https://bugs.exim.org/show_bug.cgi?id=1395
--- Comment #11 from Wolfgang Breyha <wbreyha@???> ---
It seems even my first proposed patch didn't work for DNS_NOMATCH (or other
results cached within dns.c/dns_basic_lookup()/tree_dns_fails) answers. So
neither my "ignore" option nor your TTL patch prevents negative cache entries
from being valid as long as a process is running.
The comment above the tree_search(tree_dns_fail,..) talks about "This is mainly
so that a timeout on one domain doesn't happen time and time again"...
I'm not sure if a timeout ever results in a straight DNS_NOMATCH and if it
makes sense to cache it in tree_dns_fails and which impact it would have to not
cache it in general. Not speaking of the DNS_AGAIN->DNS_NOMATCH translations.
Only direct DNS_NOMATCH results.
I'm currently thinking of possible solutions how to make negative cache TTL
aware, or if I improve my local hack by extending dns_basic_lookup with an
option to select if DNS_NOMATCH should be cached. Maybe you have even better
ideas?
--
You are receiving this mail because:
You are on the CC list for the bug.
