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

    limited-country-20626

    03/29/2016, 8:08 PM
    but it would be nice to have it back for list if it was true for list_option
  • w

    witty-crayon-22786

    03/29/2016, 8:09 PM
    @limited-country-20626, @jolly-chef-92794 : sorry for the trouble: https://rbcommons.com/s/twitter/r/3624/
  • j

    jolly-chef-92794

    03/29/2016, 8:10 PM
    Thanks!
  • w

    witty-crayon-22786

    03/29/2016, 9:00 PM
    jvm folk. i'm going to be adding an "unused dep" warning/error, which will suggest a more specific dep in cases where the direct dep is unused, but a transitive one is
  • w

    witty-crayon-22786

    03/29/2016, 9:00 PM
    would anyone object to doing that directly in zinc?
  • w

    witty-crayon-22786

    03/29/2016, 9:00 PM
    er, directly in zinc_compile.py i mean
  • w

    witty-crayon-22786

    03/29/2016, 9:02 PM
    one of the arguments to doing it there is that JvmDependencyCheck requires compilation to have finished before it runs, which was fine for undeclared direct/transitive, but not for unused
  • w

    witty-crayon-22786

    03/29/2016, 9:02 PM
    (i mean, it's fine for unused, but you might end up compiling a ton more code than you needed before you even see the error)
  • w

    witty-crayon-22786

    03/29/2016, 9:03 PM
    and yes, i recognize that we were the ones that moved JvmDependencyCheck out of zinc... i think it was the right thing to do at the time šŸ˜ƒ
  • f

    fast-megabyte-33579

    03/29/2016, 9:10 PM
    that sounds fine to us - anything that keeps the BUILD files clean is okay by me
  • f

    fast-megabyte-33579

    03/29/2016, 9:11 PM
    (we canā€™t turn on 
    strict_deps
    at the moment, thatā€™s a bit of a work in progress)
  • b

    bored-art-40741

    03/29/2016, 9:12 PM
    @witty-crayon-22786: How do you detect when a scala dep is missing due to missing information from transitive deps needed to do type inference?
  • w

    witty-crayon-22786

    03/29/2016, 9:13 PM
    @bored-art-40741 : this is "unused", not undeclared
  • w

    witty-crayon-22786

    03/29/2016, 9:13 PM
    So the zinc analysis should have accurate information on what was used (in theory)
  • b

    bored-art-40741

    03/29/2016, 9:15 PM
    Ah
  • w

    witty-crayon-22786

    03/29/2016, 9:16 PM
    Will definitely need an override/whitelist to indicate that some things shouldn't be marked (TBD what that looks like), but it should be relatively rare... about as rare as zinc incremental compile bugs =p
  • w

    witty-crayon-22786

    03/29/2016, 9:20 PM
    cc @jolly-chef-92794 @ancient-stone-50795
  • a

    ancient-stone-50795

    03/29/2016, 9:20 PM
    will it slow us down if we don't use strict_deps?
  • w

    witty-crayon-22786

    03/29/2016, 9:21 PM
    @ancient-stone-50795: unrelated to strict_deps
  • w

    witty-crayon-22786

    03/29/2016, 9:21 PM
    well, usable independently of strict_deps
  • w

    witty-crayon-22786

    03/29/2016, 9:22 PM
    this is a thing that you/we will almost certainly want on by default
  • a

    ancient-stone-50795

    03/29/2016, 9:22 PM
    I guess I just want to know if we will get a performance penalty
  • a

    ancient-stone-50795

    03/29/2016, 9:22 PM
    I think it would be good for us to know, but I'd like to be able to turn it off and on if that would be possible
  • w

    witty-crayon-22786

    03/29/2016, 9:23 PM
    yea, it'll be possible. will make sure there is no impact when it is disabled
  • c

    cold-lizard-88792

    03/29/2016, 11:23 PM
    sbt people noticed no-op is slow https://github.com/sbt/sbt/issues/2525
  • w

    witty-crayon-22786

    03/29/2016, 11:23 PM
    nice.
  • w

    witty-crayon-22786

    03/29/2016, 11:24 PM
    unfortunately we add about 10x the overhead atop of that... graph caching is coming.
  • h

    happy-kitchen-89482

    03/30/2016, 5:06 AM
    @jolly-chef-92794: I think so. The implementation uses None internally to distinguish ā€œI override the previous values with an empty listā€ from ā€œI specify no value at allā€. But once all that is done and the final value is determined to be None, it might be desirable to reflect that to the user of the option as an empty list. But Iā€™m not sure of the unintended consequences of that. There may be places where we subtly need to tell the difference.
  • h

    happy-kitchen-89482

    03/30/2016, 5:06 AM
    We could change that behavior and see if anything breaks...
  • a

    adorable-glass-89507

    03/30/2016, 9:34 AM
    would it work if I split my tests folder into tests/unit and tests/integration and have the python, java and scala folder under there?
1...110511061107...2420Latest