[XeTeX] fontspec v1.7

[Will Robertson]

On 20 Jun 2005, at 10:00 AM, Ross Moore wrote:

> On 20/06/2005, at 10:11 AM, Will Robertson wrote:
>
>> That's a very reasonable request. Always good to stress test our  
>> software ;)
>> I'll see what I can do! (Although I don't relish the idea of  
>> digging into LaTeX's maths code to figure out what it's trying to  
>> do...)
>
> In that case, maybe I should try to help.
>
> Will, send me a 1-liner that fails, loading whatever packages you can
> reduce down to.
> Let me take it from there.
>
> (I've some time for this, now that exams have started and I've almost
> 2 weeks before receiving any more scripts to mark.)

Here's as minimal as I have time to put together. No packages, just  
raw LaTeX. Obviously, this doesn't encompass the scope of the  
problem, but it should be direct enough to pinpoint the problem:

\documentclass{article}
\DeclareFontFamily{U}{osx}{}
\DeclareFontShape{U}{osx}{m}{n}{<-> "Hoefler\space Text"}{}
\DeclareSymbolFont{break}{U}{osx}{m}{n}
\DeclareMathAccent{\dot}{\mathalpha}{break}{"5F}
\begin{document}
$\dot x$
\end{document}

Actually, I suspect it's simply something along the lines that  
Hoefler Text has no meaningful character at "5F and the request for  
that glyph is confusing ATS, rather than a flaw in xdv2pdf.

> BTW, did you see my comment concerning the \fontspec commands
> within table-cells.
>
> Maybe \fontspec  should have a 'global' option that allows
> the \font command to be done as  \global\font  and set
> a global flag, rather than just a local one.

Yeah, good spotting with that.
Can you see any disadvantage to *always* using global definitions?
Different font families are used for fonts selected with different  
font feature permutations, but I guess for space reasons people may  
need to limit the number of font families they have defined at any  
one time.

Do you happen to know if it is enough to say \globaldefs=1 when  
defining the LaTeX font family to get the same effect (since I'm not  
using the \font primitive directly)? I would assume so, but time  
presses...

Cheers,
Will