Re: [Linphone-developers] sound card block size different.

[Kosta Welke]

Simon Morlat wrote:
> Yes, this big 8192 size is the reason of the sound delay, that may also cause 
> echo and feedback.

Its due to the core_api/linphonecore.c, line 1346. By removing this
check, I was able to playback sound perfectly. It seems my system simply
lies about its blocksize.

Moreover, it seems that on my embedded system, the timestamp generation
is broken. Attached is a patch that works flawlessly on my system, while
the default behavior is broken. However, I didnt try it on my PC, yet.
If you are interested in this patch, please let me know, I can make a
diff against CVS, do a command line switch, or something like that.

HTH,
Kosta

--- linphone-ori/mediastreamer/msrtpsend.c      2005-11-14 13:10:21.000000000 
+0100
+++ linphone/mediastreamer/msrtpsend.c  2006-03-08 21:02:01.000000000 +0100
@@ -83,20 +83,23 @@
 
 guint32 get_new_timestamp(MSRtpSend *r,guint32 synctime)
 {
-       guint32 clockts;
+       guint32 clockts, mints, oldclock;
+       oldclock = r->ts;
        /* use the sync system time to compute a timestamp */
        PayloadType 
*pt=rtp_profile_get_payload(r->rtpsession->profile,r->rtpsession->send_pt);
        g_return_val_if_fail(pt!=NULL,0);
        clockts=(guint32)(((double)synctime * (double)pt->clock_rate)/1000.0);
+       mints = r->ts + (r->ts_inc / 4);
        ms_trace("ms_rtp_send_process: sync->time=%i 
clock=%i",synctime,clockts);
        if (r->flags & RTPSEND_CONFIGURED){
-               if 
(RTP_TIMESTAMP_IS_STRICTLY_NEWER_THAN(clockts,r->ts+(2*r->ts_inc) )){
-                       r->ts=clockts;
-               }
-               else r->ts+=r->ts_inc;
+               if (RTP_TIMESTAMP_IS_STRICTLY_NEWER_THAN(clockts, mints)){ /* 
kosta */
+                       r->ts = clockts;
+                       }
+               else r->ts = mints;
        }else{
                r->ts=clockts;
        }
+       //fprintf(stderr, "inc: %i\n", r->ts - oldclock);
        return r->ts;
 }