Re: 2.3.99pre7-8 experiences

[Adam C Powell IV <hazelsct@mit.edu>]

Michel Dänzer wrote:

> Adam C Powell IV wrote:
>
> >    * atyfb still doesn't support Xfb with accels properly, same problems
> >      as 2.2.
>
> What are they? (Sorry if this is old news :)

(Sorry, I forgot only fbdev list members have seen this.)  Minor problems, such as:
expose events don't lead to proper redraws, little 8-pixel-wide horizontal line artifacts
appear and don't go away, some modes fail completely.  no_accel -> no problem.  Geert
sent me info on how to try and fix these problems, I've tried but no useful results yet.

http://www.mail-archive.com/linux-fbdev@vuser.vu.union.edu/msg02291.html

One other thing:

Adam C Powell IV wrote:

>    * The dmasound_awacs hunk is from Ani Joshi; built as a module it
>      can't install without nvram, so drivers/sound/Config.in should make
>      Mac DMA sound depend on CONFIG_NVRAM.  (I don't know if that will
>      make sound work, it's building with nvram at home now, I'll test it
>      tonight.)
>    * A GNOME session from gdm kills X when I move the mouse during
>      ssh-password-gnome, but this may be a sound issue, I'll test with
>      working dmasound later (hopefully).

With CONFIG_NVRAM, modprobe dmasound_awacs works, sound is good, and the gnome session
works.

I can't figure out how to make a proper patch to drivers/sound/dmasound/Config.in for
this, because you'd need to dep_tristate for both CONFIG_SOUND and CONFIG_NVRAM.  Or you
could patch dmasound_awacs.c, wrapping the two NVRAM-dependent spots in #ifdef
CONFIG_NVRAM (and setting a default volume if else).  But that would allow
CONFIG_DMASOUND_AWACS=y and CONFIG_NVRAM=m, which would fail to link.

Any ideas?

-Adam P.