CI: per-test-binary DDS domains + a ReleaseSmall lane - #76
Conversation
There was a problem hiding this comment.
Your trial has ended. Reactivate Greptile to resume code reviews.
|
@greptile review |
Greptile SummaryThe PR isolates participant-creating test binaries onto separate DDS domains, replaces a timing-dependent TCP reconnect check with generation polling, and adds LLVM-backed ReleaseSmall coverage.
Confidence Score: 5/5The PR appears safe to merge; no concrete blocking or independently actionable non-blocking issue was identified. The domain helper validates its input and is consistently propagated through the relevant networked tests, while the new reconnect check observes an actual connection-generation change and the ReleaseSmall lane is wired through the intended LLVM-backed graph. Important Files Changed
|
Codecov Report✅ All modified and coverable lines are covered by tests. 📢 Thoughts on this report? Let us know! |
Motivation
The ARM64 / DebugAllocator / TSan CI lanes flake intermittently: zig build test runs ~29 participant-creating test binaries in parallel and every one binds domain-0's fixed RTPS ports (SPDP multicast 7400, metatraffic unicast), so under a slow lane the losers hit error.BindFailed, discovery stalls, and a loopback test times out. Separately, tcp_transport_test's reconnect-generation assertion depended on a fixed 100 ms sleep that loses a TCP-settling race under load. And -OReleaseSmall was never built or run anywhere (a roadmap CI-coverage gap).
Description of changes