Skip to content

notes-copyout-browser: status-line check fails intermittently after Copy chosen design #18106

Closed

Description

@kyle-sexton

Problem

The prototype plugin's explore-directions browser suite notes-copyout-browser failed intermittently on the check "the copy is said in the status line", with the status line read as "". Failing example: run 44953095719, job Test Bash (1), 2026-11-21T21:46:46Z:

PASS: take: Copy chosen design copies the shown variant's copy-out text
PASS: take: it is the text the page's own copy-out gives
FAIL: take: FAIL the copy is said in the status line  [""]

The question is whether this is a product race on the host page (a re-render, or a state push overwriting the status, like the notes-history press bug fixed in #18091) or a test that reads the status too early.

Done when

  • The cause is named at file:line.
  • There is a fix, with either a regression check that fails before the fix and passes after it, or repeated local runs showing the flake is gone, with the count stated.

Activity

  1. kyle-sexton commented on Nov 22, 2026

    @kyle-sexton
    MemberAuthor

    The cause is in the test, not the product. The fix is already on main through #18042 (squash commit e2d9b1f, from its branch commit b131069, "wait for the copy-out status line before checking it").

    Cause

    • lib/ui-feedback/host.mjs:594 writes the copied text into #block synchronously.
    • The status line is written only once the clipboard promise settles: navigator.clipboard.writeText(text).then(done, manual) at host.mjs:597-599.
    • Run 44953095719 tested PR feat(prototype): edit and withdraw a sent note on the notes host #18042 at cd5413e, the commit before b131069. At that commit, notes-copyout-browser-take.js read #status straight after #block matched, with no wait. So it could read the status before the promise had resolved. No 2-second poll timed out: that run had no poll at all.
    • Main now waits up to 2 s for the status at notes-copyout-browser-take.js:31, then asserts it at line 32.

    Not a product race

    • render() never writes the status line.
    • The only blank write is say("") at host.mjs:734. It runs once, on the frame's first ready, which if (ready) return at line 726 guards, and it happens before the test can open the Copy menu.
    • Of the 7 failures reproduced below, 4 read "". The other 3 read "Copied the chosen design" in the failure detail, which reads the status a second time. The message arrived late; nothing overwrote it.

    Measurements

    All runs were local: 24 cores, 12 runs at a time, through the suite's own runner and playwright-cli.

    Variant Runs Failed Failure detail
    Take script without line 31 (pre-fix) 60 7 4 [""], 3 ["Copied the chosen design. …"]
    Take script on main 60 0
    • Timing. A probe that timed writeText inside the host page, over 36 runs at 12 at a time: the write settled in 6 to 19 ms, and click to status took 274 to 308 ms. Both are well inside the 2 s wait. The product writes the status correctly, and the budget does not need to be longer.
    • CI since the fix. No PR Require Checks run that failed after 2026-11-21T21:51Z has a notes-copyout-browser failure annotation. I did not count how many runs selected the suite in that window. One run failed in a different suite, notes-edit-browser (run 44967037847).

    I am closing this as completed by #18042.

  2. added
    needs-triageNot yet classified. Floor until a type and one priority tier are set.
    on Nov 22, 2026
  3. github-actions commented on Nov 22, 2026

    @github-actions
    Contributor

    Size: Small
    Complexity: Low
    Uncertainty: Known
    Flags: none
    Blocked by: none
    Blast radius: one browser test script in the prototype plugin's notes-copyout-browser suite; CI runs it and the host page is unchanged.
    Reversible: yes; revert the test edit.
    Open questions: none
    I understood this as an intermittent failure of the "copy is said in the status line" check, with the cause to be named and a fix verified. Your follow-up comment names the cause as a test that read the status before the clipboard promise settled, not a product race. It says the fix is already on main through #18042, with 60 local runs failing 7 times before the fix and 0 after. That PR is not in the recent merged-pull list I was given, so I could not confirm it here, and a maintainer should check it before closing. No open issue duplicates this one.

    Automated triage. A maintainer reviews it.

  4. removed
    needs-triageNot yet classified. Floor until a type and one priority tier are set.
    on Nov 22, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    Fields

    Priority

    medium

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions