rc.d/resolv currently not being run under certain circumstances

[Doug Barton]

Due to the fact that scripts which are ordered before 
early_late_divider in the second run are not executed (see rc), if a 
script gets reordered into the early phase as a result of including 
the local startups, it won't be run at all. On at least one of my 
systems that's happening to resolv (in both HEAD and RELENG_6). It 
goes from close to last in rcorder (after securelevel) to absolutely 
first after you include the local startups I have (nothing exciting, 
just a few ports).

Currently resolv is a little thin in terms of ordering hints. It 
contains only PROVIDE and KEYWORD, and nothing REQUIREs it. Can 
someone who knows what this script is for (brooks?) add either a 
REQUIRE in or for resolv, or both? The CVS logs seem to indicate that 
it's related to diskless booting, so I don't want to touch it if I 
don't have to. :)

FWIW, I use the attached patch to rc to debug this problem.

-- 

     This .signature sanitized for your protection
-------------- next part --------------
--- /etc/rc	Sat May  6 21:00:23 2006
+++ rc	Sat Mar 17 16:45:13 2007
@@ -85,8 +47,12 @@
 #
 files=`rcorder ${skip} /etc/rc.d/* 2>/dev/null`
 
+# XXX
+rm -f rc.early* rc.late
+
 for _rc_elem in ${files}; do
-	run_rc_script ${_rc_elem} ${_boot}
+	#run_rc_script ${_rc_elem} ${_boot}
+	echo $_rc_elem >> rc.early1
 
 	case "$_rc_elem" in
 	*/${early_late_divider})	break ;;
@@ -107,15 +73,21 @@
 _skip_early=1
 for _rc_elem in ${files}; do
 	case "$_skip_early" in
-	1)	case "$_rc_elem" in
+	1)
+		echo $_rc_elem >> rc.early2
+		case "$_rc_elem" in
 		*/${early_late_divider})	_skip_early=0 ;;
 		esac
 		continue
 		;;
 	esac
 
-	run_rc_script ${_rc_elem} ${_boot}
+	echo $_rc_elem >> rc.late
+
+	#run_rc_script ${_rc_elem} ${_boot}
 done
+
+diff -u rc.early*
 
 echo ''
 date