ports/145301: [patch] /usr/local/etc/rc.d/hald startup delay problem

[Martin Birgmeier]

>Number:         145301
>Category:       ports
>Synopsis:       [patch] /usr/local/etc/rc.d/hald startup delay problem
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    freebsd-ports-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Fri Apr 02 18:00:17 UTC 2010
>Closed-Date:
>Last-Modified:
>Originator:     Martin Birgmeier
>Release:        7.2, 7.3, 8.0
>Organization:
MBi at home
>Environment:
FreeBSD gandalf.xyzzy 7.3-RELEASE FreeBSD 7.3-RELEASE #2: Wed Mar 31 19:49:46 CEST 2010     root at gandalf.xyzzy:/d/14.1/OBJ/FreeBSD/RELENG_7_3_0_RELEASE/src/sys/XYZZY  i386
>Description:
FreeBSD's startup command for hald, /usr/local/etc/rc.d/hald, uses a custom start_cmd which waits up to 60 seconds for getty to start.

This has a bad interaction when the X server is not started from getty, but using a (custom) startup script. In this case, the X server starts *before* getty, but the hald waits until *after* getty has started. This means that the X server does not see the devices offered by hald.

Before I finally tracked down the cause, I had the following in my X startup script in order to ensure that hald was running before the X server started:

i=0
while [ $i -lt 120 ] && ! /usr/local/bin/lshal > /dev/null 2>&1
do
    if [ $i -eq 0 ]
    then
        echo -n ' (please wait)'
    else
        echo -n .
    fi
    sleep 3
    : $(( i = i + 3 ))
done

This of course only worked because it waited *longer* than the maximum delay in /etc/local/etc/rc.d/hald (fortunately, the latter *did* include such a maximum (at 60 seconds), otherwise the script above would have timed out (after 120 seconds)).

The fix is simple: simply get rid of the custom startup command in /usr/local/etc/rc.d/hald.

>How-To-Repeat:
Try to run the X server before getty starts (using a script in /usr/local/etc/rc.d, usually).

>Fix:
The bandaid patch is attached. Of course, the whole body of the start_cmd can also be deleted, but this should be checked by whoever wrote this in the first place - I do not know for what reason the wait for getty was introduced originally.

Patch attached with submission follows:

--- sysutils/hal/files/hald.in.ORIG	2010-03-27 08:59:59.000000000 +0100
+++ sysutils/hal/files/hald.in	2010-04-02 19:26:15.000000000 +0200
@@ -25,7 +25,7 @@
 
 stop_postcmd="hald_postcmd"
 start_precmd="hald_precmd"
-start_cmd="hald_start"
+# start_cmd="hald_start"
 
 local_force_depend()
 {


>Release-Note:
>Audit-Trail:
>Unformatted: