Clarification regarding the GitHub Actions pre-submit testing
ioi.lam at oracle.com
Fri Mar 19 00:28:38 UTC 2021
One problem with GHA is the frequent false negatives. I think if we can
reduce that, people will be more willing to trust the results of the GHA
 Intermittent failures due to machine configuration. Some of those
are network related (e.g., multi casting). Some are related to user
privileges (process attachment). Etc.
My feeling is that such environment-sensitive tests are a very minor
part of the overall tests. Maybe we can have a GHA-specific problem list
to exclude these tests?
If you are touching these areas, you will need to test your changes
using other means. However, I think this affects only a small number of
 Random build failures, apparently due to toolchain variation. I've
seen this quite a few times in the past week.
This doesn't seem to be specific to OpenJDK. I assume many other
projects use GHA as well. How do they handle this situation?
Any other type of issues?
More information about the jdk-dev