5.1 KiB
Row 302 (#1809): gate results
The slot ran from 08:15Z on 2026-10-23 (sage-3289), with Sage as gate owner and Dewey as reviewer. Logs: run1/gate-run.log and run2/gate-run.log. Script: run2/run-gate.sh and run2/mutants.py. The full per-file logs stay in /mnt/storage/scratch/gizmo-302-out/gate-run1/ and gate/.
Commits
| Commit | What |
|---|---|
f799e8b15 |
red first: the tests over unfixed product (d5) |
64f6558ee |
the fix |
5b63ea323 |
d6: launcher-released.test.mjs:89, Mos ruling mos-1430 (relayed in sage-3297) |
The 5b63ea323 subject still says "Mos ruling pending sage-3293; local, not pushed". I wrote it before
mos-1430 arrived, and Sage said not to amend. The ruling is mos-1430, and the commit was pushed as a
fast-forward at about 08:23Z.
Run 1 (08:18Z): stopped on a red outside the four pins
- Gate tree: 649ec67ab, a local merge of
64f6558eeonto origin/refactor1ac655005. - Red stage: as expected (the same reds as run 2).
- Green on 649ec67ab: idle-stop-race 1/1, runner.test 40/40 and compactions-launch 5/5. Then bus had 875 of 876 pass. The red was this row's own test, at launcher-released.test.mjs:89, 1 !== 0.
- It stopped at 08:19:28Z and went to Sage (gizmoduck-0388), then Mos. BUILD.md d6 has the cause.
Run 2 (08:23:41Z to 08:38:14Z, 873 s)
The gate tree was 783602e06ed92f99c7c71d1e9c016a0f9e2c9b28, a local merge (not pushed). Parent 1 is
origin/refactor 1ac655005, and parent 2 is 5b63ea323. The 12 row paths equal 5b63ea323's, and none
of them changes in 0e30a85fa..1ac655005.
Red: f799e8b15's product with the final test files from 5b63ea323
The five final test blobs:
| Blob | File |
|---|---|
52343932b2 |
packages/bus/tests/land.test.mjs |
6fb6e15c2f |
packages/bus/tests/launcher-released.test.mjs (new in d6) |
27193a745c |
packages/bus/tests/threads.test.mjs |
9dfacfea15 |
packages/cli/tests/idle-stop-race.test.mjs |
0792431b60 |
packages/harness/tests/runner.test.mjs |
The script checked each blob with hash-object on the red tree. Only launcher-released differs from
f799e8b15, and no product file differs from it.
There were seven reds, and nothing else failed:
- idle-stop-race: the released run ended with 'session stopped working: not-holder', exit 22 not 0
- runner.test landing case: pin :771, with 22 vs 0
- threads.test threadIdle case: pin :394
- land.test landRelease case: pin :464
- launcher-released: all three tests
On the bus files, 33 of 38 passed and 5 failed.
Pin :772 sits after :771 in the same test, so :771 stops the test first. :771's message is out.stderr, and that shows the old "session stopped working: not-holder" line and no "session released by the launcher" line. :89's two new lines never run on red, because the earlier refusals fail first. Run 2's green bus stage is their first green.
Green on 783602e06
| Run | Result | Wall |
|---|---|---|
| idle-stop-race | 1/1 | 2 s |
| runner.test whole | 40/40 | 12 s |
| compactions-launch | 5/5 | 14 s |
| bus | 876/876 | 24 s |
| harness | 151/151 | 13 s |
| cli | 918/918 | 615 s |
| test-auth | 15 | |
| test-conductor | 17 | |
| test-config | 55 | |
| test-discord | 81 | |
| test-extension-package | 18 | |
| test-foundation | 44 | |
| test-land-work | 86 | |
| test-queue | node 156/156, suite 28 | |
| test-release | 14 | |
| test-task | 91 |
Every scripts/test-*.sh ended with rc 0 and 0 failed. The node suites skipped nothing and left no todo. The tree was clean after the tests.
The shell scripts skipped two things:
- test-queue: "queue verify and render --check: this checkout is not the queue's canonical root". The gate runs in a worktree, not /mnt/storage/src/mosaic-stack.
- test-task: "live provider cases (row 244: no provider call in the gate)", printed twice. It printed no other skip line, so no Docker group was skipped.
Mutants on 783602e06
Each kill file ran whole, and every listed test name had to fail. After each mutant the file was checked out again and matched its sha256. At the end, broker.mjs was 0c6dbbc5… and runner.mjs was 72ec39eb…, both equal to 5b63ea323's, and the tree was clean.
| Mutant | Change | Killed by |
|---|---|---|
| M1 | no add after threadIdle's release | launcher-released (threadIdle), idle-stop-race |
| M2 | #agent always not-holder | land, launcher-released (threadIdle, landRelease), idle-stop-race, runner landing case |
| M3 | runner ignores launcher-released | idle-stop-race, runner landing case |
| M4 | a released run exits session | idle-stop-race, runner landing case |
| M5 | every lost claim reads launcher-released | launcher-released (other way), idle-stop-race |
| M7 | role.claim's D1 check removed | launcher-released (threadIdle, landRelease) |
| M8 | no add after landRelease's release | land, launcher-released (landRelease), runner landing case |
| M9 | exit releases a launcher-released run | idle-stop-race |
All eight were killed by rule, and none survived. When the run ended, no node process had its cwd in the worktree.