Re: tmda Pending Messages

On Nov 16, 2006, at 15:47, Stephen Warren wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Doug Hardie wrote:
>> Setting the HOME env variable before calling tmda-filter does  
>> work.  I
>> didn't see HOME used in the code, but it must be there somewhere.
>
> It's used internally by Python itself.
>
>> That is the easiest way to get NIS users to work.  I decided to
>> create my own wrapper for tmda-filter as procmail and others  
>> introduce
>> yet another layer of overhead with reading the entire message and
>> processing it just to set the env variables.
>
> I'm still confused why HOME isn't being automatically set correctly.
>
> Do you believe that you need to write a wrapper to set HOME because:
>
> 1) You tested invoking tmda-rfilter from sendmail and it didn't work,

Sendmail is only setting one env variable (name forgotten but its  
value is "sendmail").  It does not set HOME.

> or
> 2) You tested it on the command-line when HOME was set incorrectly,  
> and
> it didn't work?
>
> If (2), then you really need to try (1) and see if sendmail is already
> doing the right thing - it's likely quite difficult to correctly and
> accurately manually simulate the environment that sendmail creates  
> when
> it invokes tmda-rfilter.

Getting rid of all the env variables was tricky.


>
> That said, if you don't want to use procmail/maildrop, given that  
> you're
> using sendmail, then I guess you need *some* kind of wrapper between
> sendmail and tmda-rfilter to set SENDER, RECIPIENT, and EXTENSION, as
> described at:
>
> http://wiki.tmda.net/ServerConfiguration#sendmail
>
> As a first step, however, I really would suggest getting everything
> working the "standard" way using procmail, and testing everything  
> under
> for real using the MTA (if only for test email accounts), before
> spending time re-writing procmail as an optimization.

The wrapper is only about 25 lines of c code.  Very straight forward  
once you know HOME has to be set.  That should be included in the  
documentation rather than presuming its set magically by something else.

>