That's how the March 2011 trouble happened.
I think the March trouble occurred because nopublish was removed and
#testrules added and then #testrules didn't work as expected. The
code definitely does not agree with what others appear to think it
should do.
FYI, according to http://wiki.apache.org/spamassassin/SaUpdateBackend we
might be dealing with a race condition because of this criteria:
"tflags net", "userconf" or "learn" rules are always published
So we are going to have to have some criteria weight.
For example, if I have a rule that is tflags net nopublish, which takes
precent:
"tflags nopublish" rules are never published or "tflags net",
"userconf" or "learn" rules are always published
These are mutually exclusive criteria and rules can meet both criteria...
Regards,
KAM
