[Openchrome-devel] release 0.2.902 this week-end ?

[Xavier Bachelot]

Benno Schulenberg wrote:
> Xavier Bachelot wrote:
>> The AGPDMA issue consists of either screen repaint trouble or
>> machine crash on K8M800 and VM800. It seems to be related to
>> kernel and not to openchrome. According to reports, 2.6.20 is
>> fine, 2.6.23 is not.
> 
> The commenter on http://www.openchrome.org/trac/ticket/145 , 
> <rikz at yandex.ru>, has been mailing me privately, saying that he 
> hasn't found a working kernel, going as far back as 2.6.19, and 
> that he would soon try downgrading to xorg 1.3.  I've urged him to 
> post his findings to the openchrome-user mailing list, but he 
> continues to think that negative findings are no findings at all.
> 
Another commenter said it was ok for him with 2.6.20, but not ok anymore 
with 2.6.23. And I agree that off-list comments are next to useless... 
:-( Unless someone already knows where to look for this bug, we have yet 
to find a pattern. Anyone with more knowledge in how drm/dri/DDX and all 
works, and with some hints on which variables we need to chance to 
narrow the scope of the search ?

>> Given this, I'd want to revert corresponding part of changeset
>> 527 : http://openchrome.org/trac/changeset/527
> 
> The disabling does not do any harm, and it does appear to prevent 
> screen corruption and lockups on the affected chipsets.  Both of 
> those are _very_ annoying events, and should be prevented at all 
> costs.  There are few things more frustrating than rebooting after 
> a system update and getting only a black screen or seeing the 
> desktop slowly fall apart under your eyes.  When I first started to 
> use the via driver, it took me a whole month to find out that 
> off-screen pixmaps had to be disabled in order to stop corruption 
> of the background.  It would have been great if Suse had disabled 
> that by default.
>
There are some performances penalties.
imho, this is just sweeping bugs under the carpet. And there are not 
even our bugs... If no one see the bug, there's no chance for it to be 
fixed. If it was an openchrome one, we would just acknowledge we have 
it, temporarily hide it to users while continuing to work on it behind 
the scene. But this is not the case so it doesn't make sense to hide it.
Just my 0.02 ?, I will do whatever we agree on, but I want to hear other 
opinions first.

Regards,
Xavier