dovecot - Aug 2015 - Segfaults after upgrade to Debian Jessie

On Mon, 2015-08-17 at 15:23 +0200, Stephan Bosch wrote:
> Still looks like the executed extdata plugin was compiled against a 
> different pigeonhole.
Got it. There was an old version in /usr/lib/dovecot/modules/sieve/, whereas the
new one was installed into /usr/local/lib/dovecot/sieve/. That all works now,
and
no segfault. Thanks for all your help and patience.

Unfortunately my vacation rule no longer works though. I'm getting
"sieve: user's
script dict:proxy::sieve;name=active doesn't exist (trying default script
location
instead)"

I have a dict:proxy configured (/etc/dovecot/pigeonhole-sieve.dict):

connect = host=127.0.0.1 dbname=postfix user=postfix password=xxx

map {
  pattern = priv/sieve/name/$script_name
  table = virtual_users
  username_field = email
  value_field = sieve_id
  fields {
    sieve_active = $script_name
  }
}

And that added to dovecot.conf:

dict {
  sieve = mysql:/etc/dovecot/pigeonhole-sieve.dict
}

And that in 90-sieve.conf:

plugin {
  sieve = dict:proxy::sieve;name=active
}

I guess I'm doing something stupid. Any ideas?

Thanks,

Andy

Andrew Beverley schreef op 17-8-2015 om 16:50:
> On Mon, 2015-08-17 at 15:23 +0200, Stephan Bosch wrote:
>> Still looks like the executed extdata plugin was compiled against a
>> different pigeonhole.
> Got it. There was an old version in /usr/lib/dovecot/modules/sieve/,
whereas the
> new one was installed into /usr/local/lib/dovecot/sieve/. That all works
now, and
> no segfault. Thanks for all your help and patience.
>
> Unfortunately my vacation rule no longer works though. I'm getting
"sieve: user's
> script dict:proxy::sieve;name=active doesn't exist (trying default
script location
> instead)"
[...]
> I guess I'm doing something stupid. Any ideas?
Could you enable mail_debug and look at your log output? That should 
show some more details of what it is doing.

Regards,

Stephan.

On Mon, 2015-08-17 at 17:37 +0200, Stephan Bosch wrote:
> > Unfortunately my vacation rule no longer works though. I'm getting
"sieve:
> > user's script dict:proxy::sieve;name=active doesn't exist
(trying default
> > script location instead)"
> [...]
> > I guess I'm doing something stupid. Any ideas?
> 
> Could you enable mail_debug and look at your log output? That should 
> show some more details of what it is doing.
I've fixed this by changing the name of the sieve script in the database
from
"active" to "main script". I'm a bit confused by this
though.

I could see this in the log:

Aug 17 23:28:54 fieri dovecot: lmtp(10285, x at x.com): Debug:
yxlyGzZu0lUtKAAAChoztw: sieve: sieve dict backend: user=x at x.com,
uri=proxy::sieve,
script=main script
Aug 17 23:28:54 fieri dovecot: lmtp(10285, x at x.com): Debug:
yxlyGzZu0lUtKAAAChoztw: sieve: sieve dict backend: script `main script' not
found
at path priv/sieve/name/main script
Aug 17 23:28:54 fieri dovecot: lmtp(10285, x at x.com): Debug:
yxlyGzZu0lUtKAAAChoztw: sieve: user's script dict:proxy::sieve;name=active
doesn't
exist (trying default script location instead)
Aug 17 23:28:54 fieri dovecot: lmtp(10285, x at x.com): Debug:
yxlyGzZu0lUtKAAAChoztw: sieve: no default script configured for user
Aug 17 23:28:54 fieri dovecot: lmtp(10285, x at x.com): Debug:
yxlyGzZu0lUtKAAAChoztw: sieve: user has no valid location for a personal script
Aug 17 23:28:54 fieri dovecot: lmtp(10285, x at x.com): Debug:
yxlyGzZu0lUtKAAAChoztw: sieve: no scripts to execute: reverting to default
delivery.

So it's looking first for "main script", which is why it now
works. But it was
then looking for "dict:proxy::sieve;name=active". Why didn't that
work when I have
this in my sieve config?

plugin {
    sieve = dict:proxy::sieve;name=active
}

I also don't understand why it is looking for "main script" rather
than "active"
at all, and why things stopped working when I upgraded.

Sorry for all the questions - I'll be happy to update the wiki once I've
understood things better.

Thanks for all your help,

Andy