bug#49893: [PATCH] Reset mtime of a reverted buffer

[Konstantin Kharlamov <hi-angel@yandex.ru>]

On Thu, 2021-08-05 at 19:36 +0300, Eli Zaretskii wrote:
> > From: Konstantin Kharlamov <hi-angel@yandex.ru>
> > Date: Thu, 05 Aug 2021 18:28:28 +0300
> > 
> > Patch is attached. This resolves the problem reported at
> > https://github.com/emacs-evil/evil/issues/1504
> 
> Could you please describe the problem you are trying to solve,
> preferably without involving Evil?

Sure. The auto-revert-mode by default works with ‘revert-buffer-insert-file-contents--default-function’ function. This function is known to break markers in buffers, which is why recently Emacs has added a replacement function revert-buffer-insert-file-contents-delicately (the one I modify in the patch).

However, actually trying to use this new function revealed a regression in behavior of another function: the `find-file`. Basically, if you have a file `/tmp/foo` opened in Emacs (IOW Emacs has a buffer associated with this file), and then file `/tmp/foo` gets "auto-reverted", then if you execute (find-file "/tmp/foo"), the new function causes Emacs ask a user "File foo was modified, do you want to revert it? (yes/no)". It now gives that prompt always, until you make a change to the buffer.

That's a regression compared to the default behavior with `revert-buffer-insert-file-contents--default-function`. And the reason turned out to be that the function `revert-buffer-insert-file-contents--default-function` after having succesfully reverted a file, sets the buffer mtime to the mtime of the file. However the function revert-buffer-insert-file-contents-delicately didn't set mtime before that patch. I assume it is an omission from implementation, because technically that's incorrect: if the revert-buffer-insert-file-contents-delicately has successfully reverted a buffer, then we know that it has same content as the associated file, and hence it should have the same mtime.