- Shipped
- September 9, 2026 at 4:09 AM UTC
- Author
- Kamo
- Commit
- 7984561
The reduced-motion walker scanned for the `animation:` shorthand. Writing the same animation as `animationName:` plus longhand properties — the first refactor anybody reaches for — matched nothing, so the element's missing reset went unreported. Found by a concurrent session reaching it on their first attempt at the next animation, knowing the guard existed. Both forms are matched now. Proved by mutation in both directions rather than assumed: rewriting the chip flight as longhand WITH its attribute passes, and removing the attribute fails with the sentence that names the real defect — "runs holdem-chip-fly on an element with no data-holdem-* attribute, so no reduced-motion reset can reach it". The interesting part is what the mutation showed BEFORE the fix. The longhand form did not ship silently: a use the walker cannot see makes its keyframe look unused, so "declares no keyframe that no component runs" failed instead. The animation was caught; only the message was wrong, and it pointed at the opposite problem — dead code rather than an unguarded animation. That backstop is now written down, because it is the thing that makes the remaining hole survivable: factoring the shorthand into a helper still leaves no tag to walk back to, and that case will surface as a keyframe nobody runs. A guard that recognises a FORM is opted out of by an equivalent form, silently and by accident. Worth knowing before writing the next one.