handling duplicate names deterministically and adding alternative checksum algorithms

Andrew Shewmaker agshew at gmail.com
Thu Jun 16 20:01:54 GMT 2005

Thanks for the pointer to that mergesort thread and for the md5 patch.

Is rsync going to stick with qsort as the default sorting algorithm?
Would a patch to add a --mergesort option be accepted?

Even though most of my boxes are Linux and it appears that qsort usually 
runs as a mergesort, I would still like an option to enforce it so I don't get 
suprised one day when an actual qsort is used due to memory contention.

And what would be the chances of a --sha1 option patch being accepted?

Thank you again,


On 6/13/05, Wayne Davison <wayned at samba.org> wrote:
> On Thu, Jun 09, 2005 at 05:24:45PM -0600, Andrew Shewmaker wrote:
> > Our current practices would benefit if rsync were enhanced to handle
> > duplicate names deterministically as described in the todo list.
> One proposed solution to this was to use a functioned called
> mergesort().  A patch was provided that includes a compatibility
> function for systems that don't include mergesort():
>     http://lists.samba.org/archive/rsync/2004-August/010398.html
> > we would also be interested in the ability to optionally use a
> > stronger checksum algorithm.
> I just put a diff into the patches dir that implements the --md5 option
> (extracted and cleaned up from a larger patch set that was sent to me by
> Wolfgang Neuman):
>     http://rsync.samba.org/ftp/unpacked/rsync/patches/md5.diff
> As you'd expect, this tells rsync to use MD5 checksums instead of MD4.
> ..wayne..

Andrew Shewmaker

More information about the rsync mailing list