https://bz.apache.org/SpamAssassin/show_bug.cgi?id=7735
Henrik Krohns <apa...@hege.li> changed: What |Removed |Added ---------------------------------------------------------------------------- Resolution|FIXED |--- Status|RESOLVED |REOPENED --- Comment #87 from Henrik Krohns <apa...@hege.li> --- (In reply to Wolfgang Breyha from comment #86) > > In this Sample no DKIM-Sig is available... so DKIMDOMAIN and all __DKIMWL_* > and DKIMWL_* stay "unrun" and a missing dependency. > > Every meta rule which depends on them somehow is run only once in > finish_meta_tests(). And if a meta depends on a meta which depends on > DKIMDOMAIN/ff somehow.... I get random results (race condition) Ok thanks I think I get it now.. Here's a sample test case: meta __TEST_META_I (1 || TEST_DISABLED3) meta TEST_META_I __TEST_META_I Both should always hit in this case. If unrun TEST_DISABLED3 postpones things to finish_meta_tests, and the foreach with it's random hash order processes TEST_META_I first, obviously things will get wrong. I'll see if can figure out a better logic, maybe just migrate finish_meta_tests final logic into do_meta_tests somehow. -- You are receiving this mail because: You are the assignee for the bug.