[Mailman-Users] Error running bin/mailmanctl restart

[Steve Burling]

--On December 17, 2009 1:21:35 PM -0600 Paul Kleeberg <paul at FPEN.ORG> wrote:

> I am ready to tear my hair out.
>
> Thanks to Mark, I got mailman up and running except for one minor hitch.
> In my log file I see an endless stream of:
>
> 	12/17/09 11:56:49 AM com.apple.launchd[1] (org.list.mailmanctl)
> Throttling respawn: Will start in 10 seconds  	12/17/09 11:57:00 AM
> com.apple.launchd[1] (org.list.mailmanctl) Throttling respawn: Will start
> in 10 seconds
>
> Searching the archives, I found this (The thread starts at
> http://www.mail-archive.com/mailman-users@python.org/msg53952.html):
>
>
> > The only difference between my org.list.mailmanctl.plist and Apple's is
> > my verb *start*, as opposed to Apple's *startf*. I'm baffled by
> > *startf*, which is not supported by mailman and does nothing, either in
> > the CLI or in org.list.mailmanctl.plist.

At one point, I compared the stock mailmanctl with that from the one Apple 
distributes with Mac OS X Server.  The only difference was that the 'start' 
stanza in main() had been cloned as a 'startf' stanza, with a couple of 
relatively minor changes, primarily (if I remember correctly) to make it 
not daemonize, since things run from launchd aren't allowed to daemonize. 
I *think* that that's what's leading to the problem you see -- mailmanctl 
daemonizes, launchd cleans it up and it respawns, eventually triggering the 
error.

For my Mailman install on Mac OS X (client, not server) I gave up on using 
launchd to start Mailman at boot time, and just used an old-style 
StartupItem.  I did make launchd scripts for all the Mailman stuff that's 
normally run out of cron; see the archives of this list for information 
about those.

In short, if you want to use launchd to start Mailman at boot time, I think 
that you're going to have to look at Apple's version of mailmanctl, and 
make equivalent changes to the stock mailmanctl to provide a 'startf' 
variant.  If you do that, I urge you to try to do it in a cleaner way than 
Apple's hack -- my memory is that there were so few differences that it was 
a crime that they'd just duplicated the entire block of code and commented 
out parts.

NOTE:  I was doing this comparison between the mailmanctl from Mac OS X 
Server 10.5, not 10.6, so it's possible that they cleaned up their act and 
did things differently.  But a quick side-by-side compare of their 
mailmanctl and the stock one should get you pointed in the right direction.

-- 
Steve Burling                                    <mailto:srb at umich.edu>
University of Michigan, ICPSR                    Voice: +1 734 615.3779
330 Packard Street                               FAX:   +1 734 647.8700
Ann Arbor, MI 48104-2910