In article <20030115195229.GA308@drowsy.duskware.de>
martin@duskware.de wrote:

> I would suggest you look at dev/pci/machfb.c and/or dev/pci/igsfb_pci.c.
> 
> IMHO it's better to have a special frontend recognizing and setting up
> the card (and providing card specific accelerations, without the thing will
> be dog slow), and returning a higher match value so it is prefered over
> the generic VGA driver.

I agree it's _better_ . That's one of the reason why I made vga_reset()
be called from MD consinit(), not from MI VGA driver attachments.
My patch is just a hack, but I think it's still worth for some
obsolete machines. I think it's unlikely eventually someone will
write MI cirrus driver ;-p

The only changes I made in MI vga drivers are just adding some ops
to load default font from wsfonts. These ops could also be moved into
MD consinit(), but I left them in MI driver because they would be
useful even for old "genuine" VGA to replace default console font.

> Once we start gettting multiple such beasts, of course they should share
> as much code as possible, maybe even with the generic vga driver.

BTW, is there any reason why current vag_raster.c does not use MI rasops?
---
Izumi Tsutsui
tsutsui@ceres.dti.ne.jp