[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: X kills my powerbook

On Fri, 2006-03-10 at 11:34 +0100, Michel Dänzer wrote:
> On Fri, 2006-03-10 at 11:08 +0100, Michel Dänzer wrote:
> > On Thu, 2006-03-09 at 20:50 -0500, David Nusinow wrote:
> > > 
> > > I'm beginning to think that it's better to just do a snapshot of the radeon
> > > driver from HEAD. Ben's done a ton of fixes and I want to make sure I get
> > > them all. Plus, if the 7.1 timeline is to be believed, we'll be getting
> > > something similar as a real release soon enough. Thoughts from people more
> > > involved with upstream than me?
> > 
> > I'd basically agree, except that current driver HEADs will drag in EXA
> > from xserver HEAD, at least. And if you pull EXA from xserver HEAD,
> > you'll have to pull HEAD for all drivers that support EXA... The only
> > consolation I can offer for this is that these incompatible EXA changes
> > have been made precisely to avoid further incompatible changes in the
> > future.

That's known and that's also why I'm about to create a branch in the ati
driver just before the incompatible change in order to be able to do an
update release of the driver.

> Also, as great as Ben's fixes are, they seem to have caused some
> regressions of their own (e.g. with MergedFB), which is no surprise
> given their broad scope. We may have to consider creating a stable
> xf86-video-ati CVS branch.

Decision is already taken for that :) I'll do it today or tomorrow

 In the meantime, patch
> https://bugs.freedesktop.org/attachment.cgi?id=4864 from
> https://bugs.freedesktop.org/show_bug.cgi?id=5916 might be a safe way to
> address the most critical stability issues in xserver-xorg in sid.
> In the long term, I hope we can stay on top of the modular releases in
> sid and package CVS snapshots in experimental.

I think you should use the ati stable branch after I

 - have created it
 - fixed the PCI IDs issue (moved them out of the server to the ATI
 - fixed the MergedFB/Xinerama issue, this is the only remaining major
problem & regression on my radar at this point.

Also, while you should also update the kernel DRM with the new memory
map stuff (in DRM CVS only for now), my latest patch should still fix
most of the issues when using older DRMs (though at the expense of
limiting useable video memory on some cards).


Reply to: