-
-
Notifications
You must be signed in to change notification settings - Fork 2.2k
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
Failures in --pre builds #3228
Comments
Travis-CI does fail against the latest pre-release! See #3204 |
Good catch. Haven't had a chance to keep up with the PRs lately. |
@stefanv, how would you feel about allowing failures in the I have a few branches that would show how it looks like on my fork:
The idea would be that contributors can see that their builds are actually passing on supported machines, while the more advanced devs and maintainers can check to see if scikit-image is working on builds such as I don't like the fact that the color is "green", I would have preferred a colour such as "blue", but then again, maybe that isn't ideal either. |
@hmaarrfk whoa, very cool! I'm not sure how I feel about it, but it's a tentative 👍 from me. It makes a lot of sense to not allow other projects to change our status to red. It just means one more tick box for reviewers in the PR checklist. |
If we allow these failures, I am pretty sure no one will ever look at them. Perhaps we should only do the --pre installs on our nightly builds, and have those emails go out to a wider audience? |
The nightly builds seem like an appropriate place for them. |
@stefanv I have more faith in our reviewers than you do. ;) If we have a tick mark, and if we make some reviewer guidelines to always check, I think it's fine to trust people to follow process, just like we trust them to read the diff before merging. |
(Incidentally, the process would be to check allowed failures and the reviewer should raise an issue when a failure is observed) There is probably also a reasonable way to check those failures automatically and notify "the authorities". =P |
I mean, this --pre thing will come to bite us eventually, we know about it, and will work toward fixing it. I think allowing temporary failures in the --pre builds while we undergo those fixes is a good way to go about it. Disabling and enabling the "allowed failures" is only one line of travis/appveyor script. |
The issue of what to do about |
I am closing this as #3222 was merged and the discussion related to Travis is now outdated. If desired, please reopen a new issue about potentially allowing |
This first showed up in the windows builds failing.
this was due to to the fact that the --pre flag was enabled by default.
PR #3222 attempts to make this more obvious by dedicating a --pre build to the build matrix.
I have a hunch it is due to numpy 1.15.rc1
On travis this wasn't showing up for some reason. I triggered a build on my fork of master (updated) and the dedicated --pre build fails to pull in the numpy 1.15.rc1 dependency from upstream https://travis-ci.org/hmaarrfk/scikit-image/jobs/397451046#L861
When this PR builds, #3227, travis will pull in prereleases from pypi too the problem should show up for travis too.
I've found that while the --pre build compiles with numpy 1.15.rc1, it fails some tests on linux.
The text was updated successfully, but these errors were encountered: