imacarthur wrote: > On 30 Nov 2008, at 13:07, Albrecht Schlosser wrote: >> >> I tested, and it seems to work okay on Windows and Linux/Xft, but there >> seem to be some "off-by-one" errors with some glyphs at the right border >> (e.g. the capitol A in your example). However, since this obviously is >> glyph dependent, I think this must be a bug in the font itself(?). > > It is possibly in the font glyphs, but also (sometimes) I think there > are maybe "rounding errors" introduced by the anti-aliasing when the > font is rendered. I'm not sure though. > Sometimes it looks worse on this Mac (for example) than it does under > linux/XFT, using ostensibly the same font, and I assume that is because > the anti-aliasing mechanisms are different. I could be talking rubbish.
Sounds reasonable. And I don't think that we (you?) should do much more WRT this, but I wanted to have mentioned it. Maybe I didn't see a very light gray pixel, but it was there (because of anti-aliasing, as you wrote). >> Two implementation notes: >> >> (1) I wonder if the Windows-specific defines need to be in fl_font.cxx. >> IMHO they should be in fl_font_win32.cxx (without the enclosing "#ifdef >> WIN32"). > > That was my original intent, but the defines need to be set before the > inclusion of windows.h, which is done indirectly in fl_font.cxx via > "FL/x.H". > So I had to put the defines in fl_font.cxx itself - but they should have > no impact anywhere else, so I guess it is safe enough (but see below...) Okay, I see... >> (2) This update definitely breaks WindowsNT and probably Windows9x >> version >> support. I started a RFC about future Windows version support in >> fltk.development [1] - please discuss there, and we should add a summary >> or decision here later. I know that we could try to load the offending >> function "GetGlyphIndicesW" dynamically, but IMHO we should only do this >> if needed. > > Yes - this is a problem I knew I was creating, even as I created it. I > do not know what is best here. [...snip...] > If we decide we should try to, I will write the code, it should be > straightforward (I'll follow the pattern I used in screen_xywh.cxx) but > I have no means of testing it properly, so we might have to find someone > with an older Windows machine. I don't even have a Win2K box now, the > oldest I have is WinXP. I have access to Win2K and WinNT4/SP6, and I intend to build a virtual machine with Win98 - if we need it, I can probably test it. Albrecht _______________________________________________ fltk-dev mailing list [email protected] http://lists.easysw.com/mailman/listinfo/fltk-dev
