Fix for bayes rebuild bug on Solaris

[Rose, Bobby]

Ok it seems to be working.  I'm confused by the scheduling though.  I
set the rebuild to 300 secs last night so I could check adequately but
it never ran a rebuild every 5 mins.  Before anyone asks why 5 mins,
this was only to "test" the code.  

Anyway, this morning I checked the logs and the rebuld did occur.  But
it looks like it just ran twice 5mins apart after the 4hr Mailscanner
restart time.  Is this correct?

Feb  5 03:16:37 eeyore MailScanner[2023]: Bayes database rebuild is due
Feb  5 03:16:38 eeyore MailScanner[2023]: SpamAssassin Bayes database
rebuild preparing
Feb  5 03:16:43 eeyore MailScanner[2023]: SpamAssassin Bayes database
rebuild starting
Feb  5 03:22:09 eeyore MailScanner[2658]: Bayes database rebuild is due
Feb  5 03:22:11 eeyore MailScanner[2658]: SpamAssassin Bayes database
rebuild preparing
Feb  5 03:22:25 eeyore MailScanner[2658]: SpamAssassin Bayes database
rebuild starting
Feb  5 07:17:47 eeyore MailScanner[18646]: Bayes database rebuild is due
Feb  5 07:17:48 eeyore MailScanner[18646]: SpamAssassin Bayes database
rebuild preparing
Feb  5 07:18:22 eeyore MailScanner[18646]: SpamAssassin Bayes database
rebuild starting
Feb  5 07:23:02 eeyore MailScanner[19177]: Bayes database rebuild is due
Feb  5 07:23:03 eeyore MailScanner[19177]: SpamAssassin Bayes database
rebuild preparing
Feb  5 07:23:09 eeyore MailScanner[19177]: SpamAssassin Bayes database
rebuild starting


 

-----Original Message-----
From: MailScanner mailing list [mailto:MAILSCANNER at JISCMAIL.AC.UK] On
Behalf Of Julian Field
Sent: Wednesday, February 04, 2004 4:41 AM
To: MAILSCANNER at JISCMAIL.AC.UK
Subject: Fix for bayes rebuild bug on Solaris

At 18:05 02/02/2004, you wrote:
>At 17:57 02/02/2004, you wrote:
>>Gee...
>>
>>FWIW, it happened a couple of centuries ago, but I recall having 
>>serious trouble making Perl's flock() work on Solaris... same 
>>situation, all development done under linux without a hitch and 
>>Solaris ignored all the locking... and it wasn't an interoperability 
>>problem, since I was competing against my own script...
>>
>>The point is I don't quite remember what we did to solve it (we is an 
>>understatement, since it wasn't me programming, I was just the 
>>designer)... it must have been perl 5.6.x and Solaris 7, but I'm not 
>>sure either...
>>
>>Seems like you'll need a Solaris box to test it thoroughly... I 
>>wouldn't even trust Solaris-x86 to behave identically to Solaris-Sparc

>>:-(
>
>I've got an Ultra-5 so I can do a real test. If necessary, I can build 
>a
>Solaris-x86 box too. But as you say, the best place to try it is a real
sparc.

I have found the problem. Attached is a very short patch to SA.pm. This
should let you enable the "Rebuild Bayes Every" feature that does
scheduled Bayes database rebuilds.

If you turn this feature on in MailScanner.conf, you will want to set
         bayes_auto_expire 0
in your spam.assassin.prefs.conf to disable the (unsuccessful) attempts
at letting SpamAssassin rebuild its Bayes database when it feels like
it.