bug#64071: 28.2; smerge-diff creates *vc-diff* without setting it up

[Spencer Baugh <sbaugh@janestreet.com>]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Spencer Baugh <sbaugh@janestreet.com>
>> Date: Wed, 14 Jun 2023 17:41:39 -0400
>> 
>> 1. emacs -Q
>> 2. Create a buffer containing:
>> <<<<<<< left
>> foo
>> =======
>> bar
>> >>>>>>> right
>> 3. M-x smerge-mode
>> 4. M-x smerge-diff-upper-lower
>> 5. The resulting *vc-diff* buffer is in diff-mode but it's not
>> read-only, which is unusual for a *vc-diff* buffer.
>
> That command enters diff-mode, and diff-mode doesn't force read-only
> status on the current buffer.  Why should Smerge force that?

Buffer named "*vc-diff*" are universally read-only, except if they are
created through this path.  (Because they're otherwise only created by
vc, which marks them read-only.)  It breaks user expectations for a
*vc-diff* buffer to not be read-only.

>> Probably there is also other setup missing.
>
> Which ones, and to what mode they belong?

Setup done by vc when it creates *vc-diff* buffers.

After doing a little investigation, it looks like the read-only setting
is the most important divergence after all.

Maybe we should be calling diff-setup-whitespace, and maybe we should be
running vc-diff-finish-functions.  Possibly not though.

>> Relatedly, if a *vc-diff* buffer already exists, smerge-mode will use it
>> without updating the mode-line, so it will say, for example, "Diff from
>> *vc-change-log*" despite the diff being from smerge conflict resolution.
>
> This should probably be easy to fix, and the fix should be safe enough
> for the emacs-29 branch.  How about proposing such a fix, and leaving
> the more controversial aspects of this alone for now?

Hm, honestly I only care about the read-only-mode, that's by far most
confusing thing about this.  I've gotten user complaints about that
specifically being confusing, so that's what I'd like to fix.  Stefan
likes my simple patch, so what's wrong with it?