Re: [Evolution] Spamfiltering with evolution - NO LONGER WORKS

[Patrick O'Callaghan <poc usb ve>]

On Sat, 2008-01-19 at 18:05 -0500, Paul Smith wrote:
> On Thu, 2008-01-17 at 15:23 -0430, Patrick O'Callaghan wrote:
> > I think it's fairly clear that when someone says "it doesn't work" they
> > mean "it doesn't work for me", but again I apologize if you found the
> > tonen aggresive.
>
> This was merely the last in a long line of posts over the last week
> which seemed long on explaining how bad things were and short on
> providing useful information that might help resolve issues.  Taken by
> itself yours was not bad; I probably should have picked a different
> example.  Sorry about that :-(.
>
> > I considered inserting a man-in-the-middle trap in /usr/bin/bogofilter
> > to log invocations but decided against it since I don't know if Evo
> > calls the actual BF executable or has a side entrance via a library
> > (so an empty log wouldn't tell me anything).
>
> > From the info you've provided, it seems like a man-in-the-middle trap is
> the next step.  This was where I was ultimately headed when I sent my
> last email but I wanted to check some other things first; that's why I
> asked for people to enable logging, etc.
>
> I've looked at the source to the bogofilter plugin and it does indeed
> invoke /usr/bin/bogofilter with various arguments: there is no backend
> library (I don't think bogofilter provides one actually).  Further, the
> plugin doesn't provide any customization whatsoever that would allow you
> to tell it to use a different program for man-in-the-middle operation.
> So, unfortunately, you'll have to replace your /usr/bin/bogofilter with
> the man-in-the-middle script.
>
> > When did you ask for this? I must have missed it. Any way here it is:
> > the wordlist.db has mod time of 2007-11-10 07:05, so I think it's safe
> > to say it's not being touched.
>
> My suspicion, then, is that the plugin is not even being invoked.
> However, please proceed with the man-in-the-middle approach and see what
> it tells you.  You can probably guess what steps to take but enable
> logging/debug and write it to a test file.  Also print the exit code
> from bogofilter to the test file, etc.
>
> You said something about Evo logging, but did you try to enable logging
> as I described before, by setting CAMEL_DEBUG=junk (make sure you export
> it) then starting evolution (from the shell command line of course)?
> Also be sure you run "evolution --force-shutdown" first and check that
> all evolution processes are dead before doing this.
>
> If you do this you SHOULD see debug output for the bogofilter junk
> plugin for EVERY email you receive.
>
> It's possible, though, that your version of Evo was not built with
> proper debugging enabled; in that case you might not see it.  I don't
> know that much about Fedora, since I gave up on Red Hat back around Red
> Hat 5 or 6 or so, but maybe they have debug build packages you could
> install instead?
>
> > What makes this interesting is that I've just noticed that on the office
> > machine I don't have a ~/.bogofilter directory, and hence no
> > wordlist.db, but I've been merrily clicking away on the Junk icon and
> > hearing no complaint from Evo, i.e. even when BF *cannot possibly be
> > working* Evo doesn't say anything. The BF man page says it should give
> > an exit status of 3 if the wordlist.db file cannot be read. Is Evo
> > checking this?
>
> > From my reading of the code, the plugin definitely needs some love in
> the area of error handling (and customization would be nice too).
> Bogofilter's exit code is translated into a boolean: if it's true (0)
> the check function returns true (which appears to mean the message is
> spam); if it's false (any non-0 value) the check returns false, which
> means the message is not spam.  The code does NOT check for specific
> non-0 values such as 3.  Bummer.

I've upgraded to 2.12.3 (evolution-2.12.3-1.fc8 and
evolution-data-server-1.12.3-1.fc8).

Junk filtering still doesn't work. The /usr/bin/bogofilter is still not
being called.

poc