Martin Svec
2018-07-18 14:57:50 UTC
Hello,
after upgrade from Cyrus 2.5.7 to 3.0.6, we noticed that some users lost their IMAP subscriptions
and all sieve filters. After some investigation, we've found that this happens if the user's name
contains dot(s). For sieve directories and .sub files, version 2.5.x uses names with dots replaced
with "^". But 3.0.x series expects the names with dots.
Note that we've altnamespace and unixhierarchysep turned _on_ in both versions, so no namespace
transformations should be needed.
Names in 2.5.x:
/var/imap/sieve/domain/s/somedomain.com/n/name^with^dots/defaultbc
/var/imap/domain/s/somedomain.com/user/n/name^with^dots.sub
Names in 3.0.x:
/var/imap/sieve/domain/s/somedomain.com/n/name.with.dots/defaultbc
/var/imap/domain/s/somedomain.com/user/n/name.with.dots.sub
Surprisingly, other metadata files like .lock files and quota files still use the format with "^"
characters.
I've found no information regarding this incompatible change in 3.0 upgrade documentation. Is this
an expected behavior or a bug?
Best regards
Martin Svec
after upgrade from Cyrus 2.5.7 to 3.0.6, we noticed that some users lost their IMAP subscriptions
and all sieve filters. After some investigation, we've found that this happens if the user's name
contains dot(s). For sieve directories and .sub files, version 2.5.x uses names with dots replaced
with "^". But 3.0.x series expects the names with dots.
Note that we've altnamespace and unixhierarchysep turned _on_ in both versions, so no namespace
transformations should be needed.
Names in 2.5.x:
/var/imap/sieve/domain/s/somedomain.com/n/name^with^dots/defaultbc
/var/imap/domain/s/somedomain.com/user/n/name^with^dots.sub
Names in 3.0.x:
/var/imap/sieve/domain/s/somedomain.com/n/name.with.dots/defaultbc
/var/imap/domain/s/somedomain.com/user/n/name.with.dots.sub
Surprisingly, other metadata files like .lock files and quota files still use the format with "^"
characters.
I've found no information regarding this incompatible change in 3.0 upgrade documentation. Is this
an expected behavior or a bug?
Best regards
Martin Svec