[cairo] [1.10.0] Win32 redrawing issues (regression?)

[Cxx Joe]

Alright, I'll pick this up again. I have been stepping through the
code and came across this notable difference between 1.8.10 and
post-b713510f213d0a83cb0575d6870416a0f13786c9.

I would very much appreciate it if someone from the dev team could
comment on this.

1.8.10: http://img521.imageshack.us/img521/1801/workingx.png
1.9/1.10: http://img413.imageshack.us/img413/112/notworkinga.png

As you can see, in 1.9 and 1.10, src_y, mask_y and dst_y are all zero
when they were 630 with the old code.

This causes a pretty different call to pixman_image_composite which
all in all looks suspicious to me!

1.8.10: http://img440.imageshack.us/img440/6971/working2.png
1.9/1.10: http://img843.imageshack.us/img843/1813/notworking2.png

The code that invokes cairo is exactly the same, so are all the input
parameters (like cairo_win32_surface_t->extents).

So, please let me know whether I'm barking up the wrong tree, what
other things I could look at, if this seems weird to you as well...
anything! please!

Best regards
Joe