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

    witty-crayon-22786

    09/11/2015, 4:11 PM
    the old artifact =/
  • w

    witty-crayon-22786

    09/11/2015, 4:11 PM
    but no, it isn't in analysis for things like annotation processors
  • e

    enough-analyst-54434

    09/11/2015, 4:11 PM
    is this only a probelm for apt or broader ?
  • w

    witty-crayon-22786

    09/11/2015, 4:12 PM
    autovalue
  • e

    enough-analyst-54434

    09/11/2015, 4:12 PM
    so very narrow
  • e

    enough-analyst-54434

    09/11/2015, 4:12 PM
    k - that helps, thinking
  • w

    witty-crayon-22786

    09/11/2015, 4:12 PM
    and lombok, etc
  • w

    witty-crayon-22786

    09/11/2015, 4:12 PM
    i mean, it's how java "macros" work
  • e

    enough-analyst-54434

    09/11/2015, 4:12 PM
    I can't believe you let folks use lombok!
  • w

    witty-crayon-22786

    09/11/2015, 4:12 PM
    we don't
  • e

    enough-analyst-54434

    09/11/2015, 4:12 PM
    ok - very good
  • w

    witty-crayon-22786

    09/11/2015, 4:13 PM
    but i commented on the ticket: this problem would go away if compiles used 
    cache_target_dirs
  • e

    enough-analyst-54434

    09/11/2015, 4:13 PM
    So - this is "easy" to track and it knocks off a pants bug
  • w

    witty-crayon-22786

    09/11/2015, 4:13 PM
    as opposed to the manual api
  • e

    enough-analyst-54434

    09/11/2015, 4:14 PM
    k, I was going to suggest controling javac -sourcepath
  • e

    enough-analyst-54434

    09/11/2015, 4:14 PM
    this could be a clean dir and you could look at the sourcecode generated there to derive the classnames that are valid and in-play for this incremental round
  • w

    witty-crayon-22786

    09/11/2015, 4:15 PM
    oh. you mean that annotation processors also generate the source?
  • e

    enough-analyst-54434

    09/11/2015, 4:15 PM
    s/also/often only/
  • w

    witty-crayon-22786

    09/11/2015, 4:15 PM
    hm
  • e

    enough-analyst-54434

    09/11/2015, 4:15 PM
    so, apt allows your plugin to generate code
  • e

    enough-analyst-54434

    09/11/2015, 4:15 PM
    in which case another compiler round runs
  • e

    enough-analyst-54434

    09/11/2015, 4:15 PM
    until settling
  • w

    witty-crayon-22786

    09/11/2015, 4:15 PM
    gotcha.
  • w

    witty-crayon-22786

    09/11/2015, 4:17 PM
    @fancy-queen-20734: so did you go the "hook to allow a task to clean up outputs before extracting" route?
  • w

    witty-crayon-22786

    09/11/2015, 4:18 PM
    @enough-analyst-54434: but yea... zinc could almost certainly track this in its analysis if it's a javac api. probably just aren't yet =/
  • e

    enough-analyst-54434

    09/11/2015, 4:19 PM
    It doesn't make a ton of sense for it to track, maybe - I'm not sure what the java 9 built-in incremental comiler provides for transparency of results
  • e

    enough-analyst-54434

    09/11/2015, 4:19 PM
    It may be all internal though with no user-facing analysis
  • e

    enough-analyst-54434

    09/11/2015, 4:19 PM
    not sure at all
  • e

    enough-analyst-54434

    09/11/2015, 4:19 PM
    IIUC its only used to build javac itself in 8
  • w

    witty-crayon-22786

    09/11/2015, 4:19 PM
    i got the impression that the "java 9 incremental compiler" was actually a build tool, similar to make
1...341342343...2420Latest