So we have been running 0.2.2 just fine for awhile with the old pre-
plugin managesieve code. I have just installed 0.3 and activated the
managesieve plugin however when I go to to the filters tab in the
settings it never comes back and just seems to be stuck loading.
There are no logs generated and a tcpdump shows that the client starts
the sieve session on the server w/ a starttls but then seems to
stall. Is there anyway I can get roundcube to be much more verbose to
what it is doing? I just turned off TLS and it seems to work just fine.
The odd thing is that the previous non-plugin managesieve code did not
have this issue so either their is a regression in it or its
dependencies.
I am using "Cyrus timsieved v2.2.12-Invoca-RPM-2.2.12-10.el4_8.1" for
my sieve server".
Thanks, derek
Derek Yarnell UNIX Systems Administrator University of Maryland Institute for Advanced Computer Studies
List info: http://lists.roundcube.net/dev/
Followup,
I have found the culprit it is the new version of PEAR::Net_Sieve
(1.1.7) which adds this,
1166,1171c1132 < // The server should be sending a CAPABILITY response after < // negotiating TLS. Read it, and ignore if it doesn't. < $this->_doCmd();
Commenting this out allows the cyrus timesieve connection under
starttls to continue. So my guess is that the ignoring part of this
function is not doing the right thing.
-derek
On Sep 19, 2009, at 12:51 PM, Derek Yarnell wrote:
So we have been running 0.2.2 just fine for awhile with the old pre- plugin managesieve code. I have just installed 0.3 and activated
the managesieve plugin however when I go to to the filters tab in
the settings it never comes back and just seems to be stuck
loading. There are no logs generated and a tcpdump shows that the
client starts the sieve session on the server w/ a starttls but then
seems to stall. Is there anyway I can get roundcube to be much more
verbose to what it is doing? I just turned off TLS and it seems to
work just fine.The odd thing is that the previous non-plugin managesieve code did
not have this issue so either their is a regression in it or its
dependencies.I am using "Cyrus timsieved v2.2.12-Invoca-RPM-2.2.12-10.el4_8.1"
for my sieve server".Thanks, derek
Derek Yarnell UNIX Systems Administrator University of Maryland Institute for Advanced Computer Studies
Derek Yarnell UNIX Systems Administrator University of Maryland Institute for Advanced Computer Studies
List info: http://lists.roundcube.net/dev/
Derek Yarnell wrote:
I have found the culprit it is the new version of PEAR::Net_Sieve
(1.1.7) which adds this,1166,1171c1132 < // The server should be sending a CAPABILITY response after < // negotiating TLS. Read it, and ignore if it doesn't. < $this->_doCmd();
Commenting this out allows the cyrus timesieve connection under
starttls to continue. So my guess is that the ignoring part of this
function is not doing the right thing.
I was also bitten by this issue. For the record, it's not PEAR::Net_Sieve that's at fault here, it's Cyrus. Cyrus accidentally removed the CAPABILITY response after this bug: https://bugzilla.andrew.cmu.edu/show_bug.cgi?id=1338 (from 2002!) and added it back in version 2.3.10. PEAR::Net_Sieve happened to work with the broken behaviour until they fixed in this bug: http://pear.php.net/bugs/bug.php?id=14205.
At the very least, both Debian and Ubuntu have Cyrus versions with the broken Sieve behaviour. It might be a good idea to stick a README file in there with instructions on how to alter Net_Sieve to work with the broken Sieve implementations.
Francis
List info: http://lists.roundcube.net/dev/
On Sep 22, 2009, at 7:18 AM, Francis Russell wrote:
I was also bitten by this issue. For the record, it's not PEAR::Net_Sieve that's at fault here, it's Cyrus. Cyrus accidentally removed the CAPABILITY response after this bug: https://bugzilla.andrew.cmu.edu/show_bug.cgi?id=1338 (from 2002!) and added it back in version 2.3.10. PEAR::Net_Sieve happened to work with the broken behaviour until they fixed in this bug: http://pear.php.net/bugs/bug.php?id=14205.
At the very least, both Debian and Ubuntu have Cyrus versions with the broken Sieve behaviour. It might be a good idea to stick a README file in there with instructions on how to alter Net_Sieve to work with the broken Sieve implementations.
Yes I agree, after looking into this a bit more it is Cyrus fault. I
am running 2.2.12 from RHEL4 and I don't know when we will be
migrating forward at this point. But thanks for the corroboration.
Thanks, derek
Derek Yarnell UNIX Systems Administrator University of Maryland Institute for Advanced Computer Studies
List info: http://lists.roundcube.net/dev/
-
Derek Yarnell -
Francis Russell