test(signals): pin #3726 on L2 — derived store sync landing wakes parked readers (supersedes #3732) - #3789
Merged
Conversation
…ked readers Port of #3732 (brenelz) onto the hold model. The PR patched `store/next/projection.ts` to retire STATUS_UNINITIALIZED after a synchronous landing so core's #3181 settle walk would release readers parked on the superseded first flight: readers whose store node the landing left unchanged (`"length" in store`, `Object.keys`, an unchanged `length`) had no value notification to fall back on and stayed blank inside a loading boundary while a sibling read of the source updated. On L2 the core already holds, with no store-side patch: a reader of a derive with a flight up is parked on the derive itself (`pullFamily` links it, as a reader of a memo with a flight), and the derive's first commit (uninitialized → a value) is a value change for every subscriber (`recompute`: `wasUninitialized` → `insertSubs`) — the walk's uninitialized exemption is moot. The PR's three tests pass unchanged; its superseded source change is dropped (`src/store/next/` is gone). Added: the report's sequence verbatim (the source write and the superseded promise's resolution in one timer callback, under a loading boundary, microtask-driven) in signals and as a `<Loading>` web spec; a verdict probe parked on the first flight; and the case the PR listed as open — a landing a transaction holds: the parked readers re-run as the transaction's work and reveal at its commit, under a boundary and out (red on the PR's base with and without its fix; green on L2). Tests only — no changeset. RULES-INDEX regenerated (A19/A25/A29 gain the test's citations). Co-authored-by: Brenley Dueck <brenleydueck@gmail.com> Co-authored-by: Cursor <cursoragent@cursor.com>
The #3726 comment's playground, fetched: an async card source behind a memo (`createMemo(() => ready() ?? pending)`), a derived `createStore(() => source(), [])`, `createOptimisticStore` over it and a keyed `<For>`, remounted under a `<Loading>` that has already revealed; a count outside the view. Before, Activity → Board left "All cards: 1" beside an empty lane: the remounted readers were parked on the derive's first flight under the boundary's hold, and the held synchronous landing never woke them — the case PR #3732 listed as open (red on its base with and without its fix). Same root-cause family as the report, not a distinct bug. On L2 the revealed boundary holds the view switch (A29: a boundary already showing content holds like any reader) and the lane reveals with its rows as one frame — the count and the list never disagree. Pinned as the playground itself (web spec) and a signals reduction. The report's own playground, fetched too: the `Source:` hole is outside the `<Loading>` and the promise is created once — the first web test and the signals "report's playground" test now follow it exactly. Co-authored-by: Cursor <cursoragent@cursor.com>
|
Size (brotli, eager entry chunk)
Bundled with Rolldown (what Vite ships), brotli q11, decimal KB. Caps in |
Coverage Report for CI Build 37262214084Coverage remained the same at 75.991%Details
Uncovered ChangesNo uncovered changes found. Coverage RegressionsNo coverage regressions found. Coverage Stats
💛 - Coveralls |
Merging this PR will degrade performance by 8.68%
|
| Benchmark | BASE |
HEAD |
Efficiency | |
|---|---|---|---|---|
| ❌ | memo + sync render effect only (reference) |
26.9 ms | 29.4 ms | -8.68% |
Tip
Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.
Comparing port/3732-derived-sync-landing (2e24ac3) with next (1a3f87f)
This was referenced Oct 5, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Supersedes #3732 by @brenelz — his three tests are kept verbatim and he is credited with
Co-authored-byon the commits. Closes #3726.Why the fix is subsumed by L2
#3732 patched
store/next/projection.tsto retireSTATUS_UNINITIALIZEDafter a synchronous landing so core's #3181 settle walk would release readers parked on the superseded first flight — readers whose store node the landing left unchanged ("length" in store,Object.keys, an unchangedlength) had no value notification to fall back on and stayed blank inside a loading boundary. On the hold model the core already holds, with no store-side patch:pullFamily(store.ts) reads a derive with a flight up tracked, so the parked reader is linked to the firewall node itself; the sync re-derive'srecompute(fw)haswasUninitialized → valueChanged → insertSubs, which re-runs those readers in the landing's flush (they pull the settled derive untracked and the stale link trims). Under a transaction's hold the same readers re-run as the transaction's work and reveal atland— the case #3732 listed as still open. The PR's three tests pass against untouchednext; its source change is dropped (src/store/next/is gone). This is tests-only: no changeset.Verified on #3732's base (
cce43eb44): the PR's tests are red 2/3 as it reports; every new pin below is red there, and the held-landing ones stay red with the PR's fix applied. All green on L2.What this pins beyond #3732
setTimeoutwrites the source (the derive lands synchronously) and resolves the superseded promise in the same tick;Presence:under<Loading>,Source:outside; scheduler-driven (no manualflush()).isPending(() => "length" in store)) parked on the first flight: suspends while uninitialized (A19 exc. 1), answersfalseat the landing.NotReadyErroruntil then, A25/A29) — with and without a<Loading>over them (fallback until the commit, content after).createStore(fn, [])→createOptimisticStore→ keyed<For>, remounted under a<Loading>that has already revealed, with a count outside the view. It is the held-landing variant, not a distinct bug. On L2 the revealed boundary holds the view switch (A29: a boundary already showing content holds like any reader) and the lane reveals with its rows as one frame — the count and the list never disagree. Pinned as the playground itself (web spec) and a signals reduction.@solidjs/webspec rendering both playgrounds as JSX (<Loading>,<For>,<Show>,hidden).Files:
packages/signals/tests/store/derived-presence-async-3726.test.ts(8 tests),packages/web/test/derived-presence-async-3726.spec.tsx(2 tests).RULES-INDEX.mdregenerated (A19/A25/A29 rows gain the test's citations).Tests
@solidjs/signals: 4894 passed / 0 failed / 2 skipped → 4902 / 0 / 2 (+8).rules-index --checkgreen.@solidjs/web(default config): 1132 / 0 → 1134 / 0 (+2). Server 1411/0/2 and hydrate 275/0 unchanged.next@1a3f87fd1(signals: benchmarks measure the built package; the fold queue's pre-batch record (CodSpeed follow-up to #3774) #3776), no conflicts; signals dist rebuilt before the web run.Public API changes
None.