https://pantsbuild.org/ logo
Powered by
#development
  • b

    billowy-beach-68498

    09/26/2016, 5:48 PM
    this morning I was doing some testing regarding Nick’s comment on my address mapper review. I changed the BuildFileAddressMapper.scan_specs to return “set" instead of “OrderedSet" (reasoning for doing so is in the review: https://rbcommons.com/s/twitter/r/4239/, @witty-crayon-22786 may need your insights). When I run ci on this change, I saw “import error” (v2 engine is not enabled). I restart the build once, it’s green now. I am trying the third time, it is still running.
  • r

    rough-minister-58256

    09/26/2016, 5:49 PM
    hmm
  • r

    rough-minister-58256

    09/26/2016, 5:49 PM
    very strange
  • w

    witty-crayon-22786

    09/26/2016, 5:55 PM
    what import error
  • r

    rough-minister-58256

    09/26/2016, 5:57 PM
    Yujie’s been seeing an issue where pytest arbitrarily fails to import tests
  • w

    witty-crayon-22786

    09/26/2016, 5:58 PM
    is it arbitrary, or an actual undeclared import?
  • b

    billowy-beach-68498

    09/26/2016, 5:58 PM
    it is arbitrary I think. different shards failed in different runs
  • r

    rough-minister-58256

    09/26/2016, 6:00 PM
    and retrying the shard can make it green
  • b

    billowy-beach-68498

    09/26/2016, 6:01 PM
    I start to see this error when I enable v2 engine for all executions in pants (in goal_runner, always take v2 path).
  • w

    witty-crayon-22786

    09/26/2016, 6:01 PM
    to be clear, that doesn't mean that it is not a undeclared import
  • w

    witty-crayon-22786

    09/26/2016, 6:01 PM
    but non-determinism isn't fun
  • r

    rough-minister-58256

    09/26/2016, 6:02 PM
    its not failing on inner imports - but rather failing to import the test functions themselves from the test modules
  • r

    rough-minister-58256

    09/26/2016, 6:02 PM
    could potentially still be related to undeclared imports tho I suppose
  • r

    rough-minister-58256

    09/26/2016, 6:03 PM
    need a local repro
  • w

    witty-crayon-22786

    09/26/2016, 6:03 PM
    ah
  • w

    wide-energy-11069

    09/26/2016, 6:04 PM
    reminds me of cffi has conditional imports, e.g. if based on os, then we may not be able to catch that locally on osx
  • r

    rough-minister-58256

    09/26/2016, 6:17 PM
    that should be relatively deterministic tho
  • w

    wide-energy-11069

    09/26/2016, 6:18 PM
    ah that’s right
  • f

    fancy-queen-20734

    09/26/2016, 7:36 PM
    The changed integration tests fail if you have changes in your working copy. It’s a little frustrating because you can’t pass the full engine tests without committing changes.
  • w

    witty-crayon-22786

    09/26/2016, 7:38 PM
    mm. i know that there are examples of creating temporary git repos / working copies
  • r

    rough-minister-58256

    09/26/2016, 7:56 PM
    yeah.. definitely a wart
  • r

    rough-minister-58256

    09/26/2016, 9:06 PM
  • r

    rough-minister-58256

    09/26/2016, 9:11 PM
    I wonder if we should kill 
    ./pants —enable-v2-engine list
    ==
    ./pants —enable-v2-engine list ::
    now while we can
  • r

    rough-minister-58256

    09/26/2016, 9:12 PM
    and just make 
    ./pants —enable-v2-engine list
    throw a “no targets defined” error
  • r

    rough-minister-58256

    09/26/2016, 9:13 PM
    any opinions?
  • w

    witty-crayon-22786

    09/26/2016, 9:14 PM
    yes please.
  • w

    witty-crayon-22786

    09/26/2016, 9:14 PM
    as discussed with @billowy-beach-68498 , i think we should deprecate the no-args list entirely.
  • w

    witty-crayon-22786

    09/26/2016, 9:15 PM
    so, deprecate for v1 now.
  • w

    witty-crayon-22786

    09/26/2016, 9:15 PM
    and make an error for v2 immediately.
  • b

    billowy-beach-68498

    09/26/2016, 9:15 PM
    for v1 we can output empty for ./pants —enable-v2-engine list
1...173174175...906Latest