Repository navigation
flaky tests for Paramore.Brighter.InMemory.Tests #4475
Description
Describe the bug
We have flaky tests for Paramore.Brighter.InMemory.Tests
Looking at a log run https://github.com/BrighterCommand/Brighter/actions/runs/36523667172/job/109263804578?pr=4462#logs this appears to be the fault:
Failed Paramore.Brighter.InMemory.Tests.Outbox.OutboxCompactionDispatchedOnlyTests.When_compacting_only_dispatched_messages_in_outbox [2 s]
Error Message:
Oldest dispatched message should be removed by compaction
Stack Trace:
at Paramore.Brighter.InMemory.Tests.Outbox.OutboxCompactionDispatchedOnlyTests.When_compacting_only_dispatched_messages_in_outbox() in /home/runner/work/Brighter/Brighter/tests/Paramore.Brighter.InMemory.Tests/Outbox/When_compacting_only_dispatched_messages_in_outbox.cs:line 75
--- End of stack trace from previous location ---
Error: Oldest dispatched message should be removed by compaction
Since compaction relies on a background thread successfully running and removing compacted records from the Outbox, I assume this test is flaky due to timing. We should try to make this more reliable. It may have become more unreliable due to a change
To Reproduce
We have left open: https://github.com/BrighterCommand/Brighter/actions/runs/36523667172/job/109263804578?pr=4462#logs so that you can examine the logs, but many CI runs now fail
Activity
test: run Barrier-based race tests in a non-parallel collection
fix: coalesce InMemoryBox cleanup onto a single background worker
fix: keep the InMemoryBox cleanup worker going when a pass throws
test: pin InMemoryBox compaction cap at the target size
fix: wait for the InMemoryBox cleanup lock instead of dropping compac…