SA Bugzilla – Bug 6258
misc warnings on spamd restart (v3.3.0-beta1)
Last modified: 2009-12-17 09:04:23 UTC
When restarting spamd (via kill -HUP), I find the following warning messages in the log file: Dec 14 07:44:40 localhost spamd[871]: logger: removing stderr method Dec 14 07:46:10 localhost spamd[871]: Use of uninitialized value in subroutine entry at /usr/local/lib/perl5/site_perl/5.10.1/Mail/SpamAssassin/Util.pm line 323. Dec 14 07:46:10 localhost spamd[871]: Use of uninitialized value in subroutine entry at /usr/local/lib/perl5/site_perl/5.10.1/Mail/SpamAssassin/Util.pm line 324. Dec 14 07:46:10 localhost spamd[871]: child process [28905] exited or timed out without signaling production of a PID file: exit 0 at /usr/local/bin/spamd line 2535. Dec 14 07:46:29 localhost spamd[28905]: rules: meta test __NOT_SPOOFED has dependency 'DKIM_VERIFIED' with a zero score Dec 14 07:46:37 localhost spamd[28905]: spamd: server started on port 783/tcp (running version 3.3.0-beta1) SpamAssassin Server version 3.3.0-beta1 running on Perl 5.10.1 with zlib support (Compress::Zlib 2.02) Solaris 9 Sparc
Created attachment 4598 [details] avoid undef warnings in exit_status_str() Here is a rather cosmetic patch to avoid warnings on undefined status passed as argument to exit_status_str(). This only fixes a secondary problem, but does not address its origin.
(In reply to comment #0) > When restarting spamd (via kill -HUP), I find the following warning messages in > the log file: > > Dec 14 07:44:40 localhost spamd[871]: logger: removing stderr method > Dec 14 07:46:10 localhost spamd[871]: Use of uninitialized value in subroutine > entry at /usr/local/lib/perl5/site_perl/5.10.1/Mail/SpamAssassin/Util.pm line > 323. > Dec 14 07:46:10 localhost spamd[871]: Use of uninitialized value in subroutine > entry at /usr/local/lib/perl5/site_perl/5.10.1/Mail/SpamAssassin/Util.pm line > 324. These warnings are handled by the attached patch. > Dec 14 07:46:10 localhost spamd[871]: child process [28905] exited or timed out > without signaling production of a PID file: exit 0 at /usr/local/bin/spamd line > 2535. I cannot reproduce the problem on FreeBSD, nor on Solaris 11. It may be specific to Solaris 9 - can anyone supply me a VirtualBox -wrapped installed Solaris 9 with SpamAssassin's dependencies installed? It would also be useful to track down the Bug 6253. > Dec 14 07:46:29 localhost spamd[28905]: rules: meta test __NOT_SPOOFED has > dependency 'DKIM_VERIFIED' with a zero score > Dec 14 07:46:37 localhost spamd[28905]: spamd: server started on port 783/tcp > (running version 3.3.0-beta1) This is being tracked as Bug 6241 - the DKIM_VERIFIED in the __NOT_SPOOFED rule needs to be replaced by DKIM_VALID.
+1
I'm applying the warnings-suppression patch even without reaching full 3 votes, it is a trivial change, should be a no-brainer. Bug 6258: avoid warnings on undefined exit status, recognize and nicely display a HUP signal (trivial) Sending lib/Mail/SpamAssassin/Util.pm Committed revision 890963. The primary problem is still not understood nor resolved...
> This is being tracked as Bug 6241 - the DKIM_VERIFIED in the > __NOT_SPOOFED rule needs to be replaced by DKIM_VALID. revision 891224, file rulesrc/sandbox/khopesh/20_khop_bl.cf : 20_khop_bl.cf Bug 6241: comment out complex 'if's to work around a deficiency in mkrules (may be restored when fixed); 20_khop_bl.cf Bug 6258: replace DKIM_VERIFIED with DKIM_VALID to avoid a warning: "rules: meta test __NOT_SPOOFED has dependency 'DKIM_VERIFIED' with a zero score"
I cannot reproduce this on a Solaris 8 (kindly offered by Tom Schulz). Your log shows that it took a child process longer than 90 seconds (as now increased/hardwired by Bug 6191) to respond to parent with a SIGUSR1. Eventually, after 109 seconds, it did start up, so this can be considered a warning only. I can imagine that on a slow/busy box it can indeed take two minutes for a restart, so the 90 seconds timeout could still be increased, perhaps to two or three minutes. I'll close this bug as resolved (the cosmetic revision 890963 as per Comment 4 has already been applied), as well as the rules update. I'll briefly reopen the Bug 6191 and bump up the 90 second timer to two or three minutes. Thanks for testing!