April 2003 - dovecot - dovecot.org

[dovecot] 0.99.9 released
by Timo Sirainen 05 May '03

05 May '03

As promised yesterday :) BTW. I just found libsqlora8 library yesterday and I've been playing with Oracle mail storage support. Shouldn't take too long to get it working. And the NEWS file once more: v0.99.9 2003-04-28 Timo Sirainen <tss(a)iki.fi> * WARNING: mbox rewriting works now faster, but it's less likely to notice external message flag changes (it wasn't perfect before either). This also means that if you're not using index files, Dovecot may not notice changes made by another Dovecot process. This will be fixed later. + Message UIDs are now permanently stored outside index files. Deleting indexes is now possible without any kind of data loss and in-memory indexes are now usable. + mbox rewriting leaves extra space into X-Keywords header. It's shrinked or grown when updating message flag headers to avoid rewriting the rest of the file. + imap-login and pop3-login can now be started from inetd. Privilege separation is still used, so it executes dovecot and dovecot-auth processes which are left on the background. + PostgreSQL authentication support, patch by Alex Howansky - Large multiline headers (over 8kB) broke Dovecot before. Now they're parsed one line at a time and we try to avoid reading them fully into memory. - SEARCH OR was broken - Partial BODY[] fetches were broken - Timezones were still set wrong when parsing dates - Using non-synchronized literals (LITERAL+) broke APPEND - Renaming maildir subfolders inserted extra "." in the middle. - Subfolders were a bit broken with maildir - Invalid PLAIN auth request crashed auth process.

2 3

[dovecot] Corrupted index
by Mark E. Mallett 03 May '03

03 May '03

Hi- Still getting a bunch of these: Apr 30 13:45:23 mercury mem[11056]: pop3(user1): Corrupted index file /users/78/user1/Maildir/.INBOX/.imap.index: used_file_size larger than real file size (4032 vs 1608) and the occasional: Apr 30 13:08:53 mercury mem[13507]: pop3(user2): Corrupted index file (in-memory index for /users/e0/user2/Maildir): Filename mismatch for UID 1: 1051720748.28556.mercury.mv.net vs 1051715695.29601.mercury.mv.net Most of the time (but not always) the .imap* files are relatively old and apparently created by the test5 or older recent release. The errors persist in the logs until I manually remove the .imap* files, which are subsequently recreated upon the next access (without the error being logged, at least not right away). I had the impression that a "Corrupted index file" would cause the index files not to be used, and in fact recreated. Apparently not? If there is an error that causes the .imap* files to be ignored, shouldn't those files be removed? mm

2 1

[dovecot] Message causing pop3 crash
by Mark E. Mallett 03 May '03

03 May '03

Hi- Found a message that will generate a pop3 coredump when using mbox format. It seems to do with the fact that it has a mime part with a line that's 10,000 bytes long. However, that doesn't seem to be the only factor. Adjusting the size of the mail header seems to cause different kinds of crashes. I'll avoid attaching the message here but I've put it at http://www.mv.com/users/mem/x/badmail Reproducing it here on a BSD/OS system involves something like: (login to a virgin account or one you don't care about) % wget http://www.mv.com/users/mem/x/badmail % rm -rf .imap % MAIL=mbox:`pwd`/badmail % limit coredump 10m % /usr/local/libexec/dovecot/pop3 quit at which point you should get some errors and/or a coredump. You may need to do it twice (i.e. let it create the .imap/ index structure first). If you can't make it happen I can send along more info. -mm- (and yes, this was a real mail message, albeit a spam, except that I simplified it some and replaced all of the content with "x" characters)

2 1

[dovecot] Debian packages for 0.99.9
by Jaldhar H. Vyas 30 Apr '03

30 Apr '03

I uploaded the new version to sid today. Apart from the new version a big change is that it has now been split into 4 pieces. dovecot-imapd and dovecot-pop3d are the imap server and the pop3 server rspectively. They are seperate so you can install one without the other. dovecot-common contains the files they both use. dovecot is just a dummy to ensure smooth upgrades for users of older versions. As there are new packages, they will take a while to show up i the archive so I'm temporarily making them available from http://src.braincells.com/debian/sid/dovecot/ or deb http://src.braincells.com/debian sid/ deb-src http://src.braincells.com/debian sid/ via apt-get. but unless you have a burning need to upgrade, I recommend you just wait a few days. A backport to woody is also available, add deb http://src.braincells.com/debian woody/ deb-src http://src.braincells.com/debian woody/ to your sources.list. One problem I found with installing on woody is that the first time you do apt-get install dovecot, it gives you an error because /etc/dovecot.conf is no owned by dovect-common instead of dovecot. It will work correctly the second time you run the command. -- Jaldhar H. Vyas <jaldhar(a)debian.org> La Salle Debain - http://www.braincells.com/debian/

1 0

[dovecot] Dovecot RH RPM packages
by Dag Wieers 29 Apr '03

29 Apr '03

Hi, I can't remember if I already told you, but I'm building RPM packages of dovecot for RH73, RH80 and RH9. You can find these at: http://dag.wieers.com/packages/dovecot/ Or use apt as described on: http://dag.wieers.com/apt/ I also added a sysv file to the SPEC-file for use with Red Hat. If you add this SPEC-file to your tarball (probably with some custom changes) everyone can build an RPM package by issuing: rpmbuild -ta <tarball> Kind regards, -- dag wieers, dag(a)wieers.com, http://dag.wieers.com/ -- �Any errors in spelling, tact or fact are transmission errors�

1 0

[dovecot] current cvs (04.28)
by Farkas Levente 28 Apr '03

28 Apr '03

hi, after the current cvs compile: ----------------------- gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../../src/lib -I../../src/lib-mail -I../../src/lib-imap -I../../src/lib-storage -O2 -g -pipe -march=i386 -mcpu=i686 -Wall -W -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wchar-subscripts -Wformat=2 -Wbad-function-cast -c `test -f 'mail-storage-callbacks.c' || echo './'`mail-storage-callbacks.c source='main.c' object='main.o' libtool=no \ depfile='.deps/main.Po' tmpdepfile='.deps/main.TPo' \ depmode=gcc3 /bin/sh ../../depcomp \ gcc -DHAVE_CONFIG_H -I. -I. -I../.. -I../../src/lib -I../../src/lib-mail -I../../src/lib-imap -I../../src/lib-storage -O2 -g -pipe -march=i386 -mcpu=i686 -Wall -W -Wmissing-prototypes -Wmissing-declarations -Wpointer-arith -Wchar-subscripts -Wformat=2 -Wbad-function-cast -c `test -f 'main.c' || echo './'`main.c make[3]: *** No rule to make target `../lib-index/libstorage_index.a', needed by `pop3'. Stop. ----------------------- -- Levente "Si vis pacem para bellum!"

1 0

[dovecot] 0.99.9 release candidate 1
by Timo Sirainen 28 Apr '03

28 Apr '03

0.99.9 should really be released soon. There's been quite a lot of important fixes since 0.99.8.1. There's still some recent counter strangeness, but it wasn't perfect before either. mbox still doesn't necessarily notice flag changes unless they're written into index file (ie. non-Dovecots and Dovecot without index files). But it wasn't perfect before either. dovecot-uidlist file works perfectly now as far as I know. I just noticed that mbox syncing has been somewhat broken for past month. It sometimes assigned new UIDs to messages without any reason, so it was mostly just slowdown. So unless there's something really broken, I'll release 0.99.9 tomorrow. Please get 0.99.9-rc1 from http://dovecot.procontrol.fi/rc/ and see if you can break it. NEWS file: v0.99.9 2003-04-28 Timo Sirainen <tss(a)iki.fi> * WARNING: mbox rewriting works now faster, but it's less likely to notice external message flag changes (it wasn't perfect before either). This also means that if you're not using index files, Dovecot may not notice changes made by another process. This will be fixed later. + Message UIDs are now permanently stored outside index files. Deleting indexes is now possible without any kind of data loss and in-memory indexes are now usable. + mbox rewriting leaves extra space into X-Keywords header. It's shrinked or grown when updating message flag headers to avoid rewriting the rest of the file. + imap-login and pop3-login can now be started from inetd. Privilege separation is still used, so it executes dovecot and dovecot-auth processes which are left on the background. + PostgreSQL authentication support, patch by Alex Howansky - Large multiline headers (over 8kB) broke Dovecot before. Now they're parsed one line at a time and we try to avoid reading them fully into memory. - SEARCH OR was broken - Partial BODY[] fetches were broken - Timezones were still set wrong when parsing dates - Using non-synchronized literals (LITERAL+) broke APPEND - Renaming maildir subfolders inserted extra "." in the middle. - Subfolders were a bit broken with maildir - Invalid PLAIN auth request crashed auth process.

3 4

[dovecot] shared folders?
by Moe Wibble 27 Apr '03

27 Apr '03

Any plans to implement shared folders anytime soon? greetinx -- MW

5 16

[dovecot] dovecot and grsecurity (problem with resource limits)
by Heiko Zuerker 27 Apr '03

27 Apr '03

Howdy folks ! I just added Dovecot as a standard package to Devil-Linux and ran into a problem with resource limits. Grsecurity (http://www.grsecurity.net) is used in DL to prevent problems with common exploits, it also reports violations of rlimits. The following messages show up in the log, but it seems that the IMAP Server works fine: Apr 26 19:20:04 src@gate imap-login: Login: hz [192.168.0.11] Apr 26 19:20:05 src@gate kernel: grsec: attempted resource overstep by requesting 37 for RLIMIT_NPROC against limit 0 by (dovecot:10246) UID(0) EUID(0), parent (dovecot:634) UID(0) EUID(0) Apr 26 19:22:18 src@gate kernel: grsec: attempted resource overstep by requesting 37 for RLIMIT_NPROC against limit 0 by (dovecot:19802) UID(0) EUID(0), parent (dovecot:634) UID(0) EUID(0) Apr 26 19:22:18 src@gate imap-login: Login: hz [192.168.0.11] Apr 26 19:26:36 src@gate syslog-ng[606]: STATS: dropped 0 Apr 26 19:28:19 src@gate kernel: grsec: attempted resource overstep by requesting 33 for RLIMIT_NPROC against limit 0 by (dovecot:16011) UID(0) EUID(0), parent (dovecot:634) UID(0) EUID(0) How should we proceed to analyze/fix this problem? Thanks! Heiko

2 1

[dovecot] today cvs
by Farkas Levente 25 Apr '03

25 Apr '03

hi, the today cvs has a new bug during configure (both on rh 8.0 and rh 9): ----------------------------------- checking size of long long... 8 checking type of off_t... long long ./configure: line 12009: syntax error near unexpected token `else' ./configure: line 12009: `else' ----------------------------------- -- Levente "Si vis pacem para bellum!"

3 3