>>>>> "CW" == Chan Wilson <email@example.com> writes:
CW> Hmmm. While I don't mind in the least people continuing on whilst I'm
CW> swamped with work and life, I am a bit tweaked that you (Jason) didn't
CW> send me a personal note about this before blasting the whole list.
Sorry; I took a lack of response to any of my other personal messages and a
lack of response to any of my messages sent to the list as evidence that
you were out of pocket. My intent, as with the 1.94prebeta2 patch, was to
gather, integrate and test patches and ship them off to you en masse. If
no response was received, then that was a bridge that would have to be
crossed at that time.
>> My own patch to Makefile changing the default permissions on $(W_HOME)
>> to 751. This will allow Sendmail to execute the wrapper even if the
>> directory isn't in group daemon. Nobody saw any security implications
>> when we last discussed this.
>> My own patch adding -oee to the $mailer command lines, to avoid those
>> weird nonzero exit codes from Sendmail.
CW> Don't have these.
--- Makefile.orig Mon Nov 18 14:10:26 1996
+++ Makefile Mon Nov 18 14:10:36 1996
@@ -164,7 +164,7 @@
- @test -d $(W_HOME) || (mkdir $(W_HOME); chmod 750 $(W_HOME))
+ @test -d $(W_HOME) || (mkdir $(W_HOME); chmod 751 $(W_HOME))
@-$(CHOWN) $(W_CHOWN) $(W_HOME) || \
( echo ""; \
echo "Hmmm, if $(W_HOME) doesn't have an owner and group of"; \
--- sample.cf.orig Mon Nov 18 14:12:00 1996
+++ sample.cf Mon Nov 18 14:12:48 1996
@@ -50,8 +50,8 @@
# can be had by playing with this variable. It is perfectly safe to use
# $sender, but the others are insecure.
-$mailer = "$sendmail_command -oi -f\$sender";
-$bounce_mailer = "$sendmail_command -oi -f\$sender -t";
+$mailer = "$sendmail_command -oi -oee -f\$sender";
+$bounce_mailer = "$sendmail_command -oi -oee -f\$sender -t";
# You can special case the mailer used to deliver outbound mail as follows:
The rationale behind the latter patch was information from the experts on
comp.mail.sendmail to the effect that sendmail will exit with nonzero
status without giving useful information on STDOUT at various random and
unpredictable times. Setting -oee will cause sendmail to always exit zero
in these circumstances and to mail the error messages to the proper place.
CW> The one bit that none of the patch sets (Dave, Jason, or mine) address
CW> is the install bit. Dave has sent me an install.sh, and I'll rework
CW> the Makefile to use that.
I had approached it based on the minimal modification policy, but it really
does need to get fixed. Everything else uses an install.sh so there really
isn't a reason not to. (Except for the fact that someone has to rewrite
the Makefile again.)