gravitystorm left a comment (openstreetmap/openstreetmap-website#6151)
I agree with @hlfan on this. It sort of vaguely makes sense, in principle, to
cancel the other jobs when one fails, except for two key things that happen in
reality:
* Our jobs only fail at the end. We run the entire test suite to completion,
rather than failing early if a single test fails (i.e. `bundle exec rails test`
not `bundle exec rails test --fail-fast`. So, as @hlfan discusses, it's likely
that if one job fails, the other jobs are very nearly close to finishing anyway.
* Our jobs usually fail for "transient bug in test suite" reasons, much more
than "this PR has a coding mistake" reasons. I use the "re-run failed tests"
options all the time, both on PRs and on our master branch, when I see that
there was a CI failure. So when one spurious job failure cancels the others,
we're re-running jobs that were definitely nearly finished and probably were
going to pass. (One caveat is that I don't know how often the coveralls
failures would affect all three jobs completing close together, maybe the
failures will be correlated on that).
So thanks @hlfan , merged.
--
Reply to this email directly or view it on GitHub:
https://github.com/openstreetmap/openstreetmap-website/pull/6151#issuecomment-3303260329
You are receiving this because you are subscribed to this thread.
Message ID:
<openstreetmap/openstreetmap-website/pull/6151/[email protected]>
_______________________________________________
rails-dev mailing list
[email protected]
https://lists.openstreetmap.org/listinfo/rails-dev
