Re: master 400df210ce0: Fix last change of 'delete-file'

[Eli Zaretskii]

> From: Michael Albinus <michael.albinus@gmx.de>
> Cc: Robert Pluim <rpluim@gmail.com>,  emacs-devel@gnu.org, Eric S. Raymond
>  <esr@thyrsus.com>
> Date: Fri, 11 Aug 2023 09:33:09 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> Hi Eli & Robert,
> 
> >>     >> I was thinking of the opposite case, where someone has a personal 
> >> file
> >>     >> handler with a regexp with an absolute path in it.
> >> 
> >>     Eli> That'd be a buggy file handler, I think, unless it also checks
> >>     Eli> default-directory.
> >> 
> >> The file handler isnʼt buggy: it never gets a chance to run.
> >> 
> >> Let me illustrate. Suppose we have an entry of
> >> 
> >> ("\\`/var/.*\\.txt" . my-file-handler)
> >> 
> >> in `file-name-handler-alist'.
> >> 
> >> (delete-file "foo.txt") when `default-directory' is "/var" would
> >> previously result in `my-file-handler' being called. Now itʼs not
> >> called at all, and the deletion is handled by `delete-file-internal'
> >
> > But the same would be true for substitute-in-file-name, for example,
> > and for directory-file-name, and file-name-as-directory, and several
> > other primitives, which call Ffind_file_name_handler without calling
> > expand-file-name before that.
> >
> > But maybe I'm missing something here, so let's ask Michael (CC'ed) for
> > his opinion on this.
> 
> In Emacs 29, we have in Fdelete_file:
> 
> --8<---------------cut here---------------start------------->8---
>   filename = Fexpand_file_name (filename, Qnil);
> 
>   handler = Ffind_file_name_handler (filename, Qdelete_file);
>   if (!NILP (handler))
>     return call3 (handler, Qdelete_file, filename, trash);
> --8<---------------cut here---------------end--------------->8---
> 
> So the file name is always absolute. Now in Emacs 30, the file name is
> not expanded prior calling find-file-name-handler in delete-file. That
> is an error in the transition from C to Lisp.
> 
> Cc to Eric, who has done the transition.

That's my change, not Eric's.  And I still believe file handlers
should be able to deal with non-absolute file names (as they do
elsewhere).  Let's see what Michael says about this.