aboutsummaryrefslogtreecommitdiffstats
path: root/net/netselect
diff options
context:
space:
mode:
authorDoug Barton <dougb@FreeBSD.org>2012-04-13 17:16:00 +0800
committerDoug Barton <dougb@FreeBSD.org>2012-04-13 17:16:00 +0800
commitb3ab9ed23a903a9b3d5c00a7b14e53e189062515 (patch)
treee1b18b547b586276bf2bf9c235728121887e43b6 /net/netselect
parent7984dbe60654a33bd255378bb0da54faaf684656 (diff)
downloadfreebsd-ports-gnome-b3ab9ed23a903a9b3d5c00a7b14e53e189062515.tar.gz
freebsd-ports-gnome-b3ab9ed23a903a9b3d5c00a7b14e53e189062515.tar.zst
freebsd-ports-gnome-b3ab9ed23a903a9b3d5c00a7b14e53e189062515.zip
The samba rc.d script uses some clever tricks to start (up to) 3 different
services using the same script. As a result it resets rcvar several times in order to process the options for each service. The changes I made on 2012/01/14 to facilitate the removal of set_rc_var() from HEAD were effective in the case where the WINBIND option was off (the case that I tested) because that causes the related portions of the rc.d script to be removed completely on install. However, if installed from a package, or installed using the the default OPTIONS, WINBIND is on, which caused the last known rcvar to be winbind_enable. Since the common case seems to be for users to use samba_enable (which only enables smb_and nmb_ by default) the fact that rcvar=winbind_enable, but that knob is off, caused the startup script to trip on a totally unrelated portion of rc.subr. So the fix is to move processing of the winbind_ stuff first, which leaves the last known rcvar as smb_enable. Since running nmb without smb is a very unlikely scenario, this should be safe for the common case, as well as safe if the user enables winbind_. Apologies all around for not catching this sooner, and thanks to the users who reported the problem and stuck with me while I debugged it. Bump PORTREVISION since this fix is needed for the common case, as configured for the package.
Diffstat (limited to 'net/netselect')
0 files changed, 0 insertions, 0 deletions