daemon: settle the ownership keystroke before asserting a quiet report #185

Merged

natsukium merged 1 commit from daemon/mirror-report-flake into main

2026-10-19 09:54:32 +09:00

natsukium commented

2026-10-19 09:20:41 +09:00

Owner

Problem

serve::session_task::tests::a_mirrors_report_is_stored_but_the_owner_still_answers failed intermittently under load (locally during full runs, and on CI for #174) with "presentation reports are query-answering state: nothing is pushed to any window".

The test takes input ownership by writing k to the shell, and the tty echoes it back. Every handle_input ends in grid_cycle_for, so when the echo reached the grid before one of the later ColorScheme reports, that report's cycle shipped it. The stray events were exactly a RowDelta for row 0 (k) and a CursorState at column 1; the reports themselves push nothing.

Change

Wait for the echo, ship it with after_drive, and drain both receivers before the reports, the same settle step a_resize_answers_an_idle_mirrors_armed_pull_exactly_once uses. Test-only.

Verification

Running the test binary with --exact in 24 parallel loops: 9 of 360 runs failed before the fix, 0 of 720 after. session_task tests (111), clippy and prose-check are clean.

## Problem `serve::session_task::tests::a_mirrors_report_is_stored_but_the_owner_still_answers` failed intermittently under load (locally during full runs, and on CI for #174) with "presentation reports are query-answering state: nothing is pushed to any window". The test takes input ownership by writing `k` to the shell, and the tty echoes it back. Every `handle_input` ends in `grid_cycle_for`, so when the echo reached the grid before one of the later `ColorScheme` reports, that report's cycle shipped it. The stray events were exactly a `RowDelta` for row 0 (`k`) and a `CursorState` at column 1; the reports themselves push nothing. ## Change Wait for the echo, ship it with `after_drive`, and drain both receivers before the reports, the same settle step `a_resize_answers_an_idle_mirrors_armed_pull_exactly_once` uses. Test-only. ## Verification Running the test binary with `--exact` in 24 parallel loops: 9 of 360 runs failed before the fix, 0 of 720 after. `session_task` tests (111), clippy and prose-check are clean.

natsukium added 1 commit

2026-10-19 09:20:41 +09:00

daemon: settle the ownership keystroke before asserting a quiet report

darwin / detect relevant changes (pull_request) Successful in 9s

Details

darwin / darwin docs-only no-op (pull_request) Has been skipped

Details

windows / detect relevant changes (pull_request) Successful in 11s

Details

windows / windows docs-only no-op (pull_request) Has been skipped

Details

bench / instruction-count regression gate (pull_request) Successful in 49s

Details

darwin / build felis (aarch64-darwin) (pull_request) Successful in 1m11s

Details

pr / nix flake check (pull_request) Successful in 41s

Details

windows / cargo nextest (Windows) (pull_request) Successful in 3m4s

Details

pr / cargo build / clippy / test / deny (pull_request) Successful in 1m51s

Details

pr / wire schema is compatible with the base (pull_request) Successful in 2s

Details

pr / frontend smoke (x86_64-linux) (pull_request) Successful in 29s

Details

windows / frontend smoke (Windows) (pull_request) Successful in 41s

Details

windows / cargo clippy (Windows cross) (pull_request) Successful in 14s

Details

4e341b577b

a_mirrors_report_is_stored_but_the_owner_still_answers took input
ownership with a KeyBytes write, so the shell's line discipline echoed
it back. Every input ends in a grid cycle, so when the echo reached the
grid before a later ColorScheme report, that report's cycle shipped it
and the "nothing is pushed" assertion failed (9 of 360 runs under
parallel load). Wait for the echo and ship it before the reports, as
a_resize_answers_an_idle_mirrors_armed_pull_exactly_once already does;
720 of 720 runs pass under the same load.

natsukium merged commit 2ff9ac5ec5 into main

2026-10-19 09:54:32 +09:00

natsukium deleted branch daemon/mirror-report-flake

2026-10-19 09:54:32 +09:00

natsukium referenced this pull request from a commit

2026-10-19 09:54:34 +09:00

Merge pull request 'daemon: settle the ownership keystroke before asserting a quiet report' (#185) from daemon/mirror-report-flake into main