I have just upgraded to 3.01 from 2.53 (on an SGI IRIX machine) and after getting all of that completed nicely I decided to retrain the bayesian engine, so I first gave sa-learn about a thousand spam messages and the only complaint from sa-learn was about 10 lines of:
Parsing of undecoded UTF-8 will give garbage when decoding entities at /usr/local/lib/perl5/site_perl/5.8.6/Mail/SpamAssassin/HTML.pm line 182. then I fed it my mail box of about 1800 ham (~58MB) and got Out of memory! Which is not very promising at all. So why is sa-learn so upset? Rich
