One more bugfix since rc2:
- Maildir: synchronization might have sometimes set wrong flags to
messages, or crash completely
Quite rare because we were lucky before :)
And list of all the changes one more time:
v0.99.10.5 2003-12-27 Timo Sirainen tss@iki.fi
+ MySQL authentication, patch by Matthew Reimer
+ --with-moduledir configure option
- mbox: APPEND reversed given \Draft and \Deleted flags
- mbox: "LF not found" errors happened sometimes when X-IMAPbase
header was updated. Possibly corrupted mbox sometimes.
Thanks to Fabrice Bellet for finding this bug.
- Custom flags couldn't be unset
- Maildir: make sure ":2," is appended to filename when moving mails
from new/ to cur/.
- Maildir: synchronization might have sometimes set wrong flags to
messages, or crash completely
- Maildir: RENAME xx inbox.xx didn't result as uppercased ".INBOX.xx"
directory which then couldn't be accessed
- Don't crash with RAND_bytes() error messages anymore. This mostly
happened with Fedora/RedHat.
On May 27, 2004, at 11:45 AM, Timo Sirainen wrote:
And list of all the changes one more time:
I'm running 0.99.10.5 on a Mac OS X 10.2.8 system (Darwin 6.8) and I get these errors in the log:
May 27 12:20:44 dluke imap(dluke): mmap_istream.madvise(): Invalid argument
I can edit the generated config.h and undef HAVE_MADVISE, but 10.2 is supposed to have a working madvise(). I took a quick look at the dovecot code, but I don't see anything obviously incorrect.
Any suggestions on what to look for? (I could provide a non-root account on the machine if someone wants access to test.)
Thanks.
Daniel J. Luke +========================================================+ | *---------------- dluke@geeklair.net ----------------* | | *-------------- http://www.geeklair.net -------------* | +========================================================+ | Opinions expressed are mine and do not necessarily | | reflect the opinions of my employer. | +========================================================+
On 27.5.2004, at 19:24, Daniel J. Luke wrote:
I'm running 0.99.10.5 on a Mac OS X 10.2.8 system (Darwin 6.8) and I get these errors in the log:
May 27 12:20:44 dluke imap(dluke): mmap_istream.madvise(): Invalid argument
Sorry, I forgot this mail and found it only now again when checking for old "unseen" mails.
Looks like it's OS X problem here, others have had the same problem and it's still not working in 10.3.4. I don't think I'll bother writing any workarounds for this, at least yet.. You can just set mail_read_mmaped = no (which is the default). I'm not sure if there's much point in reading mails mmaped anyway.
Are you pushing these downstream to the Fedora group? And are you on the package CC list for the Fedora Bugzilla?
This hit the SpamAssassin Developer list yesterday:
------------ Forwarded Message ------------ Date: Thursday, May 20, 2004 5:06 PM -1000 From: Warren Togami wtogami@redhat.com To: spamassassin-dev@incubator.apache.org Subject: SA Devel for Fedora Core
Any spamassassin developers interested in being automatically added to all new spamassassin bugs that are reported for Fedora Core at bugzilla.redhat.com? Developers from other upstream projects like gaim have become involved with FC in this way. It has proven to be a win for both the upstream and downstream projects, as the developers often KNOW the majority of problems instantly, alert us to existing patches for the package, or even simply close stupid user reports. Otherwise they have the option of removing themselves from the CC if they simply are not interested. [1]
Any volunteers, developers or even power-users, can help as liasons to the Fedora Project. Your software is important to our distribution, so fostering communication with upstream development and making sure patches go both ways is important.
https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=123710 This is a problem tracker for spamassassin 3.0's snapshots in FC rawhide that should be solved before the release of FC3. FC3 development will expose spamassassin 3.0 snapshots and later pre-releases to thousands of additional testers, with the goal of improving the quality of both SA 3.0 and FC3.
Warren Togami wtogami@redhat.com
[1] Three upstream gaim developers are automatically added to all FC gaim bugs. As the result of the gaim partnership, several bugs were quickly killed in upstream CVS, and the FC2 package has been polished, exposing CVS fixes to wide user testing before the next upstream release. (This has even been instrumental in preventing one regression from reaching gaim-0.78, after Gentoo stole a Fedora CVS backport and they noticed an extremely subtle breakage.) As the result of this successful partnership, there are current ZERO open FC gaim bugs.
---------- End Forwarded Message ----------
On Thu, 27 May 2004, Timo Sirainen wrote:
v0.99.10.5 2003-12-27 Timo Sirainen tss@iki.fi
- MySQL authentication, patch by Matthew Reimer
- --with-moduledir configure option
- mbox: APPEND reversed given \Draft and \Deleted flags
- mbox: "LF not found" errors happened sometimes when X-IMAPbase header was updated. Possibly corrupted mbox sometimes. Thanks to Fabrice Bellet for finding this bug.
- Custom flags couldn't be unset
- Maildir: make sure ":2," is appended to filename when moving mails from new/ to cur/.
- Maildir: synchronization might have sometimes set wrong flags to messages, or crash completely
- Maildir: RENAME xx inbox.xx didn't result as uppercased ".INBOX.xx" directory which then couldn't be accessed
- Don't crash with RAND_bytes() error messages anymore. This mostly happened with Fedora/RedHat.
I just uploaded Debian packages to unstable. I've enabled mysql support and openssl (Timo: I still owe you the gnutls10 patch but I've been insanely busy recently.)
I intend to provide a woody backport but as I'm going to need to use some other backported packages, I shall probably upload it to http://www.backports.org/ rather than http://src.braincells.com/ I'll keep this list and the wiki updated.
Timo: does this release fix any of the outstanding Debian bugs? Especially #225408?
And If any Debian dovecot users would like to go through the bug list and see if there are any that can be closed, I would appreciate it. The URL is http://bugs.debian.org/cgi-bin/pkgreport.cgi?src=dovecot
-- Jaldhar H. Vyas jaldhar@debian.org La Salle Debain - http://www.braincells.com/debian/
On Thu, 2004-05-27 at 21:09, Jaldhar H. Vyas wrote:
Timo: does this release fix any of the outstanding Debian bugs? Especially #225408?
#225048 I think? I have very high hopes that it's fixed now.
And If any Debian dovecot users would like to go through the bug list and see if there are any that can be closed, I would appreciate it. The URL is http://bugs.debian.org/cgi-bin/pkgreport.cgi?src=dovecot
#221972 should be fixed too.
On Thu, 27 May 2004, Timo Sirainen wrote:
On Thu, 2004-05-27 at 21:09, Jaldhar H. Vyas wrote:
Timo: does this release fix any of the outstanding Debian bugs? Especially #225408?
#225048 I think?
Aargh I keep doing that today.
I have very high hopes that it's fixed now.
I'll close it for now. Someone will be sure to squawk loudly if it is still happening.
And If any Debian dovecot users would like to go through the bug list and see if there are any that can be closed, I would appreciate it. The URL is http://bugs.debian.org/cgi-bin/pkgreport.cgi?src=dovecot
#221972 should be fixed too.
Excellent. Thankyou very much for this wonderful program.
-- Jaldhar H. Vyas jaldhar@debian.org La Salle Debain - http://www.braincells.com/debian/
Hello,
thanks for this latest (and last ;) 0.99 version. Just to make absolutely sure. Timo, this is a drop in replacement in relation to 0.99.10.4, read no UID generation changes, so users will never know the difference?
Jaldhar, any realistic estimate when this will filter down into sarge?
Regards and thanks,
Christian Balzer
Christian Balzer Network/Systems Engineer NOC chibi@gol.com Global OnLine Japan/Fusion Network Services http://www.gol.com/
On Fri, 2004-05-28 at 01:59, Christian Balzer wrote:
thanks for this latest (and last ;) 0.99 version. Just to make absolutely sure. Timo, this is a drop in replacement in relation to 0.99.10.4, read no UID generation changes, so users will never know the difference?
Right. I doubt I'll ever change the UID generation defaults.
On Fri, 28 May 2004, Christian Balzer wrote:
Jaldhar, any realistic estimate when this will filter down into sarge?
Barring any show-stopping bugs, 10 days.
-- Jaldhar H. Vyas jaldhar@debian.org La Salle Debain - http://www.braincells.com/debian/
Hi Timo and list,
Does this release work safely over NFS with maildir? I saw an email about rc2 for nfs + mbox, which seemed to work (apart from some small dotlock warnings)?
Will it work with indexes on disk too?
Thanks,
Kind regards,
Maikel Verheijen
On Thu, 2004-05-27 at 22:15, Maikel Verheijen wrote:
Does this release work safely over NFS with maildir? I saw an email about rc2 for nfs + mbox, which seemed to work (apart from some small dotlock warnings)?
Will it work with indexes on disk too?
0.99.10.5 is only 0.99.10.4 with a few important fixes. It's no more NFS safe then .4.
Hmm. Looks like with in-memory indexes with mbox doesn't support custom flags at all. That could be fixed without too much trouble since mbox stores the flags in the mbox file itself.. Subscription file handling is also still using fcntl locking in .5.
Hi Timo and list,
On May 27, 2004, at 9:46 PM, Timo Sirainen wrote:
On Thu, 2004-05-27 at 22:15, Maikel Verheijen wrote:
Does this release work safely over NFS with maildir? I saw an email about rc2 for nfs + mbox, which seemed to work (apart from some small dotlock warnings)?
Will it work with indexes on disk too?
0.99.10.5 is only 0.99.10.4 with a few important fixes. It's no more NFS safe then .4. Ok. We will have to wait for 1.0 then :)
Kind regards,
Maikel Verheijen
participants (6)
-
Christian Balzer
-
Daniel J. Luke
-
Jaldhar H. Vyas
-
Kenneth Porter
-
Maikel Verheijen
-
Timo Sirainen