#general
- w
witty-crayon-22786
09/04/2015, 9:53 PM(chicken and egg problem... won't be able to build a working release with a broken JarBuilder!) - j
jolly-chef-92794
09/04/2015, 10:02 PM@ancient-stone-50795 said he’d publish it soon; if the double-slashes problem is causing a problem with the jar-tool publish itself I guess we can try publishing from an older version, or publish it with the unpublished version of itself >.> - w
witty-crayon-22786
09/04/2015, 10:03 PMwell, "with itself" is impossible - w
witty-crayon-22786
09/04/2015, 10:03 PMbut yea, maybe with an older version... do we know where it was introduced? - w
witty-crayon-22786
09/04/2015, 10:05 PM...hm. actually, i'm confused about the source of the bug, because JarBuilder hasn't changed in a while - e
enough-analyst-54434
09/04/2015, 10:09 PMYeah - I was pushing for a unit test to flush out the actual root cause - e
enough-analyst-54434
09/04/2015, 10:09 PMI think that's still mysterious - w
witty-crayon-22786
09/04/2015, 10:26 PM...blargh - e
enough-analyst-54434
09/04/2015, 10:28 PMSince the code is in fact old, that implies a random jar in the wild with // in its entries I think - e
enough-analyst-54434
09/04/2015, 10:28 PMJarTool is trying to add that jar to another? - e
enough-analyst-54434
09/04/2015, 10:29 PMoh no - these were internal classes - w
witty-crayon-22786
09/04/2015, 10:42 PMyea, internal classes. i ran into it trying to publish something - e
enough-analyst-54434
09/04/2015, 10:42 PMSo a bit of a wag, but not reallyCopy codeself.assertEqual('b/', fast_relpath('/c/b/', '/c/')) self.assertEqual('b', os.path.relpath('/c/b/', '/c/')) - w
witty-crayon-22786
09/04/2015, 10:43 PMmmm. could be. - e
enough-analyst-54434
09/04/2015, 10:43 PMtiming is right - w
witty-crayon-22786
09/04/2015, 10:43 PMchecking. - e
enough-analyst-54434
09/04/2015, 10:43 PMand that's in exactly the right spot - e
enough-analyst-54434
09/04/2015, 10:43 PMin products abs path with the relevant product mapping uses - w
witty-crayon-22786
09/04/2015, 10:43 PMyep. - w
witty-crayon-22786
09/04/2015, 10:43 PMer - w
witty-crayon-22786
09/04/2015, 10:43 PMwell, not quite... the double slash is internal - w
witty-crayon-22786
09/04/2015, 10:44 PMCopy codeArchive: /Users/stuhood/.m2/repository/org/pantsbuild/zinc-sbt-inc/stuhood-1441403229/zinc-sbt-inc-stuhood-1441403229.jar Length Date Time Name -------- ---- ---- ---- 0 09-04-15 14:48 META-INF/ 74 09-04-15 14:48 META-INF/MANIFEST.MF 0 09-04-15 14:48 sbt/ 0 09-04-15 14:48 sbt/inc/ 604 09-04-15 14:48 sbt/inc/ZincPrivateAnalysis$.class 789 09-04-15 14:48 sbt/inc/ZincPrivateAnalysis.class 604 09-04-15 14:48 sbt/inc//ZincPrivateAnalysis$.class 789 09-04-15 14:48 sbt/inc//ZincPrivateAnalysis.class 604 09-04-15 14:48 sbt//inc/ZincPrivateAnalysis$.class 789 09-04-15 14:48 sbt//inc/ZincPrivateAnalysis.class -------- ------- 4253 10 files - e
enough-analyst-54434
09/04/2015, 10:44 PMack - and this code is used in assembling 'classes_by_target' which is sent over to JarTool as files - e
enough-analyst-54434
09/04/2015, 10:44 PMok - that example is not from 'classes_by_target' - e
enough-analyst-54434
09/04/2015, 10:44 PMso this is likely not it - e
enough-analyst-54434
09/04/2015, 10:45 PMGarretts example in the RB looked like 'classes_by_target' classes - e
enough-analyst-54434
09/04/2015, 10:45 PMI think he just needs to hunt harder - w
witty-crayon-22786
09/04/2015, 10:45 PMwill binary search through products - w
witty-crayon-22786
09/04/2015, 10:46 PMoh... or rather than through products, can bisect on a local publish. - e
enough-analyst-54434
09/04/2015, 10:46 PMif you can reliably repro then it may be easy to just rule out fast relpath by flipping to os.path.relapth