I'm helping to administer a RedHat 5.0 i386 system. We recently updated to
Sendmail 8.9.3, using FEATURE(local_procmail). The procmail version was
probably whatever came originally with RH5.0. All worked fine at this
point, with users' .procmailrc being executed correctly when present. All
still fine after some reboots.
I then did a crop of rpm security updates using the RedHat site:
libc-5.3.12-28.i386.rpm
procmail-3.13.1-1.i386.rpm
nfs-server-clients-2.2beta40-1.i386.rpm
lpr-0.35-0.5.2.i386.rpm
pine-3.96-7.1.i386.rpm
mutt-0.95.4us-0.i386.rpm
sysklogd-1.3.31-0.5.i386.rpm
bash-1.14.7-11.i386.rpm
SysVinit-2.74-4.i386.rpm
imap-4.1.final-1.i386.rpm
Note that procmail 3.13.1 is among these. After these updates (no way to be
sure which was responsible) a user .procmailrc stopped functioning. To make
a long story short, it could be made to work only by removing group-write
permission on it. The same was true calling procmail from a .forward, but
at least that produced the "Address ... is unsafe for mailing to programs"
error, which gave me a handle with which to search the archives for this
list.
Nothing in our Sendmail .mc or .cf files mentions smrsh, so this
no-group-write requirement seems to be coming from something else that was
installed in the update sequence above. I strongly suspect it was procmail
3.13.1, but it's odd that it would be more restrictive about its rc files
than Sendmail itself.
Anyway, this is probably worth putting into an update of the Bronto-FAQ,
which currently references only smrsh as a consideration, as well as the
Original FAQ, which refers only to problems with Sendmail versions older
than 6.x.
My thanks to the maintainer of the list archive for the excellent search
engine. Please cc me on any responses, as I am no longer on this list.
--
smrsh has been superceded by ogpu, by order of Comrade Krushchev