Skip to content

Numpy-dev failures in master as of 2016-10-18 #5419

Closed

Description

@pllim

As reported by https://travis-ci.org/astropy/astropy/jobs/168654883 . I got it to pass a week or two ago after fixing Cone Search test, but it is failing from other packages now, so must be related to some PRs merged since then.

Activity

  1. added
    Affects-devPRs and issues that do not impact an existing Astropy release
    on Oct 18, 2016
  2. mhvk commented on Oct 18, 2016

    @mhvk
    Contributor

    Really weird, it looks like some of the problems are due to numbers being formatted differently, possibly because a default format gets set in some remote-data test.

  3. mhvk commented on Oct 18, 2016

    @mhvk
    Contributor

    Most errors are actually due to a change in numpy-dev, which causes record arrays to be formatted differently: numpy/numpy#8160

  4. changed the title [-]Remote data failures in master as of 2016-10-18[/-] [+]Numpy-dev failures in master as of 2016-10-18[/+] on Oct 18, 2016
  5. pllim commented on Oct 18, 2016

    @pllim
    MemberAuthor

    In that case, what is the best way to fix this? Is there a "only run this doctest if numpy is version such and such"?

  6. mhvk commented on Oct 18, 2016

    @mhvk
    Contributor

    It turns out there is another numpy-dev issue as well (allcose(np.nan, np.nan) no longer returns True), so I think best is to have a PR that fixes this. Working on it...

  7. mhvk commented on Oct 18, 2016

    @mhvk
    Contributor

    @astrofrog - on these "allowed failures" -- would it be possible to autogenerate an e-mail (or issue?) if something is merged in master that fails?

  8. mhvk commented on Oct 20, 2016

    @mhvk
    Contributor

    See #5423 for what hopefully will fix the remaining numpy-dev problems...

  9. bsipocz commented on Oct 23, 2016

    @bsipocz
    Member

    @mhvk - I believe it's possible, remind me about it please in a week time (or open an issue somewhere, I can be assigned).

  10. bsipocz commented on Oct 23, 2016

    @bsipocz
    Member

    So for the first look it's seems possible to get e-mails/gitter/slash/etc notifications, but they will be sent out based on the whole travis status and not individual builds.
    I'll contact travis to confirm this, and to point to us to a workaround if there is any.

  11. astrofrog commented on Oct 31, 2016

    @astrofrog
    Member

    @mhvk @bsipocz - what about using something like https://gist.github.com/astrofrog/061aed7312da223f76ba38e096981780 as a cron job? (maybe showing only the message when it fails)

  12. astrofrog commented on Oct 31, 2016

    @astrofrog
    Member

    Looks like this can be closed - we should open a new issue if there are other failures. Thanks @mhvk!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

Affects-devPRs and issues that do not impact an existing Astropy releasenumpy-devtesting

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions