On Mon, 9 Jun 2003, Amos Gouaux wrote: > I and a few others here have begun noticing that periodically Sieve > won't filter a message properly. Sometimes I've even resent the > message again to find that on that occasion it did get filtered > just as expected. I wonder, could that BerkeleyDB leak for which > Rob just posted those two patches have any bearing on this? Sadly, > so far I haven't been able to detect a pattern to make this report > more meaningful, but I'll keep an eye out.
The sieve parser state doesn't get reset properly after an error. Thus once a script fails to parse, all scripts will fail to parse (try to submit a good script after a bad one in a timsieved session to see this in action). If badly-formed sieve scripts could have snuck in, then this will affect lmtpd as well, leading to behavior like you describe here. The bytecode, of course, solves this problem (But that won't ever be seen in 2.1) -Rob -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- Rob Siemborski * Andrew Systems Group * Cyert Hall 207 * 412-268-7456 Research Systems Programmer * /usr/contributed Gatekeeper
