[Mailman-Developers] Re: [Mailman-Users] Wierd Mailman problem

Barry A. Warsaw bwarsaw@cnri.reston.va.us (Barry A. Warsaw)
Tue, 12 Jan 1999 15:26:54 -0500 (EST)


>>>>> "GH" == George Hartz <georgeh@blowtorch.com> writes:

    GH> The web interface seems to work okay, although I don't get
    GH> anything in the archives. Should clicking on the archives link
    GH> work before any messages have been posted to the list?

No, that's a buglet.  The archive link is not functional until at
least one message has been posted to the list.

    GH> Anyway the problem I seem to be having is that Mailman is
    GH> failing without any errors at all. User and group ID's seem to
    GH> be set correctly.  The wrapper programs aren't logging any
    GH> errors for bad UID's. (And running the programs as an
    GH> incorrect user *does* log them, so I know the logging itself
    GH> works.

I'm sure you know that if the error occurs because of a gid mismatch,
in either the CGI wrapper or the mail wrapper, errors are only logged
to syslog, not to $prefix/logs/error.  For mail, you ought to also get 
a bounce containing the error message, and for CGI -- with the next
version -- you will get a more useful Web diagnostic page.

Once control has been passed to the Python world (e.g. the `driver'
script for CGI, or `post' for mail), then errors get logged to
Mailman's log files.  Note that a misconfiguration of syslog or your
mail system might mean that those pre-Python errors are getting lost.

[This is probably more useful to mailman-developers]

Once you're in the Python world, there are several ways to log error
messages for debugging.  My favorite is, once I have a MailList
object, I like to call the LogMsg() method with kind=='debug', kind of 
like so:

    mlistobj.LogMsg('debug', 'I got to here: %s', 'foo method')

This will timestamp the message, and write it to $prefix/errors/debug.

-Barry