fix(signals): one unflushed — a staging adopted before any flush answers the committed value on every channel (spec O4)

[ryansolid]

Stacked on #3505 (its commit is included; merges cleanly after it). Closes spec O4 — the last pinned violation; the signals suite has no expected fails for the first time.

Bug (posture matrix S1)

Same-tick adoption is by design (O1): set(x, 1); action(...) in one tick makes the action own the write. A28 still governs what is visible before the flush: nothing. Inside the adopting action's body:

	latest(x)	isPending(x)
store leaf	0	false
signal (before)	1	true

The store answered correctly by coincidence (its foreign-hold rule), the signal wrongly: adoption stamped it with the transaction, and unflushedValue read a stamped node with no stash as "flushed, held". One rule, two implementations, two answers — the first concrete target of the consolidation's "one unflushed".

Fix

One definition: a node staged outside a flush and adopted before any flush has carried it is unflushed whatever its stamp.

• initTransition's adoption loop sets CONFIG_ADOPTED_UNFLUSHED when it runs outside a flush (one masked |= in the existing loop).
• unflushedValue serves the committed value for it → latest() answers the pre-write value; flushedStaged answers NOT_PENDING → the verdict sees nothing staged.
• reassignPendingTransition — the carrying flush re-stamping the transaction's pending nodes — clears the bit (one &= in the existing loop); the hold takes over.

Signal and store leaves now answer through the same arm. +56 B minified (25,193 → 25,249).

Two earlier shapes were rejected by the gate: keying on the transaction's _time === clock (re-entry re-stamps _time, so a genuinely held value read from another action read as unflushed — four pins caught it), and a sentinel in the stash plus a helper (+179 B).

Gate

• Matrix: the only changed cells are the staged, ambient state's under the adopting postures — latest/isPending now 0/false, and a memo/effect created there publishes the committed 0 and holds its 1 with the action instead of publishing the held write. 0 invariant violations across 621 cells.
• Fuzzer (test(signals): add experimental semantic fuzzer #3446, same campaign): 994 / 0 fail / 6 policy, unchanged.
• signals 2,742 (no expected fails) / solid 623 / web 825. Four brotli caps +18…+50 B.

[changeset-bot]

🦋 Changeset detected

Latest commit: e739c0b

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 11 packages

Name	Type
@solidjs/signals	Patch
test-integration	Patch
@solidjs/web	Patch
@solidjs/babel-plugin	Patch
@solidjs/compiler	Patch
@solidjs/diagnostics	Patch
@solidjs/element	Patch
@solidjs/h	Patch
@solidjs/html	Patch
solid-js	Patch
@solidjs/universal	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

[coveralls]

Coverage Report for CI Build 35191800471

Warning

No base build found for commit 3da9f43 on next.
Coverage changes can't be calculated without a base build.
If a base build is processing, this comment will update automatically when it completes.

Coverage: 71.46%

Details

• Patch coverage: No coverable lines changed in this PR.

Uncovered Changes

No uncovered changes found.

Coverage Regressions

Requires a base build to compare against. How to fix this →

---

Coverage Stats

Relevant Lines:	1018
Covered Lines:	772
Line Coverage:	75.83%
Relevant Branches:	790
Covered Branches:	520
Branch Coverage:	65.82%
Branches in Coverage %:	Yes
Coverage Strength:	14.94 hits per line

---

💛 - Coveralls

[codspeed]

Merging this PR will not alter performance

✅ 175 untouched benchmarks

---

_{Comparing fix/adopted-unflushed (e739c0b) with next (3da9f43)}