#general
- e
enough-analyst-54434
09/11/2015, 4:20 PMI don't think so - w
witty-crayon-22786
09/11/2015, 4:20 PMfor multi-module projects - f
fancy-queen-20734
09/11/2015, 4:20 PMI went the unextract route, but while it fixes the reported bug, it doesn’t feel quite right - w
witty-crayon-22786
09/11/2015, 4:21 PM@fancy-queen-20734: yea, i think i like a manual cleanup hook a bit better as a temporary solution - w
witty-crayon-22786
09/11/2015, 4:22 PMand then once we can switch compile (...and everything else?) to
, the problem goes awaycache_target_dirs - f
fancy-queen-20734
09/11/2015, 4:22 PMunextract feels like a more general solution, but I’m a little nervous about trusting the BuildInvalidator since I haven’t had much experience with it - w
witty-crayon-22786
09/11/2015, 4:23 PMthis isn't a general problem though - w
witty-crayon-22786
09/11/2015, 4:23 PMbecause the solution in any case other than compile (legacy) is to usecache_target_dirs - w
witty-crayon-22786
09/11/2015, 4:23 PM...in which case the pre extract step is "remove the target dir" - w
witty-crayon-22786
09/11/2015, 4:24 PMthe only thing that currently needs to use the manual cache API is compile - f
fancy-queen-20734
09/11/2015, 4:32 PMWhy can’t we switch compile to
?cache_target_dirs - e
enough-analyst-54434
09/11/2015, 4:32 PMI'll bow out but note sjavac's state is tracked in text files - yay!: http://hg.openjdk.java.net/jdk9/dev/langtools/file/551d797dc863/src/jdk.compiler/share/classes/com/sun/tools/sjavac/JavacState.java#l278 - w
witty-crayon-22786
09/11/2015, 4:33 PM@fancy-queen-20734: because the
block is inself.invalidated
, meaning that it affects both isolated and globaljvm_compile.py - w
witty-crayon-22786
09/11/2015, 4:33 PM... it might work, but it would be awkward. worth looking at maybe - w
witty-crayon-22786
09/11/2015, 4:34 PMtime might be better spent on something expedient, and then having you look at the memory issues @bored-art-40741 is seeing with large isolated classpaths, so that we can get them switched and put global to bed - h
happy-kitchen-89482
09/11/2015, 4:48 PM=@stuhood - h
happy-kitchen-89482
09/11/2015, 4:49 PMIf at all feasible we'd like to get off global strategy, and then expunge it from the codebase - h
happy-kitchen-89482
09/11/2015, 4:49 PMseems like a prerequisite for the new engine, among other reasons - w
witty-crayon-22786
09/11/2015, 4:55 PMand actually, if you guys are able to capture a smoking memory gun, then i will likely switch to that as well - w
witty-crayon-22786
09/11/2015, 4:56 PMbecause integrating https://github.com/sbt/sbt/compare/0.13...twitter-forks:stuhood/output-jars also gets much easier once global is out - b
bored-art-40741
09/11/2015, 4:58 PMCould output jars only be supported for isolated? - b
bored-art-40741
09/11/2015, 4:58 PMIt seems like output jars should still be optional - b
bored-art-40741
09/11/2015, 4:58 PMAt least at first - w
witty-crayon-22786
09/11/2015, 4:58 PMusing jars is already non-optional - w
witty-crayon-22786
09/11/2015, 4:58 PMfor isolated - w
witty-crayon-22786
09/11/2015, 4:59 PMbecause they're critical for reasonable runtime performance - b
bored-art-40741
09/11/2015, 4:59 PMOkay - w
witty-crayon-22786
09/11/2015, 4:59 PM(take a look in
).pants.d/compile/jvm/zinc/jars - b
bored-art-40741
09/11/2015, 4:59 PMI'm torn here. I'm tempted to suggest that we just nuke the global strategy on master and move on. At this point we are already forked and unable to upgrade because of the jvm-dep-check product/memory issues - w
witty-crayon-22786
09/11/2015, 5:00 PMhm. interesting.