http://dovecot.org/releases/dovecot-1.0.rc22.tar.gz
http://dovecot.org/releases/dovecot-1.0.rc22.tar.gz.sig
Found another bad bug in rc19 changes. Wonder why my imaptest catched
the bug only in CVS HEAD but not in branch_1_0 even though both had it.
Anyway, now the imaptest runs nicely for both, and I'm again optimistic
that the bug count is low enough for v1.0 to be released soon
BTW. My previous benchmarks which showed how mmap_disable=yes and
fsync_disable=yes give huge performance benefits are pretty
XFS-specific. I did some quick tests with ext3 and the differences were
quite small with it. I'll probably do another proper run of tests later.
+ pop3: Commit the transaction even if client didn't QUIT so cached
data gets saved.
- Fixed another indexing bug in rc19 and later which caused
transactions to be skipped in some situations, causing all kinds of
problems.
- mail_log_max_lines_per_sec was a bit broken and caused crashes with
dovecot -a
- BSD filesystem quota was counted wrong. Patch by Manuel Bouyer
- LIST: If namespace has a prefix and inbox=no, don't list
prefix.inbox if it happens to exist when listing for %.
http://dovecot.org/releases/dovecot-1.0.rc22.tar.gz.sig
Found another bad bug in rc19 changes. Wonder why my imaptest catched
the bug only in CVS HEAD but not in branch_1_0 even though both had it.
Anyway, now the imaptest runs nicely for both, and I'm again optimistic
that the bug count is low enough for v1.0 to be released soon
BTW. My previous benchmarks which showed how mmap_disable=yes and
fsync_disable=yes give huge performance benefits are pretty
XFS-specific. I did some quick tests with ext3 and the differences were
quite small with it. I'll probably do another proper run of tests later.
+ pop3: Commit the transaction even if client didn't QUIT so cached
data gets saved.
- Fixed another indexing bug in rc19 and later which caused
transactions to be skipped in some situations, causing all kinds of
problems.
- mail_log_max_lines_per_sec was a bit broken and caused crashes with
dovecot -a
- BSD filesystem quota was counted wrong. Patch by Manuel Bouyer
- LIST: If namespace has a prefix and inbox=no, don't list
prefix.inbox if it happens to exist when listing for %.
