Bug#329974: xlibmesa-dri: function __driUtilCreateScreen is freeing never allocated data

2005-11-24 Thread Julien Cristau
On Wed, Nov 23, 2005 at 11:30:07 +, Michel Dänzer wrote:

> On Tue, 2005-11-22 at 22:35 +0100, Samuel Hym wrote:
> > 
> > The __driUtilCreateScreen function (line 1357 and beyond...) is
> > freeing, at the end, framebuffer.dev_priv that has never been
> > allocated when drmOpen(NULL,BusID) fails to open
> > for instance, which must be the case in this bug report (when
> > /dev/dri/* are not readable for the user calling the function, for
> > instance).
> > 
> > Initializing framebuffer.dev_priv at NULL at the very beginning of the
> > function solved the symptom in my case but access to framebuffer.base
> > should probably not be allowed either. In a word: framebuffer should
> > be initialized to be somewhat valid in case of failures in the
> > function.
> 
> Sounds like you could provide a patch? If you do, please also submit it
> upstream at http://bugs.freedesktop.org .
> 
The code affected by this bug seems to have changed a lot between 6.8.2
and now, so I think that this is fixed upstream (and probably in
experimental). Samuel, could you try the Xorg debs from experimental and
confirm whether the bug is fixed?

Thanks,
Julien



Bug#329974: xlibmesa-dri: function __driUtilCreateScreen is freeing never allocated data

2005-11-23 Thread Michel Dänzer
On Tue, 2005-11-22 at 22:35 +0100, Samuel Hym wrote:
> 
> The __driUtilCreateScreen function (line 1357 and beyond...) is
> freeing, at the end, framebuffer.dev_priv that has never been
> allocated when drmOpen(NULL,BusID) fails to open
> for instance, which must be the case in this bug report (when
> /dev/dri/* are not readable for the user calling the function, for
> instance).
> 
> Initializing framebuffer.dev_priv at NULL at the very beginning of the
> function solved the symptom in my case but access to framebuffer.base
> should probably not be allowed either. In a word: framebuffer should
> be initialized to be somewhat valid in case of failures in the
> function.

Sounds like you could provide a patch? If you do, please also submit it
upstream at http://bugs.freedesktop.org .


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