I found this bug with my baselayout-1.12-10-r4(r5). I know baselayout-2 will be out soon. But please glance this bug for a little, and then decide whether it should be fixed. Thank you! It is in /lib/rcscript/net/arping.sh, about function "arping_start". I paste the code below: arping_start() { local iface="$1" gateways= x= conf= i= local ifvar=$(bash_variable "${iface}") einfo "Pinging gateways on ${iface} for configuration" gateways="gateways_${ifvar}[@]" if [[ -z "${!gateways}" ]] ; then eerror "No gateways have been defined (gateways_${ifvar}=( \"...\"))" return 1 fi eindent for x in ${!gateways}; do local -a a=( ${x//,/ } ) local ip="${a[0]}" mac="${a[1]}" extra= if [[ -n ${mac} ]] ; then mac=$(echo "${mac}" | tr '[:lower:]' '[:upper:]') extra=(MAC ${mac}) fi vebegin "${ip} ${extra}" if arping_address_exists "${iface}" "${ip}" "${mac}" ; then for i in ${ip//./ } ; do if [[ ${#i} == "2" ]] ; then conf="${conf}0${i}" elif [[ ${#i} == "1" ]] ; then conf="${conf}00${i}" else conf="${conf}${i}" fi done [[ -n ${mac} ]] && conf="${conf}_${mac//:/}" veend 0 eoutdent veinfo "Configuring ${iface} for ${ip} ${extra}" configure_variables "${iface}" "${conf}" # Call the system module as we've aleady passed it by .... # And it *has* to be pre_start for other things to work correctly system_pre_start "${iface}" t="config_${ifvar}[@]" # Only return if we HAVE a config that doesn't include # arping to avoid infinite recursion. if [[ " ${!t} " != *" arping "* ]] ; then config=( "${!t}" ) t="fallback_config_${ifvar}[@]" fallback_config=( "${!t}" ) t="fallback_route_${ifvar}[@]" fallback_route=( "${!t}" ) config_counter=-1 return 0 fi fi veend 1 done eoutdent return 1 } And I think bug is from this segment below: # Only return if we HAVE a config that doesn't include # arping to avoid infinite recursion. if [[ " ${!t} " != *" arping "* ]] ; then config=( "${!t}" ) t="fallback_config_${ifvar}[@]" fallback_config=( "${!t}" ) t="fallback_route_${ifvar}[@]" fallback_route=( "${!t}" ) config_counter=-1 return 0 fi The code above overrides three variables config, fallback_config and fallback_route and set config_counter to -1 in order to restart the main start loop with the new three variables, which is in iface_start, the main function for network-up in /etc/init.d/net.lo .(The three variables is used by the function) The problem is the 2nd variable may be miswritten, it should be fallback not fallback_config, because in /etc/init.d/net.lo, in iface_start, there is not a variable named fallback_config but named fallback(the other two is config and fallback_route). And in /etc/conf.d/net.example, fallback_config_xxx is not refered either, but many examples of fallback_xxx. So I think in the above code, the "fallback_config" does be "fallback". I think the users who set fallback_GATEWAY(e.g. fallback_192168000001) wouldn't get the expected result which should have been caused by fallback_GATEWAY variable. BTW: I know variable "fallback" is corresponding to "config". So in my opinion, "fallback_config" is more understandable then "fallback". However, fallback_config is a bug here. And In addition, I think "ipconfig" or "ip" is more exact than "config", because I found the all thing variable config represents is all about ip address config, either a static ip string, or dhcp, or apipa, or... And the other configs, such as dns, route, essid, there are special variables to stand for. Only config=noop is meaningful to all network config -- no operation to all network config if interface has been already up. The name "config" is a little bigger than what it represents. So..., I think in baselayout2, maybe it would have some changing. Oh, as I have seen the sources of baselayout-2 about net-script, I think it's better if adding more comments for each function, like what baselayout-1 does.
I've fixed this in our svn repo on the off chance we do another 1.12 release. Thanks!
