[cairo] 1.9.8: problem with direct access to image surface data

[cu]

I thought it might be some caching thing. Thanks, will go add that
everywhere.

Chris Wilson wrote:
>
> The documented behaviour has always been:
>
> /**
>  * cairo_surface_mark_dirty:
>  * @surface: a #cairo_surface_t
>  *
>  * Tells cairo that drawing has been done to surface using means other
>  * than cairo, and that cairo should reread any cached areas. Note
>  * that you must call cairo_surface_flush() before doing such drawing.
>  */
>
> cairo is becoming much more dependent on correct behaviour, wrt zero-copy
> snapshots and backends that need to maintain surface caches, and for
> backends that may defer rendering. (And it is always possible that the
> image surface may start deferring some its rendering...)
>
> Hope this helps.
> -ickle
>
>