(important) cyrus-imapd 2.4.16 released

Bron Gondwana brong at fastmail.fm
Thu Apr 26 12:19:16 EDT 2012

On Thu, Apr 26, 2012 at 04:12:38PM +0200, Sebastian Hagedorn wrote:
> Bron,
> sorry, somehow I forgot about this mail!
> --On 19. April 2012 20:04:04 +0200 Bron Gondwana <brong at fastmail.fm> wrote:
> >>I think I've read all the available information regarding the issue,
> >>but I'm still confused. We're still running 2.3.x for production,
> >>but I have a test system with a copy of the production data. We use
> >>fulldirhash. When I installed 2.4.14, I encountered the hashing
> >>issue. So now I ran the rehash script after updating to 2.4.16:
> >>
> >>$ time /usr/lib/cyrus-imapd/rehash -v -F /etc/imapd.conf
> >>you are using /var/lib/imap/sieve as your sieve directory.
> >>i will also hash partitions.
> >>converting configuration directory /var/lib/imap... mkdir
> >>/var/lib/imap/lock: done
> >>user quota done
> >>sieve /var/lib/imap/sieve... rename /var/lib/imap/sieve/global to
> >>/var/lib/imap/sieve/V/global: done
> >>done
> >>partition /var/spool/imap... done
> >>partition /var/spool/imap2... done
> >>partition /var/spool/imap3... done
> >>partition /var/spool/imap4... done
> >>partition /var/spool/imap5... done
> >>partition /var/spool/imap6... done
> >>
> >>real	0m7.232s
> >>user	0m5.639s
> >>sys	0m0.227s
> >>
> >>Just seven seconds? And I don't see much actual rehashing, except
> >>for the global sieve directory. What am I missing?
> >
> >Are your directories already hashed correctly?  It doesn't move anything
> >it doesn't have to!
> Well, that's the question, I suppose. I'm thinking that I may have
> misunderstood the issue. The production server running 2.3.14 is a
> 32-bit RHEL 3 system. The VM I'm using for 2.4 is a 64-bit RHEL 5
> system. Its Perl binary uses 32-bit ints. I thought previously that
> I would have to rehash the 2.3 spool for that to work, but now I
> start to think that I don't, because the hashing algorithm hasn't
> actually changed, and the Perl part has been fixed. The only problem
> would have been trying to use the broken Perl code in 2.4.14. Is
> that correct?

Yeah, could be.  The hashing algorithm in 2.4.14 was changed to 64
bit too I think.  Easiest is to test that mbpath from 2.4.16 returns
the same paths as your files on disk!


More information about the Info-cyrus mailing list