test(solid-query): pin the stale read through a memo over query.data (#11351) 🤖🤖🤖

[mgarcialeniolabs]

🎯 Changes

Adds a reproduction for #11351 to useQuery-semantics.test.tsx, in the data store face block.

A tracked computation that reaches a leaf through a createMemo(() => query.data) indirection
is not re-notified after a refetch, so its last observation stays the superseded value permanently.
A wrapper hook that gates or narrows query.data before handing it to consumers is exactly this
shape, so it reaches app code that never writes the memo explicitly.

The test is added as it.skip, following the existing convention in useQuery.test.tsx for a
documented divergence, because the fix is not in this repo — see below. Un-skipped it fails on
the last assertion only, after expect(rendered.getByText('flag: true')) has already passed: the
commit lands and the DOM swaps, and only the reader behind the memo is starved.

Why no fix here

The data node is not at fault. computeData runs exactly the expected four times across mount and
refetch, chainOnce produces one chained promise per fetch with no identity churn, the commit
lands, an untracked read through the very same memo returns the new value, and neither the memo nor
the data node is left isPending. Only the notification to one reader shape is lost.

Narrowing across reader shapes (all against the real hook):

Reader shape	Result
effect reads query.data.flag	pass
memo reads query.data.flag, effect reads the memo	pass
effect captures query.data via untrack, then reads .flag off it	pass
effect depends on a suspended memo over an unrelated async source, reads query.data.flag	pass
memo reads query.data, effect reads memo().flag	fail
memo reads query.data, effect reads both query.data and memo().flag	fail

Row 6 is the one that rules out a missing subscription: the effect reads the projection directly
and depends on the memo, and is still not notified. Row 4 rules out "any suspended memo
dependency" — the memo has to be a consumer of this projection. A memo created after the first
settle is fine; only one that suspended on its first run is affected.

The behaviour reproduces with no TanStack code involved, so I filed it upstream:
solidjs/solid#3181. 2.0.0-rc.4 is the newest published solid-js / @solidjs/signals, so
there is no version bump that fixes it.

The workarounds I tried and rejected, in case they come up in review:

• Making the committed value's identity change per commit so the memo propagates — that is
  reconcile-off, and discards the fine-grained leaf tracking pinned by 'tracks deep reads at the leaf — unrelated changes do not re-run them'.
• Returning a stable re-reading facade from get data() — works only if the getter drops its
  suspending read, which would stop query.data suspending into <Loading> and break primitive
  and array data.
• Not returning the promise on refetch, to avoid the hold — deletes the SWR hold semantics covered
  by transition.test.tsx.
• Not superseding the in-flight answer with a synchronous value at settle (a landed flag on
  chainOnce, returning the unlanded chained promise instead of wrap(state.data)) — tried and
  reverted: it does not fix the reproduction and it breaks 'keeps stale data and exposes error state when a refetch fails'.

Happy to drop the skipped test and just leave the narrowing on the issue instead, if you'd rather
not carry it.

Note on base branch: this targets solid-query-v6-pre, since the rc.1 code the issue is about
does not exist on main.

✅ Checklist

• I have followed the steps in the Contributing guide.
• I have tested code changes locally with pnpm run test:pr, or these tests do not apply to this pull request.
• I fully understand the code in this pull request, including any code generated with AI assistance.

Verified locally from the repo root:

pnpm nx run @tanstack/solid-query:test:lib      # 26 files, 346 passed, 2 skipped, no type errors
pnpm nx run @tanstack/solid-query:test:eslint   # clean

Baseline on solid-query-v6-pre before this change is 346 passed / 1 skipped, so the only delta is
the added skipped test.

🚀 Release Impact

• This change affects published code, and I have generated a changeset.
• This change is docs/CI/dev-only (no release).

Made with Cursor

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Team

Run ID: 91306f7e-2be6-455c-9168-e10da02ceae9

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}