< Previous by Date | Date Index | Next by Date > |
< Previous in Thread | Thread Index | Next in Thread > |
I didn’t see the
BaseCreator last request being updated, so I’m not sure where to find the
cseq for this re-invite. In looking at sending
the re-invite the only info I could see being updated is: mRequests[msg->header(h_CSeq).sequence()]
= msg; in Dialog::send. The problem with this is that the UAS has modified the
tag so dum can’t find any existing dialog, so the cseq of the initial
re-invite request will need to be stored somewhere at the dialogset
level? What about adding the cseq info to mLastRequest in the
BaseCreator when sending the initial re-INVITE? Thanks, -justin From: Scott Godin
[mailto:slgodin@xxxxxxxxxxxx] I don’t think
this is quite right – it will not allow multiple 200’s to an
initially forked invite request to create dialogs. I think we should
check the CSeq in response and only create a dialog if it matches the CSeq in
the base creator. This will make sure we only create multiple dialogs for
an initial UAC request, but not mid dialog requests. Scott From: Justin Matthews
[mailto:jmatthewsr@xxxxxxxxx] Further on this issue, although the patch
fixes the exception and drops the 200 for the case where dum was initially the
UAS, this scenario will still cause issues if dum was initially the UAC (since
getLastRequest will contain a valid contact). Propose the following, please note that I
am not sure that “mState == Established” is the correct check here,
please advise. In DialogSet.cxx: if (dialog == 0) {
if (msg.isRequest()
&& msg.header(h_RequestLine).method() == CANCEL)
{
dispatchToAllDialogs(msg);
return;
}
if (msg.isResponse())
{
int code =
msg.header(h_StatusLine).statusCode();
if( code > 100 && mState == Established )
{
InfoLog(<< "Cannot
create a dialog, mid-dialog fork not allowed.");
return;
} Thanks, -justin From: Scott Godin [mailto:slgodin@xxxxxxxxxxxx]
That is pretty odd behaviour by the far
end. Mid-dialog forking is not allowed (3261 sec 14.1). If we
detect a 200 with a different to tag and the dialogset is a UAS dialogset (ie.
no base creator) - then we should either: 1. Ack the 200 and then send a
bye. or 2. Just ignore the 200 - don't proceed to create a
dialog for it - but log an error. I'm leaning towards 2. Scott From:
resiprocate-devel-bounces@xxxxxxxxxxxxxxxxxxxx on behalf of Justin Matthews Exception occurs during following scenario: DUM receives a call and establishes a session as the UAS. DUM sends re-INVITE request DUM receives 200 ok response with a different to: tag then
the established session. The dialogset is found, but dialogset.cxx tries to create a
new dialog from this 200 and calls the following code: Dialog.cxx, in Dialog constructor: mLocalContact = creator->getLastRequest()->header(h_Contacts).front(); I believe that getLastRequest does not contain a
contact because this call was established from an inbound session. For now I just patched it this way, is there a
better way? If not I will commit this. ------------------- BaseCreator* creator = mDialogSet.getCreator(); //assert(creator);//
!jf! throw or something here //assert(creator->getLastRequest()->exists(h_Contacts)); //assert(!creator->getLastRequest()->header(h_Contacts).empty()); if(
NULL == creator ) { ErrLog(<< "BaseCreator is null for
DialogSet"); ErrLog(<<
response); throw Exception("BaseCreator is null for DialogSet",
__FILE__, __LINE__); }
SharedPtr<SipMessage> lastRequest(creator->getLastRequest()); if(
NULL == lastRequest ||
!lastRequest->exists(h_Contacts) ||
lastRequest->header(h_Contacts).empty()) { ErrLog(<< "No contact available for
dialogset's last request"); ErrLog(<<
response); throw Exception("No contact available for dialogset's last
request", __FILE__, __LINE__); } mLocalContact =
creator->getLastRequest()->header(h_Contacts).front(); mRemoteTarget = contact; ---------------------- Thanks, -justin |