Re: another font-selection inconsistency

[Jason Rumney]

Miles Bader wrote:
>     Then the resulting font _is_ dejavu sans mono, but is not
>     anti-aliased.  C-u C-x = says:
> ...
>     Then the resulting font is _still_ dejavu sans mono, but _is_
>     anti-aliased (it's fairly clear it's the same font, just with
>     anti-aliasing turned on).  C-u C-x = says:
> ...
> (3) If I then turn around, and do step (1) again, the font "stays"
>     anti-aliased, so clearly some state that set-frame-font changes, but
>     set-face-attribute:family:size does not, made a difference...]
>   

A couple of days ago, Kenichi Handa stated that there is currently no 
priority given to backends. I think fixing this would clear this up. 
Another thing that might help is making sure that Truetype fonts are not 
being provided to the xfont backend through your X configuration, but 
that's a workaround not a solution. Having a consistent priority for 
backends is also important on Windows, where opentype fonts are provided 
by both the uniscribe backend (with complex script support and other 
features) and the gdi backend (with some minor differences in processing 
for Windows 9x compatibility). My impression before the font-backend 
branch merge was that there was a consistent priority, even if it was 
undocumented, but recently there is definitely not.