SanjayVas
commented
3 years ago Sample failing run: https://github.com/world-federation-of-advertisers/panel-exchange-client/runs/2623308247
This appears to show a segfault in the Java Runtime Environment. The fact that this seems to be resolved by using a manual workflow with a different cache key construction implies that the issue may have been a poisoned cache. This can occur when there's been an incompatible change to the build environment that's not accounted for by neither the cache key construction nor Bazel's caching mechanisms, such as changing the C++ compiler. See #8.
If that is indeed what's causing the segfault, then the solution is to specify a new cache-version parameter to the action.
Issue: Some issue with bazel caching causing tests to fail.
Details: For this PR (https://github.com/world-federation-of-advertisers/panel-exchange-client/pull/13) (and a couple other branches), the builds inconsistently fail even when just adding comments. See the commit history for plenty of examples when it switches from failing to passing.
Solution: It appears to have been resolved after removing world-federation-of-advertisers/actions/bazel-build-test@v1 from that repo. See this commit that appears to have resolved the issue: https://github.com/world-federation-of-advertisers/panel-exchange-client/pull/13/commits/fc15b7bb58983498ed0ac6ba8e974e0c0b6bc861
It could also be that there is some other transient issue causing the problem.