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

    ancient-stone-50795

    07/28/2015, 8:42 PM
    We floated the idea and asked for advice before she started and got some hand waves on how it might be done. When she was working on it I told her to find a way to make it work, and then we would post it publicly and maybe someone would have a more concrete idea on how to make it better. That is why I asked her to write up a design document and spell out all the different proposals so we could step back and judge their merits better.
  • a

    ancient-stone-50795

    07/28/2015, 8:43 PM
    In other words, What @witty-crayon-22786 said
  • e

    enough-analyst-54434

    07/28/2015, 8:43 PM
    I don't mean to focus on @adorable-hairdresser-18353 's case except that its recent and one potentially negative aspect, this sub-RB then full RB style has been something I've found mildly troubling since fall and just wanted to bring it up
  • e

    enough-analyst-54434

    07/28/2015, 8:44 PM
    It may be good and fine
  • e

    enough-analyst-54434

    07/28/2015, 8:47 PM
    So, in short - this appears to be working for folks. I may be the only one bothered by the 1RB reviews vs 2RB reviews, and my discomfort is certainly vague, so I'm happy with you all continuing to do this. Thanks for the background.
  • a

    ancient-stone-50795

    07/28/2015, 9:00 PM
    So, we got a request from Georges Saab (chairperson of the OpenJDK governing board,) asking what kinds of topics Square might want to present at Java One and Pants came up.
  • a

    ancient-stone-50795

    07/28/2015, 9:02 PM
    Its October 25-29
  • e

    enough-analyst-54434

    07/28/2015, 9:02 PM
    Well, I would be happy to do that if you were looking to hand the invitation off, it would meet one of my summit TODOs
  • a

    ancient-stone-50795

    07/28/2015, 9:04 PM
    Depending on the kind of talk we wanted it to be, it could involve multiple companies. Presented by itself, its just going to be compared to Bazel/Buck which we really don’t have a whole lot to say about. But we do know how its solving several specific problems for 3 different large monorepo use cases.
  • h

    happy-kitchen-89482

    07/28/2015, 9:04 PM
    =@john.sirois
  • h

    happy-kitchen-89482

    07/28/2015, 9:05 PM
    I would strongly encourage mini-design docs or at least design discussions on pants-devel before coding.
  • e

    enough-analyst-54434

    07/28/2015, 9:05 PM
    @ancient-stone-50795: I think a multi-company talk is a great idea
  • e

    enough-analyst-54434

    07/28/2015, 9:06 PM
    That would definitely say something without even directly saying it
  • h

    happy-kitchen-89482

    07/28/2015, 9:06 PM
    oooh, JavaOne talk would be awesome
  • w

    witty-crayon-22786

    07/28/2015, 9:07 PM
    always happy to talk about our multi-jvm-language usecases
  • a

    ancient-stone-50795

    07/28/2015, 9:10 PM
    I could talk about pain points solved for the Java case vs. issues we had in Maven. (May end up being a lot of material!)
  • e

    enough-analyst-54434

    07/28/2015, 9:18 PM
    So should we apply for a bunch of talks, or try to coordinate one with intro + ~3 10 minute segments?
  • a

    ancient-stone-50795

    07/28/2015, 9:18 PM
    I am pretty sure we want 2 talks. One about Pants in general, and another which is the case studies.
  • a

    ancient-stone-50795

    07/28/2015, 9:19 PM
    Hopefully Bazel will also give a talk, it would just draw more attention to the entire subject of build tools for large repos. “There is no such thing as bad publicity)
  • a

    ancient-stone-50795

    07/28/2015, 9:19 PM
    Maybe more than 2 talks, but it would be good to go for at least 2
  • a

    ancient-stone-50795

    07/28/2015, 9:27 PM
    Looks like the CFP is already closed for JavaOne, but maybe we can get in through this back door...
  • j

    jolly-chef-92794

    07/28/2015, 9:31 PM
    I just noticed that at commit 
    a8a73797a50f53fc48c57e1fdaf6b8211e267c67
    (when the last codegen -- ApacheThriftGen -- was moved to SimpleCodegenTask), pants stopped working in square's java repo. Specifically, 
    ./pants clean-all compile
    fails in the round engine:
    Copy code
    Could not satisfy data dependencies for goal 'jvm' with action SingletonGroupTask_compile_jvm: u"No producers registered for 'java'"
    Though the same command works fine in the pants repo at that commit. I suspect this may be because nothing is using code_gen anymore, so nothing has 
    product_types
    of 
    java
    , and indeed adding
    Copy code
    @classmethod
  def product_types(cls):
    return ['java', 'scala', 'python']
    To simple_codegen_task.py fixes it. I can put up a patch for that, but the fact that it still apparently works fine in the pants repo makes me wonder if it's a config problem on my end.
  • w

    witty-crayon-22786

    07/28/2015, 9:32 PM
    @happy-kitchen-89482: ok, it looks like the bug is just that the 
    [cache]
    section in pants.ini doesn't apply as the 
    default
    in ./pants help, but the specific 
    [cache.compile.scala]
    (for example) section does apply in help. the options are actually applied in both cases
  • h

    happy-kitchen-89482

    07/28/2015, 9:33 PM
    hrmph
  • h

    happy-kitchen-89482

    07/28/2015, 9:33 PM
    well, it's on my doc fixit week task list to make help work properly with subsystems
    metal 1
  • w

    witty-crayon-22786

    07/28/2015, 9:35 PM
    shall i file it?
  • w

    witty-crayon-22786

    07/28/2015, 9:35 PM
    the other thing i noticed is that it's possible for options to be both advanced and recursive, but the help chooses one section or the other for them
  • e

    enough-analyst-54434

    07/28/2015, 9:37 PM
    @jolly-chef-92794: grokking...
  • e

    enough-analyst-54434

    07/28/2015, 9:40 PM
    I think the right thing to do for now is to express these product_types in the leaves, not in the baseclass - Codegen was a bad example in that regard
  • e

    enough-analyst-54434

    07/28/2015, 9:40 PM
    So ApacheThriftGen -> ['java']
1...848586...2420Latest