Re: [Fabrice Popineau] Re: Need some bug/version info

* David Kastrup (2005-02-05) writes:

> Ralf Angeli <[email protected]> writes:
>
>> After that the syntactic pass should pick up the syntax properties
>> and fontify the indicated regions.  This is done by the function
>> `font-lock-fontify-syntactically-region'.  But this function does
>> not seem to find the syntax properties.  It uses
>> `parse-partial-sexp' and all sorts of checking on the resulting
>> states which I don't understand.
>
> It appears obvious that it has been a mistake of mine not to mention
> in the release notes for AUCTeX that it generally appears to be a bad
> idea to use XEmacs with AUCTeX if you can avoid it.

I hope this is not a conclusion based on my current disability to
debug XEmacs code.  If the bug had been in Emacs instead of XEmacs I
would have gotten stuck at the same place in the code.


By the way, I just made a big leap forward with respect to the
fontification problem.  It seems that this could actually work in
XEmacs.  After several hours of poking around I found the following
causes why it hasn't been working until now:

  1) `font-lock-syntactic-keywords' doesn't get set in XEmacs.  We
     currently set the variable through `font-lock-defaults' which is
     working fine in Emacs.  While we setq `font-lock-defaults'
     directly, XEmacs font-lock.el proposes to put the respective
     value onto the major mode symbol.  I tried that and it is not
     working either.  I have to investigate further how to do this
     correctly.

  2) The syntax of the matcher for syntactic keywords differs between
     XEmacs and Emacs.  In Emacs it is something like `(1 "\"")' and
     in XEmacs the same has to be written as `(1 (7 . nil))'.

  3) Syntax properties are taken into account during sexp parsing only
     if `parse-sexp-lookup-properties' is t.  In XEmacs 21.4 the
     default value for it is nil.

While finding out all these things the following testcase evolved. You
can test it by pasting it into the *scratch* buffer and typing `C-x
C-e' at the last parenthesis.

(progn
  (require 'font-lock)
  (setq font-lock-cache-position
	(save-excursion (goto-char (point-min)) (point-marker)))
  (insert "\n\nasdf zasdfz asdf \"asdf\"")
  (add-to-list 'font-lock-syntactic-keywords
	       '("\\(z\\)[^z]+\\(z\\)" (1 (7 . ?z)) (2 (7 . ?z))))
  (setq parse-sexp-lookup-properties t)
  (font-lock-fontify-region (point-at-bol) (point-at-eol)))

The string between the "z" characters as well as the one between the
quotation marks should be highlighted.  The latter is a positive test
for checking if fontification is actually happening.

By implementing some brute force changes in my local font-latex.el I
already managed to make XEmacs fontify verbatim environments.
Verbatim commands are not working yet.

I just wonder why the interfaces of Emacs' and XEmacs' font-lock.el
have to differ so much ...

-- 
Ralf