[ubuntu-x] Mesa 7.6 go/no go criteria

Bryce Harrington bryce at canonical.com
Sat Oct 10 00:21:18 BST 2009


On Wed, Oct 07, 2009 at 02:18:56PM -0700, Rick Spencer wrote:
> 
> > 6. rickspencer3 and all: define measure for triggering roll back
> We should roll back if we encounter:
> 
> 1. We get new crashers with dupes on any hardware that we can't diagnose
> 2. We get crashers that can only be fixed by updating other components
> (i.e. -intel)
> 3. We get crashers that require any kind of kernel patches, etc...
> 4. Similar to 1-3, but for freezes and performance regressions as well
> 
> I suppose the "we get" piece could use some tightening up. Thoughts?

Status update on this...

There have been 3 confirmed regression bugs found, all against older
Radeon chipsets.  All are reasonably well diagnosed:

The first was a cosmetic issue that affected a certain chipset when
using kwin, and is now fixed in Karmic using a patch from upstream.

The second is a crasher with kwin, which is confirmed as fixed by
dropping one specific mesa patch.  I'm waiting for feedback from
upstream however this one seems to be pretty well diagnosed at this
point.  We can just go with solving it by dropping the patch I think,
but I want to give upstream a bit more time to investigate.

The third appears to affect only ancient R100 hardware, and only one
person can reproduce the problem so far.  While we don't have a fix in
hand for this, I think it should be straightforward to find.  In any
case, due to the age of the hardware in question, I'm not super
concerned about this one.


All in all, so far so good.  Aside from the three bugs above, we've been
getting entirely positive feedback on this change.

Bryce





More information about the Ubuntu-x mailing list