wpa_cli event handling (revisited)

[Kel Modderman]

Kel Modderman wrote:
> Roy Marples wrote:
>   
>> On Thursday 25 May 2006 11:21, Kel Modderman wrote:
>>   
>>     
>>> Roy Marples wrote:
>>>     
>>>       
>>>> On Thursday 25 May 2006 09:00, Kel Modderman wrote:
>>>>       
>>>>         
>>>>> The patch attached to the previous mail was not working as expected. So
>>>>> i guess what I am really asking here is; should 'wpa_cli terminate'
>>>>> cause a DISCONNECTED event if currently associated?
>>>>>         
>>>>>           
>>>> I'm guessing not as if wpa_cli is terminated then the thing that is
>>>> terminating it should be taking down the rest of the interface too.
>>>>       
>>>>         
>>> Well, wpa_cli is terminating itself (wpa_cli -i iface terminate), and if
>>> a wpa_action script is active, and the iface is associated/connected, I
>>> thought that maybe the wpa_action script should be taking down the iface
>>> too.
>>>     
>>>       
>> Then you may have a race. Gentoo init scripts can take care of this, but other 
>> distro's may not be so flexible.
>>
>> I would say that if wpa_supplicant terminates then it should send a DISCONNECT 
>> to wpa_cli causing the script to run. However, if I terminate wpa_cli I may 
>> not want the DISCONNECT even to happen as wpa_supplicant which controls the 
>> link may be running still so the link is still valid.
>>
>>   
>>     
>
> Okay, I totally agree with all of that.
>
>   


And many thanks for your advice, Roy. Your fine "software engineering" 
advice coupled with the new id_str identifier from Jouni is more than 
enough for me to implement some roaming with debian's ifupdown and 
wpa_supplicant (and ifupdown *does* have some severe limitations when 
compared to baselayout, maybe some advantages too ;-).

Thanks, Kel.