Now the unit test is included in the vie_auto_test target and executed when the automated flag is used.
TBR=mflodman
BUG=
TEST=vie_auto_test --automated --gtest_filter=FrameDropPrimitivesTest.FixOutputFileForComparison
Review URL: https://webrtc-codereview.appspot.com/381003
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1598 4adac7df-926f-26a2-2b94-8c16560cd09d
Refactoring of FrameDropHandler: It now also tracks when frames are leaving the encoder and is being sent to external transport.
Previous 'Sent' state is now renamed to 'Created'.
NOTICE: The test seems to be a little flaky on Linux so it's not ready for buildbots yet. Since this might be caused by unstable production code further investigation should be performed to clear out the flakiness. I will file an issue for this when this CL is submitted (since I don't have any code to refer to before that). Usually the flakiness is caused by a decoded/rendered callback that is left out for the last frame, but I have seen other flaky failures too, which means it's not as simple as ignoring the last frame.
These errors occur even if 400kbps bit rate and 0% PL and 0 delay is configured.
BUG=
TEST=vie_auto_test --automated --gtest_filter="ViEVideoVerificationTest.RunsFullStackWithoutErrors" in Debug+Release on Linux, Mac and Windows.
Review URL: http://webrtc-codereview.appspot.com/339005
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1597 4adac7df-926f-26a2-2b94-8c16560cd09d
and touched VoEBaseImpl::NeedMorePlayData and AudioCodingModuleImpl::PlayoutData10Ms(), for
performance reasons in Android platforms.
The two functions used about 6% of VoE originally. After the change, the percentage reduced
to about 0.2%.
Review URL: https://webrtc-codereview.appspot.com/379001
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1589 4adac7df-926f-26a2-2b94-8c16560cd09d
Investigation with corrupt payloads revealed a few places we could
be using stronger checks. These are not foolproof by any means, but
I figure the earlier we catch this the better.
BUG=242
TEST=loopback call with a hacked ViE to insert corrupt payloads, and vie_auto_test without the hacks.
Review URL: https://webrtc-codereview.appspot.com/369015
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1585 4adac7df-926f-26a2-2b94-8c16560cd09d
Style and return value changes. No impact externally, since audio_processing, audio_conference_mixer and audio_coding either already assumes 'int' as return value, assumes nothing or doesn't take care of the return value.
TESTS=vad_unittests, audioproc_unittest
Review URL: https://webrtc-codereview.appspot.com/374006
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1581 4adac7df-926f-26a2-2b94-8c16560cd09d
When _amountOfMixableParticipants == 1, we skip mixing and saturation
protection. Without this fix, an anonymous participant would only be
properly counted if it was the last added.
For example, if an anonymous participant was added first, followed by
a regular participant, _amoutOfMixableParticipants would == 1 and the
regular participant would not be mixed.
BUG=issue209
TEST=New test added to voe_auto_test to verify, and used voe_cmd_test.
Review URL: https://webrtc-codereview.appspot.com/367006
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1551 4adac7df-926f-26a2-2b94-8c16560cd09d
NetEQ now checks for too early CNG packets, and modifies the CNG
sample counter to jump forward in time if needed to combat clock
drift.
Adding a new unittest to reproduce and solve the issue. The
unittest LongCngWithClockDrift verifies that the buffer delay
before and after a long CNG period is almost constant. The test
introduces a clock drift of 25 ms/s.
BUG=http://code.google.com/p/webrtc/issues/detail?id=88
TEST=neteq_unittests NetEqDecodingTest.LongCngWithClockDrift
Review URL: https://webrtc-codereview.appspot.com/372002
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1547 4adac7df-926f-26a2-2b94-8c16560cd09d
As part of style this CL includes changing the input aggressiveness mode from int16_t to int. No other style changes made.
Impact on:
- Audio Processing: Changed return value on MapSetting().
- Function test in audio_conference_mixer already uses int. No action.
- NetEq: Function pointer changes and input parameter changes in SetVADMode() and SetVADModeInternal().
- Audio Coding: Uses enum ACMVADMode which is type independent.
- VAD: Two unit tests.
TESTS=vad_unittests, neteq_unittests, audioproc_unittest
Review URL: https://webrtc-codereview.appspot.com/373001
git-svn-id: http://webrtc.googlecode.com/svn/trunk@1544 4adac7df-926f-26a2-2b94-8c16560cd09d
