rsync - Sep 2011 - rsync 3.0.x breaks behaviour of --one-file-system --relative --delete

Hi,

I found that current rsync versions do something different when using
--one-file-system --relative and --delete. This has all worked as expected
with the old version (in our case those delivered with RHEL4 and 5) but it
doesn't work correct in the 3.0.x version.
I have opened a bug at RedHat because it's at least a regression for their
upgrade from EL5.6 to EL5.7 which updated to rsync 3.0.6.

Here the details from my BZ entry:

Steps to Reproduce:
1. on remotehost: mount /var and /var/opt on separate filesystems
2. on remotehost: cp -a /etc /var/opt/XXX
3. on target host: rsync -aHv --delete --numeric-ids --relative
--delete-excluded remotehost:/var /home/snapshots/TEST
4. on target host, repeat but with -x: rsync -aHvx --delete --numeric-ids
--relative --delete-excluded remotehost:/var /home/snapshots/TEST

Actual results:
/home/snapshots/TEST/var/opt/XXX/ is still there with its content

Expected results:
/home/snapshots/TEST/var/opt/XXX/ should have been removed

Additional info:
This has worked with rsync from EL5.6. With -xx it works but it also
removes the mount point itself (which is expected with -xx) but -x should
include the the mount point as empty directory, and therefore clean up on
the target side.
I have tested with latest rsync from upstream (3.0.8) and it also has the
bug/feature.

https://bugzilla.redhat.com/show_bug.cgi?id=735981

Maybe this is all intended and I'm just confused by the lack of sleep?

Note: I'm not subscribed, thanks for CC'ing me.

Thanks,
Simon

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

https://bugzilla.samba.org/show_bug.cgi?id=6240

On 09/06/11 12:42, Simon Matter wrote:
> Hi,
> 
> I found that current rsync versions do something different when using
> --one-file-system --relative and --delete. This has all worked as expected
> with the old version (in our case those delivered with RHEL4 and 5) but it
> doesn't work correct in the 3.0.x version.
> I have opened a bug at RedHat because it's at least a regression for
their
> upgrade from EL5.6 to EL5.7 which updated to rsync 3.0.6.
> 
> Here the details from my BZ entry:
> 
> Steps to Reproduce:
> 1. on remotehost: mount /var and /var/opt on separate filesystems
> 2. on remotehost: cp -a /etc /var/opt/XXX
> 3. on target host: rsync -aHv --delete --numeric-ids --relative
> --delete-excluded remotehost:/var /home/snapshots/TEST
> 4. on target host, repeat but with -x: rsync -aHvx --delete --numeric-ids
> --relative --delete-excluded remotehost:/var /home/snapshots/TEST
> 
> Actual results:
> /home/snapshots/TEST/var/opt/XXX/ is still there with its content
> 
> Expected results:
> /home/snapshots/TEST/var/opt/XXX/ should have been removed
> 
> Additional info:
> This has worked with rsync from EL5.6. With -xx it works but it also
> removes the mount point itself (which is expected with -xx) but -x should
> include the the mount point as empty directory, and therefore clean up on
> the target side.
> I have tested with latest rsync from upstream (3.0.8) and it also has the
> bug/feature.
> 
> https://bugzilla.redhat.com/show_bug.cgi?id=735981
> 
> Maybe this is all intended and I'm just confused by the lack of sleep?
> 
> Note: I'm not subscribed, thanks for CC'ing me.
> 
> Thanks,
> Simon
> 
- -- 
~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~
	Kevin Korb			Phone:    (407) 252-6853
	Systems Administrator		Internet:
	FutureQuest, Inc.		Kevin at FutureQuest.net  (work)
	Orlando, Florida		kmk at sanitarium.net (personal)
	Web page:			http://www.sanitarium.net/
	PGP public key available on web site.
~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~'`^`'~*-,._.,-*~
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iEYEARECAAYFAk5mTpkACgkQVKC1jlbQAQdMfQCbBBlAC7HpsdjVgZ2OguUDpsf5
wLYAoNpaUcznu6i1le6Ip81AOQI4wPdg
=7jCd
-----END PGP SIGNATURE-----

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> https://bugzilla.samba.org/show_bug.cgi?id=6240
Oops, I have checked bugzilla.samba.org but _before_ I realized it has
something to do with --one-file-system.

Sorry for the noise - of course the RedHat BZ still applies to them :)

Thanks,
Simon

>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> https://bugzilla.samba.org/show_bug.cgi?id=6240
>
> Oops, I have checked bugzilla.samba.org but _before_ I realized it has
> something to do with --one-file-system.
>
> Sorry for the noise - of course the RedHat BZ still applies to them :)
Well, now I just looked a bit closer, did you say the bug should be fixed?
As I understand the BZ you mention should be fixed in 3.0.6, which is what
RHEL5.7 ship. And as I mentioned, I have built 3.0.8 from source and it
has the same bug for me, and 3.0.8 is the latest stable.

Anyone cares to check this out on his on box?

Thanks,
Simon