Created attachment 1431534 [details] desktop with system monitor Description of problem: ======================= Even with the machine idle, mate-panel will rack up memory usage up to 4,9GB in a few hours, while using about 12-14% CPU. After it reached 4,9GB the CPU usage would drop to usual values (mostly 0%) Version-Release: ================ mate-panel-1.20.1-6.fc28.x86_64 How reproducible: ================= Consistently Steps to Reproduce: 1. Normal usage of the machine or just idling 2. Check CPU usage 3. Check memory usage after a few hours. Other info: =========== For Reference: my machine has 16GB ram, dual core AMD Athlon XP
Can you please give more info? Does it happen with previous panel version? Was is a new installation? etc.? Please provide as much infos as possible, because my cristal ball is broken ;)
Created attachment 1431744 [details] My own Mate-Panel config
(In reply to Wolfgang Ulbrich from comment #1) > Can you please give more info? Happily. :) > Does it happen with previous panel version? Never happened before. All was ok up to the latest Fedora 27 updates. > Was is a new installation? It is an upgrade from Fedora 27 to Fedora 28. > etc.? > Please provide as much infos as possible, because my cristal ball is broken > ;) I will give you something in the attachments... Hit me if you need more!
Created attachment 1431745 [details] Installed Packages
Created attachment 1431746 [details] My Full Mate Config
Created attachment 1431747 [details] fpaste --sysinfo
Please note that I did at least one "dnf upgrade" in between opening the ticket and filling this request.
Ok, you did an upgrade from f27 to f28. And was f27 already updated to MATE 1.20? I need to know which panel version was running before you noticed the issue. Because f28 started with several updates for the panel. So did you update the panel from last f27 version to current f28 panel version? 'dnf history' and 'dnf history info <id>' should give you the information.
Btw. Do you have a chance to test this with a fresh new user account ? Without changing panel conf and adding applets.
I did some additional testing and the condition causing this always-growing memory usage is the presence of at least one tray icon of a windows program (running under wine). I almost always have eMule running, so... (Time to change the description of this bug.) BTW, the new record of memory eaten by the panel is 11,6GB in 24 ours with UltraVNC in listening mode and eMule, both Active.
P.S.: I am able to reproduce consistently on a clean install on VirtualBox, qemu. Of course you need to install wine and a windows program with a tray icon, as for example eMule, UltraVNC.
I don't use wine Can you please answer my questions from https://bugzilla.redhat.com/show_bug.cgi?id=1575091#c8 ?
I changed a build configure option with MATE-1.20. For this reason i need to know the exact version from previous mate-panel build, to be sure. https://src.fedoraproject.org/cgit/rpms/mate-panel.git/commit/?h=f28&id=7978a516a6a7f5cfdd6472006b8921b0cb487744
> Ok, you did an upgrade from f27 to f28. > And was f27 already updated to MATE 1.20? > I need to know which panel version was running before you noticed the issue. > Because f28 started with several updates for the panel. > So did you update the panel from last f27 version to current f28 panel version? > 'dnf history' and > 'dnf history info <id>' > should give you the information. Unfortunately I already discarded the dnf history of f27, so instead of answering the question straight away I've done a little bit of bisecting: with mate-panel up to 1.20.0-1 memory usage is normal. The problems start with mate-panel-1.20.1-1 in both f28 and f27. In koji I found no in-betweens from 1.20.0-1 to 1.20.1-1. So I called it a day and posted the results here. But let me know if you have any intermediate version that you want me to try.
So 1.20.0-4 for f28 is fine? https://koji.fedoraproject.org/koji/buildinfo?buildID=1052868 Maybe this commit causes the problem ? https://src.fedoraproject.org/cgit/rpms/mate-panel.git/commit/?h=f28&id=7978a516a6a7f5cfdd6472006b8921b0cb487744 This was with 1.20.0-2 More builds you will find at koji https://koji.fedoraproject.org/koji/packageinfo?packageID=14801
To shorten the process. Does this build fixes the issue for you? https://koji.fedoraproject.org/koji/taskinfo?taskID=26851768
(In reply to Wolfgang Ulbrich from comment #16) > To shorten the process. Does this build fixes the issue for you? > https://koji.fedoraproject.org/koji/taskinfo?taskID=26851768 Yes. The new build fixes the issue.
This is good for you but bad in general. Switching back to configure flag --with-in-process-applets=none isn't so good if you're using a HIDPI monitor. As MATE suport this feature with 1.20 i switched to --with-in-process-applets=all Can you please file out an upstream report? This needs be fixed by them, as in-process-applets is also important for a future support of wayland. In the meanwhile i can switch back to the old build option if they found no solution.
To be be shure that it isn't caused by https://github.com/mate-desktop/mate-panel/commit/d02fcc5c27e142f7d232d5239105e87e9fe27e6e , which is the only commit for tray applet between 1.20.0 and 1.20.1. Can you please test this build please? https://koji.fedoraproject.org/koji/taskinfo?taskID=26853824 I reverted the commit in this build and use --with-in-process-applets=all again. Thank you
(In reply to Wolfgang Ulbrich from comment #19) > To be be shure that it isn't caused by > https://github.com/mate-desktop/mate-panel/commit/ > d02fcc5c27e142f7d232d5239105e87e9fe27e6e , > which is the only commit for tray applet between 1.20.0 and 1.20.1. > Can you please test this build please? > https://koji.fedoraproject.org/koji/taskinfo?taskID=26853824 > I reverted the commit in this build and use --with-in-process-applets=all Yep. This 1.20.1-9 build has the problem.
Ok, that means that memleak happens if we build the panel with --with-in-process-applets=all This is better for HIDPI resolution support. Can you please file out an upstream report? This needs be fixed by them, as in-process-applets is also important for a future support of wayland.
Hi all, Was the bug filed upstream? I think I see the same after some update: PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 1890 dtantsur 20 0 12.100g 6.286g 6340 S 14.6 56.3 193:21.75 mate-panel I can file if you tell me which information I have to provide. Note that I don't have wine though, but I do have a few apps with tray icons.
Wild guess: in my case it may be caused by the Audacious tray icon. I hadn't seen this problem before I started actively using, nor is my wife experiencing in a similar environment.
@Dmitry Tantsur Are you shure that it is caused by audacious applet? I guess audacious use Qt nowadays? Can you please add this information to https://github.com/mate-desktop/mate-panel/issues/809 ? Now we have a chance to debug this issue as audacious is a common application in all distros.
Btw. I switched the panel back to build applets out of process. This should solve the problem. f27 https://bodhi.fedoraproject.org/updates/FEDORA-2018-7cbfe6ab39 f28 https://bodhi.fedoraproject.org/updates/FEDORA-2018-f11f17bf17 But we need to debug this as building applets in process is needed for a switch to wayland in future.
mate-panel-1.20.3-1.fc27 has been submitted as an update to Fedora 27. https://bodhi.fedoraproject.org/updates/FEDORA-2018-398fd67e4d
mate-panel-1.20.3-1.fc28 has been submitted as an update to Fedora 28. https://bodhi.fedoraproject.org/updates/FEDORA-2018-b8f25d74ad
mate-panel-1.20.3-1.fc27 has been pushed to the Fedora 27 testing repository. If problems still persist, please make note of it in this bug report. See https://fedoraproject.org/wiki/QA:Updates_Testing for instructions on how to install test updates. You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2018-398fd67e4d
mate-panel-1.20.3-1.fc28 has been pushed to the Fedora 28 testing repository. If problems still persist, please make note of it in this bug report. See https://fedoraproject.org/wiki/QA:Updates_Testing for instructions on how to install test updates. You can provide feedback for this update here: https://bodhi.fedoraproject.org/updates/FEDORA-2018-b8f25d74ad
mate-panel-1.20.3-1.fc27 has been pushed to the Fedora 27 stable repository. If problems still persist, please make note of it in this bug report.
mate-panel-1.20.3-1.fc28 has been pushed to the Fedora 28 stable repository. If problems still persist, please make note of it in this bug report.