Re: Bad cp -a behaviour

[Tim Waugh]

On Wed, Aug 11, 2004 at 02:53:11AM +0100, Philip Rowlands wrote:

> On Tue, 10 Aug 2004, Tim Waugh wrote:
> 
> >How odd.  Is that a patched 5.2.1?
> 
> Possibly; it's from an unadulterated Fedora Core 2.

Yes, that's patched. :-)  But I also see the problem in Fedora Core 2.

> I tried building from source, but I get the same as before (i.e. no
> problem, no output from diff). Bizarre. Let's keep digging:
> 
> $ src/cp -afv 2/d 1
> `2/d/4' -> `1/d/4'
> `2/d/2' -> `1/d/2'
> `2/d/1' -> `1/d/1'
> `2/d/3' -> `1/d/3'
> $ ls -U1 2/d
> 4
> 2
> 1
> 3
> [ this is running on an ext3 filesystem ]
> 
> The bug might be order-specific; does your test produce the same output
> as the above with "-v"?

Thinking about it, it must be order-specific.  But here's the output I
get:

`2/d/3' -> `1/d/3'
`2/d/2' -> `1/d/2'
`2/d/1' -> `1/d/1'
`2/d/4' -> `1/d/4'

strace output shows that the only unlink is 1/d/2, here:

write(1, "`2/d/2\' -> `1/d/2\'\n", 19)  = 19
link("1/d/3", "1/d/2")                  = -1 EEXIST (File exists)
unlink("1/d/2")                         = 0
link("1/d/3", "1/d/2")                  = 0

Tim.
*/

pgpEOkDQnqXsW.pgp
Description: PGP signature