#general
- h
happy-kitchen-89482
09/22/2015, 12:24 AMAnd eyeballing the code looks like it should be OK, I don't see us adding subdirs recursively. - r
rough-minister-58256
09/22/2015, 12:34 AM - r
rough-minister-58256
09/22/2015, 12:34 AMcould the
there be the issue?for - r
rough-minister-58256
09/22/2015, 12:35 AMe.g. if someone passed in a full list of files vs just their root dirs - r
rough-minister-58256
09/22/2015, 12:37 AMassuming the add() is recursive, passing in the full list of the files you dumped above would theoretically achieve the 5x adding by way of adding the class file (1x), its parent dir (1x), grandparent (1x), great grandparent (1x), great great grandparent (1x) - r
rough-minister-58256
09/22/2015, 12:38 AMout of curiosity, do you see the same multi-stacking in the older artifact files? makes me wonder if we’ve been accumulating file size for no good reason for some time and the jar is just exacerbating - e
enough-analyst-54434
09/22/2015, 12:43 AM - e
enough-analyst-54434
09/22/2015, 12:43 AMAre you running with that fix? - h
happy-kitchen-89482
09/22/2015, 12:43 AMno, I don't see this stacking in older files - e
enough-analyst-54434
09/22/2015, 12:44 AMProbably unrelated - just similarish effects in jar files - h
happy-kitchen-89482
09/22/2015, 12:44 AMthe isolated compile strategy is doing it. - h
happy-kitchen-89482
09/22/2015, 12:44 AMIt generates all the files as "artifacts" - r
rough-minister-58256
09/22/2015, 12:44 AMah - h
happy-kitchen-89482
09/22/2015, 12:44 AMwhere it's supposed to only pass in the root dir - h
happy-kitchen-89482
09/22/2015, 12:44 AMbut that is a huge gotcha TBH - h
happy-kitchen-89482
09/22/2015, 12:45 AMtrouble is, to pick one or the other we have to go audit what every task that writes to the artifact cache is doing... - h
happy-kitchen-89482
09/22/2015, 12:50 AMWe could be robust and support both - h
happy-kitchen-89482
09/22/2015, 12:50 AMbut that would require doing our own directory walk - h
happy-kitchen-89482
09/22/2015, 12:51 AMStill something weird going on here. - h
happy-kitchen-89482
09/22/2015, 12:52 AMNow the code looks to me like it should indeed have this problem, but I can't reproduce it on an example scala target. - h
happy-kitchen-89482
09/22/2015, 12:53 AMHMMMMM - r
rough-minister-58256
09/22/2015, 12:53 AMseems like TarballArtifact.collect() should probably generally guard against this case since its fairly non-obv - h
happy-kitchen-89482
09/22/2015, 12:54 AMNo, I'm wrong - the code in jvm_isolated_compile_strategy.py is adding files, not dirs - h
happy-kitchen-89482
09/22/2015, 12:54 AMso that is fine - h
happy-kitchen-89482
09/22/2015, 12:54 AMit is collecting all the files, but not any dirs, so that's not the problem - h
happy-kitchen-89482
09/22/2015, 12:54 AMand indeed I'm not seeing this happen at HEAD - h
happy-kitchen-89482
09/22/2015, 12:54 AMlet me sync back to our currently deployed version - w
witty-crayon-22786
09/22/2015, 1:03 AM@bored-art-40741, @happy-kitchen-89482 :!!! oh - w
witty-crayon-22786
09/22/2015, 1:04 AM@bored-art-40741, @happy-kitchen-89482 : for a while there we were putting directory entries classes_by_target. ffs - h
happy-kitchen-89482
09/22/2015, 1:04 AMaha