🚀 go-pugleaf

I've been having trouble with font-lock-mode in C buffers not
fontifying strings that follow a "/**********/" type comment.  For
example, try fontifying this program:-

----------------------------------------------------------------------
/****/

#include <stdio.h>

int main (int argc, char *argv[])
{
  printf ("Hello world\n");
}
----------------------------------------------------------------------

The "Hello world\n" string doesn't get fontified.

I've traced the problem to parse-partial-sexp.  It doesn't remember in
the returned `state' value that it has seen the first character of a
2-character comment terminator, so it thinks the rest of the buffer is
within a comment.

To prove this, use M-ESC to eval the following sexps in the
c-mode-buffer.

	(setq state (parse-partial-sexp 1 6))
	=> (0 nil nil nil t nil 0 nil)

This is the state up to, but not including, the terminating "/" of the
comment.

Now find the state to the end of the first line:-

	(setq state2 (parse-partial-sexp 6 7 nil nil state))
	=> (0 nil nil nil t nil 0 nil)

Note that (nth 4 state2) is `t'.  This is wrong.




What font-lock does when it is within a comment is to go back to the
beginning, do (forward-comment 1) to skip over the comment, then
(skip-chars-backward " \t") to get to the real end of the comment.  It
then uses parse-partial-sexp to work out the state immediately after
the comment.

The problem is that `comment-start-skip' matches everything up to the
last "*" in the comment, and this hits the bug in parse-partial-sexp.

I don't know how parse-partial-sexp should be fixed.  Perhaps we need
another element in the STATE value, or perhaps (nth 4 state) could
indicate that the first character of a 2-character comment end has
been seen.


BTW what does (nth 7 state) mean?  What is a "comment of style `b'"?
The version of the elisp manual I've got (for 19.19) only describes up
to (nth 6 state).  Guess I need to track down the diffs.


--
Kevin