Are you using any special options for pyzor in your config file?
Is any of the directories in your path or where pyzor resides world-writable?

(In reply to comment #1)
> Are you using any special options for pyzor in your config file?

Here are the pyzor options from local.cf:

use_pyzor               1
pyzor_path        /usr/local/bin/pyzor
pyzor_options     --homedir /etc/mail/spamassassin/.pyzor

These settings worked fine in v3.2.5.

> Is any of the directories in your path or where pyzor resides world-writable?

No.

  Bug 6191: Plugin/Pyzor.pm, 'pyzor_options' was not untainted
Sending        lib/Mail/SpamAssassin/Plugin/Pyzor.pm
Committed revision 810895.

Created attachment 4531 [details]
untaints pyzor_options

Please try the attached patch (or grab the current SVN head).

That fixes the taint problem, but I'm still getting an error:

Sep  3 09:26:54 localhost spamd[22192]: logger: removing stderr method
Sep  3 09:27:24 localhost spamd[22192]: child process exited or timed out without signaling production of a PID file at /usr/local/bin/spamd line 2505.
Sep  3 09:27:38 localhost spamd[22194]: pyzor: [22197] error: TERMINATED, signal 15 (000f)
Sep  3 09:27:46 localhost spamd[22194]: spamd: server started on port 783/tcp (running version 3.3.0-alpha2)

I am able to run Pyzor from the command line (I get a timeout error).

> That fixes the taint problem, but I'm still getting an error:
> 
> Sep  3 09:26:54 localhost spamd[22192]: logger: removing stderr method
> Sep  3 09:27:24 localhost spamd[22192]: child process exited or timed out
> without signaling production of a PID file at /usr/local/bin/spamd line 2505.
> Sep  3 09:27:38 localhost spamd[22194]: pyzor: [22197] error: TERMINATED,
> signal 15 (000f)
> Sep  3 09:27:46 localhost spamd[22194]: spamd: server started on port 783/tcp
> (running version 3.3.0-alpha2)
> 
> I am able to run Pyzor from the command line (I get a timeout error).

Well, I don't see much that SA can do except abort the forked pyzor
process after 15 seconds if it doesn't reply in this time.

Get running the pyzor from a command line first, and only then try to
wrestle with a plugin after the basics are right. Perhaps the problem
is a firewall setting, or lack of a successful 'pyzor discover'.

closing...

(In reply to comment #6)
> > That fixes the taint problem, but I'm still getting an error:
> > 
> > Sep  3 09:26:54 localhost spamd[22192]: logger: removing stderr method
> > Sep  3 09:27:24 localhost spamd[22192]: child process exited or timed out
> > without signaling production of a PID file at /usr/local/bin/spamd line 2505.
> > Sep  3 09:27:38 localhost spamd[22194]: pyzor: [22197] error: TERMINATED,
> > signal 15 (000f)
> > Sep  3 09:27:46 localhost spamd[22194]: spamd: server started on port 783/tcp
> > (running version 3.3.0-alpha2)
> > 
> > I am able to run Pyzor from the command line (I get a timeout error).
> Well, I don't see much that SA can do except abort the forked pyzor
> process after 15 seconds if it doesn't reply in this time.
> Get running the pyzor from a command line first, and only then try to
> wrestle with a plugin after the basics are right. Perhaps the problem
> is a firewall setting, or lack of a successful 'pyzor discover'.

I got the Pyzor problem fixed (it was the firewall):

ornl71# pyzor --homedir=/usr/local/spamassassin/.pyzor ping
public.pyzor.org:24441  (200, 'OK')

I'm no longer seeing the pyzor errors but I still get the "child process exited" when starting spamd:

ornl71# /etc/init.d/spamd start
child process exited or timed out without signaling production of a PID file at /usr/local/bin/spamd line 2505.

and from the log file:

Sep 21 09:55:22 localhost spamd[24972]: logger: removing stderr method
Sep 21 09:55:52 localhost spamd[24972]: child process exited or timed out without signaling production of a PID file at /usr/local/bin/spamd line 2505.
Sep 21 09:56:41 localhost spamd[24974]: spamd: server started on port 783/tcp (running version 3.3.0-alpha2)

After further analysis, the problem is that the daemonize() function has a loop that waits 30 seconds for the server to signal that it has started.  On my box (a slow box used for development) it takes almost a minute, so the loop times out and complains.

Do we increase the timeout for the startup of spamd?

(In reply to comment #10)
> Do we increase the timeout for the startup of spamd?

sure, why not.  I can see the usefulness of that.

Thanks Larry for tracking this down.


Bug 6191: takes almost a minute to start spamd on a
slow machine, bump up the retry counter to 90 seconds
Sending        spamd/spamd.raw
Committed revision 817311.

According to Bug 6258, a forked fresh child can take 109 seconds to respond
to a parent process. I'm briefly re-opening this bug to document it,
and to bump up the timer to 180 seconds (3 minutes). I'm open to suggestions
on the exact value, I guess two minutes would probably be fine too.

This one is trivial:

Bug 6191, Bug 6258: bump the timeout for a spawned spamd start to 3 minutes
Sending spamd/spamd.raw
Committed revision 891806.