https://pantsbuild.org/ logo
Powered by
#general
  • h

    happy-kitchen-89482

    02/17/2017, 6:28 PM
    But they look OK to me
  • h

    happy-kitchen-89482

    02/17/2017, 6:32 PM
    So now we’re back to “super subtle issues"
  • e

    enough-analyst-54434

    02/17/2017, 6:33 PM
    super-duper, since pants itself uses multiple contribs which only works if this works
  • e

    enough-analyst-54434

    02/17/2017, 6:33 PM
    Could be a pex issue? IIRC - lazy here - it tries to deal with ns packages when assembling a pex chroot
  • e

    enough-analyst-54434

    02/17/2017, 6:34 PM
    - the source part of the builder
  • b

    billions-sundown-76897

    02/17/2017, 6:37 PM
    The Python import thing is my #1 confusion with Pants at the moment
  • h

    happy-kitchen-89482

    02/17/2017, 6:39 PM
    This really interacts badly with pytest’s test gathering mechanism
  • e

    enough-analyst-54434

    02/17/2017, 6:42 PM
    fwixt ~noone uses maven style layout with python tests so makes sense
  • e

    enough-analyst-54434

    02/17/2017, 6:42 PM
    this is why we switched in pants itself
  • h

    happy-kitchen-89482

    02/17/2017, 6:42 PM
    Yeah, IIRC this is why our root package for tests is 
    pants_test
    and not 
    pants
    .
  • e

    enough-analyst-54434

    02/17/2017, 6:42 PM
    yup
  • h

    happy-kitchen-89482

    02/17/2017, 6:42 PM
    But contrib is muddying the waters
  • e

    enough-analyst-54434

    02/17/2017, 6:43 PM
    Well, contrib layout is just (documented convention), I have no objection to changes
  • h

    happy-kitchen-89482

    02/17/2017, 6:44 PM
    I’ll spend another few minutes debugging this, but may end up concluding that it’s not worth the trouble of fighting with pytest
  • h

    happy-kitchen-89482

    02/17/2017, 6:45 PM
    Basically the test discovery code correctly discovers test files, but then can’t import them.
  • h

    happy-kitchen-89482

    02/17/2017, 6:45 PM
    And afaict this is only the case if those test files import from the same source root, although I have yet to verify this
  • e

    enough-analyst-54434

    02/17/2017, 6:46 PM
    contrib/findbugs/tests/python/pants_test/contrib/findbugs/
  • e

    enough-analyst-54434

    02/17/2017, 6:46 PM
    no init.py
  • e

    enough-analyst-54434

    02/17/2017, 6:46 PM
    fix and be merry
  • e

    enough-analyst-54434

    02/17/2017, 6:46 PM
    i SUSPECT
  • e

    enough-analyst-54434

    02/17/2017, 6:46 PM
    whereas contrib/python tree does
  • e

    enough-analyst-54434

    02/17/2017, 6:51 PM
    Pretty sure 
    touch contrib/findbugs/tests/python/pants_test/contrib/findbugs/__init__.py
    heals
  • e

    enough-analyst-54434

    02/17/2017, 6:51 PM
    Looks like a miss in review / initial commit
  • h

    happy-kitchen-89482

    02/17/2017, 6:51 PM
    alas no
  • h

    happy-kitchen-89482

    02/17/2017, 6:52 PM
    let me see what happens if I put a namespace package decl in there
  • e

    enough-analyst-54434

    02/17/2017, 6:52 PM
    and for sanity dump the chroot / don't delete it and inspect what's assembled
  • h

    happy-kitchen-89482

    02/17/2017, 6:52 PM
    Good eye, but no joy
  • e

    enough-analyst-54434

    02/17/2017, 6:53 PM
    to make sure its not a pex builder problem, though I just scanned that code and it looks safe
  • h

    happy-kitchen-89482

    02/17/2017, 6:53 PM
    Hmm, what happens with intermediate 
    __init__.py
    in the pex builder
  • h

    happy-kitchen-89482

    02/17/2017, 6:53 PM
    it creates them?
1...164316441645...2420Latest