[stalwart-spam-train.py] Thunderbird / mbox is recognised as 1 message only

[piscocgn]

Hi,

I tried several ways of pointing to a mbox file from Thunderbird. I also tried to export mbox again. No Way. The only way I get it working it so export via e.g. "ImportExportTools NG" as .eml files. However, native mbox would be great to train from Thunderbirds folder file directly.

Another geat way sould be to train directly via IMAP from a users folders. A user could easily sort into folders like "Learn/Spam" or "Learn/Ham" and the script could run frequently via cron.

By the way: I have recognize that if I train for a an invalid/nonexisting user with --account user101 it's process as success. Not sure if stalwart is checking the existance of the user or just silently take it.

kindly

[JimDunphy]

Hi,

Thanks for the report. I see the problem I think. A file without an .mbox extension might likely be treated as a single email file. I updated the code with a patch that is a little smarter so you should be able to reference your thunderbird folders directly without the extension. In my case, I was copying the folder and renaming it with .mbox if memory servers to differentiate between the .eml's already present and my mbox's. Note: these new changes are not tested as I only did --dry-run because we had moved to rspamd and am running that as a milter. We train with a different tool under the rspamd folder. Hopefully, this works for you.

Jim