https://bugs.winehq.org/show_bug.cgi?id=51139
--- Comment #13 from ms mst.misc@arcor.de --- (In reply to Henri Verbeet from comment #12)
Does it also help if you replace "if (isnan(t)) return TRUE;" with "if (isnan(t)) t = 0.0f;"?
i don't see any difference...
it's a couple pages of forms, the rectangles and separator lines appear to be present but the text is missing except for a few lone characters here and there.
of course that could be unrelated to the NaN problem...
there are some more traces about unimplemented methods: 0104:fixme:d2d:d2d_path_geometry_ComputeArea iface 37DC4E90, transform 01EEB644, tolerance 2.50000000e-01, area 01EEB640 stub! 0104:fixme:d2d:d2d_geometry_sink_SetSegmentFlags iface 37E22A60, flags 0 stub! 0104:fixme:d2d:d2d_device_context_DrawGeometry Ignoring stroke style 37DEC698.
In any case, from the log it looks like the reason we end up with "t" being NaN is that we have a degenerate (i.e., zero-length) line segment that intersects the curve segment. It would be good to figure out what kind of geometry reproduces that, so that we can add a test case for it.
will attach a copy-pasted test that reproduces the NaN loop...