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

    freezing-motorcycle-14222

    09/29/2015, 5:12 PM
    @bored-art-40741: The hardest part is detecting the side effects of running spurious tasks (like installing bootstrap tools it didn't need)
  • f

    freezing-motorcycle-14222

    09/29/2015, 5:12 PM
    Also, the cache makes it even harder to detect these things once the cache is warm
  • b

    bored-art-40741

    09/29/2015, 5:12 PM
    So there's one check we do where you specifically want to guard: 
    type(target) == Target
    , which is when you're specifically noting that the target is a "dependencies" target
  • b

    bored-art-40741

    09/29/2015, 5:13 PM
    The problem is that there are tasks which operate on the build graph as a pure structure, and specifically don't care about types or target internals
  • b

    bored-art-40741

    09/29/2015, 5:13 PM
    They only care about the DAG
  • w

    witty-crayon-22786

    09/29/2015, 5:13 PM
    the spurious tasks things should be a non issue post engine refactor
  • w

    witty-crayon-22786

    09/29/2015, 5:13 PM
    the reason they run now is just because the task graph is too coarse
  • w

    witty-crayon-22786

    09/29/2015, 5:13 PM
    not something to be solved here
  • b

    bored-art-40741

    09/29/2015, 5:13 PM
    Agreed, I hinted at this earlier in the conversation
  • b

    bored-art-40741

    09/29/2015, 5:13 PM
    But yeah, it's a bit too far out to discourage people from finding a more immediate solution
  • f

    freezing-motorcycle-14222

    09/29/2015, 5:13 PM
    Yeah, I wasn't planning on fixing that
  • f

    freezing-motorcycle-14222

    09/29/2015, 5:14 PM
    I was just suggesting an easy regression test (the empty target) to verify that the fix works when it's implemented
  • b

    bored-art-40741

    09/29/2015, 5:14 PM
    Gotta fly! Hopefully the plane has wifi
  • l

    limited-country-20626

    09/29/2015, 5:25 PM
    uh oh Stu is jumping to Buck
  • w

    witty-crayon-22786

    09/29/2015, 5:28 PM
    doing a bakeoff this week to check out the competition
  • w

    witty-crayon-22786

    09/29/2015, 5:28 PM
    i am also jumping to bazel and gradle =P
  • l

    limited-country-20626

    09/29/2015, 5:30 PM
    haha, I got the email and was super confused. "How can he want to add scala support to pants?"
  • h

    happy-kitchen-89482

    09/29/2015, 5:36 PM
    I've been playing with deprecated.py, and noticed that if we do the warnings.simplefilter('default') earlier than we currently do (i.e, before imports) we get the following warnings:
  • h

    happy-kitchen-89482

    09/29/2015, 5:36 PM
    Copy code
    /usr/local/Cellar/python/2.7.6/Frameworks/Python.framework/Versions/2.7/lib/python2.7/pkgutil.py:186: ImportWarning: Not importing directory '/Users/benjy/src/pants/build-support/pants_dev_deps.venv/lib/python2.7/site-packages/twitter': missing __init__.py
  file, filename, etc = imp.find_module(subname, path)
/usr/local/Cellar/python/2.7.6/Frameworks/Python.framework/Versions/2.7/lib/python2.7/pkgutil.py:186: ImportWarning: Not importing directory '/Users/benjy/src/pants/build-support/pants_dev_deps.venv/lib/python2.7/site-packages/twitter/common': missing __init__.py
  file, filename, etc = imp.find_module(subname, path)
  • h

    happy-kitchen-89482

    09/29/2015, 5:37 PM
    Shouldn't those be bundled with 
    __init__.py
    ?
  • h

    happy-kitchen-89482

    09/29/2015, 5:37 PM
    I can squelch this by being more selecting about warnings filtering, but wanted to throw out that this might be an issue.
  • w

    witty-crayon-22786

    09/29/2015, 7:33 PM
  • h

    happy-kitchen-89482

    09/29/2015, 7:35 PM
    I'm good with it.
  • h

    happy-kitchen-89482

    09/29/2015, 7:45 PM
    You reverted it?
  • w

    witty-crayon-22786

    09/29/2015, 7:46 PM
    yea... itchy trigger finger... it didn't have a green CI
  • w

    witty-crayon-22786

    09/29/2015, 8:40 PM
    (followup: it did have green CI, and is in now)
  • l

    limited-terabyte-34878

    09/29/2015, 10:45 PM
    i’ve merged all the trivial bits duplicated between junit and pytest into my mixin, and looking at sharding now, it looks like the implementation in pytest is very python-specific. is that correct or am I misreading something?
  • l

    limited-terabyte-34878

    09/29/2015, 10:46 PM
    (it creates this conftest.py which implements the sharding?)
  • w

    witty-crayon-22786

    09/29/2015, 11:01 PM
    @limited-terabyte-34878: could be, yea
  • w

    witty-crayon-22786

    09/29/2015, 11:01 PM
    might be lower benefit to merging those
1...479480481...2420Latest