Commit 8d7b7a0d71e0b89321b3341b781bc8845386def6 ("Set DamageSetReportAfterOp to true for the damage extension") causes drawing problems with rxvt-unicode running in OpenBox with xcompmgr.
These problems are:
1.) Even when the rxvt-unicode window has the input focus, its cursor is drawn as if the window didn't have the focus (ie the cursor rectangle isn't drawn solid as it should be, but only its outline is drawn).
2.) As soon as I start typing in rxvt-unicode, the cursor disappears completely.
3.) When a new rxvt-unicode is spawned (which gets the input focus automatically in my setup), the cursor isn't drawn at all in the new window.
These problems first showed up with xorg-server 1.9.0 (which was the first release to have the commit mentioned). xorg-server 1.9.3 is supposed to fix problems related to that commit AFAICT, but it doesn't fix the problems described here.
Reverting the mentioned commit fixes this.
I'm running xorg-server 1.9.3, xf86-video-intel 2.13.902 (on a i945 chip), libdrm 2.4.23, rxvt-unicode 9.10, xcompmgr 1.1.5 and OpenBox 3.4.11.
Reverting "Set DamageSetReportAfterOp to true for the damage extension" only helps with xf86-video-intel 2.13.0.
With 2.13.902, the drawing problems remain even if krh's commit to xorg-server is reverted.
I'm still seeing this issue with xserver 1.10.1, xf86-video-intel 2.15.0, libdrm 2.4.25, and Linux 2.6.38.x.
Please take a look at this problem as it affects many users and was reported many times in various distribution bugtrackers. It happens while using rxvt-unicode with xft fonts, running intel driver and using a compositor.
In my case the issue still exists while using xorg-server-1.12, xf86-video-intel-2.18 and rxvt-unicode-9.15 with Termiuns font. I have switched from xcompmgr to current cersion of compton and the problem still persists.
Here are links to some of the bug reports and relevant forum threads, maybe they will also be helpful:
Confirmed; driver bug.
Author: Chris Wilson <email@example.com>
Date: Fri Mar 23 14:56:06 2012 +0000
uxa: Remove hook for CompositeRectangles
It was broken and not flushing damage correctly. With the
improvements made to the kernel, it is no longer a significant advantage
per se and not worth its additional complexity.
Reported-by: Tilman Sauerbeck <firstname.lastname@example.org>
Signed-off-by: Chris Wilson <email@example.com>
It works now - thank you very much for your work!