Storing ownership / device nodes without root

Wayne Davison wayned at
Sat Oct 28 21:20:11 GMT 2006

On Sat, Oct 28, 2006 at 05:15:06PM +0200, Wesley W.Terpstra wrote:
> I would send patches, but I find it quite difficult to get patches  
> against patches.

I actually don't like patches against patches.  Your prior posting had a
patch against some pre-patched source, and that's the clearest way to
provide changes in my opinion, as it shows what was changed relative to
the last version.

> For that matter, how are you keeping such a large patchset consistent  
> when you modify cvs??

The nice thing about patches is that they tend to keep working as CVS
changes, so most of the patches don't need to be modified.  I check on
them using a custom perl script in the patches dir named verify-patches.

> Do you see anything else?

I have made a few more fixes, such as adding symlink handling, improving
error handling, making the code work right when --fake-super was not
specified, and fixing some compile problems when xattr support is

I also made the default owner & group for a symlink be 0:0 so that if a
system can't store any attributes in a symlink, the symlink will come
back being owned by root (which only makes a difference on an OS that
allows the owner of a symlink to be set -- linux systems would still get
the symlink created by the user running the rsync command when copying
the files back).

So, things seem to be in pretty good shape at the moment.


More information about the rsync mailing list