bug#16133: 24.3; copy-file fails on chmod when copying to FAT filesystem

[Artem Chuprina <ran@lasgalen.net>]

Glenn Morris -> Artem Chuprina  @ Sat, 14 Dec 2013 15:19:33 -0500:

 GM> I think you are right that it is a problem.
 GM> I think the problem was in documenting copy-file to always copy
 GM> permissions in the first place. But I imagine it is far too late to
 GM> change that. Not sure what the right thing to do is here...

 GM> Someone may be relying on the fact that copy-file copies permissions,
 GM> and want it it throw an error if it fails:

 GM> http://debbugs.gnu.org/cgi/bugreport.cgi?bug=8306

 GM> Others, like you, may not care.

To clarify, many others, including me, CARE of copy-file NOT throwing
error when it can copy file, but not its permissions, as it did for many
years before.  That is, the mentioned patch broke that for-many-years
successful behavior, and me personally had to return back to emacs 23 to
avoid it.

Also, it seems that backup facility also depends on copy-file (probably
not, I've not done a research, but I've seen message "cannot make a
backup file, backing up to ~/.emacs.d/%backup" (or like this) when
editing file on the same VFAT filesystem.

 GM> Since copy-file has only relatively recently starting giving an error,
 GM> maybe the least bad thing would be for it to give a message instead of
 GM> an error?

As for me, this would be OK, because I cannot imagine any scenario when
I'd prefer copied file to be left at target, but copy-file throwing an
error.  I want either success if file contents was copied successfully
(and possibly a message informing me that chmod or chown failed), or, if
I REALLY CARE OF EXACT COPY, that file itself is NOT copied at all (that
is, reverted if ok-if-exist was true).

Yes, I think that chown or copying ACLs also should not throw an error
on failure.

 GM> I don't know how we tidy this up for the future though.
 GM> Add yet another optional argument "error-if-permissions-fail"? Yuck.

I think no one will end up with that argument set true.  Probably,
someone will try to, but his users will soon ask him not to do so.

Another possible solution would be to introduce a global variable, say,
copy-file-signal-error-on-content-only :), allowing end-user (that is,
not a library author who calls copy-file, like org-mobile-push in my
case, but end user, like me) to disable throwing errors if file content
copied ok, but meta information copying failed (not only chmod, but also
chown, copying ACLs, times etc.)  I expect that most users would like
that variable to be true by default :)