarations outside tokens and font faces
Error: Test timed out in 5000ms.
 Test Files  1 failed | 116 passed (117)
      Tests  1 failed | 768 passed (769)
   Start at  03:37:34
   Duration  210.85s (transform 74%, import 10%, environment 9%, tests 5%, setup 2%)
error: script "test" exited with code 1

Web gate finding: the final `bun run test` completed with 768/769 tests passing; the existing `src/lib/styles/font-weight-token.test.ts` source scan timed out at its 5,000 ms limit and reported 15,449 ms. I left the timeout unchanged, as requested. The exact summary is: ```text FAIL |unit| src/lib/styles/font-weight-token.test.ts > font weight token guard > rejects numeric font-weight declarations outside tokens and font faces Error: Test timed out in 5000ms. Test Files 1 failed | 116 passed (117) Tests 1 failed | 768 passed (769) Start at 03:37:34 Duration 210.85s (transform 74%, import 10%, environment 9%, tests 5%, setup 2%) error: script "test" exited with code 1 ```

kayg commented

2026-09-29 02:12:15 +00:00

Author

Owner

Completed: Forgejo #291

What changed

Reduced-motion search results now trigger a compact-palette measurement after Svelte renders a changed result count. The palette retains all five destinations without clipping.
Stabilized the production motion evidence flow. It stores the selected theme before app navigation, checks the resulting theme and mode, opens the Files inspector through its real control, and keeps the confirmation toast visible during the recording.
Captured spring transitions at 390 px, 820 px, and 1440 px. The recorded transition is 2.2 seconds with frame samples at 0, 80, 180, 320, 520, and 780 ms. Paper and Tokyo Night runs passed in the focused captures.
Preserved the dev merge behavior for toast layering, sidebar shortcuts, and Calendar options.

Files changed after the dev merge:

apps/web/src/lib/components/search-dialog.svelte
apps/web/e2e/motion-spring-evidence.mjs
apps/web/src/lib/themes.test.ts

The theme test change narrows the match after the existing one-match assertion; it does not change the expectation.

Evidence

At 820 × 900, the reduced-motion palette was checked for all five destinations at 50 ms; content height stayed within 1 px of the rendered result. Focused runs passed for Paper and Tokyo Night.

Gates

bun run check — exit 0. Output:
$ svelte-kit sync && svelte-check --tsconfig ./tsconfig.json
Loading svelte-check in workspace: /home/kayg/Developer/calternal-wt/motion-spring/apps/web
Getting Svelte diagnostics...

svelte-check found 0 errors and 0 warnings

bun run test — exit 1. One source-scan test timed out; the other 768 tests passed:
FAIL |unit| src/lib/styles/font-weight-token.test.ts > font weight token guard > rejects numeric font-weight declarations outside tokens and font faces
Error: Test timed out in 5000ms.
If this is a long-running test, pass a timeout value as the last argument or configure it globally with "testTimeout".
❯ src/lib/styles/font-weight-token.test.ts:50:2
48|
49| describe('font weight token guard', () => {
50| it('rejects numeric font-weight declarations outside tokens and font …
51| const violations = sourceRoots.flatMap((root) => sourceFiles(root).f…
52| expect(violations, `Numeric weights remain:\n${violations.slice(0, 2…

 Test Files  1 failed | 116 passed (117)
      Tests  1 failed | 768 passed (769)
   Start at  03:37:34
   Duration  210.85s (transform 74%, import 10%, environment 9%, tests 5%, setup 2%)

  Transform  |component| transforming modules took 485.64s · 72% of tracked time, re-done on every run
             persist transforms across runs with fsModuleCache: true
             learn more: https://vitest.dev/guide/improving-performance#caching-between-runs

error: script "test" exited with code 1

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

cargo clippy --all-targets -- -D warnings — exit 0. Final output:
Finished dev profile [unoptimized + debuginfo] target(s) in 7m 12s

cargo test — exit 0. Exact per-target summaries from the completed workspace run:
test result: ok. 107 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 56.71s
test result: ok. 44 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 4.14s
test result: ok. 71 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 7.46s
test result: ok. 54 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.96s
test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.32s

Cleanup: cargo clean removed 16,120 files (13.8 GiB); apps/web/build and apps/web/.svelte-kit were removed.

Adversarial round and known gaps

The single time-boxed adversarial run stopped at its 15-minute limit (exit 124). Search rename concurrency and the main Rust/browser suites completed. Nested calendar/mail probes and fixture matrices could not start because CALTERNAL_SERVER_BIN was not set and their fallback binary path did not exist. The run also found an editor undo/redo history mismatch after hostile paste; evidence is on #363. A 10k-block sync took 2,279 ms and is a SLOW performance finding. The 390 px post-review inspector capture hit the existing 650 ms interaction timeout under shared-host load; the final offset was 0 px, and earlier 390 px captures passed. No timeout was raised and no source change was made for that host-timing result.

Decisions not specified in DESIGN.md

For reduced motion, measure the compact palette once after a result-count change has rendered, then let the existing resize observer keep it aligned.
Seed the appearance before opening the app route in the evidence harness so hydration cannot overwrite the intended theme.
Focus the real confirmation toast before the 2.2-second capture because the existing focus behavior pauses its four-second lifetime.
Use six fixed frame samples over the transition for the review strips.

Head: 7c7139f9dc
Branch: job/motion-spring (pushed; git push reported Everything up-to-date)

## Completed: Forgejo #291 ### What changed - Reduced-motion search results now trigger a compact-palette measurement after Svelte renders a changed result count. The palette retains all five destinations without clipping. - Stabilized the production motion evidence flow. It stores the selected theme before app navigation, checks the resulting theme and mode, opens the Files inspector through its real control, and keeps the confirmation toast visible during the recording. - Captured spring transitions at 390 px, 820 px, and 1440 px. The recorded transition is 2.2 seconds with frame samples at 0, 80, 180, 320, 520, and 780 ms. Paper and Tokyo Night runs passed in the focused captures. - Preserved the dev merge behavior for toast layering, sidebar shortcuts, and Calendar options. Files changed after the dev merge: - apps/web/src/lib/components/search-dialog.svelte - apps/web/e2e/motion-spring-evidence.mjs - apps/web/src/lib/themes.test.ts The theme test change narrows the match after the existing one-match assertion; it does not change the expectation. ### Evidence - [390 px Tokyo Night frame strip](https://git.kayg.org/attachments/1ce81f0e-46a9-48ea-8ddf-6a2c101cdbe3) - [1440 px Tokyo Night frame strip](https://git.kayg.org/attachments/4b345b8e-90e2-4f3c-b980-ccae6e603168) At 820 × 900, the reduced-motion palette was checked for all five destinations at 50 ms; content height stayed within 1 px of the rendered result. Focused runs passed for Paper and Tokyo Night. ### Gates bun run check — exit 0. Output: $ svelte-kit sync && svelte-check --tsconfig ./tsconfig.json Loading svelte-check in workspace: /home/kayg/Developer/calternal-wt/motion-spring/apps/web Getting Svelte diagnostics... svelte-check found 0 errors and 0 warnings bun run test — exit 1. One source-scan test timed out; the other 768 tests passed: FAIL |unit| src/lib/styles/font-weight-token.test.ts > font weight token guard > rejects numeric font-weight declarations outside tokens and font faces Error: Test timed out in 5000ms. If this is a long-running test, pass a timeout value as the last argument or configure it globally with "testTimeout". ❯ src/lib/styles/font-weight-token.test.ts:50:2 48| 49| describe('font weight token guard', () => { 50| it('rejects numeric font-weight declarations outside tokens and font … 51| const violations = sourceRoots.flatMap((root) => sourceFiles(root).f… 52| expect(violations, `Numeric weights remain:\n${violations.slice(0, 2… Test Files 1 failed | 116 passed (117) Tests 1 failed | 768 passed (769) Start at 03:37:34 Duration 210.85s (transform 74%, import 10%, environment 9%, tests 5%, setup 2%) Transform |component| transforming modules took 485.64s · 72% of tracked time, re-done on every run persist transforms across runs with fsModuleCache: true learn more: https://vitest.dev/guide/improving-performance#caching-between-runs error: script "test" exited with code 1 cargo fmt --check — exit 0, no output. cargo clippy --all-targets -- -D warnings — exit 0. Final output: Finished dev profile [unoptimized + debuginfo] target(s) in 7m 12s cargo test — exit 0. Exact per-target summaries from the completed workspace run: test result: ok. 107 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 56.71s test result: ok. 44 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 4.14s test result: ok. 71 passed; 0 failed; 2 ignored; 0 measured; 0 filtered out; finished in 7.46s test result: ok. 54 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.96s test result: ok. 11 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 2.32s Cleanup: cargo clean removed 16,120 files (13.8 GiB); apps/web/build and apps/web/.svelte-kit were removed. ### Adversarial round and known gaps The single time-boxed adversarial run stopped at its 15-minute limit (exit 124). Search rename concurrency and the main Rust/browser suites completed. Nested calendar/mail probes and fixture matrices could not start because CALTERNAL_SERVER_BIN was not set and their fallback binary path did not exist. The run also found an editor undo/redo history mismatch after hostile paste; evidence is on #363. A 10k-block sync took 2,279 ms and is a SLOW performance finding. The 390 px post-review inspector capture hit the existing 650 ms interaction timeout under shared-host load; the final offset was 0 px, and earlier 390 px captures passed. No timeout was raised and no source change was made for that host-timing result. ### Decisions not specified in DESIGN.md - For reduced motion, measure the compact palette once after a result-count change has rendered, then let the existing resize observer keep it aligned. - Seed the appearance before opening the app route in the evidence harness so hydration cannot overwrite the intended theme. - Focus the real confirmation toast before the 2.2-second capture because the existing focus behavior pauses its four-second lifetime. - Use six fixed frame samples over the transition for the review strips. Head: 7c7139f9dcbf264c7833ad9930ee666d566ae156 Branch: job/motion-spring (pushed; git push reported Everything up-to-date)

kayg referenced this issue from a commit

2026-09-29 02:49:03 +00:00

Merge job/motion-spring (#291: one interruptible spring app-wide, capped at 500 ms; reduced motion settles instantly)

kayg closed this issue

2026-09-29 03:37:52 +00:00

kayg referenced this issue

2026-09-29 04:41:45 +00:00

GESTURES: native macOS/iOS gestures everywhere — audit + one shared gesture layer #399

kayg referenced this issue

2026-09-29 04:50:45 +00:00

GESTURES: native macOS/iOS gestures everywhere — audit + one shared gesture layer #399

kayg referenced this issue