[PULL] bus cleanup, darwin hw/xfree86, build regression fixes

[Mark Kettenis]

> From: Jeremy Huddleston <jeremyhu at apple.com>
> Date: Sun, 2 Oct 2011 01:11:50 -0700
> 
> On Oct 2, 2011, at 00:10, Matthieu Herrb wrote:
> 
> > Was the libpciaccess issue solved ? 
> 
> You're referring to the missing support from libpciaccess, yes?  I'm still waiting on comments from my comments on those patches.
> 
> http://lists.x.org/archives/xorg-devel/2011-September/025663.html
> http://lists.x.org/archives/xorg-devel/2011-September/025664.html
> 
> Those will need to go in before my series.  If I don't hear back on
> my comments from ajax by Monday, I'll just push those patches with
> my edits if everything testes ok.

Please don't.  As you can see in the bits you quoted earlier, I asked
ajax a serious question about one of the interfaces he added.  Since
libpciaccess is a shared library it is important to get the interfaces
right.  So if ajax doesn't respond, can we at least discuss my
proposed changes to the interfaces first?

> Is that the only remaining wrinkle?

I doubt it.  An invasive change like this needs some serious testing.
I haven't found the time to do so yet.  And I first need to write some
missing libpciaccess bits for OpenBSD before I can do so.  But given
the amount of breakage that this change still causes I doubt anybody
has tested your changes.  So I'd say the proper course of action is:

1. Get the libpciaccess additions sorted out and pushed.

2. Fix at least the common drivers (intel, vesa), and perhaps one or
   two of the drivers for older hardware.

3. Ask for testing of these changes, including some non-KMS tests,
   since the changes mostly affect the modesetting paths in the
   drivers.

Step 1 & 2 are largely independent.

I really think these changes are going in the right direction, and I'm
not to worried about the fact that this will break stuff on OpenBSD
for a little while, as long as there is some solid evidence that
things work on Linux.