Sender: rich AT phekda DOT freeserve DOT co DOT uk
Message-ID: <3F2CEA9D.45657D21@phekda.freeserve.co.uk>
Date: Sun, 03 Aug 2003 11:57:33 +0100
From: Richard Dawe <rich AT phekda DOT freeserve DOT co DOT uk>
X-Mailer: Mozilla 4.77 [en] (X11; U; Linux 2.2.23 i586)
X-Accept-Language: de,fr
MIME-Version: 1.0
To: djgpp-workers AT delorie DOT com
Subject: Re: mv bug (filutils? library?)
References: <000001c35961$76c1d4a0$0101a8c0 AT acp42g>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Reply-To: djgpp-workers AT delorie DOT com

Hello.

Andrew Cottrell wrote:
> 
> FYI. WIP debugging update below.
> 
> I have traced the problem so far to the rename() in move.c which is in
> LIBC!!!! The problem line is:-
> 
>       if (rename (src_path, dst_path)== 0)
>           if (x->verbose && S_ISDIR (src_type))
>             printf ("%s -> %s\n", quote_n (0, src_path), quote_n (1,
> dst_path));
[snip]

Thanks for the work!

fileutils 4.1 from Simtel.NET (built with 2.03): Has bug.
fileutils 4.1 from 2.04 testing page: Has bug.
coreutils 5.0 built by Andrew: Doesn't have bug.

I agree with Andrew: it's a library bug. Here's why:

Using something similar to Charles's test case, is_parent in
src/libc/ansi/stdio/rename.c seems to be called with one argument with an LFN
and one argument with a SFN. So the check to see whether the source is the
parent of the destination fails.

For some reason _truename returns an LFN when called on "junk" (real_old) and
an SFN when called on "junk/junk" (real_new) or "junk\\junk":

(gdb) p _truename(real_old, 0)
$15 = 0xad0a8 "C:\\thisisalongname\\junk"
(gdb) p _truename(real_new, 0)
$16 = 0xad0c8 "C:\\THISIS~1\\JUNK\\junk"

My current idea is that we could use _lfn_gen_short_fname to collapse
"thisisalongname" to thisis~1. I thought about patching _truename, but that
seems unsatisfactory. Any other ideas?

_truename is used in a couple of places where we compare filenames (is_parent,
_rename). Perhaps we should use some kind of _truename wrapper function there,
that does the LFN collapsing using _lfn_get_short_fname?

Items to do:

1a. is_parent in src/libc/ansi/stdio/rename.c.
1b. _truename wrapper for rename & _rename?
2. Add a warning to _truename docs.
3. "What's changed in 2.04" entry.

Thanks, bye, Rich =]

-- 
Richard Dawe [ http://www.phekda.freeserve.co.uk/richdawe/ ]