Skip to content

test(Table): the rows-and-columns focus test reports the scroll again before reading rows - #12529

Open
cixzhang wants to merge 6 commits into
mainfrom
l0/table-window-test-lands-measurement
Open

cixzhang wants to merge 6 commits into
mainfrom
l0/table-window-test-lands-measurement

Conversation

@cixzhang

@cixzhang cixzhang commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

User impact

  • Who: contributors whose pull requests run the Table test suite.
  • Before: "keeps a focused cell row and column when both scroll away…" in useTableColumnsVirtualization.test.tsx failed once in CI. It read the first row window (row 3 where row 48 was due) while the focused row was still mounted. So the failure was the test's single read, not a missing focus island.
  • After: the test reports the scroll a second time and waits two more frames before it reads the rows.
  • Unchanged: no component code changes.

What changed

  • One test, test-only. After each scrollBoth, the test reports one more scroll and waits two more frames (landWrites) before reading the rows. The second scroll report publishes the scroll position again. The CI failure didn't reproduce locally, so this doesn't claim a cause. The existing late-frame case still covers a scroll that lands while a measurement is pending.
  • No retry and no timeout change. No changeset: this is a test-only change.

Evidence

  • Red: the CI failure above, which is the only reproduction. Locally, delaying every animation frame by 0, 80 or 200 ms never reproduced the stale window, and the focused row stayed mounted with focus kept in every case.
  • Green: the Table suites pass 908 of 908, three runs in a row at 24 workers.

Scope

Test-only (Table virtualization).

Owner: Astryx fix-loop lane 0. Land path: CodeBunny approval of the exact head, CI, then the merge queue.

Overlap checked: #11844, #11826, #11197 (they change column-window scroll behaviour and add their own cases in this file; this changes only the one rows-and-columns focus case read, so it stays separate and merges main after them)

…surement land

The test read the row window once after a fixed two-frame settle. In CI it
once read the first window (row 3 where row 48 was due), with the focused
row still mounted, so the failure was the read, not the island. The scroll
owner drops a scroll event while a measurement frame is pending, and the
window then follows that frame. A second scroll report and two more frames
now let every scheduled measurement land before the rows are read.
@vercel

vercel Bot commented Nov 18, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

Project Deployment Actions Updated
astryx Ready Ready Preview Nov 21, 2026 4:09pm UTC

Request Review

@cixzhang cixzhang added the lane:0 Astryx fix-loop lane 0 owns this label Nov 18, 2026
@github-actions

github-actions Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

PR Analysis Report

Current run: The current CI run concluded skipped. Current analysis is unavailable.


Generated by PR Enrichment workflow | View current CI run

@astracat-bot

astracat-bot Bot commented Nov 18, 2026 •

Copy link
Copy Markdown
Contributor

AI review status for this pull request.

Review status Updated
🔄 In progress (for maintainers only) Nov 21, 2026, 4:44 PM UTC

@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Meta Open Source bot. label Nov 18, 2026

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Changes requested: the test change is small and the case passes, but its comment and the description explain it with a cause the code does not support.

  • The new comment at useTableColumnsVirtualization.test.tsx:1021-1024, the title and the description say focusing the cell schedules a measurement frame that lands after the two-frame settle, but focusing schedules no frame and a frame pending when the scroll lands already runs in the first settle frame (the existing late-frame case asserts this), so what the change adds is a second scroll report; please reword them to what is established: the failure did not reproduce, and the second scroll report publishes the scroll position again.

Keeping the second scroll is fine once the comment says so; the late-frame case still covers a scroll that lands while a measurement is pending.

Claims and files checked

Review mode: full; reason: no-earlier-review

  • Only contributors running the Table suite are affected and no component code changes: verified (one test file, 6 added lines).
  • The case failed once in CI, reading row 3 where row 48 was due: not verifiable (no run is linked, and it did not reproduce).
  • The failure was the single read, not a lost focus island: consistent with the reported failure, where the window never moved.
  • After each scroll the test reports one more scroll and waits two more frames: verified (the existing helper, called at lines 1025 and 1035).
  • The scroll owner skips a scroll event while a measurement frame is pending: verified (useScrollableArea.ts:450-453).
  • Focusing the cell can schedule such a frame: false (focus requested no frame and made no DOM change the scroll owner observes).
  • That frame's measurement can land after the two-frame settle: false (it runs in the first settle frame and publishes the scroll position; the late-frame case at line 991 asserts the window follows).
  • No retry and no timeout change: verified (no retry option or timeout edit).
  • No changeset because the change is test-only: verified.
  • Delayed frames never reproduced the stale window locally: consistent (the case without the new wait passes with frames 80 ms late).
  • The Table suites pass, 908 of 908: the suites pass (906 of 906 in the Table folder, two runs); the exact count was not reproduced.
  • Open pull requests #11844, #11826 and #11197 also edit this file: verified.
  • packages/core/src/Table/plugins/virtualization/useTableColumnsVirtualization.test.tsx: finding (the comment at lines 1021-1024).

Automated review (maintainer link)

@cixzhang cixzhang changed the title test(Table): the rows-and-columns focus test lets every scheduled measurement land test(Table): the rows-and-columns focus test reports the scroll again before reading rows Nov 18, 2026

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Changes requested: the new comment and the title now say what the change does, but one line of the description still gives the earlier explanation.

  • The description's After line says the test "lets every scheduled measurement land before it reads the rows", but the base already did that (no measurement frame is pending when it reads the rows, and focusing the cell schedules none), so please reword it to what the change adds, for example "the test reports the scroll a second time and waits two more frames before it reads the rows".

The case passes on this head, and the earlier note on the test comment is resolved.

Claims and files checked

Review mode: incremental; reason: none

  • Only contributors running the Table suite are affected and no component code changes: verified (one test file, 6 added lines).
  • The case failed once in CI, reading row 3 where row 48 was due with the focused row still mounted: not verifiable (no run is linked, and it did not reproduce).
  • The failure was the single read, not a missing focus island: consistent with the reported failure, where the window never moved.
  • After the change, the test lets every scheduled measurement land before it reads the rows: false as a change (no measurement frame is pending at the base's read point).
  • After each scroll the test reports one more scroll and waits two more frames: verified (the existing helper, called at lines 1025 and 1035).
  • The second scroll report publishes the scroll position again: verified (the scroll owner publishes on a scroll when no measurement is pending).
  • The failure did not reproduce, so no cause is claimed: matches the new comment and the title.
  • The late-frame case still covers a scroll that lands while a measurement is pending: verified (the case at line 991).
  • No retry and no timeout change: verified.
  • No changeset because the change is test-only: verified.
  • Frames delayed by 0, 80 or 200 ms never reproduced the stale window: consistent (the case without the new wait passes at each delay).
  • The Table suites pass, 908 of 908: the suites pass (906 of 906 in the Table folder, two runs); the exact count was not reproduced.
  • Open pull requests #11844, #11826 and #11197 also edit this file: verified.
  • packages/core/src/Table/plugins/virtualization/useTableColumnsVirtualization.test.tsx: clear (the comment at lines 1021-1024 now says what the code does).

Automated review (maintainer link)

@astracat-bot astracat-bot Bot left a comment •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approved: CI passed on this head, and nothing is left for the author.

The focus case re-sends the scroll and waits two more frames before it reads the rows, its assertions are unchanged, and every description claim matches the code.

Claims and files checked

Review mode: same-head; reason: none

  • Only contributors running the Table suite are affected, and no component code changes: verified (one test file, 6 added lines).
  • The case failed once in CI, reading row 3 where row 48 was due with the focused row still mounted: not verifiable (no run is linked, and it did not reproduce).
  • The failure was the test's single read, not a missing focus island: consistent with the reported failure, where the window never moved.
  • After each scroll, the test reports the scroll a second time and waits two more frames before it reads the rows: verified (the existing helper, called at lines 1025 and 1035).
  • The second scroll report publishes the scroll position again: verified (the scroll owner publishes on a scroll when no measurement is pending).
  • The failure did not reproduce, so no cause is claimed: matches the comment at lines 1021-1024 and the title.
  • The late-frame case still covers a scroll that lands while a measurement is pending: verified (the case at line 991).
  • No retry and no timeout change: verified.
  • No changeset, because the change is test-only: verified.
  • Delaying frames by 0, 80 or 200 ms never reproduced the stale window, and focus stayed on the cell: consistent (the case without the new wait passes at each delay).
  • The Table suites pass, 908 of 908, three runs in a row: the suites pass (906 of 906 in the Table folder, and this file 40 of 40 in CI on this head); the exact count was not reproduced.
  • Open pull requests #11844, #11826 and #11197 also edit this file: verified.
  • The pull request lands after an approval of this exact head, CI, then the merge queue: consistent (CI passed on this head).
  • The title, that the test reports the scroll again before reading rows: verified.
  • packages/core/src/Table/plugins/virtualization/useTableColumnsVirtualization.test.tsx: clear.

Automated review (maintainer link)

@cixzhang

Copy link
Copy Markdown
Contributor Author

/approve-spec e0536f1

@cixzhang
cixzhang marked this pull request as ready for review November 19, 2026 12:57
@cixzhang
cixzhang marked this pull request as draft November 19, 2026 20:32
@cixzhang

Copy link
Copy Markdown
Contributor Author

/approve-spec 859463e

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approved pending CI: CI has not run on this head yet, because a draft runs no CI. Mark the pull request ready for review so CI runs; nothing else is left for the author.

The merge of main leaves the patch unchanged: the focus case still re-sends the scroll and waits two more frames before it reads the rows, its assertions are unchanged, and every description claim matches the code.

Claims and files checked

Review mode: carry-forward; reason: none

  • Only contributors running the Table suite are affected, and no component code changes: verified (one test file, 6 added lines).
  • The case failed once in CI, reading row 3 where row 48 was due with the focused row still mounted: not verifiable (no run is linked, and it did not reproduce).
  • The failure was the test's single read, not a missing focus island: consistent with the reported failure, where the window never moved.
  • After each scroll, the test reports the scroll a second time and waits two more frames before it reads the rows: verified (the existing helper, called at lines 1025 and 1035).
  • The second scroll report publishes the scroll position again: verified (the scroll owner publishes on a scroll when no measurement is pending).
  • The failure did not reproduce, so no cause is claimed: matches the comment at lines 1021-1024 and the title.
  • The late-frame case still covers a scroll that lands while a measurement is pending: verified (the case at line 991).
  • No retry and no timeout change: verified.
  • No changeset, because the change is test-only: verified.
  • Delaying frames by 0, 80 or 200 ms never reproduced the stale window, and focus stayed on the cell: consistent (the case without the new wait passes at each delay).
  • The Table suites pass, 908 of 908, three runs in a row: they passed with this same patch before the merge of main (906 of 906 in the Table folder, and this file 40 of 40 in CI); not yet run on this head.
  • Open pull requests #11844, #11826 and #11197 also edit this file: verified (all three are open).
  • The pull request lands after an approval of this exact head, CI, then the merge queue: consistent (CI has not run on this head yet).
  • The title, that the test reports the scroll again before reading rows: verified.
  • packages/core/src/Table/plugins/virtualization/useTableColumnsVirtualization.test.tsx: clear.

Automated review (maintainer link)

@cixzhang

Copy link
Copy Markdown
Contributor Author

/approve-spec 874d588

@astracat-bot astracat-bot Bot left a comment •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approved pending CI: CI has not run on this head yet, because the run that started when the pull request was marked ready found no room under the CI ceiling and was cancelled. Re-run all jobs of that CI run (or push later) so CI runs; nothing else is left for the author, and the patch is unchanged since the last review.

#11844 has since landed on main and edits this file; this head merges cleanly with it, and the focus case passes on the merged code.

Claims and files checked

Review mode: same-head; reason: none

  • Changed rules since the last review (0): none.
  • Only contributors running the Table suite are affected, and no component code changes: verified (one test file, 6 added lines).
  • The case failed once in CI, reading row 3 where row 48 was due with the focused row still mounted: not verifiable (no run is linked, and it did not reproduce).
  • The failure was the test's single read, not a missing focus island: consistent with the reported failure, where the window never moved.
  • After each scroll, the test reports the scroll a second time and waits two more frames before it reads the rows: verified (the existing helper, called at lines 1059 and 1069).
  • The second scroll report publishes the scroll position again: verified (the scroll owner publishes on a scroll when no measurement is pending).
  • The failure did not reproduce, so no cause is claimed: matches the comment at lines 1055-1058 and the title.
  • The late-frame case still covers a scroll that lands while a measurement is pending: verified (the case at line 1025).
  • No retry and no timeout change: verified.
  • No changeset, because the change is test-only: verified.
  • Delaying frames by 0, 80 or 200 ms never reproduced the stale window, and focus stayed on the cell: consistent with an earlier probe of the same case.
  • The Table suites pass, 908 of 908, three runs in a row: they pass on this head and on this head merged with the current main (937 of 937 in the Table folder there, after newer cases landed; this file 43 of 43, three runs).
  • #11844, #11826 and #11197 also edit this file, and this one merges main after them: #11826 and #11844 have landed, and this head merges cleanly with the current main and passes there; #11197 is still an open draft.
  • The pull request lands after an approval of this exact head, CI, then the merge queue: consistent (CI has not run on this head yet: its run found no room and was cancelled).
  • The title, that the test reports the scroll again before reading rows: verified.
  • packages/core/src/Table/plugins/virtualization/useTableColumnsVirtualization.test.tsx: clear.

Automated review (maintainer link)

@cixzhang
cixzhang marked this pull request as ready for review November 21, 2026 08:43
@cixzhang
cixzhang marked this pull request as draft November 21, 2026 10:18
@cixzhang

Copy link
Copy Markdown
Contributor Author

/approve-spec fabb470

@astracat-bot astracat-bot Bot left a comment •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approved pending CI: CI started on this head when the pull request was marked ready for review, and the UI test jobs that run this file have passed, but the other required jobs are still running. Nothing is left for the author.

The focus case now re-sends the scroll and waits two more frames before it reads the rows; its assertions are unchanged, and every checkable claim in the description matches the code.

Claims and files checked

Review mode: same-head; reason: none

  • Only contributors running the Table suite are affected, and no component code changes: verified (one test file, 6 added lines).
  • The case failed once in CI, reading row 3 where row 48 was due while the focused row stayed mounted: not verifiable (no run is linked, and it did not reproduce).
  • The failure came from the test's single read, not a missing focus island: consistent with the reported failure, where the row window never moved.
  • After each scroll, the test reports the scroll again and waits two more frames before it reads the rows: verified (the existing landWrites helper, called at lines 1081 and 1091).
  • The second scroll report publishes the scroll position again: verified (the scroll owner publishes on a scroll event when no measurement is pending, useScrollableArea.ts:450-496).
  • No cause is claimed for the failure: matches the comment at lines 1077-1080 and the title.
  • The late-frame case still covers a scroll that lands while a measurement is pending: verified (the case at line 1047).
  • No retry and no timeout change: verified.
  • No changeset, because the change is test-only: verified.
  • Delaying frames by 0, 80 or 200 ms never reproduced the stale window: not rerun on this head; an earlier probe of the same case agrees.
  • The Table suites pass, 908 of 908, three runs in a row: they pass on this head (CI's UI test jobs, this file 43 of 43), with newer cases landed since.
  • #11844, #11826 and #11197 also edit this file, and this one merges main after them: #11826 and #11844 have landed and this head includes both; #11197 is still an open draft.
  • The pull request lands after an approval of this exact head, CI, then the merge queue: consistent (CI is running on this head).
  • The title, that the test reports the scroll again before reading rows: verified.
  • packages/core/src/Table/plugins/virtualization/useTableColumnsVirtualization.test.tsx: clear.

Automated review (maintainer link)

@cixzhang
cixzhang marked this pull request as ready for review November 21, 2026 14:21
@cixzhang
cixzhang marked this pull request as draft November 21, 2026 15:55
@cixzhang

Copy link
Copy Markdown
Contributor Author

/approve-spec 3318894

@astracat-bot astracat-bot Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approved pending CI: this head only merges main, so the change is identical to the one approved before, but CI has not run on it because the pull request is a draft. CI runs once the pull request is marked ready for review.

The focus case re-sends the scroll and waits two more frames before it reads the rows; its assertions are unchanged, and every checkable claim in the description matches the code.

Claims and files checked

Review mode: carry-forward; reason: none

  • Only contributors running the Table suite are affected, and no component code changes: verified (one test file, 6 added lines).
  • The case failed once in CI, reading row 3 where row 48 was due while the focused row stayed mounted: not verifiable (no run is linked, and it did not reproduce).
  • The failure came from the test's single read, not a missing focus island: consistent with the reported failure, where the row window never moved.
  • After each scroll, the test reports the scroll again and waits two more frames before it reads the rows: verified (the existing landWrites helper, called at lines 1081 and 1091).
  • The second scroll report publishes the scroll position again: verified (the scroll owner publishes on a scroll event when no measurement is pending, useScrollableArea.ts:450-496).
  • No cause is claimed for the failure: matches the comment at lines 1077-1080 and the title.
  • The late-frame case still covers a scroll that lands while a measurement is pending: verified (the case at line 1047).
  • No retry and no timeout change: verified.
  • No changeset, because the change is test-only: verified.
  • Delaying frames by 0, 80 or 200 ms never reproduced the stale window: not rerun; an earlier probe of the same case agrees.
  • The Table suites pass, 908 of 908, three runs in a row: not rerun on this head; CI passed this file 43 of 43 and every test shard on the previous head, whose patch is identical.
  • #11844, #11826 and #11197 also edit this file, and this one merges main after them: #11826 and #11844 have landed and this head includes both; #11197 is still an open draft.
  • The pull request lands after an approval of this exact head, CI, then the merge queue: consistent; CI waits for the draft to be marked ready.
  • The title, that the test reports the scroll again before reading rows: verified.
  • packages/core/src/Table/plugins/virtualization/useTableColumnsVirtualization.test.tsx: clear.

Automated review (maintainer link)

@cixzhang
cixzhang marked this pull request as ready for review November 21, 2026 16:44

This branch was successfully deployed

1 active deployment
Preview — 33188948 Deployed Nov 21, 2026 by vercel[bot]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Meta Open Source bot. lane:0 Astryx fix-loop lane 0 owns this

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant