https://pantsbuild.org/ logo
Powered by
#announce
  • l

    limited-country-20626

    01/14/2019, 3:28 AM
    uninstalling the older pip package unblocked
  • w

    wide-energy-11069

    01/14/2019, 3:30 AM
    Thanks for the tip! Sounds like a bug, but having pantsbuild.pants system wide seems uncommon. Worth a ticket at least
  • l

    limited-country-20626

    01/14/2019, 3:30 AM
    I am deleting all my caches and rebuilding so 🤞
  • l

    limited-country-20626

    01/14/2019, 3:30 AM
    yeah, if I can get the release building, I will try and re-break it and then file if confirmed
    👍 3
  • l

    limited-country-20626

    01/14/2019, 3:30 AM
    it was making testing the deprecation deeply frustrating
  • w

    wide-energy-11069

    01/14/2019, 3:31 AM
    yeah I can imagine
  • a

    aloof-angle-91616

    01/14/2019, 3:40 AM
    if it’s a regression (if pantsbuild.pants pip installed used to be ok) we should probably test that moving forward. i’m working on adding testing to pre-commit hooks now, some of which fails with a pexrc active (for example) — testing these local things isn’t something to leave to whoever picks the short straw (and there shouldn’t be a short straw) and imo it seems like we really don’t want the releaser’s environment to affect the release process, for security if nothing else. i’m sorry you had to spend so much time on this
  • l

    limited-country-20626

    01/14/2019, 4:26 AM
    well, hard to attribute fully now since I have found missing deps at the heart of that stacktrace
  • l

    limited-country-20626

    01/14/2019, 4:27 AM
    not sure if this is a argument for buildgen or against it
  • l

    limited-country-20626

    01/14/2019, 4:27 AM
    anyway, will put up a PR for the dependency adjustment and see if I can repro the earlier failures after I get the graph sorted
    👍 1
  • w

    witty-crayon-22786

    01/15/2019, 7:21 PM
    @happy-kitchen-89482: re: https://github.com/pantsbuild/pants/issues/7086 ... the docker image is currently being used for the bootstrap shard, but not the 
    wheels
    shard
  • w

    witty-crayon-22786

    01/15/2019, 7:21 PM
    which means that our release wheels aren't backwards compatible
  • w

    witty-crayon-22786

    01/15/2019, 7:22 PM
    @happy-kitchen-89482: i'm thinking of just moving usage of docker to the wheels shard, and leaving the bootstrap shard without docker. @hundreds-father-404: does that affect any of your work?
  • w

    witty-crayon-22786

    01/15/2019, 7:29 PM
    ideally we could have wheel building happen in the same place where we're building the pex (because it's a prerequisite), but we can't then ship those to another shard for testing
  • w

    witty-crayon-22786

    01/15/2019, 7:29 PM
    without first putting them in s3
  • w

    witty-crayon-22786

    01/15/2019, 7:30 PM
    considering doing that anyway though. as it's definitely beneficial to be testing the same binaries we're releasing
  • h

    hundreds-father-404

    01/15/2019, 7:38 PM
    @witty-crayon-22786 yes that will conflict with what I have. How time sensitive is a fix? It looks reasonable to merge in, but the PR is already so big and has taken so many iterations I’d prefer if possible to not have to deal immediately with it After 1-2 more reviews I’m hoping to merge pants3-in-ci. Hopefully by tomorrow?
  • w

    witty-crayon-22786

    01/15/2019, 8:14 PM
    i'd be ok with that.
  • h

    hundreds-father-404

    01/15/2019, 8:30 PM
    Thank you! There are a couple followup CI changes a couple of us are working on - your fix - Danny’s centos6 changes - splitting lint from jvm tests - splitting rust osx tests from platform specific tests - changing -b to positive flag Sounds like your fix will be the first priority
  • h

    happy-kitchen-89482

    01/16/2019, 4:08 AM
    @witty-crayon-22786 Should both be built in the docker image?
  • h

    happy-kitchen-89482

    01/16/2019, 4:08 AM
    I thought the purpose of that image was to build the native code in a lowest-common-denominator environment?
  • w

    witty-crayon-22786

    01/16/2019, 5:54 PM
    @happy-kitchen-89482: it is. but the pex built on the first shards is not what gets published to the world as part of releases
  • w

    witty-crayon-22786

    01/16/2019, 5:54 PM
    the wheels are
  • w

    witty-crayon-22786

    01/16/2019, 5:54 PM
    and the wheels are rebuilt on the wheel shards
  • w

    witty-crayon-22786

    01/16/2019, 5:55 PM
    in practice, the pex that is built for the purposes of making the travis run more efficient only technically needs to be compatible with travis
  • w

    witty-crayon-22786

    01/16/2019, 5:55 PM
    but in theory, testing a thing that is close-to/exactly what will be released is preferable
  • w

    witty-crayon-22786

    01/16/2019, 5:56 PM
    so... yea. using docker for both shards is one option. alternatively, building the wheels on the same shard would also be an option (we already do in order to make the pex)
  • a

    average-vr-56795

    01/16/2019, 6:26 PM
    Do you have any idea why just zipping up the wheels takes multiple minutes? i.e. the setup-py task takes minutes, even though as far as I know it’s basically just running zip…
  • w

    witty-crayon-22786

    01/16/2019, 6:29 PM
    ...does it?
  • w

    witty-crayon-22786

    01/16/2019, 6:30 PM
    um. hm. no, i don't know what it is doing.
1...146147148...206Latest