In Cyrus 2.0.9 under Solaris 7, running reconstruct on a mailbox that was
created under 1.64 and not yet accessed under v.2 produces a syslog message
Jan 5 16:19:43 upsilon reconstruct[5695]: mailbox 'user.peanut' has old
cyrus.header
Jan 5 16:19:43 upsilon reconstruct[5695]: IOERROR: locking cache for
user.peanut: Bad file number
Jan 5 16:19:43 upsilon last message repeated 1 time
Truss shows that fcntl is being called (twice) with an fd of -1, e.g.
5695: fcntl(-1, F_SETLK, 0xFFBE81B8) Err#9 EBADF
This doesn't happen on mailboxes created under v.2, or that have been
accessed under v.2 (and therefore whose cyrus.header file has already been
converted to the new format)
Also when the problem occurs I notice that a new cyrus.header is produced,
but there is
no mailbox name at the beginning of the line with the long id-like string,
e.g.
Cyrus mailbox header
"The best thing about this system was that it had lots of goals."
--Jim Morris on Andrew
501c5e4d3a4e4a63
peanut lrswipcda
I've noticed that "normal" cyrus.header files are like
Cyrus mailbox header
"The best thing about this system was that it had lots of goals."
--Jim Morris on Andrew
user.johnh 762437913923e7b1
johnh lrswipcda
What is this data, and is the difference significant?
Thanks, John
John Holman
Head of Central Servers Team
Queen Mary and Westfield College
London, UK
