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

    happy-kitchen-89482

    07/30/2015, 8:53 PM
    yep, exactly what we do
  • e

    enough-analyst-54434

    07/30/2015, 8:54 PM
    For the global exclude approach to be sane, pants would need to provide an API indexer against the replacement log jar and then compare that index with all 3rdparty code
  • e

    enough-analyst-54434

    07/30/2015, 8:54 PM
    And at least confirm signatures match, and if not warn
  • e

    enough-analyst-54434

    07/30/2015, 8:54 PM
    And you can pray the log statements never execute
  • e

    enough-analyst-54434

    07/30/2015, 8:54 PM
    in the 3rdparty code
  • w

    witty-crayon-22786

    07/30/2015, 8:55 PM
    we have midgraph excludes
  • e

    enough-analyst-54434

    07/30/2015, 8:55 PM
    excludes are always a shot in the dark
  • w

    witty-crayon-22786

    07/30/2015, 8:55 PM
    which nick has that review out for.
  • w

    witty-crayon-22786

    07/30/2015, 8:55 PM
    we discourage them heavily in favor of excluding in 3rdparty
  • e

    enough-analyst-54434

    07/30/2015, 8:55 PM
    you just don't know unless you do symbol + flow analysis
  • h

    happy-kitchen-89482

    07/30/2015, 8:56 PM
    our pom_resolver takes a global_exclusions option, which in our pants.ini is a large list of exclusions
  • h

    happy-kitchen-89482

    07/30/2015, 8:56 PM
    and then we pin some of those back in
  • h

    happy-kitchen-89482

    07/30/2015, 8:56 PM
    we know it's correct from an api perspective because we have a global resolution
  • e

    enough-analyst-54434

    07/30/2015, 8:57 PM
    no you don't - what if your pin is wrong?
  • e

    enough-analyst-54434

    07/30/2015, 8:57 PM
    to prove the pin is right is nontrivial
  • e

    enough-analyst-54434

    07/30/2015, 8:57 PM
    it runs in production for 5 minutes is all people end up using for proof
  • e

    enough-analyst-54434

    07/30/2015, 8:58 PM
    it would be very cool to provide the symbol + flow analysis tool though!
  • h

    happy-kitchen-89482

    07/30/2015, 8:58 PM
    yes, that would be cool
  • h

    happy-kitchen-89482

    07/30/2015, 8:59 PM
    we "know" it's right because where we use this tends to be static
  • e

    enough-analyst-54434

    07/30/2015, 8:59 PM
    symbol would be much simpler, but I'm pretty sure the real world will always resolve to symbol conflicts so you need to then do the extra bit to prove the incompat is never executed
  • h

    happy-kitchen-89482

    07/30/2015, 8:59 PM
    we don't provably know it
  • h

    happy-kitchen-89482

    07/30/2015, 8:59 PM
    but our excludes of slf4j and friends work fine
  • h

    happy-kitchen-89482

    07/30/2015, 8:59 PM
    however the issue we're having now is how to add them back in for all junit_tests
  • h

    happy-kitchen-89482

    07/30/2015, 9:00 PM
    which leads me to the question: should we be able to add configurable implicit deps for every target type?
  • h

    happy-kitchen-89482

    07/30/2015, 9:00 PM
    currently we have a subsystem to do this for scala_library (to add the scala runtime)
  • e

    enough-analyst-54434

    07/30/2015, 9:00 PM
    vs overriding a target alias
  • h

    happy-kitchen-89482

    07/30/2015, 9:00 PM
    but it's a bit ad-hoc
  • h

    happy-kitchen-89482

    07/30/2015, 9:00 PM
    yes, we can hack this by overriding a target alias
  • h

    happy-kitchen-89482

    07/30/2015, 9:00 PM
    internally
  • h

    happy-kitchen-89482

    07/30/2015, 9:00 PM
    I'm wondering if this is useful enough to have as a feature
1...103104105...2420Latest