Re: Inconsistency in `string-to-number'

[Stefan Monnier]

>> Except that it will break existing code.
> <sigh> I know, as I explicitly acknowledged that fact in my message...
> Also, it was Eli who suggested it.

Don't take it personally, I was just pointing out the part which I found
to be most important.  I know I have written code which relies on this
behavior (tho I can't remember where ... hmm ... probably there's some
of it in diff-mode.el).

> Still, that code relies in undocumented behavior.  As the "1:" vs.
> "1.2:" vs. "1.2" difference demonstrates, it's just luck that
> it works.

No doubt this is a bug.  It's a long-standing one, tho, so it's not
a regression in Emacs-23.

Eli said:
> Which is why I thing we should fix it the other way around: to
> recognize "1.2:" as a number.  That cannot possibly break anything, so
> it should be safe to do before the release, IMO.

Actually it could also break code (tho it seems unlikely).  I wouldn't
mind postponing the fix to Emacs-23.2.  In the mean time, we need to
improve the docstring.


        Stefan