[ryansolid]

Summary

Builds on #4681 by @GabbeV. His commit (68be911d, unchanged, authored by him) fixes the TypeError that halted reactivity when an untracked staged-store read sat beside latest() on a store with no reactive node. This PR adds one commit on top. It extends that guard so node-less store keys also follow the #4544 ruling, instead of skipping it.

Problem with the guard alone. #4625 added the ruling in RULES-DERIVED D-25 ("an inherited posture is not a probe", #4544): a pass whose verdict posture is only inherited is served the hold's own sync write. INTERNALS-STORE-STATE R2 says stores follow signal rules by default. The guard skips holdStaging whenever stagedScreen gets no node. So on a store key with no node, an inherited-posture pass was served the committed value instead, while the same read on a signal, or on a key with a node, was served the staging. Whether the store had a node, an internal detail, changed what users saw:

• Inherited isPending: at the landing, a render effect saw a torn frame (data landed, count still 0), then the corrected one in the same flush. The signal goes straight to the corrected frame.
• Inherited latest: the pass derived 0/0 mid-hold instead of 0/1, and re-ran an extra time after the landing.

Fix. holdStaging accepts a missing node. With no node it applies only the reader-posture checks: no probe of its own (CONFIG_OWN_PROBE), not a frame, no open window or unanswered probe, and the lane condition. It skips the node-specific checks, because a node-less staging is by construction this flush's own plain sync store write. D-26 atomicity ([latest(x), x()] through a memo) is unaffected: a key that a latest() window asked about always gets a leaf node, so it reads through the node path, where _askedIn still decides. I traced this, and the D-26 case below pins it.

How did you test this change?

• New S9 block in tests/posture-store-parity.test.ts (23 tests). Six sources: signal, untracked signal, tracked store, and untracked store with a container node, with a leaf node only, or with no node. Each runs under an inherited isPending, an inherited latest, and an own-latest posture, and must match an exact trace. Each test also asserts which nodes exist, so the labels can't drift. A D-26 case pins store/signal parity for latest of the same key through a memo.
  • On fix(signals): guard node-less store staging reads #4681's head (guard only), exactly the two node-less cases fail: inherited isPending shows the torn screen 1 1/0, and inherited latest shows the 0/0 pass. All 23 pass with this commit.
• Posture oracle: visibility-oracle-posture.test.ts gains a memoInheritedUntracked reader kind (discovery coverage). Its readers are built after the write rather than re-run in the write's flush, so it does not reproduce this bug: all 160 new cells match memoUntracked, and the report is identical before and after the fix. S9 is the pin.
• Docs: D-25's 2.0.0-rc.20: a memo that calls isPending()/latest() on an unrelated signal makes a branch opened by a held update read the pre-hold world; <Loading> never settles #4544 paragraph records node-less store parity, with S9 as a pin. docs/RULES-INDEX.md is regenerated (rules-index.test.ts passes). The changeset is updated.
• Signals: build:js and types pass. Full suite after the build: 479 files, 6,689 passed, 14 expected failures, 2 skipped. The focused files (S9, nodeless-staged-verdict, verdict-memo-held-branch-4544, latest-memo-inline-parity) also pass under SIGNALS_TIER=prod and observe (102 each).
• Web, against the built signals:
  • Hydrate config (includes test/consistency, c11-trace-equals-oracle): 758 passed, no failures.
  • verdict-memo-held-branch-4544.spec.tsx: passes.
  • Client config: 1,699 passed. The 7 failures are in dist-nested-manifests: web type declarations were not built locally.
  • Server config: 2,006 passed and 1 failed, server-functions-adapter-request. It fails identically without this commit (local Node 22.20).
• Size (Rolldown 1.2.11, signals scenarios vs fix(signals): guard node-less store staging reads #4681's head): +isPending/latest +4 B minified / +7 B brotli (12,061 B vs a 12,070 B cap). The other three scenarios are unchanged. All within caps; no caps changed.
• Prettier, git diff --check, and node scripts/check-release-invariants.mjs pass.

[changeset-bot]

🦋 Changeset detected

Latest commit: c8e6ab2

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

This PR includes changesets to release 15 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
board-example	Patch
canvas-example	Patch
server-dbmon-example	Patch
todos-server-example	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

[codspeed]

Merging this PR will not alter performance

1 warning

⚠️ 1 benchmark is dominated by system calls

System calls cannot be simulated consistently and are left out of the measure. The reported value understates the real cost. Switch to the Walltime instrument to measure them.

Learn more about system calls

See affected benchmarks

✅ 201 untouched benchmarks
⏩ 5 skipped benchmarks¹

---

_{Comparing devagrawal09:tembo/nodeless-store-inherited-posture (c8e6ab2) with next (2105ffb)}

Footnotes

1. 5 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩