Re: crm doc glitches

   From: "[email protected]" <[email protected]>

   [...]

   I'm by no means as accomplished as most people on this list (so)
   reaver_cache caught me out. I completely forgot about it, probably for a
   few years or so, and many many many thousands of files sat there.

   So I set :text_cache: to //; which threw messages like

   /bin/ls: :long_cacheid:: No such file or directory
   /bin/ln: target `/known_spam/' is not a directory: No such file or directory
   ls: /known_spam/: No such file or directory
   ls: /empty/: No such file or directory

   which I happily ignored (empty and known_spam exist as dir btw.) There
   were no error messages like 'aw crud' or 'this program is seriously sick'
   or something.

   But then I noticed 'training' had no effect any more.

Huh?   Interesting.  That's a bug.

   This wasn't obvious to me (the headers for instance included
   "X-CRM114-Action: LEARNED AND CACHED SPAM" after training.)

   Am I right in thinking that you *need* reaver_cache in order to actually
   train a message cq update the .css libraries?

   (Mr Fox answered that when you hand a message back to crm to train,
   actually the corresponding message-text from reaver_cache is used, not the
   text from the message that was handed back to crm. So yes, the
   reaver_cache is needed to train.)

It _shouldn't_ be... the code is supposed to look in the cache first
and if it finds the message with the proper SFID, to use that (as that
text is "virgin" as far as not having been munged by a mail reader, etc).

It should use the text supplied in the --good or --spam training command
only if it can't find the original SFID in the reaver cache.

It's definitely a bug though.  Bugfixes genuinely appreciated...

     - Bill Yerazunis

------------------------------------------------------------------------------