Hello, Please, do not take discussion out off the list. It may be interesting for community, not for you only.
On 1/17/07, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
since november 16th 2006, I have a problem with calls from sipxtapi to sipxtapi. After 5-10 seconds I hear no sound on the headset. If I test a call from sipxtapi to the nero SIPPS softphone I have no problem. I found, that on this time if I dont't have sound on the headset, the funktion "MpdSipxPcmu::decodeIn" returns with "-1" for all RTP-packets. In my "ugly" hack, I ignore this "-1" and do so that all is OK! Now I have sound on the headset, and my last test with a connection from sipxtapi to sipxtapi with more than one hour connection time was OK!
I named this "hack" because it does not solve the root of problem. You've just removed check for out-of-order packets. As I understand it, problem is sequence number wrapping. When sequence number of RTP stream wrap to zero mLastSeqNo remain high and is always greater then iSeqNo, so '-1' is returned. Please test with patch attached to this mail. It should fix this problem. I commited this patch to sipXtapi branch, sipXtapi-media-update branch will be updated later. It is now targeted video support only and is freezen for some (I hope short) time.
The 2. "ugly" hack is, that I changed in "MPJitterBuffer.h" this: static const int JbQueueSize = (8 * (2 * 960)); // innovaphone 200 - tiptel I'm waiting for a proper solution from sipfoundry, because I don't find the problem in the RTP sources.
This is known bug: http://track.sipfoundry.org/browse/XMR-80 I have great idea how to fix it in generic way, but have no time for implementation now :( MpJitterBuffer should ask codec about length of decoded data for received RTP packet and then decide is tere enough space at the end of buffer. And if no, wrap to buffer begin. It would be great, if someone from community will write patch. -- Regards, Alexander Chemeris.
RtpSeqNumCompareFix.diff
Description: Binary data
_______________________________________________ sipxtapi-dev mailing list [email protected] List Archive: http://list.sipfoundry.org/archive/sipxtapi-dev/
