Re: procmail

[Martin Janzen <janzen@idacom.hp.com>]

tim werner <werner@mc.ab.com> writes:
>> With procmail, you can associate arbitrary actions with a match, so no
>> it would not have to be silent.
>
>Will anyone who has procmail working please send me an example of how you
>use it?  I am totally confused.


OK.  First -- just checking -- I assume you're running some version of
UNIX.  I'm using HP-UX 9.03 on an HP 9000 workstation, but everything
should work pretty much the same on other systems.


First, you have to tell the mail system that you want your mail to be
filtered through the procmail program.  You do this by creating a
one-line file called ".forward" in your home directory:

"| IFS=' '; /usr/local/bin/procmail -p"

The quotes are necessary.  Replace /usr/local/bin with the name of the
directory in which you've installed the procmail program.  Do a "chmod
644 .forward" to make sure that your mail software can read this file.


Now you can create a ".procmailrc" file in your which tells procmail how to
filter your mail.  Here are some excerpts from mine:

-------------------------------------------------------------------------------
# ~/.procmailrc
#
# Configuration file for procmail mail processing software
#
# General environment variables  (You may not need all of these; see the
# procmail(1) man page to find out what each one means.  Of course, the
# HOME variable should be set to your own home directory.)
#
HOME=		/Home/janzen
PATH=/usr/local/bin:/usr/ucb:/bin:/usr/bin:$HOME/bin
MAILDIR =	$HOME/Mail
LOGFILE =	$HOME/.procmaillog
LOCKFILE =      $HOME/Mail/procmail
TMP=		$HOME/tmp
SENDMAIL=	/usr/lib/sendmail
TMPFILE=	$TMP/procmail.$$
LOCKFILE=	$HOME/Mail/.procmail

#
# First, toss out junk mail
#

:2H
^From:.*root@idacom.hp.com
^Subject:.*cron
/dev/null

# The ":" introduces a new "recipe".  The "2" means that two expressions
# follow.  The "H" indicates that procmail should search the header only,
# ignoring case.  (The man page lists all kinds of other flags.)
#
# The first regular expression (regexp) matches a line containing "From:"
# at the start of the line ("^"), then any arbitrary characters (".*"),
# then "root@idacom.hp.com".
#
# Similarly, the second regexp matches a header line beginning with
# "Subject:" and containing the word "cron" anywhere in the subject.
#
# The line following the last expression tells procmail where to save
# the message.  In this case, I save it to /dev/null, throwing it away.
# (When you're first setting up procmail, I suggest saving unwanted
# messages to ~/junk or something, until you're satisfied that your
# recipes work as expected!)  
#
# You can also forward matched messages to another user (eg. !joeuser@host)
# or pipe them to a UNIX command (eg. |$HOME/bin/my_fancy_mail_processor).

:2H
^From:.*cypherpunks
detweiler
/dev/null

:2H
^From:.*cypherpunks
subscribe
/dev/null

#
# Now, sort mail from mailing lists into the proper folders
#
:1H
^From:.+cypherpunks
Cypherpunks

# Put all mail with a "From:" line containing the word "cypherpunks" into
# the file $HOME/Mail/Cypherpunks.

:1H
^TOcypherpunks
Cypherpunks

# "^TO" is shorthand for "^(To|Cc|Apparently-To):.*".  This is supposed to
# catch all destination addresses.

:1H
^Return-Path:.+cypherpunks
Cypherpunks

:1H
^From:.*pgpmip
PGPMIP

:1H
^From:.*wnet.edex.edu.au
PGPMIP

:1H
^From:.*Extropians
Extropians

:1H
^TOExtropians
Extropians
-------------------------------------------------------------------------------


One other thing: I like to get a summary each day, showing me what
procmail has done.  I do this by creating a script called procmailsummary.sh:

-------------------------------------------------------------------------------
#!/bin/sh
#
# Summarize the ~/.procmaillog file
#

# Use the LOGFILE variable if it's set; otherwise, use a default
LOGFILE=${LOGFILE:=$HOME/.procmaillog}

# Add a header to the message
# (This also avoids sending a message with a null body, which confuses Elm!)
echo "Subject: Procmail Summary"
echo " "

sort ${LOGFILE} | awk '
/^  Folder:/ {
        folder = $2;
        nbytes = $3;
        msgcount[folder] += 1;
        totalbytes[folder] += nbytes;
}

END {
        for (folder in msgcount)
                printf "Folder %s:\tsaved %d messages (%d bytes)\n", \
                        folder, msgcount[folder], totalbytes[folder];
}
'

if [ "$1" = "-clear" ]; then rm -f $LOGFILE; fi
-------------------------------------------------------------------------------


I then use crontab to schedule this script so that it's executed at 7AM
every morning, and mails me the result.  When I come in, I can see at a
glance which mailing lists have new messages.

(WARNING: If you're not running HP-UX, the crontab(1) command may not
exist on your system, or may work somewhat differently.  Also, on some
systems you must be root to use cron.  If you have problems, try "man
cron" or ask your sysadmin to give you a hand.)

  echo '0 7 * * * /usr/local/bin/procmailsummary.sh -clear | elm -s
    "Procmail Summary" janzen' | crontab

This should all be on one line.  Replace /usr/local/bin with the name of
the directory in which you've put the script. Make sure to do a "chmod
+x procmailsummary.sh", to make the script executable.  Finally, if you
are using a mailer other than elm, replace the "elm ... janzen" command
with something appropriate (eg. "mail janzen").

Now verify that the crontab command worked:

  crontab -l


Hope that helps...

--
Martin Janzen           janzen@idacom.hp.com
Pegasus Systems Group   c/o Hewlett-Packard, IDACOM Telecom Operation