I have a laptop which I connect to an external display, and the new xrandr
version in FC7 is a major change for me, since it allows me to switch from
built-in screen only (or screen cloning) to dual-screen mode on the fly.
However, this doesn't always work - maybe half the times I make an xrandr
resolution/display mode switch, it goes bad. There seems to be three different
ways for it to break, and I'm unsure how to debug them, or whether they're
different enough to warrant separate tickets.
1. Programs in the session die - most frequently gnome-power-manager, oftentimes
many others, popping up the bug-buddy dialog. Perhaps due to a dbus signal they
can't deal with? I tried attaching dbus-monitor and couldn't see any signals
being emitted when I change the mode, though, so perhaps it's something else.
2. The entire session (or xorg) dies, throwing me back to the gdm screen. No
clue in the log files.
3. Xorg completely hangs, forcing me to kill power. Might be upstream
gnome-desktop isn't the right component for this (it's just a tiny package with
a bad name).
since 2 and 3 seem to suggest some sort of X problem, i'm going to punt to
We may need to open up a gnome-power-manager bug if there isn't the same root
cause between all 3.
Actually, I think it's libgdk-x11 receiving the display geometry changing and
accessing a null pointer or something - sig11 anyway. Once I found the right
debuginfo rpms, the stacktrace is pretty clear on that. Unfortunately, with
everything but bug-buddy having crashed, I don't see how I copy the trace out of
So, this probably is a combination of xorg sending a geometry change event that
is not entirely correct (and sometimes crashing itself?), and gdk not validating
the structure entirely.
Created attachment 155129 [details]
bug-buddy's stracktrace of a crashing app
got lucky after enough retries - only gnome-panel crashed so I had enough of a
session left to catch and save this :)
I got this by repeating "xrandr --output VGA --off" and "xrandr --output VGA
--auto" enough times (well, twice).
ah, okay, moving to gtk2
Fixed in 2.10.12, which will go out as an update shortly after f7 gets released.
Cheers. Any test build I could provide feedback on?
You can get it out of the build system:
Thanks, I can verify that gtk2-2.10.12-1.fc7 fixed this problem for me.
Per reporters comment 8, closing as resolved in the CURRENTRELEASE