http://bugzilla.spamassassin.org/show_bug.cgi?id=4290
------- Additional Comments From [EMAIL PROTECTED] 2005-04-30 07:41 -------
I've run out of time for a while to spend on tracking this down. It appears that
something doesn't work right, in a way that depends on timing, when two
different RBL queries end up using identical DNS queries. This is caused by
dnsbl.t containing both of these two rules:
header DNSBL_TEST_TOP eval:check_rbl('test', 'dnsbltest.spamassassin.org.')
describe DNSBL_TEST_TOP DNSBL A record match
tflags DNSBL_TEST_TOP net
header DNSBL_TEST_WHITELIST eval:check_rbl('white-firsttrusted',
'dnsbltest.spamassassin.org', '127.0.0.1')
describe DNSBL_TEST_WHITELIST DNSBL whitelist match
tflags DNSBL_TEST_WHITELIST net nice
Note that they are made to be different by one saying "org." and the other "org"
without the ".", causing there to be two queries that sometimes result in the
second query not getting a response that is seen by code that is looking at the
IO::Socket. If both have the "." then the second query is never made and the
code doesn't handle that either.
My question is if this is a totally artificial aspect of the test or if in real
life we might have a reason to use one RBL in two different ways like this? If
it can arise in real use then we have a bug we have to fix. Otherwise we can
just change the test, perhaps by using a different domain for one of those.
Can anyone chime in with an answer to that?
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
