Bug 8755 - initscripts-4.83-1.i386.rpm makes sendmail, etc. fail to start
Summary: initscripts-4.83-1.i386.rpm makes sendmail, etc. fail to start
Keywords:
Status: CLOSED RAWHIDE
Alias: None
Product: Red Hat Raw Hide
Classification: Retired
Component: initscripts
Version: 1.0
Hardware: i386
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Bill Nottingham
QA Contact:
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2000-01-23 00:01 UTC by dunwoody
Modified: 2014-03-17 02:12 UTC (History)
2 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2000-01-24 05:19:39 UTC
Embargoed:


Attachments (Terms of Use)

Description dunwoody 2000-01-23 00:01:15 UTC
I installed initscripts-4.83-1.i386.rpm and after rebooting, I noticed
that some of the services started from /etc/rc.d/init.d had failed
to start, including sendmail and nfs.  I ran

	/etc/rc.d/init.d/sendmail start

manually, and it still didn't start.  Finally I ran
/usr/sbin/sendmail -bd -q1h, and sendmail worked fine.

I backed off to initscripts-4.82-1.i386.rpm and the problem went
away.  Probably something trivially wrong with one of the sourced
scripts, but I didn't have time to investigate further.

I saw this on a system running rpms that were all current with the
latest RawHide rpms as of 2000/01/22 1200PST.

Comment 1 Bill Nottingham 2000-01-24 05:19:59 UTC
Yup, known bug, fixed in initscripts-4.84.

Comment 2 Jonathan Kamens 2000-01-26 00:38:59 UTC
I suppose the fix in initscripts-4.84 is functional, but it's not how I would
have fixed it.  In fact, it's not how I *did* fix it :-).

I realize that you need to have the "for apid in $pidlist" code in daemon and
killproc because if you put it in pidofproc it's actually running in a
subprocess and therefore doesn't do the right thing (i.e., it keeps a pid in the
list which shouldn't be in the list).  But this special-case code is gross and
doesn't fix the underlying problem that pidofproc is returning bogus
information.

Here's how I fixed it:

--- functions	2000/01/25 22:38:54	1.1
+++ functions	2000/01/25 23:47:20
@@ -12,6 +12,27 @@
 # First set up a default search path.
 export PATH="/sbin:/usr/sbin:/bin:/usr/bin:/usr/X11R6/bin"

+# Unfortunately, the "pidof -x" invocations below will find this
+# script if it has the same name as the process being checked or
+# killed.  To avoid this, we want to give pidof a list of process IDs
+# to ignore.  More unfortunately, this is not as simple as just
+# telling it to ignore our PID, because we may be in a subprocess of
+# the actual script, e.g., because of a shell function run inside
+# backquotes.  Even more unfortunately, such a subprocess returns the
+# PID of the base process when $$ is evaluated, rather than returning
+# the PID of the subprocess, so there's no easy way to get the PIDs of
+# such subprocesses to exclude them.  To surmount these obstacles, use
+# some awk magic to get all the PIDs in a chain from the current PID
+# back to $$.
+
+AWK_PROG='NR == 1 { self=$1 }
+{ parent[$1] = $2 }
+END { while (self != upto) { self=parent[self]; print "-o " self } }'
+
+exclude_pids () {
+    ps -A --format '%p %P' | awk "$AWK_PROG" upto=$$ /dev/pid -
+}
+
 # Get a sane screen width
 [ -z "$COLUMNS" ] && COLUMNS=80

@@ -170,7 +191,7 @@
 	fi

 	# Next try "pidof"
-	pid=`pidof -x $1`
+	pid=$(pidof $(exclude_pids) -x $1)
 	if [ "$pid" != "" ] ; then
 	        echo $pid
 	        return 0
@@ -185,7 +206,7 @@
 	fi

 	# First try "pidof"
-	pid=`pidof -x $1`
+	pid=`pidof $(exclude_pids) -x $1`
 	if [ "$pid" != "" ] ; then
 	        echo "$1 (pid $pid) is running..."
 	        return 0


Note You need to log in before you can comment on or make changes to this bug.