Summary: | net-p2p/vuze crashes with sun-jdk-1.6 or icedtea6-bin | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | georgi |
Component: | [OLD] Java | Assignee: | Java team <java> |
Status: | RESOLVED WORKSFORME | ||
Severity: | critical | CC: | darxus, gentoo-bugzilla, gnu_andrew |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
URL: | http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=293 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | 307567, 307973 | ||
Bug Blocks: | 247140 | ||
Attachments: |
Crash error log
Crash error log 2 icedtea6-1.5 crash log sun-jdk-1.6.0.14 crash log |
Description
georgi
2009-02-22 10:21:20 UTC
Created attachment 182769 [details]
Crash error log
Here is the error log.
(In reply to comment #0) > I understand that 1.4 fixes security issues so using 1.3.1-r1 is not a good > workaround. can you reproduce this with sun-jdk? With sun-jdk-1.6.0.12 it works well and doesn't crash. (In reply to comment #0) > I understand that 1.4 fixes security issues so using 1.3.1-r1 is not a good > workaround. icedtea-bin-1.3.1-r1 has the security fixes backported so that's fine, but still not great could you try building icedtea6-1.4 (not bin) from java-overlay and see if it's any difference? I'd love to help, but I'm missing something. I added the java-overlay, but then: $ sudo layman -S * Running command "/usr/bin/svn up "/usr/local/portage/layman/java-overlay""... At revision 7595. * * Success: * ------ * * Successfully synchronized overlay "java-overlay". $ sudo emerge -s icedtea Searching... [ Results for search key : icedtea ] [ Applications found : 1 ] * dev-java/icedtea6-bin Latest version available: 1.4 Latest version installed: [ Not Installed ] Size of files: 110,676 kB Homepage: http://icedtea.classpath.org Description: A Gentoo-made binary build of the icedtea6 JDK License: GPL-2-with-linking-exception Where's icedtea? (In reply to comment #5) > > Where's icedtea? > It's icedtea6 emerge -s does substring search, and it catches icedtea6-bin. There really is no icedtea6. emerge -s icedtea6 didn't do better either. (In reply to comment #7) > emerge -s does substring search, and it catches icedtea6-bin. There really is > no icedtea6. > > emerge -s icedtea6 didn't do better either. > ~/gentoo/overlays/java-overlay/dev-java/icedtea6 $ ls ChangeLog files icedtea6-1.2-r1.ebuild icedtea6-1.3.1-r2.ebuild icedtea6-1.4.ebuild Manifest metadata.xml what about emerge -av dev-java/icedtea6 Oh, I had a layman configuration issue. It is fixed now. I will build icedtea6 and report, probably on Sunday. Created attachment 184330 [details]
Crash error log 2
The crash also happens with icedtea6. Here is the error log.
Looks like it occurs on Fedora too, adding upstream bug. Do you have the link to the upstream bug? (In reply to comment #12) > Do you have the link to the upstream bug? > see the URL field. Try changing the last line in ~/.azureus/gentoo.config to this: JAVA_OPTIONS='-XX:CompileCommand=exclude,com/aelitis/net/udp/uc/impl/PRUDPPacketHandlerImpl$5,runSupport' Note the single quotes, so that $5 is not interpreted as variable. When run from terminal window, you should see this: CompilerOracle: exclude com/aelitis/net/udp/uc/impl/PRUDPPacketHandlerImpl$5.runSupport And it should stop crashing? Please report. I'd have to downgrade to that version of Azureus and find that particular torrent which caused this. It will take me a few days. (In reply to comment #15) > I'd have to downgrade to that version of Azureus and find that particular > torrent which caused this. It will take me a few days. If you say that a newer vuze (which?) works, then I guess it's solved anyway, so no need to try :) I'm not saying it works, it's just that testing takes a while, because the crash is rare. I will actually try crashing the latest Vuze without this workaround, and then with, to see if it makes a difference. But even if it were solved with the latest Vuze, a crash in a buggy application should never bring the while virtual machine down, so there must be a bug in icedtea as well. The *whole* virtual machine, sorry for the typo. Without this fix, the latest Vuze crashes just like the old one. Without the fix, it has been stable all day today, so I assume it works. What does this line do? (In reply to comment #17) > But even if it were solved with the latest Vuze, a crash in a buggy application > should never bring the while virtual machine down, so there must be a bug in > icedtea as well. Yep, the bug is solely in Icedtea, vuze just happens to trigger it and itself is not buggy. (In reply to comment #19) > Without this fix, the latest Vuze crashes just like the old one. > > Without the fix, it has been stable all day today, so I assume it works. > > What does this line do? It disables runtime compilation of the method that was in the bug reports. Because icedtea is crashing in a compiler thread on this method. With this line, the method is interpreted, which is slower, but the method is not probably called often anyway, if the bug is so rare. So there's a bug in the icedtea that has to be fixed upstream. Users can apply this workaround (not really a fix) to vuze's config meanwhile. IcedTea6 1.4 has a newer version of HotSpot (14b08) than the sun-jdk releases. 1.5 will be out within the next week with an updated HotSpot (14b13 IIRC) so please try this and report back once the ebuild is available. I am currently using sun-jdk, so I won't notice when the new ebuild is here. If you can, please reply to this bug report once it is out, and I will test. (In reply to comment #22) > I am currently using sun-jdk, so I won't notice when the new ebuild is here. If > you can, please reply to this bug report once it is out, and I will test. icedtea6-1.5 is out for some time. icedtea6-bin not yet, could you test? Created attachment 195946 [details]
icedtea6-1.5 crash log
I tested with dev-java/icedtea6-1.5 and got the crash.
Created attachment 196034 [details]
sun-jdk-1.6.0.14 crash log
Now it seems the problem also happens with sun-jdk-1.6.0.14 (and maybe earlier versions too).
This java implementation bug can be avoided by adding this command line option to the java command: -XX:CompileCommand=exclude,com/aelitis/net/udp/uc/impl/PRUDPPacketHandlerImpl\$5,runSupport See: https://bugs.launchpad.net/ubuntu/+source/azureus/+bug/428514 I'm wondering if anyone has filed a bug with Sun to fix the JVM? (In reply to comment #26) > This java implementation bug can be avoided by adding this command line option > to the java command: > > -XX:CompileCommand=exclude,com/aelitis/net/udp/uc/impl/PRUDPPacketHandlerImpl\$5,runSupport > > See: https://bugs.launchpad.net/ubuntu/+source/azureus/+bug/428514 Yes, see comment 14 I'm adding an elog message to 4.2.0.8 *** Bug 285788 has been marked as a duplicate of this bug. *** Any difference with sun-jdk-1.6.0.18 or icedtea6-1.7? I have no crashes with 1.6.0.18 in over 24 hours, so maybe this bug has finally been fixed. Could anyone else confirm? I can confirm. With sun-jdk-1.0.6.18 vuze-4.2.0.8 is stable and no longer crashes on my system. (In reply to comment #31) > Any difference with sun-jdk-1.6.0.18 or icedtea6-1.7? dev-java/icedtea6-bin-1.7 seems to be working fine (3 days and counting) OK, all versions that seemed to be affected are now gone. |