| Summary: | dev-java/icedtea-7.2.6.1 graphical glitches relating to leaky shm segments | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Mike Nerone <mike> |
| Component: | [OLD] Java | Assignee: | Java team <java> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | alex, alexander, dimgel, felipematas, gentoo.2019, gokturk, jeremy.william.murphy, kripton, pacho, stefan.heinsen |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| URL: | https://bugs.kde.org/show_bug.cgi?id=350976 | ||
| See Also: | https://bugs.gentoo.org/show_bug.cgi?id=561104 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
files/7-xshm-fix.patch
icedtea-7.2.6.1-r1.ebuild.patch |
||
|
Description
Mike Nerone
2015-09-25 21:15:37 UTC
(In reply to Mike Nerone from comment #0) > Another user there reports that > OpenJDK-1.8 resolves the problem, but we don't have dev-java/icedtea:8 in > the java overlay at this time. Actually we do but it's called icedtea-3 because upstream wants us to use their versioning. It's a prerelease though and the current ebuild is outdated. It probably won't fetch properly now. If you hear of any other workarounds, please let us know. I'm also affected by this problem. Possible workaround is using another JDK (for example oracle-jdk-bin) Yes, that's true. To my knowledge, only OpenJDK/icedtea exhibits this issue (one person in the KDE bug claims Oracle's does, as well, but I suspect that the wrong jre/jdk is actually selected). *** Bug 561710 has been marked as a duplicate of this bug. *** (In reply to James Le Cuirot from comment #1) > (In reply to Mike Nerone from comment #0) > > Another user there reports that > > OpenJDK-1.8 resolves the problem, but we don't have dev-java/icedtea:8 in > > the java overlay at this time. > > Actually we do but it's called icedtea-3 because upstream wants us to use > their versioning. It's a prerelease though and the current ebuild is > outdated. It probably won't fetch properly now. > > If you hear of any other workarounds, please let us know. It's slotted as 8 though, so icedtea:8 works. I emerged it as that just today. The ebuild has been updated to pre06 which uses 8u60-b24. Testing is welcomed so we can make sure the final release is as good as possible. (In reply to Mike Nerone from comment #3) > Yes, that's true. To my knowledge, only OpenJDK/icedtea exhibits this issue > (one person in the KDE bug claims Oracle's does, as well, but I suspect that > the wrong jre/jdk is actually selected). As far as I can tell from https://bugs.kde.org/show_bug.cgi?id=350976, two completely different versions are being compared; OpenJDK/IcedTea 7 vs. proprietary Oracle binaries for 8. Try emerging icedtea:8 from java-overlay and see if this solves the issue. It seems likely that a fix needs backporting from 8 to 7 to fix this there, but there's not enough info here to tell what that might be. I've been using dev-java/icedtea-3.0.0_pre06 for about an hour now with no hiccups nor any glitching. I'll continue to use it and report back if it returns. I launched an IP-KVM java applet and here what I see:
$ sudo ipcs -m -p | awk '{print $3}' | sort | uniq -c | sort -n
1 1408
1 1526
1 cpid
1 Memory
2
3 1778
3 1898
3 3856
780 4339
$ ps www 4339
PID TTY STAT TIME COMMAND
4339 tty2 Sl+ 1:17 /etc/java-config-2/current-icedtea-web-vm/bin/java -Xbootclasspath/a:/usr/share/icedtea-web/netx.jar:/usr/share/icedtea-web/plugin.jar:no -classpath /opt/icedtea-bin-7.2.6.1/jre/lib/rt.jar sun.applet.PluginMain /run/user/1000/icedteaplugin-puleglot-1pHWZM/4332-icedteanp-plugin-to-appletviewer /run/user/1000/icedteaplugin-puleglot-1pHWZM/4332-icedteanp-appletviewer-to-plugin /run/user/1000/icedteaplugin-puleglot-1pHWZM/4332-icedteanp-plugin-debug-to-appletviewer
The number of shared memory segments used by this java process constantly grows:
$ sudo ipcs -m -p | awk '{print $3}' | sort | uniq -c | grep 4339
839 4339
...
$ sudo ipcs -m -p | awk '{print $3}' | sort | uniq -c | grep 4339
865 4339
The default limit for a number of shared memory segments:
$ sudo sysctl kernel.shmmni
kernel.shmmni = 4096
I applied the patch from [1] to icedtea-7.2.6.1 and java stopped leaking shm segments. ^_^ This should also fix problems with QT, alsa, etc. [1] http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2568 BTW, the same issue in other bug tracking systems: https://bugzilla.redhat.com/show_bug.cgi?id=1248530 https://bugs.archlinux.org/task/45824 https://bugzilla.opensuse.org/show_bug.cgi?id=948301 https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=799587 (In reply to Alexander Tsoy from comment #9) > I applied the patch from [1] to icedtea-7.2.6.1 and java stopped leaking shm > segments. ^_^ This should also fix problems with QT, alsa, etc. > > [1] http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2568 Thanks, I am attempting to build icedtea with a revised ebuild now and will check the result in the morning, though I'm not aware of any simple way to reproduce the problem. I don't use KDE. If everything looks good, I'll commit the result tomorrow but I don't wish to apply it to icedtea-bin. That's a lot of work and it probably won't be too long before the next release. (In reply to James Le Cuirot from comment #11) > Thanks, I am attempting to build icedtea with a revised ebuild now and will > check the result in the morning, I'll attach patches from my local overlay, just in case. > though I'm not aware of any simple way to reproduce the problem. Just start any java application with GUI and see how it's leaking shared memory. > I don't use KDE. Me too. :) Created attachment 414164 [details, diff]
files/7-xshm-fix.patch
Created attachment 414166 [details, diff]
icedtea-7.2.6.1-r1.ebuild.patch
I've tested with alexander's ebuild/patch from #12, #13, and #14, and so far, it does, indeed, appear to resolve the glitching in KDE. *** Bug 561610 has been marked as a duplicate of this bug. *** Same here, dup of https://bugs.gentoo.org/show_bug.cgi?id=560268 See also https://bugs.archlinux.org/task/45824 *** Bug 560268 has been marked as a duplicate of this bug. *** *** Bug 561104 has been marked as a duplicate of this bug. *** Sorry for the wait. I've just committed the fix to dev-java/icedtea-7.2.6.1-r1. I took the later patch from the icedtea bug report as that seems to be the full fix, though I am admittedly a little confused about what actually happened. It does apply cleanly and the problem seems to go away. I was able to observe it with just policytool, which comes with the JDK. I'll fix icedtea-bin if I have time but no promises. It takes ages to build on arm so it would only be for amd64 and x86. (In reply to James Le Cuirot from comment #20) > I've just committed the fix to dev-java/icedtea-7.2.6.1-r1. Fix confirmed working here. Thank you very much! gnu_andrew tells me the next icedtea will be out in a week or so. I'll update icedtea-bin shortly after. (In reply to James Le Cuirot from comment #20) > I've just committed the fix to dev-java/icedtea-7.2.6.1-r1. Works for me too for a week now. Thank you! :) *** Bug 564740 has been marked as a duplicate of this bug. *** Now fixed in icedtea-bin-7.2.6.2. ppc still needs to be stabilised but I'll close this now. |
