Dan:
If you look at the log file I sent originally, you'll see that sendmail
is indeed piping the message to the wrapper and a log entry gets
written to /usr/local/majordomo-1.94.5/Log.
Also, I tested the :include: directive, and that works as well.
As far as the location of perl, on Mac OS X it's in /usr/bin/perl, and
it's version 5.6.0. The majordomo.pl and config_parse.pl files have no
#!/usr/bin/perl at the top, but that's normal, isn't it? I certainly
didn't edit anything there...The majordomo file *does* have the
#!/usr/bin/perl line.
This is really confusing me, because this worked fine under Mac OS X
10.1.5. I've also tried recompiling the wrapper with GCC 2.95.x, GCC
3.1, and GCC 3.3, but nothing seens to work, and worse yet, I'm not
getting any real feedback from either majordomo or sendmail--just the
three NOQUEUE lines in the mail.log that don't really say anything
useful in English.
-Michael Amper <amper@macbus.org>
On Sunday, Sep 28, 2003, at 18:17 US/Eastern, Daniel Liston wrote:
> Can your sendmail process mail that pipes to another program?
> Can your sendmail process mail to :include: files?
>
> Sendmail must be working properly before you can link majordomo
> to it. Majordomo depends completely on the sendmail aliases.
>
> One dependency is the ability for an alias to pipe messages to
> majordomo perl scripts through the wrapper. Another dependency
> is the ability to reference a file containing email addresses
> via an alias address. The security built into sendmail must
> understand and allow (trust) majordomo to do these things.
>
> If you can test these dependencies by creating specific aliases,
> and they work, we can start to troubleshoot majordomo by checking
> ownerships, permissions, trust relationships, and the syntax of
> the majordomo aliases themselves.
>
> Can you create a file in /etc/mail with multiple email addresses,
> one per line, and call it mytest?
>
> Next edit the /etc/aliases file and add an alias to that file.
> mytest: :include:/etc/mail/mytest
> (don't forget to execute `newaliases` after editing)
>
> Now send an email to mytest@yourmachine.yourdomain.com. Do the
> email addresses in /etc/mail/mytest each receive a copy of the
> email you sent? Does everything look kosher in /var/log/mail.log?
>
> The wrapper does not really do any mailing work, it just sets
> environment variables so that perl scripts in the majordomo
> directory will execute properly. Speaking of which, do the
> majordomo, majordomo.pl, and config_parse.pl scripts have the
> correctly location for "perl" at the top of each file? There
> is a difference between /bin/perl, /usr/bin/perl, and
> /usr/local/bin/perl. (unless they are linked to the same
> executable) Do all the scripts use the same location for
> perl?
>
> Another thought: When you are logged in as the majordomo user,
> and execute `./wrapper config-test`, does the output say every
> thing is OK?
>
> Dan Liston
>
> Michael Craig Amper wrote:
>
>> Dan:
>> Thanks for the patch, but unfortunately, it didn't work. The syslog
>> reports the exact same error.
>> BTW, I'm using Sendmail 8.12.10. Also, the smrsh link was already in.
>> It doesn't really look like the message being passed to wrapper is at
>> fault, but rather that wrapper isn't generating the response >> properly.
>> -Michael Amper<amper@macbus.org>
>> On Thursday, Sep 25, 2003, at 22:56 US/Eastern, Daniel Liston wrote:
>>> Your message reveals that you are using Mac OS X 10.2.8, majordomo
>>> 1.94.5, and sendmail as your MTA, but not what version it is.
>>>
>>> If your sendmail.cf expects smrsh to add security for running pipes,
>>> you will need to create a link to "wrapper" in the <smrsh> directory.
>>> If sendmail has smmsp enabled, the submit.cf file will need to be
>>> configured to understand trusted-users, and majordomo must be trusted
>>> by both sendmail and smmsp.
>>>
>>> If the message passed to wrapper contains CRLF as EOL, majordomo
>>> will not properly recognize the blank line between headers and body
>>> of the message. This (blank) line will actually contain a single
>>> character rather than being empty. This diff cures the problem;
>>>
>>> ***majordomo.pl
>>> 59c59,60
>>> - s/\n\s+/ /g;
>>> ---
>>> + s/\015//g; # strip DOS <CR> from lines
>>> + s/\n\s+/ /g; # unfold long header lines
>>>
>>> NOTE: Octal value \015 is the same as ctrl-M or ASCII 13.
>>>
>>> Dan Liston
>>>
>>>
>>> Michael Craig Amper wrote:
>>>
>>>> I'm attempting to get majordomo up and running on Mac OS X 10.2.8.
>>>> It was previously working fine on Mac OS X 10.1.5, but we upgraded
>>>> the machine.
>>>> When I send a "help" message (or any other message) to majordomo,
>>>> the log shows that it gets received and piped to the wrapper, the
>>>> majordomo log records the message, but no output ever comes back.
>>>> Here's the output from /var/log/mail.log:
>>>> Sep 25 02:33:45 pretzel sendmail[437]: h8P6Xiwo000437:
>>>> from=<ampermc@macbus.org>, size=611, class=0, nrcpts=1,
>>>> msgid=<33F74AC6-EF22-11D7-8771-000393A4EE96@macbus.org>,
>>>> proto=ESMTP, daemon=MTA, relay=scaup.mail.pas.earthlink.net
>>>> [207.217.120.49]
>>>> Sep 25 02:33:45 pretzel sendmail[437]: h8P6Xiwo000437:
>>>> h8P6Xiwp000437: clone: owner=ampermc
>>>> Sep 25 02:33:45 pretzel sendmail[439]: NOQUEUE: 0: fl=0x0,
>>>> mode=140000: SOCK [0]->[0]
>>>> Sep 25 02:33:45 pretzel sendmail[439]: NOQUEUE: 1: fl=0x1,
>>>> mode=140000: SOCK [0]->[0]
>>>> Sep 25 02:33:45 pretzel sendmail[439]: NOQUEUE: 3: fl=0x2,
>>>> mode=140000: SOCK [0]->[[UNIX: /var/run/syslog]]
>>>> Sep 25 02:33:46 pretzel sendmail[437]: h8P6Xiwp000437:
>>>> to="|/usr/local/majordomo-1.94.5/wrapper majordomo",
>>>> ctladdr=<majordomo@pretzel.macbus.org> (1/0), delay=00:00:01,
>>>> xdelay=00:00:01, mailer=prog, pri=30605, dsn=2.0.0, stat=Sent
>>>> and the output from /usr/local/majordomo-1.94.5/Log:
>>>> Sep 25 02:33:46 pretzel.macbus.org majordomo[438] {Michael Craig
>>>> Amper <ampermc@macbus.org>} help
>>>> Looking at the pid's, it would seem that sendmail is delivering the
>>>> message to the wrapper, the wrapper is processing it, and then
>>>> attempting to resend, at which point it fails. Has anybody seen
>>>> this before?
>>>> BTW, when I call the wrapper directly from the shell and feed it
>>>> some headers, it does return the "error+help" message correctly.
>>>> Any help would be greatly appreciated!
>>>> -Michael Amper <amper@macbus.org>
References:
|
|
