https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7735
--- Comment #52 from Henrik Krohns <apa...@hege.li> --- (In reply to Wolfgang Breyha from comment #51) > I took the time and read the whole thread and bug 4549 now and understand > how "running a rule twice" should work in theory. No problem with that > anymore. > > But still, it doesn't work that way currently, because many meta rules do > not run because they get not triggered. At least I can "destroy" a working > arithmetic meta rule by adding "+ DCC_CHECK" for example if the plugin is > disabled. > > All I see is > unrun dependencies prevented meta __METATEST from running: DCC_CHECK > and no result, while without "+ DCC_CHECK" I get > rules: ran meta rule __METATEST ======> got hit (4) What does the __METATEST rule contain? Just as an example, this stock rule won't break: meta DIGEST_MULTIPLE RAZOR2_CHECK + DCC_CHECK + PYZOR_CHECK > 1 1 + 0 + 1 > 1 1 + 1 + 1 > 1 If DCC_CHECK is unrun, and RAZOR and PYZOR hit, it will work. > And this still happens with my patch I sent you last week in place, which > helps meta rules running as soon as other metas finished they depend on. So, > delaying a meta rule to the point that it is handled by finish_meta_tests() > seems to break it in some cases. Yes I see now that it's needed in case check_post_dnsbl or check_cleanup hooks finish some rules. I have to wait until I can commit changes from previous bugs first. -- You are receiving this mail because: You are the assignee for the bug.
