[cairo] _cairo_win32_display_surface_map_to_image issue

[Fred Beca]

> I'm not really sure that the data gets corrupted. It seems more likely to me
> that some GDI function doesn't behave as expected or has a different
> interpretation of some parameter. That could mean that e.g. too few bytes
> are allocated for the image surface and things go downhill when cairo/pixman
> try to access the bytes near the end of the image surface.

If you look at the details of my previous report, the crash indeed
occurs in a cairo function: data is not exactly "corrupted", but the
offset in the pixman image used for backing is wrong (the offset seems
to be applied twice. The problem for me is to find where it comes
from). I have not seen (yet) any GDI call in the path.

The problem seems to come from the fact that the backing image is
created with the dimensions of the clipping rectangle, and yet it is
considered as covering the entire original area of the HBITMAP.

Regards,

Fred.