Re: new buffer - should its mode reflect its name when the name matches auto-mode-alist?

[Stefan Monnier <monnier@iro.umontreal.ca>]

>> I'm not opposed to such a change but I think it would need to come
>> together with a cleanup of the "default-major-mode" thingy: currently,
>> the major-mode of a buffer is either set (in set-buffer-major-mode) as
>> (default-value 'major-mode) or (if that's nil) is inherited from the
>> major-mode of the current-buffer (unless that major-mode has the
>> `mode-class' property set to `special').
>> 
>> In practice, for 99% of the users, (default-value 'major-mode) is
>> fundamental-mode, so the rest is very rarely used, which can lead to
>> corner-case bugs, of course.

> Using (setq-default major-mode 'normal-mode) would be a good default.
> Currently this runs into an infinite recursion.  However, the following
> succeeds:

> (setq-default major-mode
>               (lambda () (let ((buffer-file-name (buffer-name)))
>                            (set-auto-mode))))

That's also an alternative, yes.
So how 'bout the following:
- we (setq-default major-mode ..) along the lines of the above proposal.
- we get rid of the special case "or (if that's nil) is inherited from
  the major-mode of the current-buffer (unless that major-mode has the
  `mode-class' property set to `special')."
- but we only do that for Emacs-24.

The reason why I want to delay it to Emacs-24, is because I'm worried
that it may introduce bugs in very rare corner cases which will only
show up after a long testing period.


        Stefan