Hello.  For some time, I've been experiencing occasional problems when
trying to train crm114.  I use the mailreaver setup as recommended by
the documentation, and train based on mistakes or based on whether
mailreaver is "unsure" about the state of a message.  Usually I only
train a single message at a time, though occasionally I'll train several
days' worth of "unsure" messages (typically less than 30 messages).
Sometimes when training, crm114 gets stuck in some kind of loop.  The
process tree (via ps uxww --forest) looks like this:

12999 pts/8    S+     0:00      \_ mutt
14121 pts/8    S+     0:00          \_ crm -u /home/frodo/.crm114 mailreaver.crm --good
14135 pts/8    S+     0:00              \_ crm -u /home/frodo/.crm114 mailreaver.crm --good
14136 pts/8    R+     4:59                  \_ /usr/bin/crm ./mailtrainer.crm --fileprefix= --good=reaver_cache/known_good/20100526_145754_808200_FFFFFFFFAF048C67 --spam=reaver_cache/empty/
14148 pts/8    Z+     0:00                      \_ [mailtrainer.crm] <defunct>

In this case, pid 14136 appears to be spinning, and will stay like this
forever unless terminated.  Attaching strace to it shows that it's not
making any system calls, so it's clearly not walking the files in the
cache or anything like that.  You can see from above that it's already
been running for 5 minutes, which is far more time than it should take
to train a single message.

I have around 72,000 files in .crm114/reaver_cache/texts, which doesn't
seem like an unusually large number based on what I've seen in the docs.

I've noticed that http://crm114.sourceforge.net/wiki/doku.php?id=64bit
suggests that there are issues on 64 bit systems.  I wonder if this
might be related to that?  Have these issues been addressed in the
packages in lenny?  The versions referenced on that page seem to be
newer than what's in lenny, so I suspect they haven't.

Let me know if you've got any additional questions.

Thanks.
noah

I've been using crm114 on an amd64 system without problems for several
years.  So I don't know what your problem might be.  You may try to
upgrade to the version from testing.  IIRC I've heard people had some
problems using crm114 over NFS if it happens to be your case.

Nope, no NFS.  The filesystem is ext3.  The system is a multi-processor
system, so it's possible that the problem is related to a timing issue
in such an environment.

noah

    NM> The system is a multi-processor system, so it's possible that
    NM> the problem is related to a timing issue in such an environment.

I've been using crm114 on multicore systems without problems.  I suggest
trying the current crm114 version (beware of .css format changes) and if
the problem persists then contacting upstream.