On 23.08.2010 21:43, Kevin Falcone wrote: > > This implies that RT isn't seeing Requestors for that ticket. > You may want to show a sample of the email being injected into RT > and look at the user record for the user assigned as a Requestor for > the ticket. You're also using a custom Autoreply template, and we > don't know what you've changed there. > I solved the problem by taking up your idea of dumping the raw mails and by remembering I had this problem before.
The mail is delivered using a forwarding (i...@mydomain goes to [email protected] and from there is piped into the mailgate). It seems that when sending the ticket directly to [email protected], the autoreply is triggered. When sending it to i...@mydomain, it's not. So i checked the mail contents. There is a valid From: header pointing to the requestor. The To: header points to the ticket system address. So it looks good to me, seeing that the From: header is (as far as rt--mailgate doc is concerned) is relevant for user authentication. The only difference between the working and the not working mail is that there is an empty Return-Path header in the broken one. And now it dawned to me. In lib/RT/Interface/Email.pm, RT checks for Return-Path =~ /<>/ to determine a bounce. I had patched this to make our old installation work ("return(0)"), but naturally forgot about the patch. Now, basically everything works again, but I have 2 questions: 1. does effectively disabling the routine CheckForBounce actually break something? In over 98000 tickets, I have not experienced a problem with double bounces or the likes. 2. Why does qmail forward a mail with an empty Return-path? But that's probably a whole other cup of tea. Regards, --ck RT Training in Washington DC, USA on Oct 25 & 26 2010 Last one this year -- Learn how to get the most out of RT!
