@paper1111 I've been missing that from time to time, too -- even if we manage to make the fallback logic fairly foolproof, you often don't have evidence that a post wasn't in fact missed if Smokey didn't report it
@iBug if you're going to be doing development more often, I highly recommend dual-booting. It makes life a lot easier. That's what I do on both my computers
@Mithrandir Taking auto flagging action based on a single user's decision (trusted or not) doesn't sit right with me either. Wasn't one of the conditions for auto flagging that it wasn't to be based on user input but triggered by the bot deciding what it thought was spam?
Btw @iBug I'm now trying your "don't scan posts if community bumps" issue. Do you think if searching MS for the same content with the same url would work?
@angussidney I don't need it actually, just writing a PR for smokey that requires talking with MS... Whatever I'll just search all the revisions of that URL and not care about wether it is the latest or not
@ArtOfCode are the accounts used to flag based on a pool of accounts which have a matching condition or a pool of conditions which match the criteria, and then the flag is cast on the matching account? I just thought of a possible abuse vector
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, pattern-matching website in body, +1 more: Phytolast offer by miwzqtyke on drupal.SE
I have one filter covering all sites, and one filter covering 5 sites. Those 5 have a much higher true positive rating than any others, and get a lot of spam, so I group them together to set my autoflag weight much lower on those sites.
