> > btw: I'll have some x11-fb changes in the queue, so there will
> > be at least one more patchkit. The patch need some more work
> > though. I'll want to have the x11-fb patch ready for submit
> > soon, so I can push it mainline for 2.6.13.
Uploaded to http://www.suse.de/~kraxel/uml/patches/2.6.12-rc5/
The uml-x11-fb patch depends on xen-vt-runtime
Both kernel-internal usage (fbcon) and userspace apps (map
framebuffer via mmap() and draw there) access the same pages
now, there is no copying between kernel & userspace fb pages any
more. That should fix all the funny drawing effects older
versions had. This version can also deal with multiple apps
mapping the framebuffer memory, that should help people testing
embedded stuff (using qt-embedded).
Keyboard issues (uml sometimes having a wrong idea of the
modifier key state) are fixed.
> However, as long as it does not touch anything if disabled, we
> can always ship it while marking it as BROKEN or EXPERIMENTAL.
Even if compiled in it shouldn't have any effect unless you
enable framebuffer + fbcon via "x11=<width>x<height>" on the
kernel command line. Switching between uml "stdio console" and
fbcon + vt subsystem happens at runtime now (unlike the version
which is @ sf.net where this happens at compile time).
The patch enables quite a few new drivers/subsystems:
vt subsystem, fbcon needs this.
for virtual keyboard & mouse, X11 input events are
going this way.
framebuffer core & fbcon, obviously needed for a
framebuffer driver ;)
That might have some unwanted side effects. It WorksForMe[tm]
though. I'd consider it ready for mainline now.
-mm seems unusually stable at present.
-- akpm about 2.6.12-rc3-mm3