Bug 108076
| Summary: | Java form Matlab 6.5 cannot start with NPLT or LD_ASUME_KERNEL | ||
|---|---|---|---|
| Product: | [Fedora] Fedora | Reporter: | bednar <borkows> |
| Component: | glibc | Assignee: | Jakub Jelinek <jakub> |
| Status: | CLOSED NOTABUG | QA Contact: | |
| Severity: | high | Docs Contact: | |
| Priority: | high | ||
| Version: | rawhide | CC: | david_kewley, drepper, kmilos, williama_lovaton |
| Target Milestone: | --- | ||
| Target Release: | --- | ||
| Hardware: | i686 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Fixed In Version: | Doc Type: | Bug Fix | |
| Doc Text: | Story Points: | --- | |
| Clone Of: | Environment: | ||
| Last Closed: | 2004-09-28 08:18:46 UTC | Type: | --- |
| Regression: | --- | Mount Type: | --- |
| Documentation: | --- | CRM: | |
| Verified Versions: | Category: | --- | |
| oVirt Team: | --- | RHEL 7.3 requirements from Atomic Host: | |
| Cloudforms Team: | --- | Target Upstream Version: | |
| Embargoed: | |||
Now I'm using Fedora Core 1 and java built in matlab 6.1 stiil doesn't work (I read about similar problem with java on http://www.mail-archive.com/tomcat-user@jakarta.apache.org/msg110209.html ) Here is the output from matlab: [root@localhost bin]# matlab Warning: Unable to load Java Runtime Environment: /usr/local/matlab6p1/sys/java/jre/glnx86/jre/lib/i386/native_threads/libjava.so: symbol __libc_wait, version GLIBC_2.0 not defined in file libc.so.6 with link time reference. Warning: Disabling Java support. Whot is worse, I tried export LD_ASSUME_KERNEL=2.2.5 and 2.4.1 and 2.4.19. None of these work. The problem is very important for many of people who make serious job on linux. When I changed MATLAB_JAVA environment variable to point SUN's java 1.4.2 the matlab start with java GUI but, after opening one more window of matlab (like editor or preferences) matlab freezes. I use system locale pl_:PL and have SUN's java 1.4.2 with locale for en,gb,de,es. I don't know if it could be locale problem or maybe NPTL or glibc because I cannot find any log from java crash. When I tried (FC1+Matlab 6.0) to change MATLAB_JAVA to point to SUN's Java 1.4.2 I got another error libjvm.so: cannot open shared object file: No such file or directory so I had to include the java libraries in LD_LIBRARY_PATH in order to make it start, but it does indeed freeze here as well if I open another window (this is on en_GB locale) I could only run Matlab normally on FC1 with the fix suggested in bug 110610 and the original JRE supplied with Matlab. Me too. I've also tried to change JVM to 1.4.2 with same effects. I think we should write to the Matworsk directly> I tried, because I have licenced Matlab bought at my work, but I have to ask the administartor of licence server of some number (I don't remeber its name). The licence number is not enough to send a question to the Matworks support team. I'll send the results (if there will any) after contatc with Mathworks. Ok, just upgraded to Matlab 6.5 (w/ jre1.3.1), and with the latest glibc update (glibc-2.3.2-101.4) I can run Matlab just fine without any workarounds! I hope you'll be able to confirm this too. The only problem I have now is Java complaining about some missing fonts, http://www.mathworks.com/support/solutions/data/31607.shtml which I solved by copying over font.properties.Redhat8.0 from j2re1.4.2 I can run matlab 6.5 without tricks too. But: 6.5 has (I think that its true) bug in Power Sim Blockset and jvm supplied with 6.5 has ugly big, curved, unreadable fonts. I'm seeking a solution to the fonts in JVM problem. Does abybody know how to change jvm fonts? I don't know that this bug needs to stay open; it doesn't see like a RHL / FC problem. See my comment(s) and those of Jakub in bug 110610 (which started with an identical report from bednar, and is now closed). I'll mention here in response to comment #2 that another way (if you have root privs) to avoid the libjvm.so complaint when trying J2RE 1.4.2 is to make a symlink like this: ln -s /usr/java/j2re1.4.2/lib/i386/server/libjvm.so \ $MATLAB_ROOT/bin/glnx86 This is a hack really, from doing an 'strace' on matlab and seeing where it was looking for libjvm.so. Really Mathworks need to fix their bugs. Today I filed a bug report with Mathworks regarding the fact (in my experience at least) that on RHL 9, Matlab R12.1, R13, and R13SP1 all render text lines on the Java desktop with too much whitespace between lines, wasting screen area. Interestingly for RH folks, Matlab R12.1 doesn't have this problem on RHL 8.0. This could be a bug for RH to squash, but I have no idea where to file it. I also mentioned to Mathworks some of the headaches we're all seeing regarding installing new JRE/J2RE. The first line of my bug report requested a human response; we'll see if they do reply. Also note that Matlab R14 is coming out in mid-2004, so *hopefully* it'll fix some of these issues. I'm not holding my breath though. Just an FYI: Matlab 13 with Service Pack 1 works mainly OK on Fedora, with the font fixes which Mathworks itself recommends. The exception to the "mainly OK" I've found is the Print command from the M-file editor, which doesn't work with either Java 1.3.1 or 1.4.2. I have the very same problem with Java on my fresh FC2 Test 3 system. But this time I am trying to install Oracle 9i (9.2.0.1.0). With the JVM shipped in the Oracle Universal Installer I got the very same message as bednar shows in comment #1. I got it to run in two ways 1) Using another JVM (Blackdown 1.3.1) and 2) Following the steps indicated in bug #110610 Comment #1 from Jakub Jelinek. The bad news for me is that it doesn't execute the buttons. It started but when I hit the "Next" button to start the installation process, it doesn't do any thing. It isn't blocked, it just sits there and do nothing. Others buttons works like "About..." and "Uninstall products" but the rest of the buttons doesn't. It doesn't even show a message on the terminal. Nothing!. :-( I got the same from Oracle 8.1.7. Have in mind that I just need the client installation, I'm not trying to setup a server on Linux. Oracle 9i installation used to work in RH9 (2.4.20) and it worked flawlessly in Mandrake 10 with kernel 2.6.3. Have anyone tried to install Oracle in FC2?? -William PS. I'll subscribe to fedora-test-list and post a link to this bug. BTW FC2 rocks! I close this bug. Nothing here is a problem in glibc. Yes, all JVMs had problems and got updated. |
From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.2.1) Gecko/20030225 Description of problem: The Java VM supplied with Matlab 6.1 and 6.5 cannot run with RHL 9 or Fedora because of Native Posix Threads implemented in the kernel, but in Red Hat Linux 9 it was possible to start matlab with java after setting environment variable LD_ASSUME_KERNEL=2.4.1. Now this trick doesn't work :( Version-Release number of selected component (if applicable): standard kernel of Fedora Test 3 How reproducible: Always Steps to Reproduce: 1.type in CLI: LD_ASSUME_KERNEL=2.4.1 2.run matlab 3. Actual Results: Java Support Disabled Expected Results: Working JAVA GUI of Matlab Additional info: