Re: font-backend branch

[Jason Rumney <jasonr@gnu.org>]

Kenichi Handa wrote:
> To understand why a correct font can't be used on Windows, I
> must track the code of fontset by gdb, but at the moment, I
> can't build Emacs correctly on Windows even with these gcc
> and gmake:
>   

At which point is the build failing now? It might simplify things to 
copy your lisp directory from another plaform to avoid the need to 
bootstrap, as that is where most subtle problems with Windows build 
environments come up.

I think I have fixed the font selection problem, by adding code to copy 
the entries from font_entity to the new font_object in 
w32font_open_internal.

Now there is one remaining problem that I can see. When no font is found 
for some characters, display on that line is corrupted. The first time 
such a problem is encountered, all characters in the default face on 
that line are corrupted (I think this is caused by hitting a problem 
getting the glyph code, so code in w32font_encode_char decides to switch 
to using Unicode output for that font from that point on, so any glyphs 
that have already been encoded but not yet displayed are invalidated - 
the code was designed like this because for non-truetype fonts, and on 
older versions of Windows, it is not possible to get a glyph code, but I 
did not expect it to fail for a font where it had previously succeeded). 
After that, there is always a garbage character following the boxes that 
indicate a missing font. If the cursor is anywhere on the line, then the 
garbage character changes, at the same rate the cursor is blinking at.