[Fwd: Re: [cairo] pixman crash from setting clip region on image surface]

[Maarten Breddels]

Carl Worth wrote:

>On Fri, 11 Jun 2004 00:18:43 -0600, Peter Dennis Bartok wrote:
>  
>
>>I'm resending this since nobody responded and it didn't make it into CVS 
>>either. It really does fix a problem :-)
>>    
>>
>
>Hi Peter,
>
>I'm sorry that I haven't had a chance to look at this situation
>yet. I've been pretty swamped lately, but your patch is in my queue and
>hasn't been dropped.
>
>One reason I've been waiting for a bigger chunk of time before looking
>at this is that I want to do a fairly general cleanup of memory
>ownership issues throughout the cairo and libpixman interfaces.
>
>If someone else with commit access has already looked at the patch and
>feels it's ready to commit, that would be great. Otherwise, I will get
>to it in time.
>
>-Carl
>
>  
>
I'm afraid this isn't the whole story with clipping, i've encountered
this bug also, but this wasn't the only problem. If I have some time
I'll report some bugs here. A less serious bug(in _cairo_gstate_clip)
was when setting clipping while the 'path is  small' like drawing a
really tiny circle, the tesselation of the circle results in no
trapezoids, which results in a call to
_cairo_surface_create_similar_solid with a negative height or width,
which then result in a malloc of a few gigabytes(since the int is passed
as unsigned int somewhere along the path). I'm not sure which code is
bugging, the _cairo_traps_extents I assume.

-Maarten