https://bugs.winehq.org/show_bug.cgi?id=55717
--- Comment #8 from Esme Povirk madewokherd@gmail.com --- (In reply to Jeff Smith from comment #4)
I have done a fair bit of experimenting around this, but I haven't nailed it down yet. It does NOT seem to rest on the backing context type (for example graphics->image). It may be something along the lines of the transform applying to both the region and the point under test, and in some edge cases native's result is "wrong"... The IsOutlineVisible_Line_WithGraphics test hitting on one such case.
I always thought it was strange that they rasterize the path to begin with, but given that they do it kinda makes sense to do it in device coordinates (which means transforming both of those).