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

    cool-thailand-74745

    08/05/2015, 10:11 PM
    cool
  • e

    enough-analyst-54434

    08/05/2015, 10:11 PM
    So, the deprecations tend to be end of lifed in the version bump and CHANGELOG update RB
  • e

    enough-analyst-54434

    08/05/2015, 10:11 PM
    That's a problem since no news
  • e

    enough-analyst-54434

    08/05/2015, 10:12 PM
    Right now deprecations are tied to a EOL version and they fail a compile if not removed at that version
  • e

    enough-analyst-54434

    08/05/2015, 10:12 PM
    Thus we forget about the deprecation until the version bump RB that removes them
  • e

    enough-analyst-54434

    08/05/2015, 10:13 PM
    So... the only consistent evidence will be the changelog from the version where the deprecation was introduced
  • e

    enough-analyst-54434

    08/05/2015, 10:13 PM
    Not too helpful for you
  • c

    cool-thailand-74745

    08/05/2015, 10:15 PM
    fair nuff
  • b

    bored-art-40741

    08/05/2015, 10:18 PM
    @witty-crayon-22786: what does your calling code look like?
  • w

    witty-crayon-22786

    08/05/2015, 10:21 PM
    it's a slightly older sha, but it looks like the caller is 
    classpath_util.py:66(classpath_entries)
  • w

    witty-crayon-22786

    08/05/2015, 10:22 PM
    what i expect is happening is a call with the entire list of targets in the context
  • w

    witty-crayon-22786

    08/05/2015, 10:22 PM
    such that we compute a closure for each of them
  • w

    witty-crayon-22786

    08/05/2015, 10:22 PM
    and it's roughly n^2, so
  • w

    witty-crayon-22786

    08/05/2015, 10:23 PM
    so should be fixed at the caller most likely. but it occurs to me that most of this is wasted work given that the graph isn't changing
  • w

    witty-crayon-22786

    08/05/2015, 10:27 PM
    ...also doesn't help that its being called twice. arg.
  • f

    fast-megabyte-33579

    08/05/2015, 10:28 PM
    hurrah, a 50% speedup already 😉
  • b

    bored-art-40741

    08/05/2015, 10:39 PM
    Closure can be computed all at once, too, @witty-crayon-22786
  • b

    bored-art-40741

    08/05/2015, 10:39 PM
    If possible that should reduce the complexity back to linear
  • w

    witty-crayon-22786

    08/05/2015, 10:39 PM
    yea. was thinking about what ordering would look like in that context
  • b

    bored-art-40741

    08/05/2015, 10:39 PM
    ah
  • b

    bored-art-40741

    08/05/2015, 10:39 PM
    We do have params to control that I believe
  • w

    witty-crayon-22786

    08/05/2015, 10:39 PM
    the more i think about it, the more i think this is just dumb. there is no way we need the classpath for "everything"
  • w

    witty-crayon-22786

    08/05/2015, 10:40 PM
    i think the caller is supposed to just be passing in whatever the test target is here
  • a

    astonishing-jelly-60479

    08/05/2015, 10:53 PM
    is the pants script created by the pantsbuild/setup project recommended?
  • a

    astonishing-jelly-60479

    08/05/2015, 10:53 PM
    doesn't seem to work well—or at all, rather—on OS X
  • a

    astonishing-jelly-60479

    08/05/2015, 10:53 PM
  • a

    astonishing-jelly-60479

    08/05/2015, 10:54 PM
    hmm except that's not right either
  • b

    billowy-alarm-80641

    08/05/2015, 10:55 PM
  • b

    billowy-alarm-80641

    08/05/2015, 10:56 PM
    oh, but if you're trying to setup a new repo
  • b

    billowy-alarm-80641

    08/05/2015, 10:56 PM
    there are some more things you'll need
1...146147148...2420Latest