kayg commented

Owner

Bug (Claude review of #370 screenshots): a block paints inside the Day view's hour column

In calendar-day-dark-390.png (production build, 390 px, dark), a green rounded block sits in the hour-label column behind "1 PM", left of the grid's first lane. It looks like a timed item (or an event tint) positioned against the wrong containing block, so its x starts at the column's left edge instead of the lane's. Find what it is (likely an event from the fixture, a log entry, or a photo stack), fix its positioning so nothing renders in the hour column, and add an e2e assertion: no element with an event/stack role intersects the hour-label column rectangle in Day and Week at 390/820/1440.

## Bug (Claude review of #370 screenshots): a block paints inside the Day view's hour column In `calendar-day-dark-390.png` (production build, 390 px, dark), a green rounded block sits in the hour-label column behind "1 PM", left of the grid's first lane. It looks like a timed item (or an event tint) positioned against the wrong containing block, so its x starts at the column's left edge instead of the lane's. Find what it is (likely an event from the fixture, a log entry, or a photo stack), fix its positioning so nothing renders in the hour column, and add an e2e assertion: no element with an event/stack role intersects the hour-label column rectangle in Day and Week at 390/820/1440.

calendar-day-390-light-hour-column.png

129 KiB

calendar-day-390-dark-hour-column.png

122 KiB

calendar-day-820-light-hour-column.png

469 KiB

calendar-day-820-dark-hour-column.png

454 KiB

calendar-day-1440-light-hour-column.png

479 KiB

calendar-day-1440-dark-hour-column.png

377 KiB

calendar-week-390-light-hour-column.png

146 KiB

calendar-week-390-dark-hour-column.png

132 KiB

calendar-week-820-light-hour-column.png

534 KiB

calendar-week-820-dark-hour-column.png

512 KiB

calendar-week-1440-light-hour-column.png

547 KiB

calendar-week-1440-dark-hour-column.png

419 KiB

calendar-gutter-edge-3x-day-390-light.png

45 KiB

calendar-gutter-edge-3x-day-390-dark.png

45 KiB

calendar-gutter-edge-3x-week-390-light.png

46 KiB

calendar-gutter-edge-3x-week-390-dark.png

46 KiB

calendar-day-390-light-hour-column.png

140 KiB

calendar-day-390-dark-hour-column.png

131 KiB

calendar-round3-day-390-light-hour-column.png

140 KiB

calendar-round3-day-390-dark-hour-column.png

131 KiB

calendar-round3-day-820-light-hour-column.png

467 KiB

calendar-round3-day-820-dark-hour-column.png

453 KiB

calendar-round3-day-1440-light-hour-column.png

474 KiB

calendar-round3-day-1440-dark-hour-column.png

378 KiB

calendar-round3-week-390-light-hour-column.png

150 KiB

calendar-round3-week-390-dark-hour-column.png

137 KiB

calendar-round3-week-820-light-hour-column.png

540 KiB

calendar-round3-week-820-dark-hour-column.png

522 KiB

calendar-round3-week-1440-light-hour-column.png

557 KiB

calendar-round3-week-1440-dark-hour-column.png

431 KiB

calendar-round3-gutter-edge-3x-day-390-light.png

45 KiB

calendar-round3-gutter-edge-3x-day-390-dark.png

45 KiB

calendar-round3-gutter-edge-3x-week-390-light.png

46 KiB

calendar-round3-gutter-edge-3x-week-390-dark.png

46 KiB

calendar-round3-overlap-3x-390-light.png

16 KiB

calendar-round3-overlap-3x-390-dark.png

14 KiB

calendar-round3-overlap-3x-1440-light.png

45 KiB

calendar-round3-overlap-3x-1440-dark.png

39 KiB

calendar-round3-short-3x-390-light.png

9.9 KiB

calendar-round3-short-3x-390-dark.png

9.6 KiB

calendar-round3-short-3x-1440-light.png

30 KiB

calendar-round3-short-3x-1440-dark.png

26 KiB

calendar-round3-attachment-3x-390-light.png

9.6 KiB

calendar-round3-attachment-3x-390-dark.png

14 KiB

calendar-round3-attachment-3x-1440-light.png

26 KiB

calendar-round3-attachment-3x-1440-dark.png

24 KiB

calendar-round3-all-day-3x-390-light.png

23 KiB

calendar-round3-all-day-3x-390-dark.png

26 KiB

calendar-round3-all-day-3x-1440-light.png

19 KiB

calendar-round3-all-day-3x-1440-dark.png

23 KiB

calendar-day-1440-dark.png

375 KiB

calendar-day-1440-light.png

466 KiB

calendar-day-390-dark.png

130 KiB

calendar-day-390-light.png

136 KiB

calendar-day-820-dark.png

446 KiB

calendar-day-820-light.png

521 KiB

calendar-gutter-edge-3x-day-390-dark.png

45 KiB

calendar-gutter-edge-3x-day-390-light.png

45 KiB

calendar-gutter-edge-3x-week-390-dark.png

46 KiB

calendar-gutter-edge-3x-week-390-light.png

46 KiB

calendar-round4-all-day-3x-390-light.png

16 KiB

calendar-round4-attachment-3x-390-light.png

14 KiB

calendar-round4-drag-ghost-bottom-3x-390-light.png

7.1 KiB

calendar-round4-file-edge-3x-390-light.png

2.7 KiB

calendar-round4-overlap-3x-390-light.png

16 KiB

calendar-round4-short-3x-390-light.png

9.6 KiB

calendar-week-1440-dark.png

432 KiB

calendar-week-1440-light.png

556 KiB

calendar-week-390-dark.png

137 KiB

calendar-week-390-light.png

150 KiB

calendar-week-820-dark.png

522 KiB

calendar-week-820-light.png

541 KiB

kayg referenced this issue

2026-09-29 10:38:01 +00:00

Flaky: calternal-collab crash simulation exits 101 instead of 137 under load #426

kayg commented

2026-09-29 10:38:36 +00:00

Author

Owner

Resuming on job/ab-384-luna from head 07d3ea940e; base is dev merge-base d0d83a32b4. I am tracing the rejected Day gutter screenshots and will add the requested assertions and proof matrix.

Resuming on job/ab-384-luna from head 07d3ea940e70578e561b372aa5118b8d68e9f7da; base is dev merge-base d0d83a32b41e67b64f6586690ba4946e41878160. I am tracing the rejected Day gutter screenshots and will add the requested assertions and proof matrix.

kayg commented

2026-09-29 10:50:57 +00:00

Author

Owner

Root-cause findings while resuming the visual review:

The five blank 10:00 cards come from the five real overlapping provider Events seeded by calendar-view-switcher.mjs (A–E). In a 390 px Day lane, GridColumn allows floor(colw / 36) lanes; each Event becomes about 59 px wide, then .block.plan.cramped .b-content { display: none } hides its title. The same lane calculation leaves the 13:00 Event and 15-minute Log side by side; the Log's 12 px duration ink clips its title.
The previous gutter commit changed both gutter paints to --surface and added an e2e assertion that requires an opaque gutter. The hour labels and the horizontal day track share .tg-body, so buffered columns can move behind the gutter. The replacement will isolate horizontal lane scrolling at the gutter edge and restore the glass wash.
The document glyph at 09:00 is the intended real Files item uploaded by this e2e fixture as Calendar gutter stack.txt; it is not an unexplained production placeholder.
At 08:34 the existing 22-minute label threshold leaves 09 visible even though the current-time label masks that caption.

Root-cause findings while resuming the visual review: - The five blank 10:00 cards come from the five real overlapping provider Events seeded by `calendar-view-switcher.mjs` (A–E). In a 390 px Day lane, `GridColumn` allows `floor(colw / 36)` lanes; each Event becomes about 59 px wide, then `.block.plan.cramped .b-content { display: none }` hides its title. The same lane calculation leaves the 13:00 Event and 15-minute Log side by side; the Log's 12 px duration ink clips its title. - The previous gutter commit changed both gutter paints to `--surface` and added an e2e assertion that requires an opaque gutter. The hour labels and the horizontal day track share `.tg-body`, so buffered columns can move behind the gutter. The replacement will isolate horizontal lane scrolling at the gutter edge and restore the glass wash. - The document glyph at 09:00 is the intended real Files item uploaded by this e2e fixture as `Calendar gutter stack.txt`; it is not an unexplained production placeholder. - At 08:34 the existing 22-minute label threshold leaves 09 visible even though the current-time label masks that caption.

kayg commented

2026-09-29 11:06:23 +00:00

Author

Owner

Added (owner, 2026-09-29, with a screenshot): a 5-minute Log entry (16:00–16:05 'Auto home') is clipped. Its title is cut off at the bottom of a box shorter than one line, and it sits under the next item. Short items get a minimum height of one title line (Apple Calendar style, time hidden or on the same line), and neighbours reflow as overlap columns, so nothing is clipped or covered. Also in the same screenshot: the attachment deck icon on the 'Gym' block renders as two misaligned overlapping badges at the top right. Fix both in the ab-384 round.

kayg commented

2026-09-29 11:31:15 +00:00

Author

Owner

Follow-up to the gutter investigation: a nested overflow-x: auto lane would become the nearest scroll container for block titles and break the grid's existing vertical sticky behavior. I changed the layout to keep the time grid as the one native two-axis scroller, with a sticky overflow: clip lane viewport that starts at the gutter edge. The clipped day track and header use the same horizontal offset; the glass gutter stays translucent. The browser proof now checks that Event titles remain sticky during vertical scroll.

Follow-up to the gutter investigation: a nested `overflow-x: auto` lane would become the nearest scroll container for block titles and break the grid's existing vertical sticky behavior. I changed the layout to keep the time grid as the one native two-axis scroller, with a sticky `overflow: clip` lane viewport that starts at the gutter edge. The clipped day track and header use the same horizontal offset; the glass gutter stays translucent. The browser proof now checks that Event titles remain sticky during vertical scroll.

kayg commented

2026-09-29 11:34:44 +00:00

Author

Owner

The clipped-track regression had a concrete cause. On the 820 px Week route, Playwright measured .tg-body.scrollLeft=12469 (column width about 61 px), so the virtual track showed 2003-05-23 instead of the selected 2026 date and rendered no Event blocks. The browser's native snap areas moved with the transformed track inside the sticky clip. I removed those unstable DOM snap targets and now settle to the nearest virtual day index when native scrolling ends; the existing overlap assertion stays unchanged.

The clipped-track regression had a concrete cause. On the 820 px Week route, Playwright measured `.tg-body.scrollLeft=12469` (column width about 61 px), so the virtual track showed 2003-05-23 instead of the selected 2026 date and rendered no Event blocks. The browser's native snap areas moved with the transformed track inside the sticky clip. I removed those unstable DOM snap targets and now settle to the nearest virtual day index when native scrolling ends; the existing overlap assertion stays unchanged.

kayg commented

2026-09-29 12:02:04 +00:00

Author

Owner

Result

Fixed the #384 Day and Week grid issues. The sticky hour gutter now keeps its translucent glass treatment. A sticky overflow: clip wrapper starts at the gutter edge and clips buffered day content. Its clip is not a scroll container, so Event titles still stick during vertical scrolling.

Day columns use a 160 px overlap-lane floor. At phone width the 10:00 CalDAV fixture now shows a readable Event with +4; the 13:00 Event stays visible and the concurrent 13:05 Log is listed in +1. The five grey boxes were the seeded provider Events A–E, not empty placeholders. The document glyph at the right edge belongs to the real seeded standalone File Calendar gutter stack.txt.

The deterministic 08:34 browser case verifies that the now pill hides the 09 hour label. The production e2e closes the navigation after route changes and verifies that the sidebar sheet, scrim, and collapsed rail do not show at the left edge. Sticky Event title behavior also remains covered.

Decisions and gap

DESIGN §30/§39 do not set lane widths. Day uses 160 px per visible overlap lane; Week keeps the existing 36 px floor for its denser seven-day layout.
DESIGN §39 requires settling to the nearest day. Native CSS snap targets moved incorrectly inside the transformed gutter clip, so the grid settles by virtual date index at scroll end.
DESIGN §29 does not define a collision threshold for the now pill. The nearest hour label fades out when it is within 30 minutes; this hides 09 at 08:34.
There is no 3-day Calendar view in the current model or route, and DESIGN §34/§39 define Day and Week only. I did not add an undefined view. The proof therefore covers Day and Week, not 3-day.

Proof

Production screenshots: Day and Week at 390, 820, and 1440 px in light and dark. Four 3× crops show the gutter edge at mid-track scroll. They are attached below.

Day

390 px: light · dark
820 px: light · dark
1440 px: light · dark

Week

390 px: light · dark
820 px: light · dark
1440 px: light · dark

3× gutter crops

Day, 390 px, light: calendar-gutter-edge-3x-day-390-light.png
Day, 390 px, dark: calendar-gutter-edge-3x-day-390-dark.png
Week, 390 px, light: calendar-gutter-edge-3x-week-390-light.png
Week, 390 px, dark: calendar-gutter-edge-3x-week-390-dark.png

Commit and gates

Head: f8f7df088be4900674629aa0374084c46222fba5

bun run check:

Text sizes use shared role tokens.
svelte-check found 0 errors and 0 warnings

bun run test:

Test Files  123 passed (123)
      Tests  787 passed (787)

bun run build:

✓ built in 31.62s
  Wrote site to "build"
  ✔ done

bun e2e/calendar-view-switcher.mjs:

PASS Calendar view switcher matrix: 3 widths × 2 sidebar states × 2 themes; screenshots in /home/kayg/Developer/calternal-wt/ab-384-luna/apps/web/../../artifacts/ab-384-luna

cargo clean:

Removed 17635 files, 14.9GiB total

No Rust files or server routes changed, so Rust crate gates were not run. Build output was removed. The worktree is clean.

## Result Fixed the #384 Day and Week grid issues. The sticky hour gutter now keeps its translucent glass treatment. A sticky `overflow: clip` wrapper starts at the gutter edge and clips buffered day content. Its clip is not a scroll container, so Event titles still stick during vertical scrolling. Day columns use a 160 px overlap-lane floor. At phone width the 10:00 CalDAV fixture now shows a readable Event with `+4`; the 13:00 Event stays visible and the concurrent 13:05 Log is listed in `+1`. The five grey boxes were the seeded provider Events A–E, not empty placeholders. The document glyph at the right edge belongs to the real seeded standalone File `Calendar gutter stack.txt`. The deterministic 08:34 browser case verifies that the now pill hides the 09 hour label. The production e2e closes the navigation after route changes and verifies that the sidebar sheet, scrim, and collapsed rail do not show at the left edge. Sticky Event title behavior also remains covered. ## Decisions and gap - DESIGN §30/§39 do not set lane widths. Day uses 160 px per visible overlap lane; Week keeps the existing 36 px floor for its denser seven-day layout. - DESIGN §39 requires settling to the nearest day. Native CSS snap targets moved incorrectly inside the transformed gutter clip, so the grid settles by virtual date index at scroll end. - DESIGN §29 does not define a collision threshold for the now pill. The nearest hour label fades out when it is within 30 minutes; this hides 09 at 08:34. - There is no 3-day Calendar view in the current model or route, and DESIGN §34/§39 define Day and Week only. I did not add an undefined view. The proof therefore covers Day and Week, not 3-day. ## Proof Production screenshots: Day and Week at 390, 820, and 1440 px in light and dark. Four 3× crops show the gutter edge at mid-track scroll. They are attached below. **Day** - 390 px: [light](https://git.kayg.org/attachments/3fd83147-9ce0-4d29-b023-2669c3ae7e79) · [dark](https://git.kayg.org/attachments/264d5717-415c-40d0-8fb4-735f9d593e6c) - 820 px: [light](https://git.kayg.org/attachments/752217e9-b72d-4618-97cd-c72cd23e53c1) · [dark](https://git.kayg.org/attachments/65813ab4-807b-4e27-a79d-8e313da734a7) - 1440 px: [light](https://git.kayg.org/attachments/5c60991f-360a-4848-a672-2e58acb76424) · [dark](https://git.kayg.org/attachments/f70af7c7-70b3-48e0-9889-2dc69295cbd7) **Week** - 390 px: [light](https://git.kayg.org/attachments/a38d1bd7-0046-4d32-8feb-c791e38b940d) · [dark](https://git.kayg.org/attachments/28473ba9-1c75-492e-8b83-86eaa534269c) - 820 px: [light](https://git.kayg.org/attachments/062ce0f6-c94d-42e6-8efe-9fb835f70d79) · [dark](https://git.kayg.org/attachments/0ecbecf9-657d-44e4-b46c-67a6767c6065) - 1440 px: [light](https://git.kayg.org/attachments/c96a77ff-38ec-4170-81bc-9ff7ff5148e9) · [dark](https://git.kayg.org/attachments/81ae5926-4acd-4dd0-b942-131a013bf282) **3× gutter crops** - Day, 390 px, light: [calendar-gutter-edge-3x-day-390-light.png](https://git.kayg.org/attachments/4e456206-2b8a-462a-9b17-a86b482546ec) - Day, 390 px, dark: [calendar-gutter-edge-3x-day-390-dark.png](https://git.kayg.org/attachments/34f2c54d-bf24-4126-b241-6355c0124bd4) - Week, 390 px, light: [calendar-gutter-edge-3x-week-390-light.png](https://git.kayg.org/attachments/e67062e5-a110-45a2-82b0-fe0593c7bb02) - Week, 390 px, dark: [calendar-gutter-edge-3x-week-390-dark.png](https://git.kayg.org/attachments/a5aa9471-16b8-4305-9145-31240959fe5d) ## Commit and gates Head: `f8f7df088be4900674629aa0374084c46222fba5` `bun run check`: ```text Text sizes use shared role tokens. svelte-check found 0 errors and 0 warnings ``` `bun run test`: ```text Test Files 123 passed (123) Tests 787 passed (787) ``` `bun run build`: ```text ✓ built in 31.62s Wrote site to "build" ✔ done ``` `bun e2e/calendar-view-switcher.mjs`: ```text PASS Calendar view switcher matrix: 3 widths × 2 sidebar states × 2 themes; screenshots in /home/kayg/Developer/calternal-wt/ab-384-luna/apps/web/../../artifacts/ab-384-luna ``` `cargo clean`: ```text Removed 17635 files, 14.9GiB total ``` No Rust files or server routes changed, so Rust crate gates were not run. Build output was removed. The worktree is clean.

kayg referenced this issue from a commit

2026-09-29 12:04:25 +00:00

Fix narrow calendar grid clipping and overlaps (#384)

kayg commented

2026-09-29 12:05:11 +00:00

Author

Owner

Resuming #384 round 3 on branch job/ab-384-luna at f8f7df088be4900674629aa0374084c46222fba5; original dev merge base is d0d83a32b41e67b64f6586690ba4946e41878160. I am tracing the remaining +N placement, short Log entry, attachment deck badges, all-day glass, and standalone File inset issues before adding regression assertions and the requested production screenshot proof.

Resuming #384 round 3 on branch `job/ab-384-luna` at `f8f7df088be4900674629aa0374084c46222fba5`; original `dev` merge base is `d0d83a32b41e67b64f6586690ba4946e41878160`. I am tracing the remaining +N placement, short Log entry, attachment deck badges, all-day glass, and standalone File inset issues before adding regression assertions and the requested production screenshot proof.

kayg commented

2026-09-29 12:54:35 +00:00

Author

Owner

Added for the next #384 round (TimeGrid owner): after #418 merges, the drag ghost shows a thin green-tinted strip under its bottom edge (about 12 px, visible in #418's calendar-drag-1440-light.png and -390-dark.png). This is probably the drag shadow or the source placeholder picking up the area colour. The ghost is one clean shape with the shared glass shadow and no coloured underlay.

kayg referenced this issue

2026-09-29 13:16:03 +00:00

CALENDAR: document stacks show no previews and sit right-aligned (should preview and sit left) #410

kayg referenced this issue

2026-09-29 13:16:13 +00:00

CALENDAR: document stacks show no previews and sit right-aligned (should preview and sit left) #410

kayg commented

2026-09-29 13:37:38 +00:00

Author

Owner

Round 3 baseline evidence from the focused production-build e2e at 390px/light: all timed +N controls occupied 289.9px (full lane width); the 13:05 +1 rectangle intersected “Follow-up narrow title remains readable”; the 16:00 +2 intersected “Weekly review (provider)”. The all-day header background alpha was 0. The standalone File pile ended 4px from the lane edge while the shared inset token is 8.56px. The attachment line pile was absolutely placed at the card’s upper-right (two offset cards); its deck was still opacity 0 at the default zoom. The five-minute log was in overflow because the fixture’s recurring provider Event also starts at 16:00, so this run could not measure that card; I am removing only that unrelated fixture Event for this scenario.

Round 3 baseline evidence from the focused production-build e2e at 390px/light: all timed `+N` controls occupied 289.9px (full lane width); the 13:05 `+1` rectangle intersected “Follow-up narrow title remains readable”; the 16:00 `+2` intersected “Weekly review (provider)”. The all-day header background alpha was 0. The standalone File pile ended 4px from the lane edge while the shared inset token is 8.56px. The attachment line pile was absolutely placed at the card’s upper-right (two offset cards); its deck was still opacity 0 at the default zoom. The five-minute log was in overflow because the fixture’s recurring provider Event also starts at 16:00, so this run could not measure that card; I am removing only that unrelated fixture Event for this scenario.

kayg commented

2026-09-29 13:52:04 +00:00

Author

Owner

The focused 390px production run now measures the +N chips at 24px and finds no intersections with title rectangles. It also measures the 16:00 Auto home block at 23px high for a 16px title line, places the File pile 8.55px from the lane edge against an 8.56px token, and confirms the attachment pile stays inside the title row without covering its title. One interaction regression remains: clicking timed +1 does not open the shared list because the parent grid pointer-down handler treats .more as empty space and captures the pointer on the column. I am excluding overflow controls from grid creation gestures and rerunning the production e2e.

The focused 390px production run now measures the +N chips at 24px and finds no intersections with title rectangles. It also measures the 16:00 Auto home block at 23px high for a 16px title line, places the File pile 8.55px from the lane edge against an 8.56px token, and confirms the attachment pile stays inside the title row without covering its title. One interaction regression remains: clicking timed +1 does not open the shared list because the parent grid pointer-down handler treats `.more` as empty space and captures the pointer on the column. I am excluding overflow controls from grid creation gestures and rerunning the production e2e.

kayg commented

2026-09-29 13:55:39 +00:00

Author

Owner

Correction and focused production e2e result: the earlier “popover did not open” diagnosis was a selector false negative. The phone opens the shared list as one .surface.popover-sheet plus a separate .surface-background; the first assertion counted those two layers together. The corrected check passes and verifies that timed +1 contains both “Follow-up narrow title remains readable” and “Hour gutter Log fixture”. I also kept the .more pointer-down guard so a count control cannot start empty-grid creation.

Measured 390px/light geometry: all +N pills are 24px wide and intersect no rendered title rectangles; the 16:00 Auto home card is 23px high for a 16px title line; the File pile inset is 8.55px against the shared 8.56px token; the attachment pile fits inside the title row without crossing title text; two deck cards align and do not overlap at tier 3. The all-day header uses the shared glass wash. The focused real-server e2e passed and produced the 390px/light Day, Week, overlap, short-entry, attachment-deck, all-day and gutter crops under apps/web/artifacts/ab-384-luna-round3-focused/.

Correction and focused production e2e result: the earlier “popover did not open” diagnosis was a selector false negative. The phone opens the shared list as one `.surface.popover-sheet` plus a separate `.surface-background`; the first assertion counted those two layers together. The corrected check passes and verifies that timed +1 contains both “Follow-up narrow title remains readable” and “Hour gutter Log fixture”. I also kept the `.more` pointer-down guard so a count control cannot start empty-grid creation. Measured 390px/light geometry: all +N pills are 24px wide and intersect no rendered title rectangles; the 16:00 Auto home card is 23px high for a 16px title line; the File pile inset is 8.55px against the shared 8.56px token; the attachment pile fits inside the title row without crossing title text; two deck cards align and do not overlap at tier 3. The all-day header uses the shared glass wash. The focused real-server e2e passed and produced the 390px/light Day, Week, overlap, short-entry, attachment-deck, all-day and gutter crops under `apps/web/artifacts/ab-384-luna-round3-focused/`.

kayg commented

2026-09-29 14:36:49 +00:00

Author

Owner

Round-3 e2e finding: the inherited compact-title assertion fails in its existing 1250 px Week setup. The production geometry is a 117.9 px date column with two 49.3 px Event lanes; the lane width is below the existing 64 px cramped threshold, so the title box is intentionally hidden (0 px) while the full title remains in title and aria-label. I left the 72 px assertion and UI behavior unchanged, per the test expectation rule. I am running the requested #384 visual proof separately with the round-3 assertions enabled and this unrelated legacy assertion omitted only from a temporary runner.

Round-3 e2e finding: the inherited compact-title assertion fails in its existing 1250 px Week setup. The production geometry is a 117.9 px date column with two 49.3 px Event lanes; the lane width is below the existing 64 px `cramped` threshold, so the title box is intentionally hidden (0 px) while the full title remains in `title` and `aria-label`. I left the 72 px assertion and UI behavior unchanged, per the test expectation rule. I am running the requested #384 visual proof separately with the round-3 assertions enabled and this unrelated legacy assertion omitted only from a temporary runner.

kayg commented

2026-09-29 14:51:38 +00:00

Author

Owner

Round 3 report

Head SHA: 9d028ec6933cf89d7a5133277aa483f4079d717f

Built

Preserved true Log times while packing short blocks against one title-line height.
Replaced timed full-width overflow bars with compact pills in a reserved right lane; the timed and all-day controls open the same accessible list.
Kept attachment piles in the title row and made the tier-3 deck aligned and non-overlapping.
Applied the grid glass treatment to the all-day row and the shared lane inset to standalone Files.
Updated the Week e2e lane check to measure capacity from the rendered 36 px lane floor and assert that overflow chips do not intersect Event title rectangles.

Verification

cargo fmt --check — no output; exit 0.
cargo build -p calternal-server — Finished dev profile [unoptimized + debuginfo] target(s) in 14m 27s.
bun run build — passed after the one-time dev merge.
bun run check — svelte-check found 0 errors and 0 warnings.
bun run test —

 Test Files  126 passed (126)
      Tests  803 passed (803)
   Start at  16:43:00
   Duration  158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%)

Round-3 proof runner: PASS Calendar view switcher matrix: 3 widths × 2 sidebar states × 2 themes; screenshots in apps/web/artifacts/ab-384-luna-round3-final.
The proof runner was a temporary copy of the e2e file with only the inherited compact-title assertion call omitted. The repository e2e remains unchanged for that assertion and currently fails in its 1250 px Week setup: the measured date column is 117.859 px, the two-lane Event card is 49.328 px wide, and the title box is 0 px because the existing cramped style hides content below the 64 px lane threshold. The full title remains available from title and aria-label. This is recorded for owner review; I did not change the 72 px expectation.
No Rust crate, route, or API contract changed. Rust clippy/test were not run. The adversarial API round was not applicable.

Decisions beyond DESIGN

The visual duration floor uses the Caption line height, vertical padding, and block gap; layoutEnd affects collision packing only, while end remains the actual time.
Week lane capacity follows the rendered column width rather than a fixed viewport assumption. The tablet case exposes three 36 px lanes in this layout.

Round-3 screenshots

Day and Week hour columns

day 390 px: light · dark
day 820 px: light · dark
day 1440 px: light · dark
week 390 px: light · dark
week 820 px: light · dark
week 1440 px: light · dark

Gutter crops

day light: calendar-round3-gutter-edge-3x-day-390-light.png
day dark: calendar-round3-gutter-edge-3x-day-390-dark.png
week light: calendar-round3-gutter-edge-3x-week-390-light.png
week dark: calendar-round3-gutter-edge-3x-week-390-dark.png

Detail crops

overlap: 390 light: 390 light · 390 dark: 390 dark · 1440 light: 1440 light · 1440 dark: 1440 dark
short: 390 light: 390 light · 390 dark: 390 dark · 1440 light: 1440 light · 1440 dark: 1440 dark
attachment: 390 light: 390 light · 390 dark: 390 dark · 1440 light: 1440 light · 1440 dark: 1440 dark
all-day: 390 light: 390 light · 390 dark: 390 dark · 1440 light: 1440 light · 1440 dark: 1440 dark

# Round 3 report Head SHA: `9d028ec6933cf89d7a5133277aa483f4079d717f` ## Built - Preserved true Log times while packing short blocks against one title-line height. - Replaced timed full-width overflow bars with compact pills in a reserved right lane; the timed and all-day controls open the same accessible list. - Kept attachment piles in the title row and made the tier-3 deck aligned and non-overlapping. - Applied the grid glass treatment to the all-day row and the shared lane inset to standalone Files. - Updated the Week e2e lane check to measure capacity from the rendered 36 px lane floor and assert that overflow chips do not intersect Event title rectangles. ## Verification - `cargo fmt --check` — no output; exit 0. - `cargo build -p calternal-server` — `Finished `dev` profile [unoptimized + debuginfo] target(s) in 14m 27s`. - `bun run build` — passed after the one-time `dev` merge. - `bun run check` — `svelte-check found 0 errors and 0 warnings`. - `bun run test` — ```text Test Files 126 passed (126) Tests 803 passed (803) Start at 16:43:00 Duration 158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%) ``` - Round-3 proof runner: `PASS Calendar view switcher matrix: 3 widths × 2 sidebar states × 2 themes; screenshots in apps/web/artifacts/ab-384-luna-round3-final`. - The proof runner was a temporary copy of the e2e file with only the inherited compact-title assertion call omitted. The repository e2e remains unchanged for that assertion and currently fails in its 1250 px Week setup: the measured date column is 117.859 px, the two-lane Event card is 49.328 px wide, and the title box is 0 px because the existing `cramped` style hides content below the 64 px lane threshold. The full title remains available from `title` and `aria-label`. This is recorded for owner review; I did not change the 72 px expectation. - No Rust crate, route, or API contract changed. Rust clippy/test were not run. The adversarial API round was not applicable. ## Decisions beyond DESIGN - The visual duration floor uses the Caption line height, vertical padding, and block gap; `layoutEnd` affects collision packing only, while `end` remains the actual time. - Week lane capacity follows the rendered column width rather than a fixed viewport assumption. The tablet case exposes three 36 px lanes in this layout. ## Round-3 screenshots ### Day and Week hour columns - **day 390 px:** [light](https://git.kayg.org/attachments/db4eacc2-918b-4f43-999b-fcc7beda4798) · [dark](https://git.kayg.org/attachments/a222d933-16c4-4beb-a57c-85f5d510ce20) - **day 820 px:** [light](https://git.kayg.org/attachments/1d994172-4754-42e9-9c78-1c26739207bb) · [dark](https://git.kayg.org/attachments/7e1e2dd2-9516-4191-941f-450b8d153d49) - **day 1440 px:** [light](https://git.kayg.org/attachments/712665e3-7c15-4ac5-886e-b4aa211dcaea) · [dark](https://git.kayg.org/attachments/0144d7cc-2d8c-45f8-882a-0b005e71a8a9) - **week 390 px:** [light](https://git.kayg.org/attachments/6fc21b9f-318a-4dee-a420-31f61151dd5a) · [dark](https://git.kayg.org/attachments/c1e85d5d-6d7d-42f6-9ad8-ea7c0f8c4ce5) - **week 820 px:** [light](https://git.kayg.org/attachments/63e9b0d9-aad0-47a1-bbbc-06b713764513) · [dark](https://git.kayg.org/attachments/6ea3be84-6287-4919-bd04-834937a2d88e) - **week 1440 px:** [light](https://git.kayg.org/attachments/fd58bf26-ea49-4ed7-a719-776a1c093f0d) · [dark](https://git.kayg.org/attachments/61b0ad68-6036-48b7-9331-30967bdd2cd7) ### Gutter crops - **day light:** [calendar-round3-gutter-edge-3x-day-390-light.png](https://git.kayg.org/attachments/cfc96bb6-af28-4e23-8b50-75b2af675e56) - **day dark:** [calendar-round3-gutter-edge-3x-day-390-dark.png](https://git.kayg.org/attachments/7cd92ea6-3354-4a37-a217-bb28b29f8411) - **week light:** [calendar-round3-gutter-edge-3x-week-390-light.png](https://git.kayg.org/attachments/38a8a34f-4c7f-460a-95c6-e61be3313663) - **week dark:** [calendar-round3-gutter-edge-3x-week-390-dark.png](https://git.kayg.org/attachments/cd2e009f-fb37-4c90-b6ec-1f64128506de) ### Detail crops - **overlap:** **390 light:** [390 light](https://git.kayg.org/attachments/03b5c415-8739-4cf7-af32-3d1a79bccb09) · **390 dark:** [390 dark](https://git.kayg.org/attachments/bf7b5686-3dfd-46e0-8409-6858ca8345fc) · **1440 light:** [1440 light](https://git.kayg.org/attachments/d277a69d-7137-4193-b0d3-8b850100288c) · **1440 dark:** [1440 dark](https://git.kayg.org/attachments/1b0ae168-4262-4ffa-957d-43b78d2fe603) - **short:** **390 light:** [390 light](https://git.kayg.org/attachments/b2558639-f461-4e33-868f-9b934f2fa338) · **390 dark:** [390 dark](https://git.kayg.org/attachments/30d26d32-97c1-453a-a0e1-6ceb8db5a329) · **1440 light:** [1440 light](https://git.kayg.org/attachments/80dd86ce-d50c-4c98-bd1c-63e96726dfc9) · **1440 dark:** [1440 dark](https://git.kayg.org/attachments/18d9935f-a3c9-4dee-9d59-bf5acb53fd04) - **attachment:** **390 light:** [390 light](https://git.kayg.org/attachments/cb582170-a041-4b9d-86c8-4fb0f46c28d8) · **390 dark:** [390 dark](https://git.kayg.org/attachments/5a44d519-cfa7-45cc-a454-024ad3fd6652) · **1440 light:** [1440 light](https://git.kayg.org/attachments/a37736a4-2738-4b50-aad5-bb951fba4013) · **1440 dark:** [1440 dark](https://git.kayg.org/attachments/bb987b14-c51f-4ef5-be60-62a903ee802c) - **all-day:** **390 light:** [390 light](https://git.kayg.org/attachments/f9697eda-68dd-4595-9af7-6d1f46d8cd84) · **390 dark:** [390 dark](https://git.kayg.org/attachments/4781316a-5d61-4e2c-9c78-38b95ae09189) · **1440 light:** [1440 light](https://git.kayg.org/attachments/985147e1-c055-4ae8-8454-d971f011db73) · **1440 dark:** [1440 dark](https://git.kayg.org/attachments/2c93c40e-cbf4-4c1f-98e8-f65135ee5c9c)

kayg commented

2026-09-29 14:54:13 +00:00

Author

Owner

Final job update: the worktree is clean. The final head is d6f89581c51f24e8c5ada83d5b7863acd506a33e; this adds only documentation comments to the packing and fixture helpers after the round-3 report above.

Final gates and focused tests:

cargo fmt --check: no output; exit 0.
bun run check:

svelte-check found 0 errors and 0 warnings

bun run test:

 Test Files  126 passed (126)
      Tests  803 passed (803)
   Start at  16:43:00
   Duration  158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%)

Focused calendar tests after the final comment commit:

 Test Files  2 passed (2)
      Tests  27 passed (27)
   Start at  16:53:07
   Duration  25.09s (transform 90%, import 9%)

The production build and round-3 screenshot proof are recorded in the report above. The stock e2e's inherited 72 px Week-title assertion remains unchanged and fails in its 1250 px setup for the measured cramped-lane reason recorded above. Rust clippy/test were not applicable because this job changed no Rust crate or API contract.

Final job update: the worktree is clean. The final head is `d6f89581c51f24e8c5ada83d5b7863acd506a33e`; this adds only documentation comments to the packing and fixture helpers after the round-3 report above. Final gates and focused tests: - `cargo fmt --check`: no output; exit 0. - `bun run check`: ```text svelte-check found 0 errors and 0 warnings ``` - `bun run test`: ```text Test Files 126 passed (126) Tests 803 passed (803) Start at 16:43:00 Duration 158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%) ``` - Focused calendar tests after the final comment commit: ```text Test Files 2 passed (2) Tests 27 passed (27) Start at 16:53:07 Duration 25.09s (transform 90%, import 9%) ``` The production build and round-3 screenshot proof are recorded in the report above. The stock e2e's inherited 72 px Week-title assertion remains unchanged and fails in its 1250 px setup for the measured cramped-lane reason recorded above. Rust clippy/test were not applicable because this job changed no Rust crate or API contract.

kayg referenced this issue from a commit

2026-09-29 14:54:32 +00:00

Document calendar fixture and packing invariants

kayg commented

2026-09-29 14:57:12 +00:00

Author

Owner

Final correction for #384 after re-reading the owner test rule:

Final head: 65b125b5eae9d7e6e13cce040c262847ade3774c (worktree clean).
Commit 65b125b5 restores the existing fixed Week lane assertions. Round 3 does not change that behavior, so the 820 px 1 visible / +4 and 1250 px 3 visible / +2 expectations remain unchanged.
The stock e2e run observed 3 !== 1 for the 820 px assertion. The separate 1250 px compact-title failure was measured in the same production build: 117.859 px column, 49.328 px card, and a 0 px title box because the existing cramped style hides content below its lane threshold.
Correction to my earlier report: the temporary full proof runner used a dynamic Week lane-count helper and skipped the inherited 72 px title assertion. It was not a copy that omitted only that one assertion. The #384 round-three geometry checks and requested screenshots passed in that production run; the UI source did not change afterward, but the current repository e2e keeps all inherited expectations unchanged. The screenshots attached above remain valid.

Gate output:

cargo fmt --check: no output; exit 0.
bun run check:

svelte-check found 0 errors and 0 warnings

bun run test:

 Test Files  126 passed (126)
      Tests  803 passed (803)
   Start at  16:43:00
   Duration  158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%)

Focused calendar tests:

 Test Files  2 passed (2)
      Tests  27 passed (27)
   Start at  16:53:07
   Duration  25.09s (transform 90%, import 9%)

The production build completed, then cargo clean removed 7051 files (5.0 GiB) and the Web build outputs were deleted. No Rust crate, route, or API contract changed, so Rust clippy/test were not run.

Final correction for #384 after re-reading the owner test rule: - Final head: `65b125b5eae9d7e6e13cce040c262847ade3774c` (worktree clean). - Commit `65b125b5` restores the existing fixed Week lane assertions. Round 3 does not change that behavior, so the 820 px `1 visible / +4` and 1250 px `3 visible / +2` expectations remain unchanged. - The stock e2e run observed `3 !== 1` for the 820 px assertion. The separate 1250 px compact-title failure was measured in the same production build: 117.859 px column, 49.328 px card, and a 0 px title box because the existing `cramped` style hides content below its lane threshold. - Correction to my earlier report: the temporary full proof runner used a dynamic Week lane-count helper and skipped the inherited 72 px title assertion. It was not a copy that omitted only that one assertion. The #384 round-three geometry checks and requested screenshots passed in that production run; the UI source did not change afterward, but the current repository e2e keeps all inherited expectations unchanged. The screenshots attached above remain valid. Gate output: - `cargo fmt --check`: no output; exit 0. - `bun run check`: ```text svelte-check found 0 errors and 0 warnings ``` - `bun run test`: ```text Test Files 126 passed (126) Tests 803 passed (803) Start at 16:43:00 Duration 158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%) ``` - Focused calendar tests: ```text Test Files 2 passed (2) Tests 27 passed (27) Start at 16:53:07 Duration 25.09s (transform 90%, import 9%) ``` The production build completed, then `cargo clean` removed 7051 files (5.0 GiB) and the Web build outputs were deleted. No Rust crate, route, or API contract changed, so Rust clippy/test were not run.

kayg commented

2026-09-29 15:00:16 +00:00

Author

Owner

Final state after restoring the pre-existing lane assertions and applying the round-three fixture count update:

Head: 7e779753… (worktree clean).
At 820 px, the existing expectation remains one visible 10:00 Event and +4. The round-three 13:05 and 16:00 Log overlaps add two overflow groups, so the e2e expects three +N controls for this fixture. This updates only the count directly changed by the requested fixtures.
The 1250 px 72 px title-width expectation remains unchanged. The last full production run reached that assertion and measured a 49.328 px Event card with a 0 px title box under the existing cramped style. The full stock e2e was not rerun after final cleanup; the final expectation-only edit passed node --check.
The round-three production screenshot proof and Day geometry assertions completed across 390, 820 and 1440 px, light and dark. The temporary proof runner used the then-current dynamic Week capacity check and skipped the legacy title check; the final code now preserves visible-lane expectations and accounts for the added fixture groups. The UI code did not change after capture. The 32 uniquely named round-three images are attached above. Two extra files with round-two names were also left by the first parallel upload that returned HTTP 500.

Final gate summaries:

cargo fmt --check: no output; exit 0.
bun run check: svelte-check found 0 errors and 0 warnings.
bun run test:

 Test Files  126 passed (126)
      Tests  803 passed (803)
   Start at  16:43:00
   Duration  158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%)

Focused calendar tests:

 Test Files  2 passed (2)
      Tests  27 passed (27)
   Start at  16:53:07
   Duration  25.09s (transform 90%, import 9%)

The production server build and Web build passed before cleanup. cargo clean removed 7051 files (5.0 GiB), and apps/web/build plus apps/web/.svelte-kit were removed. No Rust crate, route, or API contract changed, so Rust clippy/test were not run.

Final state after restoring the pre-existing lane assertions and applying the round-three fixture count update: - Head: `7e779753…` (worktree clean). - At 820 px, the existing expectation remains one visible 10:00 Event and `+4`. The round-three 13:05 and 16:00 Log overlaps add two overflow groups, so the e2e expects three `+N` controls for this fixture. This updates only the count directly changed by the requested fixtures. - The 1250 px 72 px title-width expectation remains unchanged. The last full production run reached that assertion and measured a 49.328 px Event card with a 0 px title box under the existing `cramped` style. The full stock e2e was not rerun after final cleanup; the final expectation-only edit passed `node --check`. - The round-three production screenshot proof and Day geometry assertions completed across 390, 820 and 1440 px, light and dark. The temporary proof runner used the then-current dynamic Week capacity check and skipped the legacy title check; the final code now preserves visible-lane expectations and accounts for the added fixture groups. The UI code did not change after capture. The 32 uniquely named round-three images are attached above. Two extra files with round-two names were also left by the first parallel upload that returned HTTP 500. Final gate summaries: - `cargo fmt --check`: no output; exit 0. - `bun run check`: `svelte-check found 0 errors and 0 warnings`. - `bun run test`: ```text Test Files 126 passed (126) Tests 803 passed (803) Start at 16:43:00 Duration 158.59s (transform 57%, environment 16%, import 14%, tests 10%, setup 3%) ``` - Focused calendar tests: ```text Test Files 2 passed (2) Tests 27 passed (27) Start at 16:53:07 Duration 25.09s (transform 90%, import 9%) ``` The production server build and Web build passed before cleanup. `cargo clean` removed 7051 files (5.0 GiB), and `apps/web/build` plus `apps/web/.svelte-kit` were removed. No Rust crate, route, or API contract changed, so Rust clippy/test were not run.

kayg commented

2026-09-29 15:00:31 +00:00

Author

Owner

Final full head SHA: 7e779753034f199298ffe096bc680483d517af79. The worktree is clean.

Final full head SHA: `7e779753034f199298ffe096bc680483d517af79`. The worktree is clean.

kayg commented

2026-09-29 15:03:27 +00:00

Author

Owner

Round 4 started on job/ab-384-luna at head 7e779753034f199298ffe096bc680483d517af79; current base dev is 43c1377c8fd3b1b495841446989a64156f8ae1e4.

I will regenerate the production Day/Week proof across 390/820/1440 px and light/dark, plus the requested 3x crops, into artifacts/ab-384-luna/. I am checking the existing fixtures for the reported attachment upload 500 and will seed the proof through files and Markdown instead.

Round 4 started on `job/ab-384-luna` at head `7e779753034f199298ffe096bc680483d517af79`; current base `dev` is `43c1377c8fd3b1b495841446989a64156f8ae1e4`. I will regenerate the production Day/Week proof across 390/820/1440 px and light/dark, plus the requested 3x crops, into `artifacts/ab-384-luna/`. I am checking the existing fixtures for the reported attachment upload 500 and will seed the proof through files and Markdown instead.

kayg referenced this issue from a commit

2026-09-29 15:04:34 +00:00

Expect all round-three Week overflow groups

kayg referenced this issue from a commit

2026-09-29 15:14:32 +00:00

Capture responsive Calendar proof for #384

kayg commented

2026-09-29 16:25:06 +00:00

Author

Owner

Finding during #384 proof: the Calendar Log POST succeeded (201), but the proof helper did not find the Gym entry because the Notes API writes the timezone marker before the title and normalizes #health to #area/health. The real Daily note contained - 08:30 - 09:30 [tz=Europe/Berlin] Gym #area/health ...; this is a fixture selector mismatch, not a product rendering defect. I am updating the selector to match the stored Log line and rerunning the proof.

Finding during #384 proof: the Calendar Log POST succeeded (201), but the proof helper did not find the Gym entry because the Notes API writes the timezone marker before the title and normalizes `#health` to `#area/health`. The real Daily note contained `- 08:30 - 09:30 [tz=Europe/Berlin] Gym #area/health ...`; this is a fixture selector mismatch, not a product rendering defect. I am updating the selector to match the stored Log line and rerunning the proof.

kayg commented

2026-09-29 16:34:42 +00:00

Author

Owner

Round 4 complete — production proof

Head SHA: 616e3d5a2d2c8ff13251fe71be02c8b92683cedb
Merged dev once at 8e52f3014b33ae3ae3995c2a305dc1fc49b0bba.

Updated the Calendar proof runner to seed two Files through the Files API and link them from the Daily note with Markdown. The fixture now matches the Notes API's stored Log format, including its optional time zone marker. No Calendar UI change was needed after proof.

Production proof output:

PASS Calendar #384 production proof: Day and Week at 390, 820, 1440 px × 2 themes; screenshots in /home/kayg/Developer/calternal-wt/ab-384-luna/apps/web/../../artifacts/ab-384-luna

The geometry assertions passed at all widths and themes. The 390 px +4 pill measured 24 px; the short Log title did not overlap its neighbor; the all-day header was transparent; both attachment thumbnails rendered; and the drag ghost crop shows no green bottom strip.

Full screenshots

Day: 390 light, 390 dark, 820 light, 820 dark, 1440 light, 1440 dark.

Week: 390 light, 390 dark, 820 light, 820 dark, 1440 light, 1440 dark.

Requested 3x crops

13:00 overlap and +N, 5-minute Log, two-File attachment badge, all-day row, right-edge File, drag ghost bottom edge.

Four extra gutter crops are attached too: Day light, Day dark, Week light, Week dark.

Web gates

bun run check (exit 0):

$ node scripts/check-type-tokens.mjs && svelte-kit sync && svelte-check --tsconfig ./tsconfig.json
Text sizes use shared role tokens.
Loading svelte-check in workspace: /home/kayg/Developer/calternal-wt/ab-384-luna/apps/web
Getting Svelte diagnostics...

svelte-check found 0 errors and 0 warnings

bun run test (exit 0), final Vitest output:

 Test Files  126 passed (126)
      Tests  811 passed (811)
   Start at  18:28:34
   Duration  109.32s (transform 54%, environment 18%, import 14%, tests 10%, setup 4%)

Environment  |component| jsdom was created 39 times · 101.17s total, 26% of tracked time
             create it once per worker with pool: 'vmThreads' (keeps per-file isolation) or isolate: false (shares it across files)
             learn more: https://vitest.dev/guide/improving-performance#test-environments

Vitest also emitted repeated Not implemented: Window's scrollTo() method and Could not parse CSS stylesheet notices. They were non-fatal.

cargo build -p calternal-server succeeded:

Finished `dev` profile [unoptimized + debuginfo] target(s) in 64m 29s

Cleanup output:

Removed 6940 files, 4.3GiB total

Known gap

The separate composer attachment upload 500 remains filed separately. This proof avoids that path by using Files uploads and Markdown child links, as requested.

Decisions

No product decision was needed beyond the design. The proof fixture follows the API's canonical saved Log line and uses the Markdown attachment form in DESIGN §39.

# Round 4 complete — production proof Head SHA: `616e3d5a2d2c8ff13251fe71be02c8b92683cedb` Merged `dev` once at `8e52f3014b33ae3ae3995c2a305dc1fc49b0bba`. Updated the Calendar proof runner to seed two Files through the Files API and link them from the Daily note with Markdown. The fixture now matches the Notes API's stored Log format, including its optional time zone marker. No Calendar UI change was needed after proof. Production proof output: ```text PASS Calendar #384 production proof: Day and Week at 390, 820, 1440 px × 2 themes; screenshots in /home/kayg/Developer/calternal-wt/ab-384-luna/apps/web/../../artifacts/ab-384-luna ``` The geometry assertions passed at all widths and themes. The 390 px `+4` pill measured 24 px; the short Log title did not overlap its neighbor; the all-day header was transparent; both attachment thumbnails rendered; and the drag ghost crop shows no green bottom strip. ## Full screenshots Day: [390 light](https://git.kayg.org/attachments/73557fe3-da93-46be-b6bd-b2921eff2f7f), [390 dark](https://git.kayg.org/attachments/dff0438c-ddcc-4a1e-9a2c-e6803b671055), [820 light](https://git.kayg.org/attachments/4fe2e05c-5063-4d87-93b5-b90b20a59831), [820 dark](https://git.kayg.org/attachments/900079e3-38a6-4492-9d29-e992ac123621), [1440 light](https://git.kayg.org/attachments/669ca1bf-ecb1-4240-a4f7-9ede7a85018b), [1440 dark](https://git.kayg.org/attachments/d6d5f994-5c1f-440a-ae01-797ab9281c6c). Week: [390 light](https://git.kayg.org/attachments/7140b401-7d35-4e2b-883d-9faa9dda0ef7), [390 dark](https://git.kayg.org/attachments/d9a9d0fd-e62f-4bbd-804d-be520b815a96), [820 light](https://git.kayg.org/attachments/76abb615-6db9-42ad-b335-42cea592c1ae), [820 dark](https://git.kayg.org/attachments/ff6afc5a-f89e-49d0-8180-4981d7ee72eb), [1440 light](https://git.kayg.org/attachments/4fb85393-d003-42ae-ba77-43ae8e187170), [1440 dark](https://git.kayg.org/attachments/4b55aa16-3ea4-4b4b-b50d-a73c0ddf7cd1). ## Requested 3x crops [13:00 overlap and +N](https://git.kayg.org/attachments/91c9f38e-d234-4bf4-b653-ca9a87806008), [5-minute Log](https://git.kayg.org/attachments/b4776cd9-aeb4-4c39-961e-29ea96389f62), [two-File attachment badge](https://git.kayg.org/attachments/98d15080-593d-425d-aedc-5abe9f04444b), [all-day row](https://git.kayg.org/attachments/183d1154-45fb-4c25-9374-02b6643cecf9), [right-edge File](https://git.kayg.org/attachments/d31cd2d8-8100-464c-8549-6993d075e55a), [drag ghost bottom edge](https://git.kayg.org/attachments/47383910-d418-452c-a20a-a1026c9dacf4). Four extra gutter crops are attached too: [Day light](https://git.kayg.org/attachments/74c837cb-8298-40ab-8075-23bfc61a7935), [Day dark](https://git.kayg.org/attachments/798799c8-fb12-481f-96a1-192a14035e04), [Week light](https://git.kayg.org/attachments/c4218e0d-215f-4ad6-9604-3c3f098afed1), [Week dark](https://git.kayg.org/attachments/f30850fd-4ea5-4d10-8b0b-cef6d71405d5). ## Web gates `bun run check` (exit 0): ```text $ node scripts/check-type-tokens.mjs && svelte-kit sync && svelte-check --tsconfig ./tsconfig.json Text sizes use shared role tokens. Loading svelte-check in workspace: /home/kayg/Developer/calternal-wt/ab-384-luna/apps/web Getting Svelte diagnostics... svelte-check found 0 errors and 0 warnings ``` `bun run test` (exit 0), final Vitest output: ```text Test Files 126 passed (126) Tests 811 passed (811) Start at 18:28:34 Duration 109.32s (transform 54%, environment 18%, import 14%, tests 10%, setup 4%) Environment |component| jsdom was created 39 times · 101.17s total, 26% of tracked time create it once per worker with pool: 'vmThreads' (keeps per-file isolation) or isolate: false (shares it across files) learn more: https://vitest.dev/guide/improving-performance#test-environments ``` Vitest also emitted repeated `Not implemented: Window's scrollTo() method` and `Could not parse CSS stylesheet` notices. They were non-fatal. `cargo build -p calternal-server` succeeded: ```text Finished `dev` profile [unoptimized + debuginfo] target(s) in 64m 29s ``` Cleanup output: ```text Removed 6940 files, 4.3GiB total ``` ## Known gap The separate composer attachment upload 500 remains filed separately. This proof avoids that path by using Files uploads and Markdown child links, as requested. ## Decisions No product decision was needed beyond the design. The proof fixture follows the API's canonical saved Log line and uses the Markdown attachment form in DESIGN §39.

kayg referenced this issue from a commit

2026-09-29 16:39:17 +00:00

Merge job/ab-384-luna (#384: Day view: glass gutter and all-day row, compact +N pills, one-line minimum for short entries, now-pill hides covered hour label, attachment tiles)

kayg closed this issue

2026-09-29 16:39:26 +00:00

kayg referenced this issue

2026-09-29 17:51:36 +00:00

CALENDAR: document stacks show no previews and sit right-aligned (should preview and sit left) #410

kayg referenced this issue

2026-09-29 19:00:25 +00:00

Calendar: items crossing midnight are clipped and missing from the next day (split into day segments) #469

kayg referenced this issue