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

    witty-crayon-22786

    04/11/2016, 7:27 PM
    @happy-kitchen-89482: re: worker_count... in our case, it's memory bound rather than cpu bound, unfortunately
  • l

    limited-country-20626

    04/11/2016, 7:34 PM
    I wonder if that default is sensible. It matches multiprocessing.cpu_count, which includes hyperthreading. On my OSX laptop that is 8. We see all the CPUs pegged with the worker_count set to 4 as it is.
  • w

    witty-crayon-22786

    04/11/2016, 7:45 PM
    @ancient-stone-50795: re https://github.com/pantsbuild/pantsbuild.github.io/pull/6 ... did that go in pantsbuild/pants as well?
  • a

    ancient-stone-50795

    04/11/2016, 8:03 PM
    @witty-crayon-22786: I couldn't find that readme checked in to pantsbuild/pants. I figured its just a placeholder for someone that casually comes across that repo
  • w

    witty-crayon-22786

    04/11/2016, 8:13 PM
    ah, you're right. sorry, looked at that from my phone initially and didn't realize that wasn't a doc page
  • h

    happy-kitchen-89482

    04/11/2016, 9:16 PM
    Yeah, maybe that should be cpu_count() / 2 or something
  • w

    witty-crayon-22786

    04/11/2016, 9:18 PM
    the other trouble is that when you run in a VM environment, sometimes you'll get something ridiculous like 24 CPUs, even though you only have access to a few of them.
  • b

    bored-art-40741

    04/11/2016, 9:24 PM
    I wonder if there's a good library for getting the heuristically "right" answer to this question
  • l

    limited-country-20626

    04/11/2016, 9:32 PM
    Android OS builds use (cpu//2 + 1). Which sounds too high for us still.
  • f

    fast-megabyte-33579

    04/12/2016, 3:51 PM
    does anyone use `annotation_processor`s? I'm trying to work out the right incantations
  • f

    fast-megabyte-33579

    04/12/2016, 3:51 PM
    in particular, if I'm using one in a third-party jar, do I define the annotation_processor in the BUILD file in 3rdparty?
  • f

    fast-megabyte-33579

    04/12/2016, 3:51 PM
    or does it go next to my application sources?
  • f

    fast-megabyte-33579

    04/12/2016, 4:11 PM
    all the examples I can find have the 'jvm_binary' with the dependency on the processor but I need to generate some java sources for another jvm library to depend on
  • f

    fast-megabyte-33579

    04/12/2016, 4:15 PM
    but then it's not clear if I add a dependency on the processor in the child library or the parent one
  • e

    enough-analyst-54434

    04/12/2016, 4:22 PM
    annotation_processor
    is widely misunderstood I think - that target is only if you have local java code implementing an annotation processor, it has nothing to do with using a 3rdparty annotation processor.
  • e

    enough-analyst-54434

    04/12/2016, 4:23 PM
    When using a 3rdparty annotation processor, generally the dependencies are all correct if the code that uses that annotation the processor processes has a dep on the jar providing the annotation - everything else should just work.
  • e

    enough-analyst-54434

    04/12/2016, 4:24 PM
    The only time this wouldn't be the case is for a standard annotation - say provided by the jre lib. In those cases you might need an explicit dep on an annotation processor 3rdparty lib, and - IMO - that dep belongs in all libraries that use the annotation to be processed.
  • f

    fast-megabyte-33579

    04/12/2016, 4:28 PM
    thanks john.sirois - I'm using the 
    mysema
    QueryDSL annotation processor that reads hibernate annotations on a class called Foo (in java_library named 
    entity
    ) and produces a source for a class called QFoo. This QFoo is then depended on by some other library (scala_library named 
    repository
    ). In this case, does the repository just have a dependency on the annotation processor jar while the entities have a dependency on hibernate?
  • e

    enough-analyst-54434

    04/12/2016, 4:28 PM
    Its personal taste to be sure
  • f

    fast-megabyte-33579

    04/12/2016, 4:28 PM
    (We're converting a maven project that ran this annotation processor and wrote into generated-sources)
  • e

    enough-analyst-54434

    04/12/2016, 4:30 PM
    Since the entities don't need to be processed from their point of view - I think thats probably uncontroversial - they are just structs after all - I'd place the dep on the annotation processor in the code that actually needs the enhancements the annotation processor provides
  • f

    fast-megabyte-33579

    04/12/2016, 4:30 PM
    so how does it know to run the annotation processor?
  • e

    enough-analyst-54434

    04/12/2016, 4:30 PM
    magic
  • f

    fast-megabyte-33579

    04/12/2016, 4:30 PM
    I don't see any lines in the pants output
  • e

    enough-analyst-54434

    04/12/2016, 4:31 PM
    java 6+ looks for a manifest file on the classpath
  • e

    enough-analyst-54434

    04/12/2016, 4:31 PM
    if it finds them, all noted annotation processors run
  • e

    enough-analyst-54434

    04/12/2016, 4:31 PM
    Yeah - pants is unaware of the java-level manifest detection and activation
  • f

    fast-megabyte-33579

    04/12/2016, 4:31 PM
    ah interesting
  • e

    enough-analyst-54434

    04/12/2016, 4:32 PM
    You can run javac so that it doesn't use automagic, but pants does not do that today
  • f

    fast-megabyte-33579

    04/12/2016, 4:33 PM
    I guess that means the processor doesn't spit out the classes into a separate jar
1...115211531154...2420Latest