New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Migrate CI to GH Actions #2964
Migrate CI to GH Actions #2964
Conversation
This one is actually broken: http://nihroadmap.nih.gov/bioinformatics |
Per @jaimergp, I've observed websites that have blacklisted IP ranges for CI services like Travis. The PDB is actually one of them, I think. |
I'm getting back to the OpenCL tests. I think the error in TestOpenCLNonbondedForce is just an overly strict tolerance. The accuracy of energy accumulation depends on the number of threads. For CPU OpenCL, it only uses a very small number of threads. When that's combined with single precision, the accuracy comes out less than what the test is expecting. So that just leaves TestOpenCLDrudeSCFIntegrator to figure out. |
Figured it out. The error didn't happen on computers with more than two cores, which is what made it hard to reproduce. The fixes for both are in #3003. |
devtools/ci/gh-actions/scripts/run_steps_inside_docker_image.sh
Outdated
Show resolved
Hide resolved
The fix is merged. |
I think this is now https://commonfund.nih.gov/bioinformatics, as hinted by archive.org. |
Ohhh it's all green! 😍 |
Yay! Shall I merge this? |
Well, depends on whether you want to deal with this now or in the future:
|
Oh, and I guess we should also delete Azure and Travis from here, or do you want to run them side by side for a while? |
They should get pushed to the
That seems fine.
Up to you what you think is best. |
🎉🎉🎉 |
I don't think we need to run Travis and Azure in parallel for long, but let's wait until this is merged and working before removing them. |
Ok, that's the behavior I implemented, perfect! I'll need you to add these GitHub secrets (S3 authentication) for this to work. They are added here as plain text, then GitHub will take care of the rest.
I'll give it a thought, but for sure this is not blocking now. I'll open an issue so I can have it assigned and I'll revisit it in the future. Same for Gromacs. |
Where exactly do I get those values from? I only have limited experience with AWS. I think Is |
Thanks! I'll go ahead and merge this, and we can see what happens on the next PR. |
Yay! 🎉 |
Looks like S3 deployed without errors, but there are two link checks that are rejected:
They should be added to the ignore list. The Continuum one might need to be replaced with by https://docs.anaconda.org/anaconda/install/. |
It looks like this worked correctly on #3018, including pushing the docs to AWS when it was merged. Shall I go ahead and turn off the Travis and Azure builds? |
I think so! We can always roll back from history if needed. |
I guess that the |
Done! For the moment I haven't changed any files. I just disabled the Travis webhook and the Azure app. If we need to reenable them, it's just a few clicks. |
Phew, this has been fun 😄! Hopefully it provides a solid start to end up with a CI that is useful, comprehensive and extensible to future needs. Sorry about the insane amount of commits, but it took a bit to get all the pieces in place!
This will close #2919
Summary
This PR provides CI for the following OS / architectures:
The matrix gets expanded with:
Other features:
Cron jobs run nightly, by default on all matrix entries. This can be changed if needed.
Comments
CACHE_VERSION
: needed in case cache needs to be cleared. Do not set to something like0
because GHA will mask it believing it's sensitive material, making the output very difficult to read. Better use something likeplkghj
; short, meaningless and hard to find in a useful message. Cache expires after a week anyway, so no need to keep a correlative scheme here.developer.rst
).Before merge
conda-forge
should useconda-forge-pinning
version restraints2018.*
. This should be addressed in a different PR so we can unpin it.