On Mon, 2016-01-18 at 13:06 +0530, arun at accosted.net wrote: > From: Arun Raghavan <git at arunraghavan.net> > > Needed for upcoming beamforming code. > --- >  src/modules/echo-cancel/echo-cancel.h |  2 +- >  src/modules/echo-cancel/webrtc.cc     | 14 ++++++++------ >  2 files changed, 9 insertions(+), 7 deletions(-) > > diff --git a/src/modules/echo-cancel/echo-cancel.h b/src/modules/echo-cancel/echo-cancel.h > index 4693516..613f7e3 100644 > --- a/src/modules/echo-cancel/echo-cancel.h > +++ b/src/modules/echo-cancel/echo-cancel.h > @@ -65,7 +65,7 @@ struct pa_echo_canceller_params { >               * to C++ linkage. apm is a pointer to an AudioProcessing object */ >              void *apm; >              int32_t blocksize; /* in frames */ > -            pa_sample_spec rec_ss, play_ss; > +            pa_sample_spec rec_ss, play_ss, out_ss; >              bool agc; >              bool trace; >              bool first; > diff --git a/src/modules/echo-cancel/webrtc.cc b/src/modules/echo-cancel/webrtc.cc > index 5741f45..5f00286 100644 > --- a/src/modules/echo-cancel/webrtc.cc > +++ b/src/modules/echo-cancel/webrtc.cc > @@ -333,6 +333,7 @@ bool pa_webrtc_ec_init(pa_core *c, pa_echo_canceller *ec, >      ec->params.webrtc.apm = apm; >      ec->params.webrtc.rec_ss = *rec_ss; >      ec->params.webrtc.play_ss = *play_ss; > +    ec->params.webrtc.out_ss = *out_ss; >      ec->params.webrtc.blocksize = >          (uint64_t) (pa_bytes_per_second(out_ss) / pa_frame_size(out_ss)) * BLOCK_SIZE_US / PA_USEC_PER_SEC; >      *nframes = ec->params.webrtc.blocksize; > @@ -372,17 +373,18 @@ void pa_webrtc_ec_play(pa_echo_canceller *ec, const uint8_t *play) { >  void pa_webrtc_ec_record(pa_echo_canceller *ec, const uint8_t *rec, uint8_t *out) { >      webrtc::AudioProcessing *apm = (webrtc::AudioProcessing*)ec->params.webrtc.apm; >      webrtc::AudioFrame out_frame; > -    const pa_sample_spec *ss = &ec->params.webrtc.rec_ss; > +    const pa_sample_spec *rec_ss = &ec->params.webrtc.rec_ss; > +    const pa_sample_spec *out_ss = &ec->params.webrtc.out_ss; >      pa_cvolume v; >      int old_volume, new_volume; >  > -    out_frame.num_channels_ = ss->channels; > -    out_frame.sample_rate_hz_ = ss->rate; > +    out_frame.num_channels_ = rec_ss->channels; > +    out_frame.sample_rate_hz_ = rec_ss->rate; >      out_frame.interleaved_ = true; >      out_frame.samples_per_channel_ = ec->params.webrtc.blocksize; >  >      pa_assert(out_frame.samples_per_channel_ <= webrtc::AudioFrame::kMaxDataSizeSamples); Not directly related to this patch, but this assertion seems wrong. We compare kMaxDataSizeSamples to the number of frames, but kMaxDataSizeSamples is calculated as frames * channels. Also, kMaxDataSizeSamples assumes that the maximum sample spec is 16- bit, stereo, 32 kHz, and that the maximum buffer size is 60 ms. However, we seem to allow using 48 kHz, and I didn't find any limitations for the channel count. We seem to use 10 ms buffers, so that gives some headroom. webrtc_ec_fixate_spec() should check all this, and make sure that our parameters don't exceed kMaxDataSizeSamples. The changes in this patch all look good. -- Tanu