[Bro-Dev] non-deterministic notice_policy order

Jonathan Siwek jsiwek at ncsa.illinois.edu
Tue Sep 27 14:57:17 PDT 2011


I saw that the suppression-disable.bro test periodically failed because the order of elements in Notice::ordered_policy varies between runs (run `bro -e ''` multiple times, looking at notice_policy.log each time for differences).  And I think the reason for that variance is because the table/set hashing is partly computed based on the memory address of PolicyItem$pred functions.

If that's right, is there an idea of how to make hash computations for functions deterministic?

- Jon


More information about the bro-dev mailing list