Skip to content

fix(meeting-recap): keep the document under budget at every width, not three - #28

Merged
scheemunai merged 1 commit into
mainfrom
fix/meeting-recap-mid-width
Sep 5, 2026
Merged

fix(meeting-recap): keep the document under budget at every width, not three#28
scheemunai merged 1 commit into
mainfrom
fix/meeting-recap-mid-width

Conversation

@scheemunai

Copy link
Copy Markdown
Owner

The defect

meeting-recap exceeded the 3,000px fail line at widths the contract never sampled.

BEFORE   peak 3,385px @ 320
         over 2,900 in two bands:  320-372   and   545-615

The contract samples 390, 768 and 1440 — and all three sit inside a compact layout. 390 is inside the card collapse, 768 is past the wide breakpoint, and nobody looked at the band between. The 545-615 band is the collapse ending at 34rem (544px) while the action table still renders tall until ~640.

The fix

  • Collapse breakpoint 34rem -> 40rem so the action table stays in card form until it has the width to render a table compactly. That closes 545-615.
  • A 24rem band for 320-384, which was over budget even with the collapse on, because every line wraps more. It spends width and vertical rhythm, never content: the decision numeral gutter is dropped so the text gets the full measure, and the seven collapsed table rows — every one paying the same padding — lose spacing.

The only type change is the h1 at <=24rem, from its 1.75rem clamp floor to 1.5rem (24px), still far above every floor in the contract (13px prose / 12px metadata / 11px labels). Nothing structural is removed.

Verification — a sweep, not three samples

Stepped the width and asserted window.innerWidth before each measurement; the harness refuses rather than records if the viewport does not read back.

320-700  step 1  (381 widths)   PEAK 2,820 @ 320   over 2,900: none   overflow>0: none
700-1920 step 2  (611 widths)   PEAK 2,328 @ 704   over 2,900: none   overflow>0: none
width before after
peak 3,385 @ 320 2,820 @ 320
390 2,821 2,774
768 2,082 2,082 (unchanged)
1,440 1,866 1,866 (unchanged)

The two sampled widths that were already compliant did not move.

Other contract checks

bytes             15,082  (cap 18,432)
banned constructs 0 of 10 patterns
overflow-x        0 at every width swept

One note on the diff

Four of my own rules in the new narrow band were larger than the corresponding rule in the wider band — the narrow band adding space back at the width with the least margin. Reading the diff found three; a check asserting narrow <= wide for every shared property found the fourth (.tags{margin-top}). That check reports 0 inversions, 0 duplicate declarations on the committed file.

🤖 Generated with Claude Code

…t three

The contract sampled 390, 768 and 1440. All three sit inside a compact
layout, so two over-budget bands went unmeasured: 320-372 (peak 3,385px)
and 545-615 (peak 3,174px), the second caused by the card collapse ending
at 34rem while the table still renders tall until ~640.

Move the collapse breakpoint to 40rem so the action table stays in card
form until it has the width to render compactly, and add a 24rem band that
spends width and vertical rhythm at the narrowest sizes: the decision
numeral gutter is dropped so the text gets the full measure, and the
collapsed table rows -- seven identical cards, the largest spacing lever
in the document -- lose padding.

Nothing structural is removed. The only type change is the h1 at <=24rem,
from a 1.75rem clamp floor to 1.5rem, which is still far above every floor
in the contract.

Swept 320-700 at step 1 and 700-1920 at step 2, asserting innerWidth before
each measurement:

  peak  3,385px @ 320  ->  2,820px @ 320
  390   2,821       ->  2,774
  768   2,082       ->  2,082   (unchanged)
  1,440 1,866       ->  1,866   (unchanged)
  over 2,900        ->  none at any width
  overflow-x        ->  0 at every width

Co-Authored-By: CRHQ <noreply@crhq.ai>
@scheemunai
scheemunai merged commit 131fb42 into main Sep 5, 2026
5 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants