Skip to content
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

Skip failing tests to allow 0.18.1 release on conda-forge #40

Merged
merged 1 commit into from
Mar 28, 2017

Conversation

lesteve
Copy link
Member

@lesteve lesteve commented Mar 23, 2017

From a scikit-learn developer this is what makes the most sense. It was already discussed in #39 and #24 but to sum up:

  • skip the test on OSX to allow 0.18.1 to be available on conda-forge
  • do not modify the main code so that 0.18.1 has the same behaviour wherever you installed it from (PyPI, conda without conda-forge, conda with conda-forge).

ping @jakirkham @jschueller @ogrisel @amueller.

fixes #38.
closes #39 closes #24.

I added @ocefpaf (since he added himself in #24) and myself to the recipe maintainers.

@lesteve lesteve mentioned this pull request Mar 23, 2017
@lesteve lesteve changed the title Skip failing tests on OSX to allow 0.18.1 release on conda-forge Skip failing tests to allow 0.18.1 release on conda-forge Mar 23, 2017
@conda-forge-linter
Copy link

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe) and found it was in an excellent condition.

recipe/meta.yaml Outdated
sha256: 6f871ab07657becb3cca158dc4af96a1a14c0a360add04664fba325da43537fa
sha256: f861b988089b0ccc74c0be10a105ed2fe34a2a608d6f19e67147b4da26e39ac3
patches:
- disable-0.18.1-failing-tests-on-osx.patch # [osx]
Copy link
Contributor

@astaric astaric Mar 23, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

If I remember correctly, ridge instability was also problematic on linux.

Copy link
Member Author

@lesteve lesteve Mar 23, 2017

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah I had a few screw-ups sorry about that, I think it should be fine now but let's see what the CIs have to say about this.

@lesteve
Copy link
Member Author

lesteve commented Mar 23, 2017

Hmmm the commit message is slightly misleading because that was when I still thought that the tests were failing only on OSX. Please edit it if you use "squash and merge" ...

Alternatively I could amend the commit but I would rather wait to make sure that the CI are green before doing so.

@lesteve
Copy link
Member Author

lesteve commented Mar 23, 2017

Just for the record, the CIs are green Circle Travis.

I am going to amend the commit message and force push.

Add ocefpaf and lesteve as recipe maintainers.
@jschueller
Copy link

Eventhough I not fully agree on not patching a known resolved bug, I think this should go in. Maintainers ?

This was referenced Mar 28, 2017
@jakirkham
Copy link
Member

We need another PR like this for branch master_np1.11. So we can build it against NumPy 1.11.

@lesteve
Copy link
Member Author

lesteve commented Mar 28, 2017

We need another PR like this for branch master_np1.11. So we can build it against NumPy 1.11.

Done in #41.

@jakirkham
Copy link
Member

Thanks.

@amueller
Copy link
Contributor

@jschueller I don't want to insist on this too much, but you are doing exactly that. There are many known fixed bugs in master, and trying to patch them here will probably lead to more bugs given the intricacies of the codebase and interdependencies. So you're patching one of them selectively. Alright. Why this one and not the rest?

@jschueller
Copy link

jschueller commented Mar 28, 2017

@amueller because this one messes with the unit tests that are run during the package build : if I were to patch I'd prefer to actually fixing the bug if possible instead of creating patching to skip it (and that's less work as the patch already exists). Again, I'm also fine with skipping them, that's just not how I usually do things.

@amueller
Copy link
Contributor

It's probably fine to path it, but it's still not as clear-cut.

@lesteve
Copy link
Member Author

lesteve commented Mar 28, 2017

Can someone with the necessary rights please merge this one as well as #41? From the recipe, I am guessing that @amueller and @ogrisel have the necessary rights. If not @jakirkham maybe?

@amueller amueller merged commit f138c0c into conda-forge:master Mar 28, 2017
@lesteve lesteve deleted the update-to-0.18.1 branch March 28, 2017 16:46
@jakirkham
Copy link
Member

FYI you should get rights to this repo in ~24hrs (or less) @lesteve. If you don't, please ping core.

@lesteve
Copy link
Member Author

lesteve commented Mar 28, 2017

Thanks @amueller for merging this one!

FYI you should get rights to this repo in ~24hrs (or less) @lesteve. If you don't, please ping core.

Good to know, thanks @jakirkham.

@jakirkham
Copy link
Member

To put things in perspective w.r.t. releases and patching, I think the real issue that we are running into is not so much which kind of patches we use, but that we are now left to do this after the release has occurred. Maybe we should be doing some more rcs here so we can also see if a build is going to work at conda-forge in addition to allowing downstream users try out new features in scikit-learn. We do have branches setup for this very process. I don't know if this is something you would be open to trying for patch releases. Just something to think about (especially with 0.19 around the corner).

@amueller
Copy link
Contributor

Doing RCs with conda-forge would be really really good.
I'm not sure how close around the corner 0.19 is given how busy Joel and me are right now, but it is something we should do most certainly.

@lesteve
Copy link
Member Author

lesteve commented Mar 28, 2017

Maybe we should be doing some more rcs here so we can also see if a build is going to work at conda-forge in addition to allowing downstream users try out new features in scikit-learn.

Completely agree with this as I commented elsewhere and we will do our best to do this for the 0.19 release.

@lesteve
Copy link
Member Author

lesteve commented Apr 5, 2017

FYI you should get rights to this repo in ~24hrs (or less) @lesteve. If you don't, please ping core.

@conda-forge/core for some reason it does not look like I have the rights on this repo although I was added to the recipe maintainers. Let me know if there is something I need to do.

@jakirkham
Copy link
Member

The team update script has been super flaky of late. Mainly a consequence of growing pains. Restarted it to see if we can get all the way through to scikit-learn (was getting rate limited by GitHub before that).

When I have some more time, I'm planning on replacing it with a web service. That should scale better with our demands, provide better feedback to end users, run more quickly, and be less vulnerable. Was able to add the needed features to conda-smithy and test them. Just need the time to write a web service around it.

xref: conda-forge/conda-forge-webservices#63

@lesteve
Copy link
Member Author

lesteve commented Apr 5, 2017

OK thanks a lot @jakirkham for the details! For completeness there is no hurry, I just happened to check today whether I had the rights.

@jakirkham
Copy link
Member

Thanks for understanding @lesteve. If it comes to it, I can always add you manually. Though I'm hoping we can fix the tooling first.

@lesteve lesteve mentioned this pull request Nov 16, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

update to 0.18.1
6 participants