fix(signals): guard reconcile's array paths against array/object shape changes

[brenelz]

The bug

reconcile()'s object diff refuses to recurse into a pair whose container kinds disagree — Array.isArray(previous) !== Array.isArray(next) replaces the slot instead of merging into it. The array paths reach the same recursion through keyedMatch and positional pairing, and never applied that rule: two keyless wrappables "match" because keyFn returns undefined for both, regardless of whether they're arrays or objects.

So an incoming array gets merged into a proxy whose target is an object. Array.isArray on a Proxy inspects the target, which is fixed at wrap time, so the slot is wrong forever after:

const [state, setState] = createStore({ list: [{ x: 1 }] });
setState(reconcile({ list: [[10, 20]] }, "id"));

Array.isArray(state.list[0]);       // false
JSON.stringify(state.list[0]);      // {"0":10,"1":20}

Spread, .map, <For each> and JSON.stringify all see an object. The reverse direction is worse — an array-shaped target receiving an object reads length off that object and the store presents [], silently dropping the data.

The fix

Five call sites shared the gap: the keyed prefix loop and the positional loop in both applyStateFast and applyStateSlow, plus applyArrayItem (which covers the keyed diff's trailing and moved slots). They now share a recursablePair() helper that folds the existing markRaw check together with the container-kind check, so a kind change replaces the slot by reference — exactly what the object diff and descendInto already do.

Net effect on the hot paths is neutral: recursablePair subsumes checks those sites already performed inline.

Tests

Four cases added to tests/store/reconcile.test.ts, covering both directions, the keyed-array trailing slot (applyArrayItem), and positional key: null merging. All four fail on next and pass with the fix.

Found by differential fuzzing (random trees reconciled against a plain-object oracle, checking both direct reads and effect-observed values). Full suite green: 19/19 turbo tasks, 1150 @solidjs/signals tests.

🤖 Generated with Claude Code

[changeset-bot]

🦋 Changeset detected

Latest commit: b19d23a

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
solid-js	Patch
babel-preset-solid	Patch
@solidjs/web	Patch
@solidjs/html	Patch
@solidjs/h	Patch
@solidjs/universal	Patch
@solidjs/element	Patch
hackernews-spa-example	Patch
hackernews-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-hq]

Merging this PR will not alter performance

✅ 7 untouched benchmarks
⏩ 117 skipped benchmarks¹

---

_{Comparing brenelz:fix/reconcile-array-object-shape-guard (b19d23a) with next (63cd066)}

Footnotes

1. 117 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. ↩

[ryansolid]

Merged — verified the diagnosis independently: descendInto (reconcile.ts:275) has always applied the container-kind rule, and keyedMatch really does let two keyless wrappables pair across kinds. Also confirmed both rewritten call sites are reached only with both sides wrappable (the keyedMatch loop guarantees it), so folding the checks into recursablePair is behaviour-preserving apart from the intended kind guard. All four new tests fail on next and pass with the fix; full suite and CodSpeed green.