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

    witty-crayon-22786

    05/22/2017, 5:03 PM
    yep, thanks... that should be enough to go on!
  • w

    witty-crayon-22786

    05/22/2017, 5:12 PM
    interesting... the trivial repro isn't there.
  • w

    witty-crayon-22786

    05/22/2017, 5:12 PM
    @straight-waiter-86090 : was one of the machines in the "correct" order?
  • w

    witty-crayon-22786

    05/22/2017, 5:12 PM
    ie, the literally specified order?
  • w

    witty-crayon-22786

    05/22/2017, 5:13 PM
    ie, the following results in the expected literal order in both cases:
    Copy code
    $ cat testprojects/src/resources/org/pantsbuild/testproject/ordering/BUILD
SOURCES=['p', 'a', 'n', 't', 's', 'b', 'u', 'i', 'l', 'd', 'p', 'a', 'n', 't', 's']

resources(
  name='literal',
  sources=SOURCES,
)

resources(
  name='globs',
  sources=globs(*SOURCES),
)
  • w

    witty-crayon-22786

    05/22/2017, 5:13 PM
    (on OSX, afaict)
  • w

    witty-crayon-22786

    05/22/2017, 5:15 PM
    (seeing whether it breaks on Linux: https://github.com/pantsbuild/pants/pull/4612)
  • s

    straight-waiter-86090

    05/22/2017, 5:15 PM
    if I remove the 
    sorted
    from my test, it prints the filenames in the order specified in the globs.
  • w

    witty-crayon-22786

    05/22/2017, 5:15 PM
    oh.
  • w

    witty-crayon-22786

    05/22/2017, 5:15 PM
    on both platforms?
  • w

    witty-crayon-22786

    05/22/2017, 5:15 PM
    that would be expected.
  • w

    witty-crayon-22786

    05/22/2017, 5:16 PM
    if so, the bug would be elsewhere
  • w

    witty-crayon-22786

    05/22/2017, 5:16 PM
    ?
  • s

    straight-waiter-86090

    05/22/2017, 5:16 PM
    yes, the same on both platforms.
  • w

    witty-crayon-22786

    05/22/2017, 5:16 PM
    ok, thanks... so it's somewhere deeper.
  • w

    witty-crayon-22786

    05/22/2017, 5:16 PM
    then the sorting is definitely not necessary there, yea?
  • w

    witty-crayon-22786

    05/22/2017, 5:16 PM
    (at least with v2, it should not be necessary. v1 it might be.)
  • s

    straight-waiter-86090

    05/22/2017, 5:18 PM
    I’m not sure if it’s necessary. I used that 
    sorted
    code from the v1 fingerprint code, mostly to make sure the sha1 of the files were the same.
  • s

    straight-waiter-86090

    05/22/2017, 5:18 PM
    Is there a way to create a Snapshot directly as a test?
  • s

    straight-waiter-86090

    05/22/2017, 5:20 PM
    I see that the fingerprint comes from the Snapshot that is passed into 
    EagerFilesetWithSpec
    .
  • w

    witty-crayon-22786

    05/22/2017, 5:24 PM
    @straight-waiter-86090 : not from within a Task probably... it's a brand new API
  • w

    witty-crayon-22786

    05/22/2017, 5:24 PM
    but yea, the snapshot code is going to be the next potential culprit.
  • w

    witty-crayon-22786

    05/22/2017, 5:29 PM
    @straight-waiter-86090 : the Snapshots are represented as tar files under 
    .pants.d/snapshots
  • w

    witty-crayon-22786

    05/22/2017, 5:29 PM
    so it's possible that we're ending up with mismatches in those files. my next step is going to be to compare the snapshots for a target across platforms.
  • s

    straight-waiter-86090

    05/22/2017, 5:46 PM
    I located the two snapshot files if that’s helpful to you. they have different names, but when I untar them, the files seem to the same to me.
  • w

    witty-crayon-22786

    05/22/2017, 5:47 PM
    if they have different names, then they have different fingerprints
  • w

    witty-crayon-22786

    05/22/2017, 5:47 PM
    @straight-waiter-86090 : i think i'm a few seconds away from having the same thing, so hold on to those in case i don't repro
  • w

    witty-crayon-22786

    05/22/2017, 5:47 PM
    and thank you.
  • s

    straight-waiter-86090

    05/22/2017, 5:47 PM
    Sure.
  • s

    straight-waiter-86090

    05/22/2017, 5:48 PM
    Copy code
    ➜  scratch ls *.tar
4e6156c80841df389f1ec3e75da31bbbe0239281aeae7a17fc388916cdb1e08d.tar 5028e8fbc2062fa1ab7b5b3aec8822eba987eda65909bbb64738f7f3511fd075.tar
➜  scratch ls -l *.tar
-rw-r--r--  1 dyoung  SKYDIVE\Domain Users  124416 May 22 13:42 4e6156c80841df389f1ec3e75da31bbbe0239281aeae7a17fc388916cdb1e08d.tar
-rw-r--r--  1 dyoung  SKYDIVE\Domain Users  124416 May 22 13:44 5028e8fbc2062fa1ab7b5b3aec8822eba987eda65909bbb64738f7f3511fd075.tar
➜  scratch md5 *.tar  
MD5 (4e6156c80841df389f1ec3e75da31bbbe0239281aeae7a17fc388916cdb1e08d.tar) = 947386a1d5faedea424672986b6200c7
MD5 (5028e8fbc2062fa1ab7b5b3aec8822eba987eda65909bbb64738f7f3511fd075.tar) = 451e8c6b0f1c41ed211ed0da53d19d2c
1...174117421743...2420Latest