I've been unavailable to maintain this port for quite a while and thanks to some very capable people this port hasn't died of horrible bitrot. My hat's off to everyone who stepped in and kept this monstrosity up to date. I'm providing a mega-patch which updates the port to 0.44.2 and fixes the issue of LDAP support not getting built. Thus the following PRs can be closed: http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/60727 http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/60748 http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/61082 There's quite a few cosmetic changes to the port as well. I had filed <http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/33929> almost two years ago to register the UID/GID for 'courier'. <http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/33929> is still marked open and has led to the conflict with security/pf reported in <http://www.freebsd.org/cgi/query-pr.cgi?pr=ports/58115>. I had considered using mailnull when I was writing this port but rejected that idea beacuse that's the sendmail queue user. Installing a port with lots of suid binaries set to a user of an entirely different MTA is a Bad Thing IMHO. Presumably mailnull was thought up with the idea that no binaries would ever be owned by that user. Besides mail/courier/pkg-deinstall depends on the home directory of the user courier is installed as to be courier's queue directory not sendmail's. mail/courier/pkg-deinstall asks the user if the /var/spool/courier/ can be removed upon deinstall. And changing mailnull's home directory to accommodate courier would be a Bad Thing too. There's another unadvertised "feature" in having user <courier> with it's own home dir. /var/spool/courier/.noident is installed and provides a bit of security by obscurity by not advertising to the world which user courier is running as whe then inetd(8) has the auth service turned on. So how about we register in the Porter's Handbook UID/GID 465 as courier (taken from /etc/services smtp over TLS/SSL port number). files/patch-configure and files/patch-courier::configure are new with this patch and files/patch-courier::configure.in is removed. CONFLICTS has been expanded. WITH_GHOSTSCRIPT_AFPL support has been changed to honor GSPORT The WITH_LDAP issue happened because I was setting: CONFIGURE_ENV+= LDFLAGS="${LDFLAGS}" which was getting clobbered by <bsd.openssl.mk> doing the same thing so that in the end CONFIGURE_ENV would expand to: LDFLAGS="-L${LOCALBASE}/lib" LDFLAGS="-rpath=/usr/lib:${LOCALBASE}/lib" instead of the intended: LDFLAGS="-L${LOCALBASE}/lib -rpath=/usr/lib:${LOCALBASE}/lib" The subtlety which led to this bug was that I never actually declared LDFLAGS in a Makefile -- only in scripts/configure.courier Perhaps the behaviour in <bsd.openssl.mk> regarding the interaction between LDFLAGS and CONFIGURE_ENV should be documented if it isn't already. While I was at it I got rid of OpenLDAP 1.x support keeping only the OpenLDAP 2.x and now honor the WANT_OPENLDAP_VER tunable. WITH_TRANSPORT+=uucp changed since uucp is no longer part of the base system and NOUUCP is no longer a /etc/make.conf tunable for buildworld. WITH_EXPECT not being set now adds --disable-changepass which causes the suid authdaemon.passwd to not be installed. LIB_DEPENDS= fam.0:${PORTSDIR}/devel/fam is non-optional since the port will use the library if it finds it regardless of one's intention to use it or not. For instance if libfam is installed and one doe not want it used by courier but chosses the WITH_LDAP support than libfam will get sucked in as well. Then the user deletes fam and courier breaks because the user had no idea courier depended on it. So until courier gets a configure --without-fam tunable this should be a mandatory dependency. and last but not least etc/rc.d/courier.sh startup script has been rewritten to work with rc.subr which allows one to move it to /etc/rc.d/courier and have it just work. :) How-To-Repeat: cd /usr/ports/mail/courier make WITH_LDAP=yes install
Responsible Changed From-To: freebsd-ports-bugs->sergei I'll handle it.
State Changed From-To: open->closed Committed, thanks!