zmtrainsa + dspam = broken on 5.0.16GA network edition

[shankwc]

I was on 5.0.6 before I upgraded to 5.0.16. I used to be able run the zmtrainsa tool and point it to individual mailboxes. This seems to be broken now if you are using dspam. You can still run zmtrainsa with no arguments and have it train on the system accounts. But if you try to give it a user mailbox it will hang forever on dspam training. I looked into this further and I can see that the system account is handled differently than user mailboxes.

In the autoTrainSystem() section if dspam is enabled it calls:
dspam_train zimbra ${spamdir} ${hamdir}
where the spamdir and hamdir variables are tmp directories where the contents of the spam and ham system accounts have been unrolled into.

If you look at the trainAccountFolder() section you will see that when doing dspam training - that section actually calls dspam directly. That is the part that it hangs on.

Perusing the dspam_train script, it seems to me the problem is with the extraction and/or user flag. Also, we have a multi-server configuration with 2 MTAs to handle the spam processing load. I wonder if there is no extraction routine that pulls the spam out of the users Junk folder on the Message Store server and into a tmp location on the MTA - and that is the issue?

So my question is this: Is anyone using dspam with Zimbra 5.0.16 and are you able to manually run zmtrainsa against a user account?

[ericortego]

I am seeing zmtrainsa behave the same way in 7.1 with dspam enabled when pointed at a user's folder.