fix: stabilize two CI test failures (Paris today, worker cancel race) #307

Closed

nessar wants to merge 3 commits from feature/ci-test-stability into develop

pull from: feature/ci-test-stability

merge into: nessar:develop

nessar:develop

nessar:feature/obligation-solution-ranking

nessar:feature/bun-migration

nessar:feature/planning-architecture

nessar:feature/incremental-exact-scoring

nessar:feature/benchmark-performance

nessar:feature/verify-trace-free-planning

nessar:feature/remove-decision-trace

nessar:feature/gitea-actions

nessar:feature/gitflow-ci

nessar:main

nessar:select-duration

nessar:dependabot/npm_and_yarn/typescript-6.0.3

nessar:dependabot/github_actions/docker/build-push-action-7

nessar:dependabot/github_actions/docker/login-action-4

nessar:dependabot/github_actions/docker/metadata-action-6

nessar:dependabot/github_actions/docker/setup-buildx-action-4

nessar:dependabot/github_actions/actions/upload-artifact-7

nessar:copilot/sub-pr-61-another-one

nessar:copilot/sub-pr-61-again

nessar:copilot/sub-pr-61

nessar:copilot/sub-pr-49

nessar:copilot/sub-pr-37

nessar:copilot/update-ci-cd-pipeline

nessar:copilot/refactor-duplicated-code

Conversation 1 Commits 3 Files Changed 3

+19 -9

nessar commented 2026-10-19 09:52:24 +02:00

Owner

Copy link

Copy source

No description provided.

nessar added 2 commits 2026-10-19 09:52:24 +02:00

fix(calendar): mark today on the Europe/Paris date ... f458a83599

The calendar highlighted today from the browser's local date while the
default planning horizon starts on the Europe/Paris date, so the two
disagreed for a few hours a day (and always, for users in another time
zone). The horizon's first day was then not marked as today. This also
failed the CI test run in UTC containers between 22:00 and 24:00 UTC.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

test(planning): hold the worker until the preparation cancel arrives ... af20c96f01

The worker reads its queue only between event-loop turns, so the
calendar preparation could finish before the cancel sent on its second
progress update arrived, and the test received a complete plan. Use the
real-worker harness hold already applied to the checkpoint cancellation
test, and cancel on the first preparation update so the hold is released.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

nessar referenced this pull request 2026-10-19 10:11:27 +02:00

test(benchmark): calibrate again after the protocol and judge against the slower value #308

nessar added 1 commit 2026-10-19 10:21:55 +02:00

test(calendar): pin the clock for the Paris today marker ... 036b13613c

The test of the today marker read the wall clock, so it only exercised
the Paris/local date mismatch between 22:00 and 24:00 UTC and passed
with or without the fix the rest of the day. Pin the clock to 22:16 UTC,
already the next day in Paris, so a UTC runner fails without the fix at
any time.

Refs #301

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

nessar referenced this pull request 2026-10-19 10:28:59 +02:00

Today marker and default planning horizon disagree near midnight outside Europe/Paris #301

nessar referenced this pull request 2026-10-19 10:28:59 +02:00

Complete and verify the Bun CI cutover #293

nessar commented 2026-10-19 10:29:07 +02:00

Author

Owner

Copy link

Copy source

Closing without merging: #301 will be done properly. The findings were moved to the tickets:

• the Paris "today" marker and the pinned-clock test: #301
• the worker cancellation race and the benchmark gate noise: #293

The branch feature/ci-test-stability is kept for reference (f458a83 the Paris fix, af20c96 the worker cancellation test, 036b136 the pinned-clock test). The benchmark calibration change was independent and went in separately as #308. #304 and #305 can still hit the two test failures until those fixes land in develop some other way.

Closing without merging: #301 will be done properly. The findings were moved to the tickets: - the Paris "today" marker and the pinned-clock test: [#301](https://gitea.nessar.fr/nessar/vac-optimizer/issues/301#issuecomment-5448) - the worker cancellation race and the benchmark gate noise: [#293](https://gitea.nessar.fr/nessar/vac-optimizer/issues/293#issuecomment-5453) The branch `feature/ci-test-stability` is kept for reference (f458a83 the Paris fix, af20c96 the worker cancellation test, 036b136 the pinned-clock test). The benchmark calibration change was independent and went in separately as #308. #304 and #305 can still hit the two test failures until those fixes land in `develop` some other way.

nessar closed this pull request 2026-10-19 10:29:08 +02:00

nessar referenced this pull request 2026-10-19 10:43:44 +02:00

ci: run push workflows only on develop, main and release tags #309

Pull request closed

Please reopen this pull request to perform a merge.

Sign in to join this conversation.

Reviewers

No Reviewers

Labels

Clear labels

blocked-on-bun-release

Cannot be acted on until a new Bun release ships

bug

dependencies

docker

documentation

duplicate

enhancement

github_actions

good first issue

help wanted

invalid

javascript

needs-info

needs-triage

question

ready-for-agent

wontfix

No labels

Milestone

No items

No Milestone

Projects

Clear projects

No projects

Assignees

Clear assignees

renovate (renovate) Bot nessar (nessar)

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: nessar/vac-optimizer#307