Re: Patch for bug #1912 applied incorrectly to 6.8 branch

On Wed, 2005-02-09 at 20:53 -0800, Alan Coopersmith wrote:
> Adam Jackson wrote:
> > On Wednesday 09 February 2005 02:39, Michel Dänzer wrote:
> > 
> >>It looks like instead of the second hunk of

Scratch that, BTW, the second hunk for radeon_reg.h was applied
correctly.

> >>https://bugs.freedesktop.org/attachment.cgi?id=1433
> >>
> >>a RADEONSetFBLocation() call was added to RADEONAdjustFrame().
> >>RADEONSetFBLocation() should be called as rarely as possible because it
> >>messes with the guts of the GPU memory controller.
> 
> According to CVSweb, the RADEONSetFBLocation in RADEONAdjustFrame() came
> from this patch, which was checked into 6.8.2 in Decemeber (rev 1.19.2.7):
>    * xc/programs/Xserver/hw/xfree86/drivers/ati/radeon_driver.c
>    Bugzilla #1220 (https://bugs.freedesktop.org/show_bug.cgi?id=1220)
>    attachment #980 (https://bugs.freedesktop.org/attachment.cgi?id=980):
>    Fix garbage screen on radeon cards which may appear after a system
>    suspend/resume cycle.
>    Patch by Matthias Hopf <[email protected]>
> 
> It appears a previous patch adding RADEONSetFBLocation to fix this same
> bug  [...]

Same bug? This commit was supposed to be attachment #1433 for bug #1912
(as the subject says, and the commit message for radeon_reg.h),
attachment #980 was indeed for bug #1220, but as you'll notice, it's for
RADEONEnterVT(), not RADEONAdjustFrame()... Still, this being
incorrectly applied again looks like the most plausible explanation for
this accident.


>  > I think we need to get this backed out before the release, please.
> 
> Isn't the release supposed to be done now?

It shouldn't be released like this. The incorrect line is superfluous at
best, potentially harmful.


-- 
Earthling Michel Dänzer      |     Debian (powerpc), X and DRI developer
Libre software enthusiast    |   http://svcs.affero.net/rm.php?r=daenzer