This Bugzilla instance is a read-only archive of historic NetBeans bug reports. To report a bug in NetBeans please follow the project's instructions for reporting issues.
I'm using the RC1 and trying to diff a source with it's cvs repository corrispondent, memory usare grows from a medium of 70MB to up 254MB. Textual diff is as fast as previous version
Scheduled for promo D.
Errata Corrige: "usare" stands for "usage"
Confirming that there is a leak when using graphical diff. You don't even need a CVS controlled file to reproduce it. I simply took two local ide.log files (one of 92kB, the other one of 131kB) and did 10 Tools/Diff operations on these two files. Memory meter has shown me the following progress: before diffing - 11,6 after closing 1st diff - 14,9 after closing 2nd diff - 15,4 after closing 5th diff - 15,6 after closing 10th diff- 19,0 As you can see there is a growth, though it is not constant, eg. between 2nd and 5th diffing there was a small increase. I have been watching the increase in profiler and I saw quite strange numbers for the objects of org.netbeans.modules.diff.* classes: For the 1st, 2nd and up to 5th diffing I saw 1 instance of all the provider and visualizer diff objects and for the 10th diffing I saw 2 instances of all these objects. (see the two attached screenshots of jprofiler window). Also note that the visualizer objects should have had no instances because the diff window was always closed at the time I took the heap snapshot.
Created attachment 14277 [details] After doing 5th diffing and closing the window
Created attachment 14278 [details] After doing 10th diffing and closing the window
Raising to P2.
Souldn't be this issues assigned to mentlicher@netbeans.org instead of martine@netbeans.org (or are they the same)?
Dont' have a clue. Maybe Martin does have some reason to use a pseudonym. :o)
It was originally assigned to "Martine@netbeans.org". I have no idea who is it, but it's not me ;-) So I'm reassigning it to myself, however if "Martine" would like to solve this I would welcome that.
Tonda, I cannot reproduce on local files. Profiler shows only one DiffAction and one PatchAction instance. All other instances are bogus, profiler cannot find path to GC root. Is there any property to disable Timeout* references. GC is not enough ane have to wait several minutes.
Davide, could you please provide more details? What diff action have you used? Does really textual diff do not cause the leak? I have not found any leak in Tools -> Diff action, so I'll close as WORKSFORME until having better reproduction steps. Thank you
I cannot reproduce. There is no leak after multiple invocations (closing previous results). See attachement.
Created attachment 16446 [details] Leak detector output
Verified.