Re: [Linux-fbdev-devel] [PATCH] s3fb: mode selection fixes

[Ondrej Z. <san...@cr...>]

On Mon, Oct 01, 2007 at 07:43:45PM +0200, Krzysztof Helt wrote:
> > > - sets r,g,b length field from the bits_per_pixel value otherwise
> > >   the fbset fails in simple case like switching depths:=20
> > >   8bpp ->32bpp -> 8bpp
> >=20
> > When some depth and some color lengths are requested then this driver
> > (or precisely svga_match_format() ) only changes them up (which is corr=
ect
> > behavior according to some previous email from A. Daplas).
> >=20
> > But if you do some partial change of mode using fbset, then fbset uses
> > previous values for unspecified values of var structure in
> > FBIOPUT_VSCREENINFO call. For example if you use 32bpp and do
> > fbset -depth 16, then fbset calls FBIOPUT_VSCREENINFO with arguments
> > requesting depth 16 but color length 8 for each channel. Better way
> > is to put zeroes to color lengths (which is what fbset does when you
> > do something like 'fbset 640x480-60 -depth 16') - workaround is to use
> > argument -rgba 0,0,0,0  with -depth change using fbset.
> >=20
>=20
> What is a difference if certain color depths allow only specific color le=
ngths?

No difference. svga_match_format() interprets color lengths in
FBIOPUT_VSCREENINFO as minimal requests.

Perhaps svga_match_format() could be changed to choose any mode with
matching depth in case there is no mode with matching depth and color
lengths. I will send patch.

--=20
Elen sila lumenn' omentielvo

Ondrej 'SanTiago' Zajicek (email: san...@cr..., jabber: santiago@=
njs.netlab.cz)
OpenPGP encrypted e-mails preferred (KeyID 0x11DEADC3, wwwkeys.pgp.net)
"To err is human -- to blame it on a computer is even more so."