[milters] Archive

Lists Index Date Thread Search

Article: 396
From: Anthony Howe
Date: 2005-03-17 03:13:02 -0500
Subject: Re: milter-ahead runs fine, then dies on OpenBSD

Removal...........: milters-request@milter.info?subject=remove
More information..: http://www.milter.info/#Support

Todd Pytel wrote:
> I've got milter-ahead 0.8 and libsnert 1.40 installed and, apparently,
> correctly configured on an OpenBSD 3.5 system. Mail logs indicate
> that the milter operates properly, caching results and generally
> doing what it should. But after a short time (usually less than an
> hour) the milter-ahead process dies. It sounds an awful lot like the
> problem on FreeBSD mentioned in the thread started here:

I suspect is this problem:

FreeBSD, OpenBSD, and I presume NetBSD insist on using really old 
Berkeley DB 1.85, which Keith Bostic wrote and he has told me has 
several design problems and is no longer supported by him.

All my efforts to narrow down the BDB cache problem in milter-sender 
(and my other milters that use a cache) points to the BDB 1.85 library, 
not the milter. I've grown so feedup with *BSD systems refusal to move 
forward in time in this one aspect, that milter-sender's default caching 
mode for *BSD is now "flatfile".

This should have been the case in milter-ahead 0.8 and libsnert 1.40. So 
use the -z option to switch the cache type and path. For example:

	-z flatfile,/path/to/cache.txt

However, I later found two serious Cache flatfile related bugs in 
LibSnert 1.40 that are fixed in 1.41, which has yet to be released. 
Contact me to get access to a pre-release libsnert 1.41 and a small 
patch for milter-ahead.c.

Anthony C Howe                                 +33 6 11 89 73 78
http://www.snert.com/       ICQ:
7116561         AIM: Sir Wumpus

"held in my arms / his sun washed face / eyes closed" - Anthony

Lists Index Date Thread Search