GNOME Bugzilla – Bug 736069
Evolution window content not drawn
Last modified: 2018-09-18 13:09:25 UTC
With Evolution 3.12.2 and 3.12.5 I experienced that suddently the Evolution window as not drawn anymore. Only the red stop button showed sometimes. There is nothing in the Linux kernel logs. Evolution-Data-Server seems to run fine in the background and the notification of new messages are shown. Pressing Ctrl + n or something else does not have any effect. I first thought it was a Radeon driver problem, reported as bug 81239 in the freedesktop.org Bugzilla [1], but owen in #gtk+@irc.gnome.org told me that is very unlikely and looks more like an Evolution problem with WebKitGTK+. So I am creating a new ticket here and closing it in the fdo bug tracker. Unfortunately I am *not* able to reproduce it reliably. Here is the backtrace with Evolution 3.12.2.
+ Trace 234053
[1] https://bugs.freedesktop.org/show_bug.cgi?id=81239
Today I was able to reproduce it with Evolution 3.12.6. I’ll attach two screenshots captured with GIMP.
Created attachment 286424 [details] Blank window with red button
Created attachment 286425 [details] Blank window without red button The red button is the only thing that appears after some seconds. The message area is normally filled with the image of the virtual desktop I was before, that means is not redrawn. Messages are still received as I get the notifications about new messages.
It happened again, when closing the preference window. The preference window border was still show. Here is the backtrace.
+ Trace 234104
Michael, could you please check the last backtrace again to verify that it is unrelated to the Radeon driver? That’d be awesome.
I don't see anything directly pointing to the Radeon drivers. radeon_drm_cs_emit_ioctl is a helper thread routine for submitting OpenGL command streams to the kernel radeon driver. Since no other thread is waiting in any OpenGL related code AFAICT, it seems unlikely that the problem is OpenGL related. Unless maybe one of the poll()s is waiting for activity from the X11 display file descriptor?
(In reply to comment #6) […] > Unless maybe one of the poll()s is waiting for activity from the X11 display > file descriptor? Is there a way to check that next time I experience this problem?
Thanks for a bug report. I do not see anything odd in the backtraces, just usual threads, most of them from your IMAPx accounts, but none really stuck. Even the main thread is not stuck, which means the drawing should be fine. The webkitgtk part in evolution is only with the message preview (and other view's previews), otherwise the webkitgtk is not used (it'll be also used in the message composer in 3.16.0). I do not know what to do with this within evolution itself. I can give you a little application with only gtk and webkitgtk being involved, but you can probably use devhelp the same, to test whether you can reproduce the issue out of evolution.
I'm closing this, because it had been a long time since the last comment and because many things changed meanwhile, in the drivers, in Evolution (most notably port from webkit1gtk to webkit2gtk) and in webkitgtk+ itself as well.