bug#14434: 24.3; lisp mode's comment start seems bad

bug#14434: 24.3; lisp mode's comment start seems bad

[Leo Liu]

After fixing bug#14303, we can observe the following bug:

1. Emacs -q
2. insert in a emacs lisp mode buffer:
   (
     ;;
   )
3. Move point to be end of ;;
4. M-j

You should see:

   (
     ;;
     ;
   )

This seems to be due to bad comment-start-skip. Is the following fix correct?

--- lisp/emacs-lisp/lisp-mode.el.gz
+++ #<buffer lisp-mode.el.gz>
@@ -219,7 +219,7 @@
   (make-local-variable 'comment-start-skip)
   ;; Look within the line for a ; following an even number of backslashes
   ;; after either a non-backslash or the line beginning.
-  (setq comment-start-skip "\\(\\(^\\|[^\\\\\n]\\)\\(\\\\\\\\\\)*\\);+ *")
+  (setq comment-start-skip "\\(\\(^\\|[^\\\\\n;]\\)\\(\\\\\\\\\\)*\\);+ *")
   (make-local-variable 'font-lock-comment-start-skip)
   ;; Font lock mode uses this only when it KNOWS a comment is starting.
   (setq font-lock-comment-start-skip ";+ *")

Diff finished.  Wed May 22 12:24:18 2013

bug#14434: 24.3; lisp mode's comment start seems bad

[Andreas Röhler]

Am 22.05.2013 06:26, schrieb Leo Liu:
> After fixing bug#14303, we can observe the following bug:
>
> 1. Emacs -q
> 2. insert in a emacs lisp mode buffer:
>     (
>       ;;
>     )
> 3. Move point to be end of ;;
> 4. M-j
>
> You should see:
>
>     (
>       ;;
>       ;
>     )
>
> This seems to be due to bad comment-start-skip. Is the following fix correct?
>
> --- lisp/emacs-lisp/lisp-mode.el.gz
> +++ #<buffer lisp-mode.el.gz>
> @@ -219,7 +219,7 @@
>     (make-local-variable 'comment-start-skip)
>     ;; Look within the line for a ; following an even number of backslashes
>     ;; after either a non-backslash or the line beginning.
> -  (setq comment-start-skip "\\(\\(^\\|[^\\\\\n]\\)\\(\\\\\\\\\\)*\\);+ *")
> +  (setq comment-start-skip "\\(\\(^\\|[^\\\\\n;]\\)\\(\\\\\\\\\\)*\\);+ *")
>     (make-local-variable 'font-lock-comment-start-skip)
>     ;; Font lock mode uses this only when it KNOWS a comment is starting.
>     (setq font-lock-comment-start-skip ";+ *")
>
> Diff finished.  Wed May 22 12:24:18 2013
>
>
>
>

Maybe it would be worth to re-consider that comment-start-skip.
What about to drop it?

All we need is the position of comment start and it's end. What to do with code before resp. after, let the modes decide.

AFAIU you can't check for an even number of backslashes with a regexp, never.
So that will remain buggy.

Cheers,

andreas

bug#14434: 24.3; lisp mode's comment start seems bad

[Stefan Monnier]

> After fixing bug#14303, we can observe the following bug:
> 1. Emacs -q
> 2. insert in a emacs lisp mode buffer:
>    (
>      ;;
>    )
> 3. Move point to be end of ;;
> 4. M-j

> You should see:

>    (
>      ;;
>      ;
>    )

> This seems to be due to bad comment-start-skip.

That's one way to see it.  Another is that it's a regression introduced
by the fix to bug#14303.  It probably affects more than
emacs-lisp-mode.  So it'd be better to fix it by improving your fix to
bug#14303.  E.g. after the new while loop which finds the comment start, do
a looking-at to make sure the match-data extends as far as possible.

Tho maybe, we should take a new look at bug#14303 to see why it failed,
since after all the old comment-search-backward did already pay
attention to comment-use-global-state (via comment-search-forward), so
maybe bug#14303 was really due to a bug in comment-search-forward.


        Stefan

bug#14434: 24.3; lisp mode's comment start seems bad

[Leo Liu]

On 2013-05-23 05:32 +0800, Stefan Monnier wrote:
> That's one way to see it.  Another is that it's a regression introduced
> by the fix to bug#14303.  It probably affects more than
> emacs-lisp-mode.  So it'd be better to fix it by improving your fix to
> bug#14303.  E.g. after the new while loop which finds the comment start, do
> a looking-at to make sure the match-data extends as far as possible.
>
> Tho maybe, we should take a new look at bug#14303 to see why it failed,
> since after all the old comment-search-backward did already pay
> attention to comment-use-global-state (via comment-search-forward), so
> maybe bug#14303 was really due to a bug in comment-search-forward.

Thanks, Stefan, for the hints.

The fix for bug#14303 seems bad so I have reverted it and instead set
comment-use-global-state to t in octave-mode.

Leo

bug#14434: 24.3; lisp mode's comment start seems bad

[Stefan Monnier]

> The fix for bug#14303 seems bad so I have reverted it and instead set
> comment-use-global-state to t in octave-mode.

Great, thank you,


        Stefan