[asterisk-users] Simultaneous caller/callee hangup; hangup extensions execute only once; unable to determine if destination channel up

[John Hurst]

Hello

This is a question regarding whether there's any way within hangup 
extensions to determine whether the caller or callee leg (or both) of a 
bridged call has hung up.  The test case I have is running under 
Asterisk 1.8.17.0, but the behaviour is observed in 1.8.18.0 (and also 
1.6.2.18).

Within the dialplan, the Dial() application with the "F" flag, so that 
once the caller hangs up, the dialplan jumps to a new priority which 
enables the called party to enter some digits which describe the outcome 
of the call.  Also, the "g" flag is used to attempt to continue 
execution of the dialplan if the called party hangs up.

Minimally, the dialplan is covered by the following:

[test]
exten => _1000,1,Set(_CALLER_HUNGUP=false)
exten => _1000,2,Set(_CALLEE_HUNGUP=false)
exten => _1000,3,Dial(SIP/${EXTEN},60,CgF(test^1000^10))
exten => _1000,4,Set(_CALLEE_HUNGUP=true)

exten => _1000,10,Set(_CALLER_HUNGUP=true)
exten => _1000,11,AGI(afterCallWork.agi)

exten => h,1,NoOp(${CALLER_HUNGUP})
exten => h,2,NoOp(${CALLEE_HUNGUP})
exten => h,3,AGI(postCall.agi)


Normally, the hangup extensions execute twice: once when the caller 
hangs up, then once more when the called party hangs up, either during 
or after the execution of afterCallWork.agi.  This second call is 
important so that clean up can be performed.

However, if the two parties hang up simultaneously (or within a 
split-second of each other), I often see only one execution of the 
hangup extensions.  Stranger still, the hangups can occur so close to 
each other that execution of the hangup extension occurs without the 
either the priority 4 or priority 10 steps being executed (it can be 
difficult replicate this, but inserting a Wait(1) call at priority 4 and 
another at priority 10 can help here).

In such cases, I see the output from the two NoOps as false and false.  
(This is difficult to replicate because of the precise timing it 
requires - it is easy if you insert Wait(1) at priority 4 and 10, but 
whether this is valid or not is debatable.  I can replicate this issue 
with just the dialplan above on a slowish server).

So I need to be able to query the status of the other channel from 
within postCall.agi, because if both parties have hung up, I may only 
get one execution of the hangup extensions, and I can go ahead and 
perform the cleanup.  Is this possible?  I've tried calling CHANNEL 
STATUS for the destination channel within postCall.agi, but even when 
the destination channel is definitely still up, the call returns an 
error "511 Command Not Permitted on a dead channel" (presumably because 
the current (caller) channel has hung up).

I can't find anything that I can use within the execution of the hangup 
extensions for the caller to determine whether the destination channel 
is still up.

Is it a bug that I only get the one call to the hangup extensions when 
both caller and callee channels hangup so close to each other that 
neither the "F" nor "g" flags have the desired effect?

Thanks in advance for any help.

John Hurst