Description of problem: I have recently installed a Rocky 8.7 workstation with the latest Mate packages from EPEL (e.g. mate-panel-1.26.2-4.el8.x86_64) but am seeing various errors in the logs while using the desktop. Firstly the fairly innocuous: Nov 30 18:05:11 host journal[1260042]: Theme parsing error: gtk.css:3286:25: 'font-feature-settings' is not a valid property name Nov 30 18:05:11 host journal[1260042]: Theme parsing error: gtk.css:3748:23: 'font-feature-settings' is not a valid property name /usr/share/themes/YaruOk/gtk-3.0/gtk.css 490:spinbutton { font-feature-settings: "tnum"; } 1306:scale marks, scale value { color: alpha(currentColor,0.55); font-feature-settings: "tnum"; } 1469:progressbar { font-size: smaller; color: rgba(0, 0, 0, 0.4); font-feature-settings: "tnum"; } Then these types of errors are more worrying: Nov 30 18:16:34 host journal[838427]: g_object_add_weak_pointer: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: g_object_add_weak_pointer: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: gtk_widget_hide: assertion 'GTK_IS_WIDGET (widget)' failed Nov 30 18:16:34 host journal[838427]: gtk_toggle_button_set_active: assertion 'GTK_IS_TOGGLE_BUTTON (toggle_button)' failed Nov 30 18:16:34 host journal[838427]: g_object_set_data: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: g_object_set_data: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: g_object_set_data: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: gtk_widget_hide: assertion 'GTK_IS_WIDGET (widget)' failed Nov 30 18:16:34 host journal[838427]: gtk_toggle_button_set_active: assertion 'GTK_IS_TOGGLE_BUTTON (toggle_button)' failed Nov 30 18:16:34 host journal[838427]: g_object_set_data: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: g_object_set_data: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: g_object_set_data: assertion 'G_IS_OBJECT (object)' failed Nov 30 18:16:34 host journal[838427]: invalid (NULL) pointer instance Nov 30 18:16:34 host journal[838427]: g_signal_connect_data: assertion 'G_TYPE_CHECK_INSTANCE (instance)' failed I have setup the exact same dekstop on Ricket 9.0 and everything works fine and without any of these errors. I have not tried with Rocky 8.6 to see if this is just a case of Mate needing to be updated for EL 8.7? Version-Release number of selected component (if applicable): mate-panel-1.26.2-4.el8.x86_64 mate-menus-libs-1.26.0-3.el8.x86_64 .. .. etc. How reproducible: Steps to Reproduce: 1. Start standard Mate desktop and open "Appearance" settings to see the gtk.css errors 2. Add the "Window List" applet to the panel and try second mouse to open the "Preference" - the window doesn't open and there are lots of g_object and gtk_widget errors in the logs. 3. Actual results: While the desktop mostly seems stable, the amount of logging is worrying and the Expected results: The EPEL Mate packages works great on Rocket 9.0 and I don't see these errors. Additional info:
I am thinking YaruOk is to modern for rhel8 and deriviates. Use a theme which is supported by rhel8, or file out an upstream Bug, or delete this entry in gtk.css:3286:25 and gtk.css:3748:23 For the other errors i am thinking rhel8 needs a general update of gtk3 and glib. And i am thinking more and more it was a mistake to push Mate to rhel8. Sorry i don't want and i don't want spend more unpaid spare time to downgrade parts of Mate to versions which are more compatible to glib and gtk versions. Wait for an general rhel8 update or upgrade to rhel9.
Wolfgang, Thanks for the fast reply. I agree that it is very difficult to maintain Mate on RHEL8 - I know because I did end up compiling it myself last year and ended up having to update glib2 and gtk3 packages just to get things working. I had hoped that the EPEL8 Mate packages had found a way around this but I guess there are still issues. It seems pretty unlikely that RHEL8 will ever update glib2 - they haven't changed it at all between 8.0 & 8.7. When I updated the OS versions of glib2 & gtk3, everything *seemed* to continue working, but obviously there is a ton of software compiled against the older versions. Personally I would love to see a fully functioning Mate in EL8 because it gives us a nice upgrade path from EL7 -> EL8 -> EL9 with a consistent user experience (gnome changes every other week). Anyway, thanks again for the work you have put in so far. I'll see if I can get any further by updating glib2 & gtk3 and recompiling your Mate packages. Daire
I can reproduce the issue with the preference window of windows.list I think the problem is here in the glade.ui file for the windows.list https://github.com/mate-desktop/mate-panel/commit/fecd94f75922fbcc1cbe19598152616320fd740c ``` <requires lib="gtk+" version="3.22"/> <requires lib="gtk+" version="3.24"/> ``` [root@rock8 ~]# rpm -qa gtk3 gtk3-3.22.30-11.el8.x86_64 You can try to revert this commit. It's from 1.26 branch Let me know if that helps
Simple reverting the commit doesn't work. You will run in a conflict with one later commit https://github.com/mate-desktop/mate-panel/commit/0a5006eba63a8214e6448b491b22663d0c13e54e
Simply reducing the gtk3 version in glade file fixes the the issue with windows-list-preferences in my test VM. Can you please try out this scratch build? https://koji.fedoraproject.org/koji/taskinfo?taskID=94835242 The warnings in journal i couldn't reproduce with my VM.
Yes! Confirmed that these RPMs fix the preferences for the windows-list applet. I now get window and no errors in the journal. Thanks for that - much appreciated! So the remaining errors that I see can be reproduced by opening a mate-terminal and then closing it again: Dec 3 10:08:42 host journal[7109]: g_object_unref: assertion 'G_IS_OBJECT (object)' failed Dec 3 10:08:42 host journal[7109]: g_object_unref: assertion 'G_IS_OBJECT (object)' failed Dec 3 10:08:42 host journal[7109]: g_object_unref: assertion 'G_IS_OBJECT (object)' failed Dec 3 10:08:42 host journal[7109]: g_object_unref: assertion 'G_IS_OBJECT (object)' failed Dec 3 10:08:42 host journal[7109]: g_object_unref: assertion 'G_IS_OBJECT (object)' failed .. .. Now, these don't seem to cause any issues (the mate-terminal exits just fine), but there is a lot of this spam in the logs. I only did a quick test but it seems like most applications exhibit this when exiting. But if you don't see those errors, then I need to dig into the particulars of my install to make sure I don't have some weird environment when running Mate. Thanks again, Daire
Actually, on closer inspection, those errors are all coming from the mate-settings-daemon. I ran it with debug (/usr/libexec/mate-settings-daemon --debug --replace) but it did not add anything extra to the logs. I also tried with "G_MESSAGES_DEBUG=all /usr/libexec/mate-settings-daemon" but again nothing extra in the output. I'm not sure is glib2 or mate-setting-daemon needs to be compiled with extra debugging. Daire
I can trigger this warnings in a terminal running `/usr/libexec/mate-settings-daemon --replace` and than closing another open terminal or another application. But i don't see them in journal running `journalctl -b0 -f --no-tail`. Also i can trigger this warning in f36. Here running 2 terminals with `/usr/libexec/mate-settings-daemon --replace`. Every time when starting a new daemon in one terminal, the daemon in the other terminal closed and i see the warnings. But i don't see those warnings when closing another window. Running m-s-d with gdb gives me the warnings but no breakpoint. So i can`t create a stacktrace. No idea to debug this more or to fix it. You can try to report it upstream. Well, i will create a new mate-panel build with the fix for windows list applet.
FEDORA-EPEL-2022-9b4191fb18 has been submitted as an update to Fedora EPEL 8. https://bodhi.fedoraproject.org/updates/FEDORA-EPEL-2022-9b4191fb18
Yea, I don't know why you can't see the warnings in the journal while I do (Rocky 8.7). I also tested the same on Rocky 9.0 + EPEL9 Mate and I don't see the warnings in the journal or in the console when running "/usr/libexec/mate-settings-daemon --replace" and closing various desktop windows/applications. I suspect if I mention it upstream (mate) they will likely just say "use a newer glib2" and/or upgrade from EL8 -> EL9. Anyway, I can modify the desktop xdg to redirect the mate-settings-daemon output to /dev/null to keep the harmless log spam for every window close at bay. Thanks again, Daire
FEDORA-EPEL-2022-9b4191fb18 has been pushed to the Fedora EPEL 8 testing repository. You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-EPEL-2022-9b4191fb18 See also https://fedoraproject.org/wiki/QA:Updates_Testing for more information on how to test updates.
FEDORA-EPEL-2022-9b4191fb18 has been pushed to the Fedora EPEL 8 stable repository. If problem still persists, please make note of it in this bug report.