Skip to content

test(sync): observe receiver admission after cancellation - #80

Closed
rldyourmnd wants to merge 1 commit into
mainfrom
fix/sync-resume-release-fence-20261002
Closed

rldyourmnd wants to merge 1 commit into
mainfrom
fix/sync-resume-release-fence-20261002

Conversation

@rldyourmnd

@rldyourmnd rldyourmnd commented Oct 2, 2026 •

Copy link
Copy Markdown
Contributor

The canceled-transfer resume fixture slept 150 ms and assumed the receiver journal had been released; full Ubuntu CI correctly refused the immediate reopened journal while a filesystem operation retained its lock. Observe typed receiver admission within three seconds before the single resumed transfer, join sender cancellation, require hash-named part progress and retain byte-identity/reduced-refetch assertions.

A real-lock regression shows that 150 ms elapsed is insufficient and verifies admission after actual release. macOS sync E2E: 15 passed, 0 failed, 0 ignored; strict all-target/all-feature sync lint and formatting passed. Native Linux sync E2E also passed: 15 passed, 0 failed, 0 ignored; strict all-target/all-feature sync lint and formatting passed. Runtime/wire/automatic operation replay are unchanged. The original failure and measurement boundary are retained in docs/reports/rds-sync-release-fence-20261002.md.

Fixes #79.

Full Ubuntu and both native CI lanes passed. Full macOS CI failed in the existing real-clock admission_pause_does_not_report_encoder_starvation fixture (session.rs:1871, one miss versus expected zero): original job. This PR remains open until that separate clock boundary is repaired and all required checks pass. The failure is not suppressed or counted as green.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

sync cancellation fixture assumes journal release after a fixed sleep

1 participant