bug#4533: 23.1: reverting fails to update line ending mode line

[Chong Yidong]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> I think the use of buffer-file-coding-system-explicit in revert-buffer
>> is bogus, and should be removed---see below.  What do you think?
>
> Its use is not bogus.  It's so that when a file's coding-system is
> incorrectly auto-detected, the user can force the use of a correct
> coding-system and subsequent revert-buffers won't disregard it.
>
> So maybe the problem is that C-x C-s should not set
> buffer-file-coding-system-explicit (unless the C-x C-s prompted the user
> to choose  a coding-system, I guess).

I see.  The comments in mule.el say that

;; This variable is set in these three cases:
;;   (1) A file is read by a coding system specified explicitly.
;;       after-insert-file-set-coding sets the car of this value to
;;       coding-system-for-read, and sets the cdr to nil.
;;   (2) A buffer is saved.
;;       After writing, basic-save-buffer-1 sets the car of this value
;;       to last-coding-system-used.
;;   (3) set-buffer-file-coding-system is called.
;;       The cdr of this value is set to the specified coding system.
;; This variable is used for decoding in revert-buffer and encoding in
;; select-safe-coding-system.

Indeed, this seems to imply that (2) can be omitted, as you suggest,
since "force selecting" a coding system should trigger (1) and (3).  Is
there any reason that (2) was originally included?