bug#60146: file-exists-in-trash-p needs better name or semantics

[Eli Zaretskii]

> Date: Fri, 16 Dec 2022 21:17:13 -0800
> From: Paul Eggert <eggert@cs.ucla.edu>
> 
> The recently-added function (file-exists-in-trash-p FILE) is poorly 
> named since it's not really related to trash - it's simply checking for 
> the existence of a directory entry named FILE.
> 
> How about extending file-exists-p instead? (file-exists-p FILE t) would 
> be like (file-exists-p FILE) except it would not follow symlinks. This 
> extension can be implemented via a single system call on POSIX systems, 
> and this would be more efficient and would avoid a race in the current 
> implementation of file-exists-in-trash-p. (Though of course pretty much 
> any use of this new function makes one vulnerable to races....)
> 
> If extending file-exists-p is too much, at least please rename 
> file-exists-in-trash-p to something like files--exists-nofollow-p, to 
> indicate that it's private to files.el and to say better what it means.

I don't really mind renaming that function, but the reason I called it
like I did was that apparently no one needed such a functionality in
Emacs until now, except in this obscure use case of moving to trash.
That seems to tell me that extending file-exists-p would be a solution
waiting for the problem, something that we don't like doing.

In any case, if we do decide to extend file-exists-p, it would need to
be done on master, as that is not a trivial change, which has to be
done in C.

Lars, Stefan, WDYT?