Hi all, I've just begun to look into pjsip and have encountered a problem that I'm trying to understand. Some users of our voip application reports missing audio in one direction in some rare cases during a phone call. If I compare the logs from the faulty calls with logs from calls that works as expected, I get these differences: === A call that works === 2015-06-01 10:57:45,900 |5| 10:57:45.900 pa_dev.c !Recorder thread started 2015-06-01 10:57:45,901 |5| 10:57:45.900 ec0x7fcfa42e07 Prefetching.. 2015-06-01 10:57:45,901 |5| 10:57:45.901 pa_dev.c !Player thread started 2015-06-01 10:57:45,918 |5| 10:57:45.918 ec0x7fcfa42e07 Prefetching.. === A call that fails === 2015-06-02 09:14:31,948 |5| 09:14:31.948 pa_dev.c !Player thread started ?... 2015-06-02 09:14:32,218 |5| 09:14:32.218 ec00BE72C0 55 samples reduced, buf_cnt=1226 2015-06-02 09:14:32,238 |5| 09:14:32.238 ec00BE72C0 112 samples reduced, buf_cnt=1274 2015-06-02 09:14:32,258 |5| 09:14:32.258 ec00BE72C0 288 samples reduced, buf_cnt=1146 2015-06-02 09:14:32,278 |5| 09:14:32.278 ec00BE72C0 26 samples reduced, buf_cnt=1280 2015-06-02 09:14:32,298 |5| 09:14:32.298 ec00BE72C0 319 samples reduced, buf_cnt=1121 2015-06-02 09:14:32,308 |5| 09:14:32.308 ec00BE72C0 55 samples reduced, buf_cnt=1226 2015-06-02 09:14:32,338 |5| 09:14:32.338 ec00BE72C0 112 samples reduced, buf_cnt=1274 2015-06-02 09:14:32,358 |5| 09:14:32.358 ec00BE72C0 288 samples reduced, buf_cnt=1146 2015-06-02 09:14:32,378 |5| 09:14:32.378 ec00BE72C0 26 samples reduced, buf_cnt=1280 ? ?(hundreds of equal lines) ?... 2015-06-02 09:14:45,377 |5| 09:14:45.377 ec00BE72C0 319 samples reduced, buf_cnt=1121 2015-06-02 09:14:45,387 |N| 09:14:45.387 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,387 |5| 09:14:45.387 ec00BE72C0 1 samples reduced, buf_cnt=1280 2015-06-02 09:14:45,417 |N| 09:14:45.417 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,417 |5| 09:14:45.417 ec00BE72C0 319 samples reduced, buf_cnt=1121 2015-06-02 09:14:45,437 |N| 09:14:45.437 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,437 |5| 09:14:45.437 ec00BE72C0 1 samples reduced, buf_cnt=1280 2015-06-02 09:14:45,457 |N| 09:14:45.457 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,457 |5| 09:14:45.457 ec00BE72C0 319 samples reduced, buf_cnt=1121 2015-06-02 09:14:45,477 |N| 09:14:45.477 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,477 |5| 09:14:45.477 ec00BE72C0 1 samples reduced, buf_cnt=1280 2015-06-02 09:14:45,497 |N| 09:14:45.497 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,497 |5| 09:14:45.497 ec00BE72C0 319 samples reduced, buf_cnt=1121 2015-06-02 09:14:45,507 |N| 09:14:45.507 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,507 |5| 09:14:45.507 ec00BE72C0 1 samples reduced, buf_cnt=1280 2015-06-02 09:14:45,537 |N| 09:14:45.537 Master/sound Underflow, buf_cnt=0, will generate 1 frame 2015-06-02 09:14:45,537 |5| 09:14:45.537 ec00BE72C0 291 samples reduced, buf_cnt=1149 ?(hundreds of equal lines) ?... I'm curious about the missing line "Recorder thread started" from the faulty call log; from the logs it looks like there is a recorder that triggers the "samples reduces" lines. Also, it looks like the recorder callback is writing frames to the delay buffer for 'ec00BE72C0' and the player callback is fetching frames from the delay buffer for 'Master/sound'? Has anybody experience from this behaviour? Best regards, Martin -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.pjsip.org/pipermail/pjsip_lists.pjsip.org/attachments/20150604/a81192f6/attachment.html>
