[XeTeX] Embedding of fonts from included graphics

[Jonathan Kew]

On 1 Jun 2006, at 4:59 pm, Bruno Voisin wrote:
>
> Try this:
>
> %!TEX TS-program = xelatex
> %!TEX encoding = UTF-8 Unicode
> \documentclass[12pt]{article}
> \usepackage{graphicx}
> \begin{document}
> off \\
> \includegraphics{obstboth6}
> \end{document}
>
> and look at the result in Reader or Acrobat Professional 7: empty  
> boxes!

Yes, I can verify that I see problems (viewing with Acrobat Reader 5,  
my preferred version when I have to use an Adobe viewer).

> Now comment out the line containing the word "off": no more boxes! Or
> replace obstboth6 by obstboth5: no more boxes either!
>
> The problem seems to be triggered by the combination of 3 factors:
>
> - inclusion of the PDF file obstboth6.pdf, converted from EPS using
> GhostScript (as opposed to Apple's Distiller);
>
> - presence of the exact word "off" before the inclusion of the PDF  
> file;
>
> - viewing of XeTeX's PDF output in Reader or Acrobat 7.
>
> Does this mean "off" is incorporated without some form of protection
> inside the PDF stream, and interpreted as a PDF "keyword" by the
> viewer (if that is indeed the appropriate term)?

No. You'll find that "iff" has the same effect, or even "jiffy". Or  
"office", but not "orifice". As far as I can tell from a few  
experiments, any "ff", "ffi", or "ffl" ligature causes the problem.  
But "fi" and "fl" don't. Bizarre.

>
> In any case, and if that is reproducible on your setup, then that
> sounds like a bug, from either XeTeX, GhostScript or Acrobat.

It sure sounds like a bug *somewhere*.

Personally, I suspect Acrobat, as I've seen it garble fonts in PDFs  
before, when every other indication was that they're valid. But  
proving it (one way or another) isn't going to be easy.

> And then: why is it always me, that is faced with such things?

Because the software has a built-in deadline detector! :)

JK