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

    rough-minister-58256

    06/01/2016, 5:59 PM
    seems like 
    follow_links=True
    is the default behavior for wrapped globs - would it be sufficient here to just pop the 
    follow_links
    kwargs and discard that for now?
  • r

    rough-minister-58256

    06/01/2016, 6:00 PM
    or should we actually thread kwargs?
  • w

    witty-crayon-22786

    06/01/2016, 6:00 PM
    it's all a bit hairy... but yea, if you can confirm that that is the default for rglobs, we can probably drop it
  • r

    rough-minister-58256

    06/01/2016, 6:00 PM
  • w

    witty-crayon-22786

    06/01/2016, 6:01 PM
    @rough-minister-58256: i don't think there is a plan to support that setting
  • w

    witty-crayon-22786

    06/01/2016, 6:01 PM
    @rough-minister-58256: for now, might be easiest to pop it though, with a warning maybe?
  • r

    rough-minister-58256

    06/01/2016, 6:01 PM
    ah yeah that sounds good
  • r

    rough-minister-58256

    06/01/2016, 6:02 PM
    yeah, this would be purely for BUILD compatibility across the v1/v2 gap
  • r

    rough-minister-58256

    06/01/2016, 6:02 PM
    then once we’re fully on v2 we’d kill the syntax in favor of taht being default behavior
  • r

    rough-minister-58256

    06/01/2016, 6:42 PM
  • r

    rough-minister-58256

    06/02/2016, 8:29 PM
    hah
  • r

    rough-minister-58256

    06/02/2016, 8:30 PM
    looks like testing 
    list
    on one of the macaw targets generated a 1.6GB traceback with 8.5M nested `Throw`’s
  • w

    witty-crayon-22786

    06/02/2016, 8:31 PM
    haha
  • r

    rough-minister-58256

    06/02/2016, 8:41 PM
    go/phugget/P6020
  • r

    rough-minister-58256

    06/02/2016, 8:42 PM
    new one I noticed last night in there: 
    x/y/z is not a prefix of w/y/z
  • r

    rough-minister-58256

    06/02/2016, 8:43 PM
    the rest are all the 
    TargetMacro.Factory
    miss thing
  • r

    rough-minister-58256

    06/02/2016, 8:43 PM
    will file one for the prefix one
  • w

    witty-crayon-22786

    06/02/2016, 8:50 PM
    sounds good
  • w

    witty-crayon-22786

    06/02/2016, 8:50 PM
    i suspect it's a bad path
  • w

    witty-crayon-22786

    06/02/2016, 8:50 PM
    might be worth just eyeballing first
  • r

    rough-minister-58256

    06/02/2016, 9:06 PM
    ~600k instances of that excluding the ones in the macaw output - seems pathological.
  • w

    witty-crayon-22786

    06/02/2016, 9:08 PM
    k
  • w

    witty-crayon-22786

    06/02/2016, 11:45 PM
    @enough-analyst-54434: hey: i've been thinking about the semantics of providing names to targets in BUILD files
  • w

    witty-crayon-22786

    06/02/2016, 11:46 PM
    i'd like to move to a case where an object name is optional if it is the only object in the file
  • w

    witty-crayon-22786

    06/02/2016, 11:46 PM
    since that handles the 
    go_remote_library
    case, and a bunch of other cases where we currently need to add a name only in order to indicate that something is the default target
  • e

    enough-analyst-54434

    06/02/2016, 11:47 PM
    I think you can scap only and just let dup name error handle many
  • e

    enough-analyst-54434

    06/02/2016, 11:47 PM
    for one
  • w

    witty-crayon-22786

    06/02/2016, 11:48 PM
    yea, exactly
  • w

    witty-crayon-22786

    06/02/2016, 11:48 PM
    the only downside i can see is that if you forget to give something a name, it "disappears"
  • e

    enough-analyst-54434

    06/02/2016, 11:48 PM
    and ProjectTree is needed to get an abstract basename
1...100101102...906Latest