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

    ancient-stone-50795

    08/25/2015, 5:15 PM
    So if you specify tests-jar, it pulls down 
    <artifactId>-<version>-tests.jar
  • a

    ancient-stone-50795

    08/25/2015, 5:15 PM
    Yes, that is true, it can pull in other test-jar artifacts transitively. We had a problem with that in one artifact. It was an internal artifact so we ended up just moving some crap around.
  • b

    bored-art-40741

    08/25/2015, 5:16 PM
    Or maybe better said: I'm concerned that wanting a single test jar which I can at the top level just transform to be a classifier won't work, because that test jar is assuming that its dependencies are resolved with scope 
    test
  • b

    bored-art-40741

    08/25/2015, 5:16 PM
    Maybe this is not an issue in practice
  • a

    ancient-stone-50795

    08/25/2015, 5:17 PM
    If you pull in a jar with test-jar, you probably need to process all of the dependencies in scope <test> transitively when you download that pom.xml file.
  • a

    ancient-stone-50795

    08/25/2015, 5:17 PM
    This doesn’t work in Pants right now (with Ivy).
  • b

    bored-art-40741

    08/25/2015, 5:18 PM
    Fascinating, it doesn't?
  • b

    bored-art-40741

    08/25/2015, 5:19 PM
    I'm not sure if I'm more tempted to feature-leapfrog Ivy or to declare that this whole idea is silly and ban depending on third party test jars internally
  • a

    ancient-stone-50795

    08/25/2015, 5:20 PM
    It doesn’t because of just the reason you say. Ivy doesn’t know you are pulling in that one jar in ‘test scope’ so doesn’t necessarily pull in the transitive dependencies you need.
  • a

    ancient-stone-50795

    08/25/2015, 5:20 PM
    Our workaround for this one jar was to just explicitly list those dependencies in the target in the repo.
  • b

    bored-art-40741

    08/25/2015, 5:22 PM
    Hey
  • b

    bored-art-40741

    08/25/2015, 5:22 PM
    I love that solution
  • b

    bored-art-40741

    08/25/2015, 5:22 PM
    ++
  • b

    bored-art-40741

    08/25/2015, 7:14 PM
    @witty-crayon-22786 and other twitter folk: 
    worker_count
    is only for the isolated strategy
  • b

    bored-art-40741

    08/25/2015, 7:14 PM
    We're using zinc to compile java+scala, but we're not running the isolated strategy
  • b

    bored-art-40741

    08/25/2015, 7:14 PM
    People are seeing only a single core being used to compile
  • b

    bored-art-40741

    08/25/2015, 7:14 PM
    We used to have an option on the global strategy that told zinc how many workers to use
  • b

    bored-art-40741

    08/25/2015, 7:14 PM
    Has that been lost?
  • j

    jolly-chef-92794

    08/25/2015, 7:21 PM
    @enough-analyst-54434: regarding https://github.com/pantsbuild/pants/issues/2057 and your last comment on https://rbcommons.com/s/twitter/r/2677/: 
    DistributionLocator
    no longer depends on 
    JVM
    , but 
    Ivy
    and 
    Executor
    still use 
    DistributionLocator
    without declaring a subsystem dependency (because 
    Ivy
    and 
    Executor
    aren’t tasks or subsystems, so they have no such concept). Should I just require 
    Distribution
    be passed explicitly as an argument all the way down the chain, so that the tasks that ultimately use 
    Ivy
    have to be responsible for declaring a 
    DistributionLocator
    dependency?
  • w

    witty-crayon-22786

    08/25/2015, 7:33 PM
    @bored-art-40741: hm, pretty sure that didn't exist.
  • w

    witty-crayon-22786

    08/25/2015, 7:34 PM
    @bored-art-40741: scalac is single threaded... unless you invoke it multiple times, you don't get parallelism
  • h

    happy-kitchen-89482

    08/25/2015, 7:34 PM
    I see no evidence of it as far back as 0.0.38
  • h

    happy-kitchen-89482

    08/25/2015, 7:34 PM
    Maybe zinc ran parallel scalacs
  • w

    witty-crayon-22786

    08/25/2015, 7:34 PM
    only if you invoke it multiple times
  • w

    witty-crayon-22786

    08/25/2015, 7:34 PM
    have looked at this code quite a bit
  • h

    happy-kitchen-89482

    08/25/2015, 7:34 PM
    we're definitely getting a massive slowdown post upgrade
  • b

    bored-art-40741

    08/25/2015, 7:35 PM
    We used to pass 
    worker_count: 4
    in pants.ini despite the fact that it no longer applies to the global strategy
  • w

    witty-crayon-22786

    08/25/2015, 7:35 PM
    it never applied to the global strategy
  • w

    witty-crayon-22786

    08/25/2015, 7:35 PM
    it is applied to the execution_graph, which is used by isolation
  • h

    happy-kitchen-89482

    08/25/2015, 7:35 PM
    @bored-art-40741: I added worker_count: 4 last week, so people could opt-in to the isolated strategy, iirc
1...217218219...2420Latest