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

    enough-analyst-54434

    07/30/2015, 9:07 PM
    Can't you break your rule and have a junit tool dep different from the user-dep?
  • h

    happy-kitchen-89482

    07/30/2015, 9:07 PM
    not sure I understand
  • e

    enough-analyst-54434

    07/30/2015, 9:08 PM
    ... which implies users don't write ':junit' as a dep in junit_tests
  • e

    enough-analyst-54434

    07/30/2015, 9:08 PM
    gotcha
  • e

    enough-analyst-54434

    07/30/2015, 9:10 PM
    So, one thing that does make sense to me - maybe - is running with extra fixed classpath sets
  • e

    enough-analyst-54434

    07/30/2015, 9:10 PM
    ie, i want to run/test with a set of deps that inject an agent
  • e

    enough-analyst-54434

    07/30/2015, 9:10 PM
    and that sort of affordance would cover this logging case however it was spelled in config/CLI
  • e

    enough-analyst-54434

    07/30/2015, 9:11 PM
    today to do this you need to write some extra targets pointing at a given target that add agent deps for example
  • e

    enough-analyst-54434

    07/30/2015, 9:11 PM
    which is a bit awkward for a 1-off run with perfagent
  • h

    happy-kitchen-89482

    07/30/2015, 9:14 PM
    yeah
  • e

    enough-analyst-54434

    07/30/2015, 9:14 PM
    But that might just be the ability to select a jvm profile easily
  • h

    happy-kitchen-89482

    07/30/2015, 9:14 PM
    junit is actually a weird case
  • h

    happy-kitchen-89482

    07/30/2015, 9:14 PM
    because it is both a tool and a compile-time dep
  • h

    happy-kitchen-89482

    07/30/2015, 9:15 PM
    actually, now that I think of it
  • h

    happy-kitchen-89482

    07/30/2015, 9:15 PM
    where does the compile-time junit dep come from?
  • h

    happy-kitchen-89482

    07/30/2015, 9:15 PM
    I don't remember
  • e

    enough-analyst-54434

    07/30/2015, 9:15 PM
    some tasks did injection ad-hoc ...
  • e

    enough-analyst-54434

    07/30/2015, 9:16 PM
    but in any case, since the junit api stays unshaded, it leaks though anyway into the runtime classpath allowing targets to get away w/o declaring the dep
  • h

    happy-kitchen-89482

    07/30/2015, 9:16 PM
    yes
  • h

    happy-kitchen-89482

    07/30/2015, 9:16 PM
    at test running time
  • h

    happy-kitchen-89482

    07/30/2015, 9:16 PM
    but at compile time?
  • h

    happy-kitchen-89482

    07/30/2015, 9:17 PM
    what is putting junit on the compile-time classpath?
  • b

    bored-art-40741

    07/30/2015, 9:17 PM
    I think we have it in both 3rdparty/BUILD and BUILD.tools
  • b

    bored-art-40741

    07/30/2015, 9:18 PM
    Yeah, we basically have separate notions of junit the tool and junit the compile time dep
  • e

    enough-analyst-54434

    07/30/2015, 9:19 PM
    which is good
  • e

    enough-analyst-54434

    07/30/2015, 9:19 PM
    it looks like no open source code injects the junit dep into JavaTests targets
  • e

    enough-analyst-54434

    07/30/2015, 9:19 PM
    it must be manually added to the dependencies list in the BUILD
  • e

    enough-analyst-54434

    07/30/2015, 9:20 PM
    So maybe Tom just edited the wrong junit dep and should have been mucking in 3rdparty/BUILD instead
  • e

    enough-analyst-54434

    07/30/2015, 9:22 PM
    And/or de-aliasing those 2 deps depending on how you have it set up
  • w

    witty-crayon-22786

    07/30/2015, 9:22 PM
    we manually add it internally, yea.
1...105106107...2420Latest