search

conda-forge / conda-smithy

The tool for managing conda-forge feedstocks.
https://conda-forge.org/
BSD 3-Clause "New" or "Revised" License
146 stars 169 forks source link

Update Travis CI template #1942

Closed mfansler closed 1 month ago

mfansler commented 1 month ago

Travis CI linux-ppc64le jobs are currently failing (https://github.com/conda-forge/status/issues/178). Travis support recommended updating the config (#1940). This PR implements those changes.

Checklist

Closes #1940. Should resolve https://github.com/conda-forge/status/issues/178.

Testing

These changes have been tested in https://github.com/conda-forge/r-backbone-feedstock/pull/15. Some relevant documentation of the partial success (and previous failure):

Also testing in https://github.com/conda-forge/r-pki-feedstock/pull/23. See Travis CI jobs for the build. Only 1/2 ppc64le jobs ran on first try. However, restarting individual failed job led to successful start on second try.

Control Tests Just to make sure this wasn't a fluke of timing (maybe the changes weren't the cause?), I also restarted 4 failed jobs without the changes and they still all failed.

Summary

These changes appear to only provide partial resolution. Half of the linux-ppc64le jobs pass, but that is a far better situation than 0% that has been the situation up to now. This seems like returning to the previous status quo where users may need to manually restart them.

I'd recommend we merge this since it at least appears to improve the situation.

h-vetinari commented 1 month ago

Since you report the changes to be working/effective, is there a reason this PR is still a draft?

mfansler commented 1 month ago

@h-vetinari thanks for having an eye out. I updated with the results. Seems like it doesn't fully resolve the situation, but it does at least get some jobs passing again.

jakirkham commented 1 month ago

I've had massive issues with the Travis UI recently (cannot even access the CI logs without tedious workarounds), so any improvement here is welcome.

Interesting. Have been dealing with the same issue for ~1 month now. Reported to Travis, but have no leads yet. Maybe they can find some similarities between our situations to narrow down the problem. Have you already reported this issue?

h-vetinari commented 1 month ago

Have you already reported this issue?

No. I really have no confidence in Travis and spending any time on it makes my hair stand on end. I prefer to switch affected feedstocks to cross-compilation or emulation rather than dealing with Travis any more than absolutely necessary.

That's not to say that I don't welcome improvements there (hence kudos to @mfansler), but it's unlikely to be me pursuing these.

jakirkham commented 1 month ago

Understandable. Am taking the same approach at this point