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

    millions-controller-72965

    03/10/2016, 8:25 PM
    so it looks like it's only extending abis for python 3+
  • m

    millions-controller-72965

    03/10/2016, 8:25 PM
    and not doing the same for python 2+
  • r

    rough-minister-58256

    03/10/2016, 8:27 PM
    yeh that seems problematic - if you’re in a position to fix/test/submit a PR would be happy to take a look.
  • m

    millions-controller-72965

    03/10/2016, 8:27 PM
    ok, i'll take a look at it
  • m

    millions-controller-72965

    03/10/2016, 8:29 PM
    any easy way of configuring pants.ini to use a custom build of pex for testing or would i have to do a custom pants build as well?
  • r

    rough-minister-58256

    03/10/2016, 8:34 PM
    its probably easiest to work entirely in the pex repo w/ the cli using the test model I pasted above
  • r

    rough-minister-58256

    03/10/2016, 8:35 PM
    you can 
    pip install pex
    , and then from the pex repo you can pex pex itself for testing: 
    pex -o /tmp/pex-test.pex requests . -c pex
    and then 
    /tmp/pex-test.pex
    in place of where you’d normally invoke 
    pex
  • r

    rough-minister-58256

    03/10/2016, 8:36 PM
    then once it’s fixed in 
    pex
    , would just be a matter of doing a pex release and bumping the version in pants’ 
    3rdparty/python/requirements.txt
  • r

    rough-minister-58256

    03/10/2016, 10:06 PM
    btw, looks like someone already filed an issue for this exact thing last week: https://github.com/pantsbuild/pex/issues/213
  • f

    fancy-queen-20734

    03/11/2016, 1:02 AM
    @happy-kitchen-89482: you were right that clearing up type coercion rules might break a few incorrectly specified things simple smile https://rbcommons.com/s/twitter/r/3561/
  • h

    happy-kitchen-89482

    03/11/2016, 1:33 AM
    That was @ancient-stone-50795’s insight actually
  • l

    limited-country-20626

    03/11/2016, 3:24 PM
    @adorable-glass-89507 we are going to do a brief info dump of what we learned setting up our new artifact cache. We will also probably share some gists. Give us a little bit to pull it together.
  • a

    adorable-glass-89507

    03/11/2016, 3:33 PM
    @limited-country-20626: thanks, I appreciate it
  • l

    limited-country-20626

    03/11/2016, 3:33 PM
    Sure - we already had it in mind so it is good to have someone interested to keep us motivated simple smile
  • a

    adorable-glass-89507

    03/11/2016, 3:35 PM
    I promise to put it at good use :)
  • l

    limited-country-20626

    03/11/2016, 3:35 PM
    Does our option scoping allow for goal-wide options? Specifically I am thinking for something like declaring 
    --allow-empty
    for all codegen tasks at once instead of having to do it once per subclass. Benjy is that possible?
  • b

    bored-art-40741

    03/11/2016, 7:52 PM
    I'm going to start the release. Ping me if there are any late breaking commits you desperately need to get in!
  • g

    gentle-potato-32845

    03/11/2016, 8:01 PM
    hi friends. i had a question. how does java (maybe ivy?) work when there are classes of identical name and package, but from different 3rdparty jars (like different 
    org
    and 
    name
    )
  • g

    gentle-potato-32845

    03/11/2016, 8:01 PM
    i think keep getting problems from this because of things like jersey/jetty that switch maintainers
  • b

    bored-art-40741

    03/11/2016, 8:07 PM
    Broadly speaking, it doesn't
  • b

    bored-art-40741

    03/11/2016, 8:07 PM
    Ivy resolution and classpath construction is not aware of the contents of jars, only the coordinates of those jars
  • b

    bored-art-40741

    03/11/2016, 8:07 PM
    So if two or more jars provide a symbol on the classpath, the first one (from left to right) to appear on the classpath is the one that gets classloaded
  • b

    bored-art-40741

    03/11/2016, 8:08 PM
    And you generally end up with incredibly scary non-determinism, since this can break at any time if the classpath happens to change order
  • g

    gentle-potato-32845

    03/11/2016, 8:08 PM
    oh wow good to know. i wonder if maven was just doing it in a different order so it worked (in my case)
  • g

    gentle-potato-32845

    03/11/2016, 8:08 PM
    and then it broke in pants
  • g

    gentle-potato-32845

    03/11/2016, 8:08 PM
    its very scary because you get runtime errors :<
  • b

    bored-art-40741

    03/11/2016, 8:08 PM
    Yup
  • b

    bored-art-40741

    03/11/2016, 8:08 PM
    In these circumstances it's very advisable to find such collisions and eliminate them if possible
  • b

    bored-art-40741

    03/11/2016, 8:09 PM
    If you run 
    ./pants bundle <that binary>
    , it has a step that barfs out all of the collisions in yellow
  • g

    gentle-potato-32845

    03/11/2016, 8:09 PM
    tiiight
1...105010511052...2420Latest