Xorg crash in xf86ActivateDevice() with nvidia driver 9629

[Daniel Stone]

On Sun, Dec 03, 2006 at 03:55:47PM +0400, Mitch wrote:
> The structure seemed all but correct to me. The question is why is Xorg 
> treating the Damage Extension Manger as an input device.

Because the NVIDIA driver adds it as an input device.  Up until
recently, this was the only way to be able to listen on an fd.

> Seems to me 
> like an Xorg bug. Also changing the ABI without enabling versioning 
> dependencies seems like Xorg is breaking the API.

We have an input ABI version, which the NVIDIA driver does not check.

> I would expect Xorg to 
> refure to load the driver than crashing.

It does, but does not currently cope with video and input drivers in the
same driver.

> Crashing is always the wrong 
> thing to do, rather warn and exit than SEGV.  Are the closed binary 
> vendors supposed to be monitoring the groups for ABI changes ?

Yes, they are, and they are aware of this particular ABI change.

Cheers,
Daniel