> Oh, totally agreed - trust me, I'm not in any hurry to push these > changes onto my production servers. I was mostly just curious if using > refs would make any notable performance difference. > > Forgot to mention it, but anyone should feel free to suggest other > things for me to try or provide; every single machine I have > SpamAssassin on so exhibits the same behavior so it would appear that > I'm particularly blessed with an ability to reproduce the problem. Not > sure when I can get to each request, but I usually have time here and > there. At this point my next idle task is to try and find a box where > the problem *doesn't emerge* and go from there.
I just managed to reproduce it - on a virtual CentOS 5 with SA 3.2.5 and perl 5.8.8. Let me see if I can isolate the problem to a small Perl test case... Mark
