Re: Parens in column 1 equals defun fiasco

Pardon the stream-of-consciousness post. -- Hrvoje Niksic <hniksic(a)arsdigita.com&gt; spake thusly: I haven't researched this fully, but I suspect this bug is in the way font-lock.c operates. However, in addition to porting the syntax-table code from GNU Emacs, I also ported `font-lock-fontify-syntactically-region' to use `parse-partial-sexp' instead of `syntactically-sectionize' (the font-lock.c method of syntaxification). So if I'm correct, this bug didn't exist on GNU Emacs because they don't have font-lock.c, and this bug is now fixed because we're not using it. I copied your sample code twice into an empty tcl-mode and did a `font-lock-fontify-buffer', and it fontified correctly. I also tried the following in the scratch buffer: --- (defun emacs-lisp-byte-compile () "Byte compile (the file containing the current buffer." (interactive) (if buffer-file-name ;; XEmacs change. Force buffer save first (progn (save-buffer) (byte-compile-file buffer-file-name)) (error "The buffer must be saved in a file first."))) --- And the '(' I stuck in the comment didn't seem to screw up fontification. However, indenting still seems to be somewhat hosed (`indent-sexp' works properly expect for the comment, but `indent-for-tab-command' doesn't), and fontification doesn't happen properly if a line is changed, although a fontify-buffer will fix it. So the bug appears to be half-fixed, and the main problem appears to be `beginning-of-defun', which `syntactically-sectionize' uses to set up its caches, and which is incorrectly reporting the embedded parenthesis as the start of defun. For what it's worth, `beginning-of-defun' is broken on GNU Emacs, but their fontification isn't, although indenting is broken in the same way. I suspect the reason fontification works is that they use a different method of refontifying which hides the problem. I can research this a bit further and try to fix it. I suspect that there are at least a couple solutions. One of them would be to apply custom syntax tables to all strings to mark opening parentheses as punctuation, but that seems overkill. It might also be possible to fix `beginning-of-defun' to use `parse-partial-sexp' instead of just doing a regex search. I've also been wanting to redo lazy-lock-mode to bring it more in line with GNU Emacs', which is more sophisticated and has more elegant behavior. I realize there are some issues with XEmacs not having the same hooks that GNU Emacs' lazy-lock uses, but I think those issues can be resolved. I also think that if proper behavior can be achieved, it might be worth rolling the lazy-lock functionality into font-lock-mode, as long as there's no opposition to it.