solidjs / solid / 41190189489

78%

main: 89%

push

github

fix(signals): keep an error boundary's fallback until a held list recovery's frame reveals (#4526) (#4532)

* fix(signals): keep an error boundary's fallback until a held list recovery's frame reveals (#4526)

When the error's replacement landed held with an action (a guess on the
errored store re-asked it), the list's re-map was a parked pass that
disposed the old rows' effects, readers the boundary had caught the error
on. The in-heap clear of the list's effect saw those dead readers still in
the set and skipped the redraw that joins the hold; the seam then dropped
them and revealed outside the hold, showing the pre-error rows. Judge the
last reader as the seam does (`prune`), so the reveal joins the hold.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015hPxMkC3jg3Uivh65gvaf2

* perf(signals): keep an error boundary's recovery linear in its readers (#4526)

The #4526 fix asked `prune` on every reader clear, and `prune` walked every
reader still waiting: a recovery of N readers was N²/2 steps (8000 render
effects under <Errored>: ~820 ms vs ~40 ms). `catchStatus` now asks with
`pass` 0, which ends the walk at the first reader kept by its status — one
is left, so this clear is not the last — and leaves the readers behind it
to a later prune. The seam, `ready` and the output's pass walk as before.

A deterministic pin counts the Map iteration steps of a 1000-reader
recovery (linear: ~N; quadratic: ~N²/2).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015hPxMkC3jg3Uivh65gvaf2

---------

Co-authored-by: Claude <noreply@anthropic.com>

797 of 1098 branches covered (72.59%)

Branch coverage included in aggregate %.

1129 of 1367 relevant lines covered (82.59%)

30.08 hits per line