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

    rough-minister-58256

    07/29/2015, 7:13 PM
    oh it exists alright
  • r

    rough-minister-58256

    07/29/2015, 7:14 PM
    hopefully not for long tho
  • w

    witty-crayon-22786

    07/29/2015, 7:14 PM
    aye.
  • e

    enough-analyst-54434

    07/29/2015, 7:14 PM
    Excellent!
  • w

    witty-crayon-22786

    07/29/2015, 7:22 PM
    ...it's 100% reproducible with
  • w

    witty-crayon-22786

    07/29/2015, 7:22 PM
    Copy code
    -interpreter_requirement: CPython>=2.7,<3
+interpreter_requirement: CPython>=2.7.7,<3
  • e

    enough-analyst-54434

    07/29/2015, 7:25 PM
    If you can turn up PEX_VERBOSE=5 and provide before/after logs in an issue, this should make it clear what is getting added to the chroot path in each python version and help debug this.
  • e

    enough-analyst-54434

    07/29/2015, 7:26 PM
    Pex logs the environment it creates and runs with in quite good detail
  • w

    witty-crayon-22786

    07/29/2015, 7:28 PM
    PEX_VERBOSE=5 doesn't seem to output anything during pytest.run.chroot, but PEX_VERBOSE=1 does... sec
  • e

    enough-analyst-54434

    07/29/2015, 7:29 PM
    That is surprising
  • w

    witty-crayon-22786

    07/29/2015, 7:29 PM
    ...hm, no. neither does.
  • e

    enough-analyst-54434

    07/29/2015, 7:30 PM
    SO, to narrow - this is in fact an issue with python test chroots?
  • w

    witty-crayon-22786

    07/29/2015, 7:30 PM
    yes
  • w

    witty-crayon-22786

    07/29/2015, 7:30 PM
    and thrift
  • w

    witty-crayon-22786

    07/29/2015, 7:31 PM
    i was very, very close to releasing 25c930b85cb14eed177c97a6eb4e73c2a2c5d2af, but was seeing non-determinism in whether the chroot was set up correctly for pytest
  • e

    enough-analyst-54434

    07/29/2015, 7:34 PM
    pants master certainly logs under test runnig with PEX_VERBOSE
  • e

    enough-analyst-54434

    07/29/2015, 7:34 PM
    so we need to solve that discrepency 1st
  • w

    witty-crayon-22786

    07/29/2015, 7:34 PM
    ok... could be stripped somewhere in our script. will look.
  • e

    enough-analyst-54434

    07/29/2015, 7:34 PM
    PEX_VERBOSE=5 ./pants test tests/python/pants_test/option/
  • e

    enough-analyst-54434

    07/29/2015, 7:34 PM
    run that on pants master
  • w

    witty-crayon-22786

    07/29/2015, 7:36 PM
    yep
  • w

    witty-crayon-22786

    07/29/2015, 7:39 PM
    still looking into where that's getting stripped.
  • e

    enough-analyst-54434

    07/29/2015, 7:39 PM
    The pex 1.0.1 upgrade from pex OLD specifically needed to change PytestRun in a way that if not merged correctly, would squelch PEX*
  • w

    witty-crayon-22786

    07/29/2015, 7:40 PM
    hm
  • e

    enough-analyst-54434

    07/29/2015, 7:40 PM
  • w

    witty-crayon-22786

    07/29/2015, 7:40 PM
    we're using 
    os.execv
    , which as far as i can tell passes environment variables through
  • w

    witty-crayon-22786

    07/29/2015, 7:40 PM
    looking
  • e

    enough-analyst-54434

    07/29/2015, 7:40 PM
  • w

    witty-crayon-22786

    07/29/2015, 7:41 PM
    this is our first upgrade since 1.0.1
  • e

    enough-analyst-54434

    07/29/2015, 7:42 PM
    OK - as long as your code has the manual subprocess.Popen I pointed to there, it should be good on passing PEX_* through
1...909192...2420Latest