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

    enough-analyst-54434

    09/25/2015, 11:16 PM
    so - that says something right there!
  • e

    enough-analyst-54434

    09/25/2015, 11:17 PM
    if the Zinc thing looked at scala_plugin metadata
  • e

    enough-analyst-54434

    09/25/2015, 11:17 PM
    it could post process compile of one into adding the plugin to downstream rounds that dep on the target
  • e

    enough-analyst-54434

    09/25/2015, 11:18 PM
    Maybe you don't even need to do this? javac automagically sees apt plugins classes - no work needed
  • e

    enough-analyst-54434

    09/25/2015, 11:18 PM
    For scalac there is or was the manifest xml
  • e

    enough-analyst-54434

    09/25/2015, 11:18 PM
    so that would at least need to be on the cp emitted by compilation of the scalac_plugin
  • e

    enough-analyst-54434

    09/25/2015, 11:19 PM
    I thought though it was both manifest xml on cp + flags
  • e

    enough-analyst-54434

    09/25/2015, 11:19 PM
    unlike javac 6+ with apt automagic
  • e

    enough-analyst-54434

    09/25/2015, 11:19 PM
    actually its worse last I remember and scalac insists on a jar
  • e

    enough-analyst-54434

    09/25/2015, 11:20 PM
    but now isolated emits jars, so might work
  • e

    enough-analyst-54434

    09/25/2015, 11:21 PM
    It looks like somewhere along the way, the handling for ScalacPlugin was lost
  • e

    enough-analyst-54434

    09/25/2015, 11:22 PM
    The target still exists, ZincCompile would just need to honor it and deal with it
  • e

    enough-analyst-54434

    09/25/2015, 11:23 PM
    Basically - I bet you have a <100 line patch - tests to enable generic scalac plugin bootstrap for pants
  • e

    enough-analyst-54434

    09/25/2015, 11:23 PM
    w/o GroupTask
  • e

    enough-analyst-54434

    09/25/2015, 11:27 PM
    Looking at modern scalac - I think loose classes are accepted, but I'm not sure the plugin cp is re-scanned, so zinc's cached compilers would just need to be honoring changed pluginpath as a cached compiler invalidator
  • e

    enough-analyst-54434

    09/25/2015, 11:27 PM
    Not sure what speed issues this would present on cached compiler evictions
  • e

    enough-analyst-54434

    09/25/2015, 11:28 PM
    Its that niggly bit that could blow this up >100 lines, everything else looks straight forward
  • b

    bored-art-40741

    09/25/2015, 11:33 PM
    Promising!
  • b

    bored-art-40741

    09/26/2015, 12:06 AM
    Foursquare is now unforked and deployed straight off of master (the commit after release 0.0.50)
  • b

    bored-art-40741

    09/26/2015, 12:06 AM
    Yay!
  • w

    witty-crayon-22786

    09/26/2015, 12:07 AM
    huzzah!
  • a

    ancient-stone-50795

    09/26/2015, 12:08 AM
    congratulations!
  • w

    witty-crayon-22786

    09/26/2015, 12:08 AM
    that also means john's refactorings didn't completely break everything
  • a

    ancient-stone-50795

    09/26/2015, 12:08 AM
    The only fork we have is the moment is special logic for the aforementioned development with wire that is now broken anyway, so we might as well not be using it.
  • w

    witty-crayon-22786

    09/26/2015, 12:08 AM
    (edit)
  • b

    bored-art-40741

    09/26/2015, 12:11 AM
    Oh one small break
  • b

    bored-art-40741

    09/26/2015, 12:11 AM
    Copy code
    target_alias_map = self.context.build_file_parser.registered_aliases().targets

Exception message: 'BuildFileAliases' object has no attribute 'targets'
  • b

    bored-art-40741

    09/26/2015, 12:11 AM
    What's the new API for this?
  • b

    bored-art-40741

    09/26/2015, 12:12 AM
    The method seems to still exist
  • b

    bored-art-40741

    09/26/2015, 12:12 AM
    But I think it's been broken upstream
1...468469470...2420Latest