Community
Participate
Working Groups
Created attachment 121593 [details] trace dump from one hang I've been hitting lots of hangs lately, after importing tons of stuff (once from the file system, once from cvs). After importing, the build starts, and hangs around 21% or 24%. Each time I did ctrl-break to capture the dumps. Normally I'd write it off, but I have just started using a new computer ... come "Intel quad core 2" thing, with Windows Vista :( [Santa Claus put it under my tree] ... so, I'm just wondering if there's some new threading problems being exposed.
Created attachment 121594 [details] trace dump from another hang I just wanted to document these traces ... sorry I can't interpret them more ... but wanted to get them documented, thinking if others are seeing similar then the more traces the better?
Oddly enough, in the second case, I was able to "cancel" the build (tap the red square), it started up again, and then completed. I tried that in the first case, and it did cancel, and did start to rebuild, but then hung again around 80 percent ... and that time _everything_ hung (UI, etc.). HTH
At first glance, the thread dump from comment 0 doesn't show any hang (e.g. thread "Compiler Source File Reader" is still running). Did you observe zero CPU activity when it happened? The thread dump from comment 1 shows that an external builder is running. Maybe this builder did hang but I cannot say from the thread dump.
(In reply to comment #3) > At first glance, the thread dump from comment 0 doesn't show any hang (e.g. > thread "Compiler Source File Reader" is still running). Did you observe zero > CPU activity when it happened? > > The thread dump from comment 1 shows that an external builder is running. Maybe > this builder did hang but I cannot say from the thread dump. > I didn't think to look at the cpu usage directly ... but after it set there on same percentage for 10 minutes, I did one thread dump, then did another thread dump 3 minutes later, and they were exactly the same (except for time stamps, and the dump file name).
Kent, any idea?
I agree with Jerome about the first thread dump. There is some contention between 5 threads over access to the EclipsePreferences (to get the File.getCharSet()), but its well protected & appears thread safe. David, if this can be easily reproduced, then it would help to know what the CPU usage is when the progress seems to stall. But be aware that not all builders update the progress incrementally, some just add 100% of their alloted time when they're done. So progress can jump from 20% to 34% after a long pause.
Is it ok to close my own bug as "worksforme"? Feel free to change to "invalid" if that's your convention. I tried three times to reproduce (starting with a fresh install) and nothing hung (or even appeared to) so I'm fine closing this, and at least the dumps are there if anyone finds them helpful. (Each attempt to reproduce takes 2 or 3 hours, letting it import and build 413 projects!) I'll do a better job of watching CPU next time, if there is a next time. Thanks for taking a look.
I think WORKSFORME is the correct resolution since this cannot be reproduced. In JDT/Core we also set the target milestone when resolving a bug.
Verified for 3.5M5.
