the current preview-latex code base for XEmacs loads the image data
immediately into image specifiers instead of just specifying the file
name. The reason was some horrible bug that arose in connection with
Mule when loading dired.el or dired-x.el that caused image files to
get loaded with some character or EOL conversion applied (which
I believe this this might have been addressed at that time in
development. It turns out that the current installation procedure of
preview-latex for XEmacs has some chicken-and-hen-problem due to this
workaround code we used within preview-latex itself. The requirements
for up-to-date XEmacs versions for current AUCTeX versions are pretty
rigid, anyway, so I'd like to remove the workaround. This requires
a) making sure that indeed the current code base of XEmacs will load
image files without conversions (character set _and_ line endings) on
both MULE and non-MULE XEmacsen.
b) it would be _very_ nice to know just _what_ version of XEmacs
started doing this correctly so that we can mention it in the
installation prerequisites and dependencies.
If anybody can get a hold on this information, I'd be pretty glad.
The workaround appeared in preview-latex with the following ChangeLog
entry, so I expect that at a similar time an XEmacs fix, _if_ there
has been a fix, would have been concocted.
date: 2002/04/16 23:35:28; author: nixsf; state: Exp; lines: +19 -5
(preview-create-icon): Cater for file-handlers (as in dired) forcing
coding-system autodetection at glyph-instantiation time if the glyph must
be read from a file; read it by hand.
Thanks for any input you can give me on that. For one thing, I'd like
the version info to be correct, and for another, of course I want to
know if this bug _really_ is gone.
David Kastrup, Kriemhildstr. 15, 44793 Bochum