Re: copy-directory

Re: copy-directory

[Stefan Monnier]

> Please don't rely on the `chown' command being available for the
> preserve-uid-gid option.

Especially since chown only works for root anyway, and since I don't
think we should cater specially for root-uses of Emacs (I rather
discourage the use of Emacs when running as root), I'd rather we don't
touch the uid.


        Stefan

Re: copy-directory

[Michael Albinus]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Please don't rely on the `chown' command being available for the
>> preserve-uid-gid option.
>
> Especially since chown only works for root anyway, and since I don't
> think we should cater specially for root-uses of Emacs (I rather
> discourage the use of Emacs when running as root), I'd rather we don't
> touch the uid.

I've tried to be as close as possible to `copy-file'. There is the code

#ifndef MSDOS
  /* Preserve the original file modes, and if requested, also its
     owner and group.  */
  if (input_file_statable_p)
    {
      if (! NILP (preserve_uid_gid))
	fchown (ofd, st.st_uid, st.st_gid);
      fchmod (ofd, st.st_mode & 07777);
    }
#endif	/* not MSDOS */

Shall we add a subroutine function for this, as proposed by Eli? Or
shall we remove PRESERVE-UID-GID from `copy-directory'?

>         Stefan

Best regards, Michael.

Re: copy-directory

[Stefan Monnier]

> Or shall we remove PRESERVE-UID-GID from `copy-directory'?

I guess we can remove it for now,


        Stefan

Re: copy-directory

[Michael Albinus]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Or shall we remove PRESERVE-UID-GID from `copy-directory'?
>
> I guess we can remove it for now,

OK, I'll do so.

>         Stefan

Best regards, Michael.

Re: copy-directory

[David Kastrup]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Or shall we remove PRESERVE-UID-GID from `copy-directory'?
>
> I guess we can remove it for now,

I think that UID/GID should transparently be preserved when copying
directories whenever possible (in particular when using /sudo:: methods
or similar).  I don't think that there is an important usage case for
"don't preserve UID/GID even if you could": it would likely imply
copying a tree and turning it into root/root possession on today's
systems.

I would like to add that before the time of disk quotas, the owner of a
file was always able to do chown/chgrp on files (this cleared the setuid
bit for security reasons, but that was it).  It was the usual way of
passing data to somebody else (and nobody other).

-- 
David Kastrup

Re: copy-directory

[Michael Albinus]

David Kastrup <dak@gnu.org> writes:

> Stefan Monnier <monnier@iro.umontreal.ca> writes:
>
>>> Or shall we remove PRESERVE-UID-GID from `copy-directory'?
>>
>> I guess we can remove it for now,
>
> I think that UID/GID should transparently be preserved when copying
> directories whenever possible (in particular when using /sudo:: methods
> or similar).  I don't think that there is an important usage case for
> "don't preserve UID/GID even if you could": it would likely imply
> copying a tree and turning it into root/root possession on today's
> systems.

Although it is not on my top priority, I would second David. A primitive
function `set-file-owner' would not harm, if it is explicitely said in
the docstring, that it works only in the cases the underlying OS
supports it.

And a Tramp file name handler could be called for this, useful at least
for "su" and "sudo" connections.

What I'm curious: does change ownership work for other systems, like
Win32? The smbclient, used in Tramp for "smb" connections, offers a
chown subcommand ... maybe it is applicable only, if there is a Samba
server at the other end.

Best regards, Michael.

Re: copy-directory

[Eli Zaretskii]

> From: Michael Albinus <michael.albinus@gmx.de>
> Date: Tue, 06 Oct 2009 19:26:36 +0200
> 
> What I'm curious: does change ownership work for other systems, like
> Win32?

Yes, it does, if the process who calls the appropriate APIs has the
required privileges.

Re: copy-directory

[Michael Albinus]

Eli Zaretskii <eliz@gnu.org> writes:

> Please don't rely on the `chown' command being available for the
> preserve-uid-gid option.  It is only available on Posix platforms.
> Instead, please add a primitive to do this from within Emacs.

I know that. But I do not know how important it is on W32.

> And one more comment about the code:
>
>       (mapc
>        (lambda (file)
> 	 (if (file-directory-p file)
> 	     (copy-directory file newname keep-time preserve-uid-gid parents)
> 	   (copy-file file newname t keep-time preserve-uid-gid)))
>        ;; We do not want to delete "." and "..".
>        (directory-files
> 	directory 'full "^\\([^.]\\|\\.\\([^.]\\|\\..\\)\\).*"))
>
> Instead of this complicated regexp, isn't it easier to just filter out
> "." and ".." by adding a suitable comparison to the function you map
> of the files?  I find it not easy to convince myself that the regexp
> indeed matches every valid file name but those two.

I've stolen it from `dired-re-no-dot'. One option would be to add a
similar regexp to files.el, given that I have used it also in
`delete-directory'? (And I will need it also in Tramp)

Best regards, Michael.

Re: copy-directory

[Eli Zaretskii]

> From: Michael Albinus <michael.albinus@gmx.de>
> Cc: emacs-devel@gnu.org
> Date: Sun, 04 Oct 2009 21:58:51 +0200
> 
> Eli Zaretskii <eliz@gnu.org> writes:
> 
> > Please don't rely on the `chown' command being available for the
> > preserve-uid-gid option.  It is only available on Posix platforms.
> > Instead, please add a primitive to do this from within Emacs.
> 
> I know that. But I do not know how important it is on W32.

NTFS, the native filesystem of Windows 2K/XP/Vista supports file
ownership, so it's at least as important as on Posix.