solidjs / solid / 41715245178

78%

main: 89%

push

github

fix(signals): isPending() in a held pass is final there and re-derives at the landing; a first value staged this flush probes as its read (#4648) (#4667)

* fix(signals): an isPending() probe in a held pass is final there and re-derives at the landing; a first value staged this flush probes as its read (#4648)

An untracked isPending() in a pass a hold holds still reads the committed
value (ruling on #4662), but no longer answers true beside the fresh value
the pass's plain read saw (D-23): the pass is the hold's work, shown at the
landing, and a stale reader of what the probe read (A15), so it re-derives
on the landed world (F-1). An async memo is not re-asked (F-9). A probe of
a first value staged this flush reads the staging, as its plain read does
(D-23 exc. 1), instead of undefined. Replaces #4662.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0117BoWNSSgtemsSYsaX2hXE

* fix(signals): a late reader's isPending() probe settles with the hold too (#4648)

Two shapes the first commit left: a pass that first reads the held node in
a later flush, and one that probed only in the flush that staged the write,
before the hold existed. Each read the committed value through its probe;
neither was held yet, so it landed `p: true` beside the fresh value (D-23)
or stayed on `inner: 0` after everything settled (F-1). The probe now
watches the seam: if the flush parks with the pass, it runs again next
round, held, where its probe is final and it re-derives at the landing
(A15). Async memos and first passes are not re-run (F-9). Both read orders
give the same frames; `inner: 0` stays while held.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0117BoWNSSgtemsSYsaX2hXE

* fix(signals): the hold lands in one frame, and an in-flush probe reads the staging (#4648)

A held memo that re-derives at a landing now holds the release of the
hold's stashed runs, as... (continued)

807 of 1108 branches covered (72.83%)

Branch coverage included in aggregate %.

1132 of 1370 relevant lines covered (82.63%)

30.31 hits per line