adminandClaude Opus 5.5
2333f66ba2
Suite 755 for b17a924 was red in three places. Both e2e 65 and e2e 67
were green on their own. Two of the new checks did not fit into the
full run. The product is not at fault.
- e2e 66 (66s): the per-frame pixel read never stopped. splice(0)
emptied the array, so the length bound was never reached. Each
pixelAt leaves the stage shifted to its pixel until the next
drawing. The editor, idle after Stop, kept it at (-710, -376), so
the Play click of e2e 67 hit no object. The loop now ends on a flag
while the run still draws each frame.
- e2e 65 (65k2, 65k3): the console panel keeps 500 lines and is never
cleared in the suite. Lines of earlier steps pushed the first beats
out, so the beats began at 1.3. The checks now read the harness's
console log since Play, which has no limit.
- 65k2 also writes panelBusiest, the three most frequent other lines
in the panel. The next CI run names the module that logs that much.
Tests (melektaus-atx):
- e2e 66+67 before the fix: 67 breaks off as in CI.
- e2e 66+67 with the fix: ALL PASSED, 52 checks.
- e2e 65 under a flood of 100 lines/s: ALL PASSED, 76 checks.
The old reading under the same flood shows the CI picture.
- e2e 65+66+67: ALL PASSED, 126 checks (a first run had the known
e2e 67 firstAt wobble, 0.1).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>