Re: master 10a7615b5d4: Separate filename-deletion mechanism from policy.

[Andrea Corallo <acorallo@gnu.org>]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Po Lu <luangruo@yahoo.com>
>> Cc: "Eric S. Raymond" <esr@thyrsus.com>
>> Date: Sun, 06 Aug 2023 21:37:21 +0800
>> 
>> "Eric S. Raymond" <esr@thyrsus.com> writes:
>> 
>> >     This is a pure refactoring step, delete-file's behavior is
>> >     unchanged. But the C core is a little simpler now.
>> > ---
>> >  lisp/files.el | 20 ++++++++++++++++++++
>> >  src/fileio.c  | 40 +++++++++-------------------------------
>> >  2 files changed, 29 insertions(+), 31 deletions(-)
>> >
>> > diff --git a/lisp/files.el b/lisp/files.el
>> > index f8867432000..84a8c308b09 100644
>> > --- a/lisp/files.el
>> > +++ b/lisp/files.el
>> > @@ -6352,6 +6352,26 @@ non-nil and if FN fails due to a missing file or directory."
>> >        (apply fn args)
>> >      (file-missing (or no-such (signal (car err) (cdr err))))))
>> >  
>> > +(defun delete-file (filename &optional trash)
>> > +  "Delete file named FILENAME.  If it is a symlink, remove the symlink.
>> > +If file has multiple names, it continues to exist with the other names.q
>>                                                                           ^
>> Typo alert!
>> 
>> Thanks.  While I'm not enthusiastic about moving functions from C to
>> Lisp, others seem to appreciate the gesture, so I won't mention this
>> subject now.
>
> This kind of changes should have been discussed before it was
> installed.  There are several issues I can see here:
>
>   . delete-file-internal must call expand-file-name on its argument,
>     as all primitives that interface to the filesystem must; I fixed
>     that;
>   . Emacs will now be unable to call delete-file or rename-file during
>     loadup, until files.el is loaded -- not sure if this is a problem,
>     but I guess we will see;
>   . the changeset failed to adjust internal_delete_file, which still
>     called Fdelete_file; I fixed that (I hope, see below)
>
> Andrea, please take a look at the last bullet above: I made
> internal_delete_file call Fdelete_file_internal, which means it no
> longer supports remote files.  If that is a problem, we should call
> Qdelete_file there, but then I wonder whether calling Lisp at that
> place, even though delete-file is preloaded, can cause any trouble?

I think calling Qdelete_file *should* work as we guard against circular
dependency compilations commanded by the native compiler, but it's
impossible to be sure of that without testing the patch.

Andrea