solidjs / solid / 36379141978

74%

main: 89%

push

github

fix(signals): woken transactions do not re-enter over a staged bump; matrix derived source composes truth (F4/F6) (#3684)

* test(signals): matrix derived source composes truth through the functional setter (F4); F7 pinned

The derived source's `commitTruth` read the truth with `untrack(truth)`
and wrote it back plain. A plain read in an action body sees the
COMMITTED truth (A28), and two same-key actions entangle into one
transaction, so a second action's confirm landed its mutation on the
pre-first base — the "stray frame" F4 pinned was the truth the harness
itself committed. The functional setter composes on the held value.

Every F4 row passes with the harness fixed: the F4 pins are removed, the
`F1/F4` index rows are retagged `F1`, and the two derived index rows
(`across two actions: insert tail / delete tail`) that F4 alone covered
pass and are unpinned. The one remaining non-F1 row — chained source,
keyed mapArray, `insert head + delete tail` resolved B then A — is
retagged F7 with the traced mechanism: B's base write commits at once
and the base drops key 5 under A's armed positional overrides, so
`length` (6) and `view[5]` (the override) disagree with `5 in view`
(false) and mapArray's `slice` leaves a hole. Left pinned; needs a rule.

Matrix: 716 pass / 184 expected fail / 0 fail -> 6 F6 rows fail until
the scheduler fix in the next commit; 779 / 121 / 0 with it.
RULES-INDEX regenerated (test citation counts only).

Agent: Claude via Cursor
Co-authored-by: Claude via Cursor <noreply@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>

* fix(signals): a woken transaction is not re-entered over a staged ambient bump (F6)

The flush re-enters a parked transaction that `wakeParked()` queued only
from an idle pass, because entering adopts the ambient batch. The full
pass judged idle by `scheduled` — an empty dirty heap and no active
transition (scheduler.ts ~996) — which missed a write the finalize staged
in the ambient batch with NO subs... (continued)

625 of 914 branches covered (68.38%)

Branch coverage included in aggregate %.

922 of 1183 relevant lines covered (77.94%)

26.93 hits per line