Subject: Re: Speaking of the Multia... X anyone?
To: None <port-alpha@NetBSD.ORG, that1guy@azstarnet.com>
From: Ross Harvey <ross@teraflop.com>
List: port-alpha
Date: 04/21/1998 00:49:16
> Has anyone gotten X to run on a TGA multia system?  The X binaries I
> downloaded with the rest of the installation of 1.3 seemed to be missing
> an X server.  I tried installing the version of X from the 1.2
> installation only to get an xinit shutdown due to an error reading the
> keyboard.
> 
> If this sounds familiar...  yes, I did post asking about this along with
> some details.

I haven't looked at X support at all yet, but IIRC this is the story:
(I know this isn't exactly the answer you requested, but other people
have asked more general questions recently.)

	1. The only X server that NetBSD/alpha ever had was a TGA
	   (not TGA2) that was tested primarily on the Multia.

	2. The diffs were to an older version of Xfree86 and they
	   don't directly apply to the current versions

	3. No one has hacked on this for a while, so it doesn't
	   work any more. (I presume it could be resurrected without
	   much trouble, though.)

	4. We just switched the low-level console code so if it
	   wasn't broken before it probably is right now in
	   -current.

Part of the problem was the unfortunate way that Xfree86 configured the
frame buffers: it assumed there was just one PCI bus, just one ISA bus,
that it could magically just do inb() and outb(), and in general it
worked on linux because linux didn't care about the lack of a proper
interface as long as they could sort of make the alphas they ran on
look like PC's.  (And linux never ran on the TC and TL systems like
NetBSD does, so it was easier for them to just forget about machine-
independent interfaces. These days, the assumption of PCI isn't as
serious, though many other problems due to lack of interface remain.)

The many poor abstractions in this non-interface and the fact that this
prevents sharing low-level event and framebuffer code kind of made
NetBSD developers (think "Chris") want to work on other things. Also,
the two places that fund NetBSD alpha right now (NASA and Avalon) don't
really care in an official way about X (the big alphas don't even have
displays), so we aren't working on it.

At this point, some volunteers willing to run -current and hack on X
might be helpful...

Ross