From: Jon M. Taylor <taylorj@ecs.csus.edu>
  To  : ggi-develop@eskimo.com
  Date: Sat, 1 Aug 1998 20:58:40 -0700 (PDT)

Re: Fwd: Re: kgicon, etc

On Sat, 1 Aug 1998, Ursetto, James F. wrote:

> This apparently didn't go through before.  I had found a "solution" for
> the every-scanline-starts-in-the-middle-of-the-screen problem, in case
> no one else figured it out yet (still reading new messages).
> 
> ----- Forwarded message from Jim Ursetto <ursetto@uiuc.edu> -----
> Date: Mon, 27 Jul 1998 17:59:32 -0500
> From: Jim Ursetto <ursetto@uiuc.edu>
> To: ggi-develop@eskimo.com
> Subject: Re: kgicon, etc
> 
> At 03:32 PM on 1998 July 27, Emmanuel Marty did write:
> 
> > This said, I'd love to hear from Stefan Mars, Jim Ursetto, Teunis
> Peters, > Matthias Grimrath .. :) Now that KGI drivers have high odds to
> fold into > the kernel, that'd be a good time to work on them uh ;)
> 
> First off, let me say many thanks to Andrew for porting the ET6000
> driver to kgicon.  I've been trying out kgicon for the last few kernel
> revisions, and here are some things I've come across so far.
> 
> - After you insert the kgicon driver, but before the con2fbmap, you're
> still writing to the original console.  But every time you scroll one
> (invisible) line in this mode, the display offset register gets changed.
> That means the text starts in the middle of the screen.  To fix this
> I've had to either hit enter an extra time before the con2fbmap, or
> remove the kgi driver, remap the consoles, reinsert, and remap again.

	Fundamentally, the problem is that we are trying to overload
drivers in the kernel.  That is, the kgicon driver is programming the same
hardware as the built-in vgacon driver, but there is no handoff from one
driver to the other.  The proper thing to do is to compile the kgicon
driver into the kernel instead of vgacon.

> - Yes, I said remove the KGI driver.  I haven't downloaded CVS for a
> while, but at last check the KGI driver was non-removable.  I slapped
> together a small patch to fix this.  Please holler if you'd like it.
> Unfortunately, although the driver unloads fine, the text font is not
> restored, and the mode might be set incorrectly.

	Most of the S3 drivers do not restore settings correctly when
unloading.  A lot of those functions are stubs.  I don't know about the
other drivers, but maybe now is the time to finally code that stuff (we
never really needed it before). 

> - But under vgafb, you can't restore the font!  Because the setfont
> program doesn't work. :( It works under vesafb, but that can't coexist
> with kgicon. Though 80x25T can be set via stm, the font is garbled.

	Dunno.

> I'm waiting until vgafb supports resizable consoles and font changes (will
> it ever?) 

	Probably not, seeing as how it will be going away for good in the
next kernel revision or two. 

> before I do any major work on the driver.  Anyway, the driver   
> works fine, it's just the kgicon restriction of one resolution and bit    
> depth which was the motivation of making the module removable.

	Andrew said he fixed this.  Have you updated your CVS?

Jon

---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed