https://bugs.winehq.org/show_bug.cgi?id=47872
--- Comment #10 from Vincent Povirk madewokherd@gmail.com --- I was hoping that hard-coding "MS Shell Dlg", instead of using the unexpected value we get out of DEFAULT_GUI_FONT, would fix the remaining failures. I still think this change is a step in the right direction - we shouldn't be testing parts of gdi32 that are already tested elsewhere.
We're still sort of testing gdi32 behavior here, as it substitutes another font for "MS Shell Dlg". But it's also testing that gdiplus succeeds in this case with the real font (detected using GetTextFaceW) and not MS Shell Dlg.
But now I'm confused, because test_nonexistent_font in gdi32:font also seems to test for substitution of "MS Shell Dlg" but it expects the face name to be unchanged.
Perhaps what we really want to test is that we get the same result from gdiplus that we would through gdi32 with the LOGFONT structure, but that basically means duplicating code from gdiplus into the tests.