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

    enough-analyst-54434

    09/04/2015, 5:29 PM
    That reminds me - I had an idea on the problem @jolly-chef-92794 is facing on the ride home last night
  • e

    enough-analyst-54434

    09/04/2015, 5:29 PM
    I really think there needs to be a new task that dependens on ivy resolve products and produces shaded jars
  • e

    enough-analyst-54434

    09/04/2015, 5:29 PM
    And then those jars are used by BinaryCreate and BundleCreate
  • e

    enough-analyst-54434

    09/04/2015, 5:30 PM
    in addition to the manual shading they do via base class on the code that lives in the repo
  • w

    witty-crayon-22786

    09/04/2015, 5:30 PM
    that would also solve our primary usecase for shading (inbound)
  • e

    enough-analyst-54434

    09/04/2015, 5:30 PM
    Bonus - artifact cache
  • e

    enough-analyst-54434

    09/04/2015, 5:30 PM
    because this stuff is expensive
  • j

    jolly-chef-92794

    09/04/2015, 5:30 PM
    Definitely expensive.
  • w

    witty-crayon-22786

    09/04/2015, 5:30 PM
    ...because actually, by definition, if it's source living in our current repo, does it need to be shaded?
  • w

    witty-crayon-22786

    09/04/2015, 5:31 PM
    (scratch "by definition")
  • e

    enough-analyst-54434

    09/04/2015, 5:31 PM
    for export perhaps
  • e

    enough-analyst-54434

    09/04/2015, 5:31 PM
    but agreed - unlikely for local binaries
  • e

    enough-analyst-54434

    09/04/2015, 5:31 PM
    s/unlikely/crazytown if so?/
  • e

    enough-analyst-54434

    09/04/2015, 5:31 PM
    I can't imagine the case where you have to shade local code from itself
  • w

    witty-crayon-22786

    09/04/2015, 5:32 PM
    although, running on a hadoop cluster that uses libraries from your repo might be common.
  • e

    enough-analyst-54434

    09/04/2015, 5:32 PM
    but that's moot - you need to run the shader against it to update its references to shaded 3rdparty
  • w

    witty-crayon-22786

    09/04/2015, 5:32 PM
    er
  • e

    enough-analyst-54434

    09/04/2015, 5:32 PM
    ack on that - good example
  • w

    witty-crayon-22786

    09/04/2015, 5:32 PM
    yea
  • a

    ancient-stone-50795

    09/04/2015, 5:33 PM
    That would certainly be a selling point for pants in our CI.
  • b

    bored-art-40741

    09/04/2015, 5:34 PM
    I'd love to be able to specify shading in our 3rdparty/BUILD
  • b

    bored-art-40741

    09/04/2015, 5:34 PM
    Specifically for several misbehaved hadoop dependencies (looking at you hive-exec)
  • j

    jolly-chef-92794

    09/04/2015, 5:34 PM
    So I was actually fixing up my patch to only shade external jar dependencies individually if they (1) aren’t bundled into the output jar, and (2) their shading pattern depends on the jar id. Because it actually is about 2x faster if you can run one shading step against the whole monolithic jar. A separate shading step for individual external dependencies could thus potentially slow things down, but the artifact cache would obviously help a lot on incremental builds.
  • e

    enough-analyst-54434

    09/04/2015, 5:34 PM
    yes
  • e

    enough-analyst-54434

    09/04/2015, 5:35 PM
    I think the artifact cache saves things
  • e

    enough-analyst-54434

    09/04/2015, 5:35 PM
    I would not be opposed to a land current with fixes, backpedal with new task in follow up RB
  • e

    enough-analyst-54434

    09/04/2015, 5:35 PM
    I only really care master is not broken by the current approach
  • j

    jolly-chef-92794

    09/04/2015, 5:36 PM
    Okay. I’ll push more changes on it soon then; I’m just debugging some test failures first.
  • w

    witty-crayon-22786

    09/04/2015, 5:39 PM
    virtual cupcakes for @adventurous-twilight-44110 's last day: 🍰
  • e

    enough-analyst-54434

    09/04/2015, 5:39 PM
    Really impressive breadth and depth in your internship Cody
1...291292293...2420Latest