-
Notifications
You must be signed in to change notification settings - Fork 133
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
Update dependency pinnings to latest versions and enable circleci #253
Conversation
@@ -1,7 +1,7 @@ | |||
anaconda-client=1.6.* | |||
argh=0.26.* | |||
beautifulsoup4=4.6.* | |||
conda=4.3.32 | |||
conda=4.3.31 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could be that conda-forge/conda-feedstock#36 is only waiting on Travis CI..
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Indeed.
This will automatically cause Python 3.6 to be used for the base build system -- not a bad thing, just an FYI. |
Yes, I have discovered this as well. I think that should be fine. |
I will make all those test cases that fail now due to 3.6 agnostic of the python version. Just waiting for all tests to finish first. |
Actually, I think this won't use 3.6 automatically. Instead, installing the package should use whatever is the default python version in the given conda installation. |
Correct, but since we use |
Ok, so somebody must have updated it. Originally, I had set it to the latest Py35 version such that no new python has to be installed when setting up travis jobs. Good catch. One more reason to make bioconda-utils flexible in this regard. |
That would be me in #204, due to |
So, prior to #204 we had Currently we have After this PR we would have So overall, since we prioritize |
Ah, of course the channel priorities :-). You are right, it does not really make a difference. |
At some point in the future we could be using some As for this PR: |
Thanks for the ping @mbargull. Yeah I hope that installer will be useful for speeding up CIs. Would be interested in any feedback any of you may have. Ultimately will want to move it to an org. Have tried out some version of them on all OSes. There is some speculation as to whether it (or something like it) could be a community installer. Though it is a bit controversial. Would point to the thread following this comment for more details. |
For speeding up the CI, we could also make use of travis caching mechanisms. Seems to me this would be equivalent to using an installer. Am I missing something? |
When using the macOS 10.10 image, caching incurs 3-5mins of setup time in my experience. That occurs even if the cache is empty. Have seen it go upwards of that as well. My last build using the cache on miniforge used ~5mins to setup the cache. Installing and upgrading Miniconda from scratch takes ~2mins. So caching is actually slightly worse than just using Miniconda and upgrading. Have pressed Travis CI repeatedly to rebuild the image to avoid this overhead, but they don't appear to be interested. Issue ( travis-ci/travis-ci#9027 ) is relevant here. Caching still runs into the same download count issues that another installer would except that we don't have a good alternative metric for how often we have restored the installer from cache. So there isn't a good way to return that info to Anaconda. IOW even if the time improved, there is a downside to this approach vs. building an installer and using it. |
@daler @bgruening I am unsure who of you added the those secret variables:
to the build-docs.sh script. They are defined via the Travis CI interface. As I am unable to migrate them to CircleCI, would you mind putting them here? |
@daler, @bgruening, we also need |
@johanneskoester I had added that, that's for the encrypted private key generated by travis-ci CLI (key.enc for pushing docs. The encryption label is https://github.com/bioconda/bioconda-utils/blob/master/.travis.yml#L16, but since it was originally encrypted using the travis-ci CLI I will have to make changes for circle-ci. Looks to be mostly similar to the travis-ci setup, https://github.com/circleci/encrypted-files. I will get to this later today. |
Never mind, I am working on this already. Almost done! |
Thanks! Sorry I've been so busy lately, I've been following all your work though. |
.circleci/config.yml
Outdated
- *setup_docker | ||
- add_ssh_keys: | ||
fingerprints: | ||
- f8:26:86:86:f8:d3:a5:66:ea:7d:f6:42:2e:5c:7a:82 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@daler actually using custom SSH keys seems pretty easy in circleci compared to what we had before. Very convenient.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, that is really convenient!
The only previously failing test case passes now. Hence, I will merge this now fast, for reasons outlined in our @bioconda/core gitter chat. |
Further, this PR removes the Python pinning, such that we can build for all supported version of python 3