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

    bored-art-40741

    04/28/2016, 10:29 PM
    Otherwise the cycle exception won't trigger
  • b

    bored-art-40741

    04/28/2016, 10:30 PM
    We have a task early on that does this as part of rule validation
  • b

    bored-art-40741

    04/28/2016, 10:30 PM
    But for performance reasons we don't set it in many other places
  • j

    jolly-chef-92794

    04/28/2016, 10:30 PM
    I’m seeing the infinite recursion trigger at the start of 
    simple_codegen_task
    when it tries to do the invalidation check
  • j

    jolly-chef-92794

    04/28/2016, 10:31 PM
  • j

    jolly-chef-92794

    04/28/2016, 10:31 PM
    I’m somewhat surprised there’s no topsorts before codegen
  • j

    jolly-chef-92794

    04/28/2016, 10:32 PM
    it infinite loops in the invalidation check when it tries to compute the transitive invalidation hash
  • j

    jolly-chef-92794

    04/28/2016, 10:33 PM
    I’ll post an issue I think; at the very least we can just call topsort at the top of codegen
  • j

    jolly-chef-92794

    04/28/2016, 10:36 PM
  • e

    enough-analyst-54434

    04/28/2016, 10:37 PM
    Does the cycle have synthetic targets in the mix?
  • e

    enough-analyst-54434

    04/28/2016, 10:38 PM
    I too have the expectation a sort with cycle check happens very early, but that would not cover injected targets ... though I think insertion in the graph covers those
  • e

    enough-analyst-54434

    04/28/2016, 10:38 PM
    tis odd
  • b

    bored-art-40741

    04/28/2016, 10:40 PM
    Probably fingerprint computation should pass through a chain so it can detect a cycle
  • w

    witty-crayon-22786

    04/28/2016, 10:55 PM
    hm... https://github.com/pantsbuild/pants/pull/3295/commits/54d139687d2ad86c0f9fa4a4cd312561ed4bb183 seems like it should be sufficient to get travis to pick up jdk8, but it doesn't seem to be.
  • w

    witty-crayon-22786

    04/28/2016, 10:55 PM
    that doesn't need to be committed to master, right? the edit in my branch should be sufficient?
  • e

    enough-analyst-54434

    04/28/2016, 10:55 PM
    Do you know for a fact travis linux has jdk8?
  • e

    enough-analyst-54434

    04/28/2016, 10:56 PM
    plus 
    language: python
  • e

    enough-analyst-54434

    04/28/2016, 10:57 PM
    I think you can only say travis-order me up XvY for primary environment X
  • e

    enough-analyst-54434

    04/28/2016, 10:57 PM
    You might try apt-get - @limited-country-20626 pioneered use of this exp feature for android
  • e

    enough-analyst-54434

    04/28/2016, 10:58 PM
    ... and most importantly, I think it makes more sense to chip-in on Jenkisfile / jenkins.pantsbuild.org instead
  • e

    enough-analyst-54434

    04/28/2016, 10:59 PM
    its basically ready for complete cutover - @bored-art-40741 not sure if you agree
  • w

    witty-crayon-22786

    04/28/2016, 10:59 PM
  • w

    witty-crayon-22786

    04/28/2016, 10:59 PM
    will iterate.
  • e

    enough-analyst-54434

    04/28/2016, 10:59 PM
    Yes, but that is primary lang
  • e

    enough-analyst-54434

    04/28/2016, 10:59 PM
    I think thats the issue you're hitting
  • w

    witty-crayon-22786

    04/28/2016, 10:59 PM
    i see
  • e

    enough-analyst-54434

    04/28/2016, 10:59 PM
    we mark python primary
  • e

    enough-analyst-54434

    04/28/2016, 11:00 PM
    ie: on the osx travis ci
  • e

    enough-analyst-54434

    04/28/2016, 11:00 PM
    I need to manually upgrade python
  • w

    witty-crayon-22786

    04/28/2016, 11:00 PM
    ok... i'm fine with editing jenkins if i can find it.
1...121412151216...2420Latest