cmd/bd has outgrown Go's implicit 10m per-package default. Measured through this wrapper it takes 572s, leaving under 5% margin, so runner variance pushes it over: it failed at 600.343s and 600.728s on two consecutive runs of #6090 while #6038 landed just under. No single test is responsible, so the timeout panic names whichever test was in flight and reads like a hang rather than the package running out of clock. 30m matches every other -race lane in the repo for this same code: main.yml's integration package and cmd/bd shards, ci-measurements.yml, and nightly.yml. Splitting or de-slowing cmd/bd is the real fix and belongs on main; this only stops the clock from being the failure. Co-authored-by: Claude Fable 5 <noreply@anthropic.com> |
||
|---|---|---|
| .. | ||
| go-list-test-names | ||
| lib | ||
| detect-package-gates.sh | ||
| fmt-check.sh | ||
| gen_proxied_shard_manifest.py | ||
| go-test-shard-names.sh | ||
| go-test-shard-packages.sh | ||
| install-dolt.sh | ||
| package-mcp.sh | ||
| package-npm.sh | ||
| pr-core.sh | ||
| pr-lint.sh | ||
| pr-policy.sh | ||
| pull-dolt-image.sh | ||