On 6.7.2014, at 12.07, Toni Mattila toni@solu.fi wrote:
Getting following panics with traces: Jul 6 08:11:52 dovecot: imap(user@domain.tld): Error: Cached message size smaller than expected (1782 < 1872) Jul 6 08:11:52 dovecot: imap(user@domain.tld): Error: Maildir filename has wrong S value, renamed the file from /path/cencored/Maildir/cur/1201713258.3136.host.cencored.fi,S=1782:2,S to /path/cencored/Maildir/cur/1201713258.3136.host.cencored.fi,S=1872:2,S Jul 6 08:11:52 dovecot: imap(user@domain.tld): Error: Corrupted index cache file /path/cencored/Maildir/dovecot.index.cache: Broken physical size for mail UID 5758 Jul 6 08:11:52 dovecot: imap(user@domain.tld): Error: read(/path/cencored/Maildir/cur/1201713258.3136.host.cencored.fi,S=1782:2,S) failed: Invalid argument Jul 6 08:11:52 dovecot: imap(user@domain.tld): Error: read(/path/cencored/Maildir/cur/1201713258.3136.host.cencored.fi,S=1782:2,S) failed: Invalid argument (uid=5758)
This is intended.
Jul 6 08:11:52 dovecot: imap(user@domain.tld): Panic: file istream.c: line 167 (i_stream_read): assertion failed: (old_size <= _stream->pos - _stream->skip)
This isn't intended, but I can't seem to be able to reproduce the crash. What's your doveconf -n output?
These can of course be fixed with maildir-size-fix.pl and deleting dovecot index files, but there are thousands like these. Dovecot 1.2 didn't have any issues with these old maildirs transfered over from courierimap.
maildir_broken_filename_sizes=yes restores the old behavior.