> I'm interested, but it is good to post the explanation so it becomes
> part of the public record google can find.
Well, that wouldn't be the first time I report a bug indirectly; so here goes:
When font dictionaries are altered (by changing FontMatrix, etc ...)
in PostScript, it is customary to keep a copy of the original
dictionary, which appears in the modified dictionary as a object
associated with the key OrigFont. Now, if a programmer wants to access
the original dictionary, he / she must of course expect to search the
dictionary recursively; hence the original definition of getorigfont
(which wasn't that bad actually):
% Here the stack must contain a font dictionary
/getorigfont
{
dup /OrigFont known
{
/OrigFont get getorigfont
}
if
}
def
This definition could have worked in other languages ... but not in
PostScript, because of the nature of dictionaries: they're so-called
“composite objects”, meaning that if you copy this object on the stack you get
eaxctly the same physical object (this is a bit like pointers in C); for
further details see the PostScript Language Reference Manual, section 3.3.1
(http://www.adobe.com/products/postscript/pdfs/PLRM.pdf). Anyway, this means
that a dictionary can contain itself, and this was the case with the font at
stake (LMSans by the Latin Modern team, converted from cminch by Knuth). So
retrieving OrigFont yielded the exact same dictionary at each iteration, and
this created the endless loop.
This (somewhat weird) definition of the font only started with
Ghostscript 8.56; this is why the bug didn't happen with older versions
of Ghostscript.
The real bugfix Wolfgang Glunz (author of pstoedit) did was to check
at each iteration if the retrieved OrigFont dictionary was actually
different, thus avoiding the loop, but still looking for the “most
original” font dictionary. As I said, the new code will be incorporated in
pstoedit 3.45 and looks like:
% font getorigfont font
/getorigfont
{ 210 pscover
%
% retrieves the father font of the input font if possible
% returns a font.
%
dup /OrigFont known
% OrigFont is provided at least if Font was generated with makefont
{ 211 pscover
dup % for comparison
dup /OrigFont get ne % ne is a weak check against recursion
% but is was needed from gs8.56
{
/OrigFont get getorigfont
} if
} if
}
def
For people wanting to experiment with pstoedit, you will certainly find the
-pscover switch most useful: it writes to the file “pscover.txt” a unique
number corresponding to a particular chunk of code in the PostScript prologue
(the psinXXXXXX file); look for “xyz pscover”, where xyz is a integral number;
this is how I found the bug in the first place.
That's about all. Hope you had fun reading this,
Arthur
___________________________________________________________________________________
If your question is of interest to others as well, please add an entry to the
Wiki!
maillist : ntg-context@ntg.nl / http://www.ntg.nl/mailman/listinfo/ntg-context
webpage : http://www.pragma-ade.nl / http://tex.aanhet.net
archive : https://foundry.supelec.fr/projects/contextrev/
wiki : http://contextgarden.net
___________________________________________________________________________________
