SA Bugzilla – Bug 1567
allow_user_rules doesnt work with some tests, some of the time
Last modified: 2003-05-11 10:41:21 UTC
Anyways... I fixed this bug a while ago, but I guess I never committed it. Oops. Here's a patch. It basically allows users to define header:exists rules, and some rbl rules when allow_user_rules is on.
Created attachment 693 [details] patch for review for 2.50
for 2.50
ARgh... I forgot to CC: sa-devel. This is intended for 2.50, so if anyone would like to review it... Here's the summary: Anyways... I fixed this bug a while ago, but I guess I never committed it. Oops. Here's a patch. It basically allows users to define header:exists rules, and some rbl rules when allow_user_rules is on.
Duncan, This is a significant change for 2.50 series, but looks okay to me. Just one question: have you tested each of these sub-categories to make sure they work okay and don't completely bog down spamd if the user does something funky like add a huge number of bad RBL rules?
Subject: Re: allow_user_rules doesnt work with some tests, some of the time Actually, I haven't tested it nearly enough. I imagine that the RBL ones aren't 100% necessary. I'd be happy if we just added allow_user_rules to work with the header:exists tests. That fix has been tested for a while in Debian packages.
I'm okay with allowing the header:exists tests.
Another option I would consider to allow the RBL tests would be to have RBL tests require allow_user_rules to be set to >= 2 instead of == 1, then the other places the option is tested would have to test for >= 1 as well, but I think I wwill be okay with it as long as every place got changed.
Subject: Re: [SAdev] allow_user_rules doesnt work with some tests, some of the time IIRC, allow_user_rules with spamd _could_ (depending on perl) give users access as the user spamd is run as. (Using /(?{code})/) perlre says this is not possible if run-time evaluation of variables happens. Someone should probably actually test this. Of course, this issue makes the addition of DNS tests to be a triviality.
Duncan: Are you sure this works? There's a "next" on the lines before the ones you added... :o)
reassigning to Duncan
Created attachment 845 [details] Patch for review.
umm... right you are. this should work better
Actually... some of the RBL concerns may be valid, and I have not tested that. Maybe we should hold off until 2.60. (Still... someone should test that...)
alright... since this isnt entirely going in to 2.54, and its already committed for 2.60, we'll let this be closed... the RBL stuff can be reported separately if it's a problem. (And I like having no bugs assigned to me!)