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

    rough-minister-58256

    09/15/2015, 7:57 PM
    "IMO, this is too tightly coupled to a particular service discovery strategy. As sirois said, an mdns resolver would a reasonable choice as well, and a second resolver implementation should require new flags to use. Instead, I'd recommend something like --resolver with choices ['none', 'pinger', 'rest-hostlist'], where 'none' disables pinging, and 'rest-hostlist' uses the hostnames from the the --read-to and --write-from as its inputs.” <— assuming this
  • w

    witty-crayon-22786

    09/15/2015, 7:58 PM
    @enough-analyst-54434:
    Copy code
    List<> entries = library.mkEntries();
  • w

    witty-crayon-22786

    09/15/2015, 7:58 PM
    is i guess what i was thinking of
  • w

    witty-crayon-22786

    09/15/2015, 7:59 PM
    or perhaps the inference that happens for the return types of static methods
  • w

    witty-crayon-22786

    09/15/2015, 7:59 PM
    ...but unless those are captured somewhere, it's hard to say you're "using" them
  • w

    witty-crayon-22786

    09/15/2015, 8:00 PM
    your heuristic about action at a distance is the right one though
  • w

    witty-crayon-22786

    09/15/2015, 8:00 PM
    (inference generally reduces action at a distance though)
  • w

    witty-crayon-22786

    09/15/2015, 8:01 PM
    (or at least the kind that means you have to update declarations)
  • f

    fancy-queen-20734

    09/15/2015, 8:04 PM
    Have we considered putting together a http proxy api instead? That’d be another way to make it pluggable without requiring that pants know about how resolution happens.
  • w

    witty-crayon-22786

    09/15/2015, 8:04 PM
    if pants forks that proxy, then pants needs to know how resolution happens
  • b

    bored-art-40741

    09/15/2015, 8:04 PM
    @witty-crayon-22786: I have a proper profile graph of this noop compile badness
  • w

    witty-crayon-22786

    09/15/2015, 8:05 PM
    hence, pluggable resolver
  • b

    bored-art-40741

    09/15/2015, 8:05 PM
    The action is in add_products_by_target / add_abs_paths
  • b

    bored-art-40741

    09/15/2015, 8:05 PM
    There's a relpath that isn't fast_rel;path
  • w

    witty-crayon-22786

    09/15/2015, 8:05 PM
    oh, nice.
  • b

    bored-art-40741

    09/15/2015, 8:07 PM
    JvmCompileIsolatedStrategy.compute_classes_by_source
  • b

    bored-art-40741

    09/15/2015, 8:07 PM
    ...
  • b

    bored-art-40741

    09/15/2015, 8:07 PM
    Copy code
    if os.path.exists(compile_context.analysis_file):
        products = self._analysis_parser.parse_products_from_path(compile_context.analysis_file,
                                                                  compile_context.classes_dir)
        for src, classes in products.items():
          relsrc = os.path.relpath(src, buildroot)
          classes_by_src[relsrc] = classes
          unclaimed_classes.difference_update(classes)
  • w

    witty-crayon-22786

    09/15/2015, 8:11 PM
    ...argh.
  • w

    witty-crayon-22786

    09/15/2015, 8:12 PM
    thanks
  • b

    bored-art-40741

    09/15/2015, 8:16 PM
    Fixing now and double checking it picks it up
  • w

    witty-crayon-22786

    09/15/2015, 8:19 PM
    @bored-art-40741: that one is probably in global as well
  • f

    fancy-queen-20734

    09/15/2015, 8:26 PM
    I have to say, trying to debug subproc callbacks is trickier than I would like
  • f

    fancy-queen-20734

    09/15/2015, 8:29 PM
    also subproc_map forks aren’t killed by a keyboard interrupt
  • f

    fancy-queen-20734

    09/15/2015, 8:29 PM
    as far as I can tell
  • f

    fancy-queen-20734

    09/15/2015, 8:29 PM
    or rather, I’m running tests which hang, and then leave subproc forks running on Ctrl-C
  • w

    witty-crayon-22786

    09/15/2015, 8:30 PM
    @bored-art-40741: i fixed zinc to do lazy loading of analysis, and it looks like it does only use a fraction of it in many cases
  • w

    witty-crayon-22786

    09/15/2015, 8:32 PM
    i'm fixing one more issue with it
  • b

    bored-art-40741

    09/15/2015, 8:40 PM
    Sweet
  • b

    bored-art-40741

    09/15/2015, 8:49 PM
    @fancy-queen-20734: In pants? Which worker pool?
1...365366367...2420Latest