Re: ANN: Emacs on GNUstep / OS X 8.0-rc1

[Adrian Robert]

On Mar 8, 2005, at 10:15 AM, M. Uli Kusterer wrote:

> At 16:42 Uhr -0500 07.03.2005, Adrian Robert wrote:
> > http://emacs-on-aqua.sf.net/
> >
> > Status respecting backends on GNUstep is unchanged: works on Xlib, has
> > rendering anomalies under Art.  Specifically, menus are rendered blank
> > when they change, and the letter 'f' is rendered in proportional 
> > spacing
> > (from NSString -drawAtPoint:withAttributes:) in all fonts tried (in 
> > some
> > cases the character after the 'f' must be 'i' or another 'f').
> > "defaults write Emacs GSBackend libgnustep-xlib" is necessary.
>
>  Not sure whether you already know the source of the "f" problem and 
> simplified the explanation, or whether you don't know where this comes 
> from, but this looks to me like ligatures are turned on where they 
> shouldn't. Ligatures are a typographical fanciness for representing 
> certain character pairs with a single character that looks like an 
> amalgamation of the two. Often done for fi, fl and a few others.
>
>  I think there should be a way to turn of ligatures, but you may have 
> to use the Cocoa text system directly instead of using NSString's 
> drawAtPoint:.

Yes, the rendering issue did have to do with ligature, but the fault is 
not in Art, but in GUI.  Art just reveals the issue because it 
implements ligature whereas Xlib does not.

The problem is in [GSLayoutManager -_run_cache_attributes], where it 
currently sets a glyph run's 'ligature' attribute by default to '1'.  
Either it should set this to '0' if the font is monospaced, or some 
lower level should ignore ligature=1 for fixed fonts.

This raises the question of how to detect monospaced fonts.. if you ask 
[NSFont -isFixedPitch], you usually get NO even for fonts that are 
"supposed to be" monospaced.  This is a known issue even on OS X, and 
relates to the fact that most fonts seem to have a few characters 
somewhere with a different width.  (This is presumably why, e.g., not 
all fixed-width fonts on the system show up under that category in the 
OS X font panel.)

One quick-and-dirty hack found on the web is to check if 'i' and 'm' in 
the font have the same width.  On Art, this test passes for FreeMono, 
which does indeed render as a fixed-width font, and fails for 
BitstreamVeraSansMono-Roman.  This is correct, as the latter seems not 
to be a fixed-width font.. see attached screenshot, in which the top 
row is a "natural" rendering (in TextEdit), while the bottom is a 
rendering in emacs when spacing is forced to be fixed (by moving the 
cursor over the row).  Paradoxically, in the natural rendering we see 
'm' is actually rendered _skinnier_ than 'i'.

Although correct in this case, the 'i'/'m' hack seems sloppy and also 
might be inefficient and/or inaccurate.  Other recommendations are 
welcome.

Also, can anyone recommend a good fixed-width font usable under Art?  
FreeMono looks greyed-out on my system...

Thanks.