Replication is broken with modseq issue in 2.3.6

[Bron Gondwana]

On Thu, 06 Jul 2006 17:17:39 -0400, "David S. Madole" <david at madole.net> said:
> On 05 Jul 2006, at 02:54, Bron Gondwana wrote:
> 
> >If you're running replication on any post 2.3.3 cyrus, your replica
> >contains indexes with '0' as the modseq value.  This means that your
> >messages will not be fetchable until you either fix that value or
> >patch Cyrus.  Reconstruct from 2.3.6+ is one option for fixing it
> 
> Interesting, I discovered the same problem right around the same time
> you did.
> 
> I submitted a patch that does actually replicate the modseq value,
> although unfortunately my first effort was done a little too quickly
> and I neglected conversion to/from network byte order as used in the
> index files, meaning that the values are not replicated correctly on
> x86 machines.
> 
> Ken did fix my patch, however, and put it into CVS so I expect this
> will be working in the next release. I also fixed my own patch with
> the needed ntonl and ntohl etc and re-uploaded into Bugzilla for 
> anyone who wants it before then:
> 
> https://bugzilla.andrew.cmu.edu/show_bug.cgi?id=2859
> 
> Make sure you get the fixed version.

Hi David,

We don't actually care about the modseq value because we're not
using it - we just care about being able to fetch.  That said,
we're running both my patches now, so modseq will be set to 1 in 
the replica from now on.

One downside of your patch - it changes the network protocol of
replication so that 2.3.6 and 2.3.7 won't be able to replicate
to each other in either direction - you'll have to upgrade both
your machines at once.

Ken has already committed your patch to CVS (I read the cyrus-cvs
commits mailing list now so I can keep track of changes), so I could
pull it from there as well.

Thanks,

Bron.
-- 
  Bron Gondwana
  brong at fastmail.fm