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.
Actual: - memory increases and decreases Expected: - memory should continuously increases.
Created attachment 82969 [details] screen-shot
Please check it against show stopper criteria
The problem is confirmed. Several issues found: 1) DTraceDataCollector indicatorBuffer is not cleared in If clause when collector isSlave. 2) a lot of probes are droped (as the example given has really intensive memory allocation
Item #1 identified by Maria definitely should be fixed in 6.7. Fix is safe and simple. I'm investigating item #2 currently.
I played with bufsize, dynvarsize and cleanrate options, but could not get rid of drops completely. (BTW, we already have issue 158372 about them.) Incorrect decreasing heap size is coming from DTrace. I wonder if it's the consequence of drops or the bug mentioned here: http://www.mail-archive.com/dtrace-discuss@opensolaris.org/msg01210.html Don't have a solution yet.
Item #1 fixed in http://hg.netbeans.org/cnd-main/rev/4b1dd71c389e
If it is known DTrace issue that is not fixed yet, then do not worry. Just waive this bug or downgrade it. I do not know if the already fixed part is worth integrating to 6.7, is it? It does not look as a showstopper, the dtrace issue does, but it is not under our control.
Integrated into 'main-golden', will be available in build *200906020201* on http://bits.netbeans.org/dev/nightly/ (upload may still be in progress) Changeset: http://hg.netbeans.org/main-golden/rev/4b1dd71c389e User: Alexey Vladykin <alexey_vladykin@netbeans.org> Log: fixing #166245 wrong memory indicator with DTrace provider
Decreasing heap size on indicator was caused by shuffled DTrace output on multiprocessor system, as described at http://www.solarisinternals.com/wiki/index.php/DTrace_Topics_Hints_Tips#Post-sort This is fixed in http://hg.netbeans.org/cnd-main/rev/656457b78520 Mentioned changeset also tweaks some DTrace options to have less drops. But the drops inevitable in such extreme cases as program doing malloc(2) in a loop. I vote for integrating both changesets in release67. Item #1 is a memory and performance problem. Item #2 causes strong user confusion.
Alexander, please evaluate this issue for showstopper status.
I think it IS a stopper. I have reviewed Alexey's revisions and fixes are safe and correct
I agree that it's a showstopper as working with DTrace is very important for Solaris. I've verified the fix in dev build, please integrate it into release67 branch.
Fixed in release67: http://hg.netbeans.org/release67/rev/45cd51dd4b2d
Integrated into 'main-golden', will be available in build *200906030252* on http://bits.netbeans.org/dev/nightly/ (upload may still be in progress) Changeset: http://hg.netbeans.org/main-golden/rev/656457b78520 User: Alexey Vladykin <alexey_vladykin@netbeans.org> Log: fixing #166245 wrong memory indicator with DTrace provider
verified in RC2 build 2009-06-03_00-01-07