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

    billowy-minister-38128

    01/25/2017, 1:23 AM
    also, running on the CI agent, the issue mostly occurs in the temp PR branch… from the CLI, it is much harder to repro.
  • w

    witty-crayon-22786

    01/25/2017, 1:23 AM
    ...hm
  • w

    witty-crayon-22786

    01/25/2017, 1:24 AM
    so, the issue is potentially related to usage of symlinks for these paths.
  • w

    witty-crayon-22786

    01/25/2017, 1:24 AM
    is it possible that docker is doing something non-standard with the symlinks?
  • b

    billowy-minister-38128

    01/25/2017, 1:24 AM
    i’m using an agent-wide compiler cache at ~/.pants-cache
  • b

    billowy-minister-38128

    01/25/2017, 1:25 AM
    my first thought was that symlinks were causing this.
  • w

    witty-crayon-22786

    01/25/2017, 1:25 AM
    nah, not there. rather, within 
    .pants.d
  • w

    witty-crayon-22786

    01/25/2017, 1:25 AM
    (is where the symlinks are)
  • b

    billowy-minister-38128

    01/25/2017, 1:25 AM
    yes, i typed that before you
  • b

    billowy-minister-38128

    01/25/2017, 1:26 AM
    yes, everything gets copied to .pants.d
  • w

    witty-crayon-22786

    01/25/2017, 1:26 AM
    er
  • w

    witty-crayon-22786

    01/25/2017, 1:26 AM
    what?
  • w

    witty-crayon-22786

    01/25/2017, 1:26 AM
    you mean when you hit the cache?
  • w

    witty-crayon-22786

    01/25/2017, 1:26 AM
    or are you explicitly copying?
  • b

    billowy-minister-38128

    01/25/2017, 1:27 AM
    i’m not explicitly copying anything.. when i hit the cache.
  • w

    witty-crayon-22786

    01/25/2017, 1:27 AM
    ok
  • b

    billowy-minister-38128

    01/25/2017, 1:27 AM
    seems the problem occurs more often if .pants.d does not exist.
  • w

    witty-crayon-22786

    01/25/2017, 1:27 AM
    the review peiyu linked before was my first attempt to stabilize these paths
  • w

    witty-crayon-22786

    01/25/2017, 1:29 AM
    the symlinks are within 
    .pants.d/compile/zinc/${taskhash}/${targetid}/
    , between 
    ./{$targethash}
    and 
    ./current
  • b

    billowy-minister-38128

    01/25/2017, 1:29 AM
    yes, i actually found this earlier when researching the problem… the review, that is.
  • b

    billowy-minister-38128

    01/25/2017, 1:29 AM
    yep, very aware of the symlinks.
  • f

    full-kilobyte-93429

    01/25/2017, 1:31 AM
    i have some saved logs (from last month), just check there was this bug as well, didn't pay attention at the time, so ... may not be to do with docker
  • w

    witty-crayon-22786

    01/25/2017, 1:31 AM
    so you said you can repro this 1 in 20 times on an OSX machine, and 9 in 10 in docker?
  • w

    witty-crayon-22786

    01/25/2017, 1:31 AM
    @full-kilobyte-93429 : that would have been pre-zinc upgrade though, no?
  • f

    full-kilobyte-93429

    01/25/2017, 1:32 AM
    after zinc upgrade for the strict-deps thing
  • w

    witty-crayon-22786

    01/25/2017, 1:33 AM
    hm.
  • f

    full-kilobyte-93429

    01/25/2017, 1:33 AM
    i have no idea if i can repro but they were using new zinc...
  • b

    billowy-minister-38128

    01/25/2017, 1:33 AM
    @witty-crayon-22786 1/20 on OSX, 9/10 on CentOS, never in Docker CentOS
  • f

    full-kilobyte-93429

    01/25/2017, 1:34 AM
    we can probably drop some debug statements in the zinc wrapper
  • b

    billowy-minister-38128

    01/25/2017, 1:34 AM
    since i can repro so reliably on CentOS (7 btw), i can give 1.3.0.dev7 a try
1...160116021603...2420Latest