On Fre, 2008-11-14 at 18:29 +0100, Wolfgang Pichler wrote:
> Michael Neuhauser schrieb:   
> > I'm pretty sure that the wrong User information layer 1 byte is to
> > blame. Especially since I've tested it with my libpri (version 1.2.2 +
> > bristuff 0.3.0.pre1n, already pretty old) and for a digital call, the
> > user-layer 1 byte is *not* there (Bearer Capability is only 4 bytes
> > long, not 5 as in your SETUP). Libpri contains the following code:
> >
> >         static FUNC_SEND(transmit_bearer_capability)
> >         ...
> >         	if ((tc & PRI_TRANS_CAP_DIGITAL)&&(pri->switchtype == PRI_SWITCH_EUROISDN_E1)) {
> >         		/* Apparently EuroISDN switches don't seem to like user layer 2/3 */
> >         		return 4;
> >         	}
> >         	/* add userlayer 1 info */
> >
> i do have asterisk 1.4.22 and libpri-1.4.7 - so quit new...
> i can see this code in q931.c in the FUNC_SEND - but the user-layer-1 is 
> still there...

You have to look closer, the code is slightly different in libpri-1.4.7:

                 if ((tc & PRI_TRANS_CAP_DIGITAL) && (pri->switchtype == PRI_SWITCH_EUROISDN_E1) &&
                                (call->transmoderate == TRANS_MODE_PACKET)) {
                                /* Apparently EuroISDN switches don't seem to like user layer 2/3 */

There is an extra check for (call->transmoderate == TRANS_MODE_PACKET)
and since your call has a 'circuit' transfer mode, the condition is not
met and the userl1 value is added to the bearer caps. But from the
discussion with Klaus about video calls I now think that this check is
not the cause/cure for your problem.

In transmit_bearer_capability() the userl2/l3 are not added if they are
invalid (< 0), but the code for userl1 is different because it adds the
userl1 no matter what its value is, as long as transmoderate !=
TRANS_MODE_PACKET.

> i think it is not a good idea to put a debug statement into there to see 
> what is going on...

Why not? That's what pri_message() is for (among other things). But I
think it is not needed.

> how can i get rid of the user-layer-1 bit ?

In my opinion, you should check if not adding a userl1 value < 0 in
transmit_bearer_capability() solves your problem, i.e., something like:

         	if (call->transmoderate != TRANS_MODE_PACKET) {
        +		if (call->userl1 == -1) /* Do not add an invalid user-info-layer-1 value. */
        +			return pos + 2;
         		/* If you have an AT&T 4ESS, you don't send any more info */

Even if this is not valid for all possible cases (can there be a
userl2/l3 value if userl1 is invalid?), it should allow to test if the
userl1 value is the cause of the rejected call.

Regards,
	Mike

-- 
Dr. Michael Neuhauser                              mailto:mike@firmix.at
Firmix Software GmbH                                  sip:mike@firmix.at
Vienna/Austria/Europe                               tel:+43-1-7890849-30
Linux Development and Services                     http://www.firmix.at/