Skip to content

Lift the answer-quality popover above the follow-up cards; redraw the quality icons - #88

Merged
jaketracey merged 1 commit into
mainfrom
fix/answer-quality-popover-stacking
Sep 1, 2026
Merged

Lift the answer-quality popover above the follow-up cards; redraw the quality icons#88
jaketracey merged 1 commit into
mainfrom
fix/answer-quality-popover-stacking

Conversation

@jaketracey

@jaketracey jaketracey commented Sep 1, 2026

Copy link
Copy Markdown
Contributor

What

  • Stacking: the answer-quality dropdown was painted under the "Ask next" follow-up cards (see Jake's screenshot). Every late-arriving block fades up with rp-answer-tail, whose entrance animation used animation-fill-mode: both. A filled-forward opacity/transform interpolation leaves each block a permanent stacking context, so the z-30 popover inside the actions row could never rise above the sibling blocks that follow it. The keyframes only declare from, so backwards ends on the same pixels and releases the context once the entrance is over. Same one-word change to rp-answer-in (the answer body) for the same reason.
  • Icons: the four confidence glyphs move to the 24-unit grid and 1.7 stroke of the neighbouring answer-action icons, with the stroke thinning as the box grows so every size lands at the same optical weight. The exclamation dots are zero-length round-capped strokes instead of 0.15-radius filled circles, which rendered below a pixel and left the triangle reading as a bare outline.

Verified

  • deno task check green (typecheck, lint, fmt, 215 tests).
  • Chrome, local build, real GRDC answer with a Low confidence score and three follow-ups:
    • 1280px layout viewport: dropdown hit-tests on top of all three follow-up cards; the actions row reports transform: none after its entrance. Forcing animation-fill-mode: both back on reproduces the trapped panel (row keeps a transform, the block below hit-tests on top).
    • True 390px layout viewport (iframe probe, header relaxed locally and reverted): bottom sheet opens at z-90, scrollWidth equals innerWidth.
    • Both on the default palette and the Observatory dark palette.
  • Isolated probe page confirms the mechanism in Chrome: both lets a later sibling paint over a z-30 child popover; backwards does not.

🤖 Generated with Claude Code

https://claude.ai/code/session_01GYuuU1yPFeGKZpkwW2mLs6


View with [code]smith Autofix with [code]smith
Need help on this PR? Tag @codesmith-bot with what you need. Autofix is disabled.

… quality icons

The dropdown that opens from the answer-quality trigger was painted under
the "Ask next" cards that arrive after it, whatever z-index it carried.
Every late-arriving block fades up with `rp-answer-tail`, whose animation
filled `both`; a filled-forward opacity/transform interpolation leaves the
block a permanent stacking context, so the popover inside the actions row
could never rise above the sibling blocks below it. The keyframes only
declare `from`, so `backwards` ends on identical pixels and releases the
context once the entrance is over. Same change to `rp-answer-in`, the
answer body, for the same reason.

The confidence icons move to the 24-unit grid and 1.7 stroke of the
answer-action icons beside them, with the stroke thinning as the box grows
so each size lands at the same optical weight. The exclamation dots are
zero-length round-capped strokes rather than 0.15-radius filled circles,
which rendered below a pixel and left the warning triangle reading as a
bare outline.

Verified in Chrome on a local build: dropdown hit-tested on top of all
three follow-up cards at 1280px, bottom sheet at a true 390px layout
viewport with no horizontal overflow, both on the default and Observatory
(dark) palettes; forcing the old fill mode back on reproduces the trapped
panel.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GYuuU1yPFeGKZpkwW2mLs6
@jaketracey
jaketracey merged commit 70d6505 into main Sep 1, 2026
2 checks passed
@jaketracey
jaketracey deleted the fix/answer-quality-popover-stacking branch September 1, 2026 21:38
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.

1 participant