bug#6902: 23.2; indent-line-to possible devide by zero error bug

bug#6902: 23.2; indent-line-to possible devide by zero error bug

[Arik Mitschang]

In indent.el the indent-line-to function has a comparison that devides
the current column count by tab-width, whose value is an integerp and
allowed to be zero (and in my experience is often 0). I noticed the
issue in specific with org mode which utilizes this function for a lot
of interesting tasks.

I can provide the following patch which makes a check to avoid an
arithmetic error:

=== modified file 'lisp/indent.el'
--- lisp/indent.el	2010-05-19 03:06:48 +0000
+++ lisp/indent.el	2010-08-24 15:20:39 +0000
@@ -177,7 +177,9 @@
   (back-to-indentation)
   (let ((cur-col (current-column)))
     (cond ((< cur-col column)
-	   (if (>= (- column (* (/ cur-col tab-width) tab-width)) tab-width)
+	   (if (and (> tab-width 0)
+		    (>= (- column (* (/ cur-col tab-width) tab-width)) 
+			tab-width))
 	       (delete-region (point)
 			      (progn (skip-chars-backward " ") (point))))
 	   (indent-to column))

Thanks,
~Arik

bug#6902: 23.2; indent-line-to possible devide by zero error bug

[Glenn Morris]

Arik Mitschang wrote:

> In indent.el the indent-line-to function has a comparison that devides
> the current column count by tab-width, whose value is an integerp and
> allowed to be zero (and in my experience is often 0).

Why do you want to set tab-width to 0?

A quick grep for tab-width shows other places that will break.

bug#6902: 23.2; indent-line-to possible devide by zero error bug

[Arik Mitschang]

Glenn Morris writes:
 > Arik Mitschang wrote:
 > 
 > > In indent.el the indent-line-to function has a comparison that devides
 > > the current column count by tab-width, whose value is an integerp and
 > > allowed to be zero (and in my experience is often 0).
 > 
 > Why do you want to set tab-width to 0?

Honestly I cannot remember, but I had it set there and I assume there
was a reason.

 > 
 > A quick grep for tab-width shows other places that will break.

The fact that it's value can be zero (without even warning in the doc
string) and this value causes functionality to break doesn't seem good
to me. If nobody but me ever had it set to zero, then perhaps just
updating its doc-string to warn users away from that value would be
nice.

Thanks,
~Arik

bug#6902: 23.2; indent-line-to possible devide by zero error bug

[Chong Yidong]

Arik Mitschang <arik.mitschang@gmail.com> writes:

>  > A quick grep for tab-width shows other places that will break.
>
> The fact that it's value can be zero (without even warning in the doc
> string) and this value causes functionality to break doesn't seem good
> to me. If nobody but me ever had it set to zero, then perhaps just
> updating its doc-string to warn users away from that value would be
> nice.

I have made such an update to the doc-string.  It doesn't seem
worthwhile to try to handle zero tab-widths in all the places where it
could be problematic.  Thanks.