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

    ancient-stone-50795

    03/08/2016, 6:25 PM
    The entire annotation processor / classpath issue I know. I was hoping that the provided_dependency change (or whatever you want to call it) will help remove annotation processors from the runtime classpath.
  • w

    witty-crayon-22786

    03/08/2016, 6:26 PM
    it might, but there is a more specific fix in the case of annotation processors.
  • a

    ancient-stone-50795

    03/08/2016, 6:26 PM
    Or, as John suggested, remove running annotation processors entirely from the regular compile and then force users to describe annotation processors in specific targes for a separate invocation of javac that just runs annotation processors.
  • w

    witty-crayon-22786

    03/08/2016, 6:26 PM
    that's another flavor of what i suggested above, yea.
  • b

    bored-art-40741

    03/08/2016, 6:26 PM
    Okay, if there are concrete 
    resources()
    targets that are specifically supposed to affect invalidation (and none of them are synthetic), then I don't see the issue with a subclass
  • w

    witty-crayon-22786

    03/08/2016, 6:27 PM
    @bored-art-40741: it's a half measure.
  • b

    bored-art-40741

    03/08/2016, 6:27 PM
    It's a half measure that fully fixes my problem and Eric's problem with master as it is now
  • b

    bored-art-40741

    03/08/2016, 6:27 PM
    And it's easy to back out for a full-measure
  • w

    witty-crayon-22786

    03/08/2016, 6:27 PM
    no, it does not fully fix the problem
  • b

    bored-art-40741

    03/08/2016, 6:28 PM
    I think we have different definitions of "fully fix" here
  • b

    bored-art-40741

    03/08/2016, 6:28 PM
    I am fully fixed if resources targets never invalidate anything except stuff that feeds bundling
  • w

    witty-crayon-22786

    03/08/2016, 6:28 PM
    sure... that's your review above.
  • b

    bored-art-40741

    03/08/2016, 6:29 PM
    I believe Eric is fully fixed in a somewhat unsatisfying way if he has a way to describe resources targets that should not be opted out of zinc fingerprinting, even though it's not the most granular way to do it
  • b

    bored-art-40741

    03/08/2016, 6:29 PM
    Anyway, I'm not going to pull hairs about this--master is how I want it with that patch landed, and I have the ships
  • b

    bored-art-40741

    03/08/2016, 6:30 PM
    If you guys want to focus on getting a more complete treatment of annotation processors before you get an intermediate solution, that's fine with me
  • b

    bored-art-40741

    03/08/2016, 6:30 PM
    Please do put me on any such review
  • b

    bored-art-40741

    03/08/2016, 6:38 PM
    Okay, here's an example of a situation where just changing the resource, because of the mechanics of 
    TaskIdentityFingerprintStrategy
    , is probably too aggressive about invalidation: 
    jvm-platform-validate
  • b

    bored-art-40741

    03/08/2016, 6:38 PM
    That task should almost certainly have a scope restricted fingerprint strategy
  • b

    bored-art-40741

    03/08/2016, 7:39 PM
    Whoa, we don't start uploading zinc artifacts until the entire compile is finished?
  • r

    rough-minister-58256

    03/08/2016, 8:21 PM
    @limited-terabyte-34878: looks like the pytest timeout tests might be broken on OSX CI: https://travis-ci.org/pantsbuild-osx/pants/jobs/114603975
  • a

    ancient-stone-50795

    03/08/2016, 8:50 PM
    @happy-kitchen-89482: I just had a double take:
    Copy code
    // Author: <mailto:benjy@google.com|benjy@google.com> (Benjy Weinberger)
//  Based on original Protocol Buffers design by
//  Sanjay Ghemawat, Jeff Dean, and others.
//
// A proto file used to test the "custom options" feature of google.protobuf.
    I'm working on a patch to protoc
  • r

    rough-minister-58256

    03/08/2016, 8:51 PM
    hah nice
  • h

    happy-kitchen-89482

    03/08/2016, 8:59 PM
    Oh yeah, I implemented custom options in proto2 back in the day
  • h

    happy-kitchen-89482

    03/08/2016, 8:59 PM
    It was a nightmare
  • a

    ancient-stone-50795

    03/08/2016, 9:01 PM
    The ruby generator doesn't emit code for them... I've got a hack week project to try and add support
  • f

    flat-kitchen-27511

    03/08/2016, 9:08 PM
    Some one here just using vim with pants or is it just idea all teh way?
  • a

    ancient-stone-50795

    03/08/2016, 9:28 PM
    I think @mammoth-pharmacist-54215 uses vim
  • m

    mammoth-pharmacist-54215

    03/08/2016, 9:28 PM
    sometimes, I often use pycharm though
  • m

    mammoth-pharmacist-54215

    03/08/2016, 9:28 PM
    depends on which is more convenient
  • m

    mammoth-pharmacist-54215

    03/08/2016, 9:29 PM
    @rough-minister-58256 and @witty-crayon-22786 both use vim I believe.
1...104610471048...2420Latest