WebRTC: Solving the audio explosion issue in RTMP to RTC.

[PieerePi]

WebRTC: Solving the audio explosion issue in RTMP to RTC.

---

TRANS_BY_GPT3

[codecov-io]

Codecov Report

Merging #2011 into 4.0release will increase coverage by 0.00%.
The diff coverage is 0.00%.

@@             Coverage Diff             @@
##           4.0release    #2011   +/-   ##
===========================================
  Coverage       57.17%   57.17%           
===========================================
  Files             125      125           
  Lines           52707    52707           
===========================================
+ Hits            30134    30135    +1     
+ Misses          22573    22572    -1     

Impacted Files	Coverage Δ
trunk/src/app/srs_app_rtc_codec.cpp	0.00% <0.00%> (ø)
trunk/src/protocol/srs_service_utility.cpp	72.97% <0.00%> (+0.54%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
| Δ = absolute <relative> (impact), ø = not affected, ? = missing data |
Powered by Codecov. Last update a28f985...aca4f09. Read the comment docs.

TRANS_BY_GPT3

[wnpllrzodiac]

old code only support mono audio?

[PieerePi]

@wnpllrzodiac

No, the incoming audio is stereo.

It's a very common case, you can reproduce this issue easily. Use OBSStudio to publish stream to SRS, and use WebRTC to play.

(gdb) p *frame_
$7 = {data = {
0x2199b00 ""...,
0x219bb10 ""..., 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}, linesize = {8192, 0, 0, 0, 0, 0, 0, 0}, extended_data = 0x20e4ec0, width = 0,
height = 0, nb_samples = 1024, format = 8, key_frame = 1, pict_type = AV_PICTURE_TYPE_NONE, sample_aspect_ratio = {num = 0, den = 1},

The size of first plane is 8192, but only has 1024 samples, that is 10244(FLTP) = 4096 bytes in total.
The size of second plane is 8192, but only has 1024 samples, that is 10244(FLTP) = 4096 bytes in total.

Since decoded audio frame is planar format, the linesize is greater than the size of useful data. In old code, the second channel data is lost and filled by another half useless data in first plane (that's noise).

@winlinvip Please have a look and merge this PR, thanks!

[winlinvip]

Thank you very much, we have already reviewed this PR and there is indeed this issue. I will look into merging it as soon as possible.

TRANS_BY_GPT3

[PieerePi]

Okay, thank you!
There is also a PR #2012, can you help take a look when you have time?

Also, you mentioned there is a discussion group for GB28181. How can I join it?

TRANS_BY_GPT3

[winlinvip]

See also #2172

[winlinvip]

For #2106