Skip to content

shell: e2e — D60's bar waited for whole, not read after a fixed pause (#850) - #851

Merged
napkinstack-agent[bot] merged 3 commits into
mainfrom
shell-850-confidential-kept-flake
Nov 15, 2026
Merged

napkinstack-agent[bot] merged 3 commits into
mainfrom
shell-850-confidential-kept-flake

Conversation

@kitokey-marc

@kitokey-marc kitokey-marc Bot commented Nov 15, 2026 •

Copy link
Copy Markdown
Contributor

shell: e2e — D60's bar waited for whole, not read after a fixed pause

Closes #850.


Module

modules/shell — exactly one.

  • This pull request touches a single module
  • ⚠️ Cross-module — the cross-module label is required, justification below

Deliverable

Out of cycle: #850. The window's e2e scenario confidential_kept (D60) failed in #825's attest, and it runs in every shell e2e, so it can block any attest. It is outside the freeze for that reason (the coordinator, 2026-11-15).

What it changes

Only the e2e, scripts/e2e_window.py, in confidential_kept:

  • The defect. Turning a meeting confidential adds a row to the bar, and the bar's window grows to it asynchronously. Two checks read the window's height (fits: scrollHeight against clientHeight) after a fixed time.sleep(0.3): the French « the bar holds whole » and the English « … on, the window whole ». A loaded machine outruns the pause.
  • The fix. Each waits for the state instead: whole, the content within the window, with wait_for up to 5 s. The screenshots that follow each check therefore start from a bar at rest. This is tests.md §5's rule: wait for an observable state, never a duration.
  • The English check says which part failed. Its detail was the row alone. It is now {marked, row, fits}: the mark written, the row's words, and the window's heights.

Oracle

Kind of oracle: the e2e itself, its race taken out.


Review budget

  • Lines changed (excluding generated): 13 (9 added, 4 removed)
  • Files touched: 1
  • Within budget

Contracts

  • No contract affected

Definition of Done


Test sheet

Verifier: session 92ded11e-58ef-492d-ba56-c820b9011de8

# Given · when · then Kind Result Evidence Commit Confirmed
S1 Given confidential_kept at this head, when it runs alone, then both height checks pass, waiting for the bar to be whole, with no fixed pause before either automated passed The coordinator's run under the lock, KITOKEY_E2E_ONLY=confidential_kept at 862cf88, its e2e.txt read here: 38 ok, no FAIL; « on, with its context, the bar holds whole — 440 wide, as tall as it shows » ok (l.9), « in English, … on, the window whole » ok (l.31); « its checks passed — not the module's e2e ». Read at the head: no time.sleep before either check, each waits on whole up to 5 s Run again at 37236d3 (main merged, the diff unchanged): 38 ok, no FAIL, the same two checks ok (l.9, l.31). 37236d3 —
S2 Given the full shell e2e at this head, under the attest's load, when it runs, then confidential_kept passes, its English check and the screenshot after it included automated passed The attest's full shell e2e at 37236d3, under its load, its e2e.txt read here: 1793 ok, no FAIL, « e2e: passed ». In confidential_kept: « on, with its context, the bar holds whole — 440 wide, as tall as it shows » ok (l.655), « in English, « Confidential meeting — nothing leaves this computer », on, the window whole » ok (l.677), and the screenshot after it, « d60-2-small-on-dark: the screenshot is dark », ok (l.678) — where #825's attest had timed out 37236d3 —
S3 Given a bar that never fits (a mutant making whole false), when the scenario runs, then both checks fail, the English one showing {marked, row, fits} automated passed Same, with whole = lambda b: False (numstat 1 1): 36 ok, exactly two FAIL, the two checks: the first got: ([271, 271], [440, 271]), the English one got: {'marked': True, 'row': 'Confidential meetingNothing leaves this computer', 'fits': [271, 271]}; « FAILED » Run again at 37236d3, same mutant (1 1): 36 ok, the same two FAIL and nothing else, the same got:. 37236d3 —

Verifier's remarks: (1) The two time.sleep left in confidential_kept (0.3 s in seen, 0.4 s before the first screenshot) come before screenshots, not checks: out of #850's scope unless a screenshot raced. (2) whole reads fits twice per look; harmless, the content's height does not shrink while the window grows, so a look passes no earlier than it should. (3) S3's English fits [271, 271] holds on a real window: the mutant forces the check, as asked; the window itself was whole.


Summary

DONE

  • The bar's height waited for as a state, whole, within 5 s, before both of D60's checks that read it.
  • The English check's detail says the mark, the row and the heights.

VERIFIED

ASSUMED

  • The red was the height (fits), not the mark. The row's words were there, and the height is the one read after a fixed pause; the detail now says which if it comes back.

NOT VERIFIED

RISKS

  • A bar that takes more than 5 s to grow still fails, loudly, with its heights shown.

Signals to report

  • I had to look at another module's implementation
  • A kernel or playbook rule got in my way for no good reason
  • A quality gate blocked without improving quality
  • A rule applied by hand would be worth automating

🤖 Generated with Claude Code

kitokey-marc Bot and others added 2 commits November 15, 2026 03:26
…e row, the width (#850)

Its row read « Confidential meetingNothing leaves this computer » when it failed: both words
there, so the mark not written within 3 s or the window too narrow failed it, which the
row alone did not say. The check is the same; what it shows when it fails is all of it.

Agent-Session: 14f73956-90b9-46a3-9fcc-52dc45c43948
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…#850)

Turned confidential, the bar adds a row and its window grows to it, asynchronously. Its two
checks read the window's height 0.3 s after the gesture: a loaded machine outruns that, and
the check failed with every word of the row there. Now waited for as a state, `whole`, within
five seconds, before each check — and so before the screenshot that follows it.

Agent-Session: 14f73956-90b9-46a3-9fcc-52dc45c43948
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@kitokey-marc kitokey-marc Bot added the out-of-cycle Work outside the current cycle: an incident, a production defect; justified in the PR label Nov 15, 2026
kitokey-marc Bot added a commit that referenced this pull request Nov 15, 2026
… cancelled one does (#845)

The same race, in a_transcript_ended_by_kitokey_ends_with_what_it_started_and_is_not_written:
its child checked the instant the answer came, with `kill -0`, which a zombie not yet reaped
still answers. It failed #851's attest and #836's. Now waited for with `gone`, zombie-aware,
within half a second. Without the group's KILL in `end`, it still fails.

Agent-Session: 14f73956-90b9-46a3-9fcc-52dc45c43948
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
napkinstack-agent Bot pushed a commit that referenced this pull request Nov 15, 2026
…checked the instant it is answered (#845) (#846)

* shell: a cancelled transcription's deaf child is waited for as gone, not checked the instant it is answered (#845)

The KILL sent, the child is gone once the kernel has delivered it and reaped it: on a
loaded runner, after the answer (#812's CI, run 43596085957). Now polled as a state for
half a second after the answer; the two seconds' bound kept on the answer itself, read
before. Without the group's KILL, it still fails: « still runs half a second after the
answer ».

Agent-Session: 14f73956-90b9-46a3-9fcc-52dc45c43948
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* shell: a transcript ended by Kitokey waits for its child gone, as the cancelled one does (#845)

The same race, in a_transcript_ended_by_kitokey_ends_with_what_it_started_and_is_not_written:
its child checked the instant the answer came, with `kill -0`, which a zombie not yet reaped
still answers. It failed #851's attest and #836's. Now waited for with `gone`, zombie-aware,
within half a second. Without the group's KILL in `end`, it still fails.

Agent-Session: 14f73956-90b9-46a3-9fcc-52dc45c43948
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: kitokey-marc[bot] <356215105+kitokey-marc[bot]@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
The transcription tests' children waited for as gone on main: this branch's attest no longer
carries #845's race.

Agent-Session: 14f73956-90b9-46a3-9fcc-52dc45c43948
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
@napkinstack-agent
napkinstack-agent Bot merged commit f3a35b7 into main Nov 15, 2026
14 of 17 checks passed
@napkinstack-agent
napkinstack-agent Bot deleted the shell-850-confidential-kept-flake branch November 15, 2026 07:18
kitokey-marc Bot added a commit that referenced this pull request Nov 15, 2026
…eciders-key

D60's bar waited for whole on main: confidential_kept no longer reads its height after a fixed
pause in this branch's e2e.

Agent-Session: 14f73956-90b9-46a3-9fcc-52dc45c43948
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

out-of-cycle Work outside the current cycle: an incident, a production defect; justified in the PR

Projects

None yet

Development

Successfully merging this pull request may close these issues.

shell e2e: confidential_kept (D60) reads the bar's height after a fixed pause, and failed in a local attest while CI is green

0 participants