Empathy 3.25.90-7.mga8 crashes on startup. It had been working under Mageia 7, most likely also later on Mageia 8. It crashed somewhere at nouveau/dri under kernel 5.15.46-desktop586-1.mga8. If it would not also crash under kernel 5.5.15 I´d suspected it to be a kernel related issue. Perhaps anyone can have a look. The main report is at https://gitlab.freedesktop.org/drm/nouveau/-/issues/174 . I don´t know if Empathy is at all suspected to work on a Pentium IV system with very old graphics card. As it seems there have still been updates concerning dri because before Empathy could be run without any problems.
Also gdb complains a bit in spite of all debuginfo & source being installed.
It still works as before if I specify export LIBGL_ALWAYS_SOFTWARE=1 on the command line. Under Mageia 7 I had this as a default and likely that is the cause why it had been working.
I have seen that mcc exhibits a very similar dmesg on startup. Could it be that mcc just ignores LIBGL_ALWAYS_SOFTWARE=1 and that it would run normally if it wouldn't ignore LIBGL_ALWAYS_SOFTWARE=1?
Summary: Empathy has started to crash on startup since the latest updates on my Pentium 4 => make MCC work with LIBGL_ALWAYS_SOFTWARE=1
I have now tested a full Gnome Session with LIBGL_ALWAYS_SOFTWARE=1 in /usr/share/xsessions/gnome.desktop. The result did surprise me. It booted without problems and the majority of programs work without any error: gnome-terminal, the gnome filemanager, rhythmbox, empathy, shotwell. From what I have tested before, I know that pidgin, firefox and thunderbird also work without LIBGL_ALWAYS_SOFTWARE. The only program which did still expose problems was the email client evince when trying to view the content of an email (webkit renderer crashed). All other functionality of evince did seem to run without any problem. The other gtk3 program which does not work on my Pentium 4 is mcc, the Mageia Control Center. As programs incompatible with LIBGL_ALWAYS_SOFTWARE are the real exception I'd believe that you can change mcc in a way that it works without OpenGL. That should really be doable for a plain configuration program. Most subshells like edit-urpm-sources.pl do work out of the box without any setting like GLSW on my NVIDIA Corporation NV17M [GeForce4 420 Go]. It is just the right icon view of the main menu which requires a closer look. The menu text strings at the left are viewed correctly.
Excuse me, the mail client is called evolution, not evince (afaik this would be the document viewer).
It would be good if all programs worked at least with LIBGL_ALWAYS_SOFTWARE=1. I don’t know what is causing these errors. Perhaps someone can look at the backtraces I have posted at the freedesktop thread. If useful or necessary I will provide more.
Thank you for the report (sorry for not replying sooner), and all the work you have put into diagnosing the cause. Your other bug report cited in comment 0 is helpful. Assigning this to the 'tools' team for MCC; CCing tmb for the openGL aspect in case he can add something.
CC: (none) => tmbAssignee: bugsquad => mageiatools
The main task for me now would be to compile mesa/mesa!10752 fixes and test whether that fixes this issue. If yes I´d suggest to enable it at least for i386. Unfortunately I don´t have too much spare time this weekend so I may have to postpone this task. It would be really great help if someone could offer a ready compiled i386 mesa/mesa!10752 for Mageia 8. It is hard to do since I can not make that on my i586. The machine is simply too old to compile such a big software. On the other hand everything would need to fit into Mageia 8 / i386.
Great news: I have just tested the mesa/mesa!10752 from Karol Herbst and mcc works very well with it. You can find the packages at http://download.opensuse.org/repositories/home:/estellnb:/mageiaupdtst/Mageia_8_backports/i586/, an install instruction at https://www.elstel.org/software/external-repositories.html.en and the source rpm at https://build.opensuse.org/package/show/home:estellnb:mageiaupdtst/mesa.
Also other gtk(3) programs have worked without problems like Meld, Filezilla, Libreoffice, Gnumeric. Only Empathy seems to require the nouveau dri-driver which was not to be compiled for that commit.
There appears to be a suspend/s2ram issue either related to the kernel or to the patches of Karol Herbst: Bug 31227
obs build bug / Mageia 9/Cauldron - truncated binaries: https://bugzilla.opensuse.org/show_bug.cgi?id=1204672
Summary: make MCC work with LIBGL_ALWAYS_SOFTWARE=1 => Mesa: acquire the patches of Karol Herbst
Is there any prospect of getting these patches into Mageia 9? The obs has now translated Cauldron/i586 while there is currently still an internal obs error for Cauldron/x86_64. Do you think it would be enough to test i586 in advance? As Karol Herbst said these are mere fixes of threading issues and should not break anything. They pose a huge improvement for anyone who uses i586 and especially i586/gtk3.
(In reply to Elmar Stellnberger from comment #13) > Is there any prospect of getting these patches into Mageia 9? They are already in mesa 22.3 branch we have in cauldron
Are they also in the alpha1 release of Mageia 9?
MCC works well with the new patches. However I am afraid that I have not tested them fully with evince, yet. It says "A WebKitWebProcess crashed when displaying the message" as soon as you wanna view an email. Now I have collected core files but cannot get the debuginfo for them: > gdb Missing separate debuginfo for the main executable file Try: yum --enablerepo='*debug*' install /usr/lib/debug/.build-id/e6/a890a468bb0b9d98e614b6296a80a858a0da09 Core was generated by `/usr/libexec/webkit2gtk-4.0/WebKitNetworkProcess 657 37'. Program terminated with signal SIGSEGV, Segmentation fault. > urpmf /usr/lib/debug/.build-id/e6/a890a468bb0b9d98e614b6296a80a858a0da09 returns nothing Packages webkit2-debuginfo-2.38.2-1.mga8, webkit2-jsc-debuginfo-2.38.2-1.mga8, libwebkit2gtk4.0_37-debuginfo-2.38.2-1.mga8 are installed. Can anybody be of help?
Oops, meant evolution the mail client, not evince the document viewer.
Here is the Gnome bug for webkit2: https://gitlab.gnome.org/tasali/webkit2gtk/-/issues/1
The debuginfo fallacy is now an own bug: https://bugs.mageia.org/show_bug.cgi?id=31296