1 // Copyright 2014 The Chromium Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style license that can be
3 // found in the LICENSE file.
8 #include "base/format_macros.h"
10 #include "base/message_loop/message_loop.h"
11 #include "base/run_loop.h"
12 #include "base/strings/stringprintf.h"
13 #include "base/sys_byteorder.h"
14 #include "build/build_config.h"
15 #include "media/base/audio_buffer.h"
16 #include "media/base/audio_bus.h"
17 #include "media/base/audio_hash.h"
18 #include "media/base/decoder_buffer.h"
19 #include "media/base/test_data_util.h"
20 #include "media/base/test_helpers.h"
21 #include "media/base/timestamp_constants.h"
22 #include "media/ffmpeg/ffmpeg_common.h"
23 #include "media/filters/audio_file_reader.h"
24 #include "media/filters/ffmpeg_audio_decoder.h"
25 #include "media/filters/in_memory_url_protocol.h"
26 #include "media/filters/opus_audio_decoder.h"
27 #include "testing/gtest/include/gtest/gtest.h"
31 // The number of packets to read and then decode from each file.
32 static const size_t kDecodeRuns
= 3;
33 static const uint8_t kOpusExtraData
[] = {
34 0x4f, 0x70, 0x75, 0x73, 0x48, 0x65, 0x61, 0x64, 0x01, 0x02,
35 // The next two bytes represent the codec delay.
36 0x00, 0x00, 0x80, 0xbb, 0x00, 0x00, 0x00, 0x00, 0x00};
38 enum AudioDecoderType
{
43 struct DecodedBufferExpectations
{
44 const int64 timestamp
;
49 struct DecoderTestData
{
50 const AudioDecoderType decoder_type
;
51 const AudioCodec codec
;
53 const DecodedBufferExpectations
* expectations
;
54 const int first_packet_pts
;
55 const int samples_per_second
;
56 const ChannelLayout channel_layout
;
59 // Tells gtest how to print our DecoderTestData structure.
60 std::ostream
& operator<<(std::ostream
& os
, const DecoderTestData
& data
) {
61 return os
<< data
.filename
;
64 // Marks negative timestamp buffers for discard or transfers FFmpeg's built in
65 // discard metadata in favor of setting DiscardPadding on the DecoderBuffer.
66 // Allows better testing of AudioDiscardHelper usage.
67 static void SetDiscardPadding(AVPacket
* packet
,
68 const scoped_refptr
<DecoderBuffer
> buffer
,
69 double samples_per_second
) {
70 // Discard negative timestamps.
71 if (buffer
->timestamp() + buffer
->duration() < base::TimeDelta()) {
72 buffer
->set_discard_padding(
73 std::make_pair(kInfiniteDuration(), base::TimeDelta()));
76 if (buffer
->timestamp() < base::TimeDelta()) {
77 buffer
->set_discard_padding(
78 std::make_pair(-buffer
->timestamp(), base::TimeDelta()));
82 // If the timestamp is positive, try to use FFmpeg's discard data.
83 int skip_samples_size
= 0;
84 const uint32
* skip_samples_ptr
=
85 reinterpret_cast<const uint32
*>(av_packet_get_side_data(
86 packet
, AV_PKT_DATA_SKIP_SAMPLES
, &skip_samples_size
));
87 if (skip_samples_size
< 4)
89 buffer
->set_discard_padding(std::make_pair(
90 base::TimeDelta::FromSecondsD(base::ByteSwapToLE32(*skip_samples_ptr
) /
95 class AudioDecoderTest
: public testing::TestWithParam
<DecoderTestData
> {
98 : pending_decode_(false),
99 pending_reset_(false),
100 last_decode_status_(AudioDecoder::kDecodeError
) {
101 switch (GetParam().decoder_type
) {
103 decoder_
.reset(new FFmpegAudioDecoder(message_loop_
.task_runner(),
108 new OpusAudioDecoder(message_loop_
.task_runner()));
113 virtual ~AudioDecoderTest() {
114 EXPECT_FALSE(pending_decode_
);
115 EXPECT_FALSE(pending_reset_
);
119 void DecodeBuffer(const scoped_refptr
<DecoderBuffer
>& buffer
) {
120 ASSERT_FALSE(pending_decode_
);
121 pending_decode_
= true;
122 last_decode_status_
= AudioDecoder::kDecodeError
;
125 base::Bind(&AudioDecoderTest::DecodeFinished
, base::Unretained(this)));
126 base::RunLoop().RunUntilIdle();
127 ASSERT_FALSE(pending_decode_
);
130 void SendEndOfStream() {
131 DecodeBuffer(DecoderBuffer::CreateEOSBuffer());
135 // Load the test data file.
136 data_
= ReadTestDataFile(GetParam().filename
);
138 new InMemoryUrlProtocol(data_
->data(), data_
->data_size(), false));
139 reader_
.reset(new AudioFileReader(protocol_
.get()));
140 ASSERT_TRUE(reader_
->OpenDemuxerForTesting());
142 // Load the first packet and check its timestamp.
144 ASSERT_TRUE(reader_
->ReadPacketForTesting(&packet
));
145 EXPECT_EQ(GetParam().first_packet_pts
, packet
.pts
);
146 start_timestamp_
= ConvertFromTimeBase(
147 reader_
->GetAVStreamForTesting()->time_base
, packet
.pts
);
148 av_free_packet(&packet
);
150 // Seek back to the beginning.
151 ASSERT_TRUE(reader_
->SeekForTesting(start_timestamp_
));
153 AudioDecoderConfig config
;
154 AVCodecContextToAudioDecoderConfig(reader_
->codec_context_for_testing(),
157 EXPECT_EQ(GetParam().codec
, config
.codec());
158 EXPECT_EQ(GetParam().samples_per_second
, config
.samples_per_second());
159 EXPECT_EQ(GetParam().channel_layout
, config
.channel_layout());
161 InitializeDecoder(config
);
164 void InitializeDecoder(const AudioDecoderConfig
& config
) {
165 InitializeDecoderWithResult(config
, true);
168 void InitializeDecoderWithResult(const AudioDecoderConfig
& config
,
170 decoder_
->Initialize(
171 config
, NewExpectedBoolCB(success
),
172 base::Bind(&AudioDecoderTest::OnDecoderOutput
, base::Unretained(this)));
173 base::RunLoop().RunUntilIdle();
178 ASSERT_TRUE(reader_
->ReadPacketForTesting(&packet
));
180 // Split out packet metadata before making a copy.
181 av_packet_split_side_data(&packet
);
183 scoped_refptr
<DecoderBuffer
> buffer
=
184 DecoderBuffer::CopyFrom(packet
.data
, packet
.size
);
185 buffer
->set_timestamp(ConvertFromTimeBase(
186 reader_
->GetAVStreamForTesting()->time_base
, packet
.pts
));
187 buffer
->set_duration(ConvertFromTimeBase(
188 reader_
->GetAVStreamForTesting()->time_base
, packet
.duration
));
189 if (packet
.flags
& AV_PKT_FLAG_KEY
)
190 buffer
->set_is_key_frame(true);
192 // Don't set discard padding for Opus, it already has discard behavior set
193 // based on the codec delay in the AudioDecoderConfig.
194 if (GetParam().decoder_type
== FFMPEG
)
195 SetDiscardPadding(&packet
, buffer
, GetParam().samples_per_second
);
197 // DecodeBuffer() shouldn't need the original packet since it uses the copy.
198 av_free_packet(&packet
);
199 DecodeBuffer(buffer
);
203 ASSERT_FALSE(pending_reset_
);
204 pending_reset_
= true;
206 base::Bind(&AudioDecoderTest::ResetFinished
, base::Unretained(this)));
207 base::RunLoop().RunUntilIdle();
208 ASSERT_FALSE(pending_reset_
);
211 void Seek(base::TimeDelta seek_time
) {
213 decoded_audio_
.clear();
214 ASSERT_TRUE(reader_
->SeekForTesting(seek_time
));
217 void OnDecoderOutput(const scoped_refptr
<AudioBuffer
>& buffer
) {
218 EXPECT_FALSE(buffer
->end_of_stream());
219 decoded_audio_
.push_back(buffer
);
222 void DecodeFinished(AudioDecoder::Status status
) {
223 EXPECT_TRUE(pending_decode_
);
224 EXPECT_FALSE(pending_reset_
);
225 pending_decode_
= false;
226 last_decode_status_
= status
;
229 void ResetFinished() {
230 EXPECT_TRUE(pending_reset_
);
231 EXPECT_FALSE(pending_decode_
);
232 pending_reset_
= false;
235 // Generates an MD5 hash of the audio signal. Should not be used for checks
236 // across platforms as audio varies slightly across platforms.
237 std::string
GetDecodedAudioMD5(size_t i
) {
238 CHECK_LT(i
, decoded_audio_
.size());
239 const scoped_refptr
<AudioBuffer
>& buffer
= decoded_audio_
[i
];
241 scoped_ptr
<AudioBus
> output
=
242 AudioBus::Create(buffer
->channel_count(), buffer
->frame_count());
243 buffer
->ReadFrames(buffer
->frame_count(), 0, 0, output
.get());
245 base::MD5Context context
;
246 base::MD5Init(&context
);
247 for (int ch
= 0; ch
< output
->channels(); ++ch
) {
250 base::StringPiece(reinterpret_cast<char*>(output
->channel(ch
)),
251 output
->frames() * sizeof(*output
->channel(ch
))));
253 base::MD5Digest digest
;
254 base::MD5Final(&digest
, &context
);
255 return base::MD5DigestToBase16(digest
);
258 void ExpectDecodedAudio(size_t i
, const std::string
& exact_hash
) {
259 CHECK_LT(i
, decoded_audio_
.size());
260 const scoped_refptr
<AudioBuffer
>& buffer
= decoded_audio_
[i
];
262 const DecodedBufferExpectations
& sample_info
= GetParam().expectations
[i
];
263 EXPECT_EQ(sample_info
.timestamp
, buffer
->timestamp().InMicroseconds());
264 EXPECT_EQ(sample_info
.duration
, buffer
->duration().InMicroseconds());
265 EXPECT_FALSE(buffer
->end_of_stream());
267 scoped_ptr
<AudioBus
> output
=
268 AudioBus::Create(buffer
->channel_count(), buffer
->frame_count());
269 buffer
->ReadFrames(buffer
->frame_count(), 0, 0, output
.get());
271 // Generate a lossy hash of the audio used for comparison across platforms.
272 AudioHash audio_hash
;
273 audio_hash
.Update(output
.get(), output
->frames());
274 EXPECT_EQ(sample_info
.hash
, audio_hash
.ToString());
276 if (!exact_hash
.empty()) {
277 EXPECT_EQ(exact_hash
, GetDecodedAudioMD5(i
));
279 // Verify different hashes are being generated. None of our test data
280 // files have audio that hashes out exactly the same.
282 EXPECT_NE(exact_hash
, GetDecodedAudioMD5(i
- 1));
286 size_t decoded_audio_size() const { return decoded_audio_
.size(); }
287 base::TimeDelta
start_timestamp() const { return start_timestamp_
; }
288 const scoped_refptr
<AudioBuffer
>& decoded_audio(size_t i
) {
289 return decoded_audio_
[i
];
291 AudioDecoder::Status
last_decode_status() const {
292 return last_decode_status_
;
296 base::MessageLoop message_loop_
;
297 scoped_refptr
<DecoderBuffer
> data_
;
298 scoped_ptr
<InMemoryUrlProtocol
> protocol_
;
299 scoped_ptr
<AudioFileReader
> reader_
;
301 scoped_ptr
<AudioDecoder
> decoder_
;
302 bool pending_decode_
;
304 AudioDecoder::Status last_decode_status_
;
306 std::deque
<scoped_refptr
<AudioBuffer
> > decoded_audio_
;
307 base::TimeDelta start_timestamp_
;
309 DISALLOW_COPY_AND_ASSIGN(AudioDecoderTest
);
312 class OpusAudioDecoderBehavioralTest
: public AudioDecoderTest
{};
313 class FFmpegAudioDecoderBehavioralTest
: public AudioDecoderTest
{};
315 TEST_P(AudioDecoderTest
, Initialize
) {
316 ASSERT_NO_FATAL_FAILURE(Initialize());
319 // Verifies decode audio as well as the Decode() -> Reset() sequence.
320 TEST_P(AudioDecoderTest
, ProduceAudioSamples
) {
321 ASSERT_NO_FATAL_FAILURE(Initialize());
323 // Run the test multiple times with a seek back to the beginning in between.
324 std::vector
<std::string
> decoded_audio_md5_hashes
;
325 for (int i
= 0; i
< 2; ++i
) {
326 for (size_t j
= 0; j
< kDecodeRuns
; ++j
) {
329 ASSERT_EQ(last_decode_status(), AudioDecoder::kOk
);
330 // Some codecs have a multiple buffer delay and require an extra
331 // Decode() step to extract the desired number of output buffers.
332 } while (j
== 0 && decoded_audio_size() == 0);
334 // On the first pass record the exact MD5 hash for each decoded buffer.
336 decoded_audio_md5_hashes
.push_back(GetDecodedAudioMD5(j
));
339 ASSERT_EQ(kDecodeRuns
, decoded_audio_size());
341 // On the first pass verify the basic audio hash and sample info. On the
342 // second, verify the exact MD5 sum for each packet. It shouldn't change.
343 for (size_t j
= 0; j
< kDecodeRuns
; ++j
) {
344 SCOPED_TRACE(base::StringPrintf("i = %d, j = %" PRIuS
, i
, j
));
345 ExpectDecodedAudio(j
, i
== 0 ? "" : decoded_audio_md5_hashes
[j
]);
349 ASSERT_EQ(kDecodeRuns
, decoded_audio_size());
351 // Seek back to the beginning. Calls Reset() on the decoder.
352 Seek(start_timestamp());
356 TEST_P(AudioDecoderTest
, Decode
) {
357 ASSERT_NO_FATAL_FAILURE(Initialize());
359 EXPECT_EQ(AudioDecoder::kOk
, last_decode_status());
362 TEST_P(AudioDecoderTest
, Reset
) {
363 ASSERT_NO_FATAL_FAILURE(Initialize());
367 TEST_P(AudioDecoderTest
, NoTimestamp
) {
368 ASSERT_NO_FATAL_FAILURE(Initialize());
369 scoped_refptr
<DecoderBuffer
> buffer(new DecoderBuffer(0));
370 buffer
->set_timestamp(kNoTimestamp());
371 DecodeBuffer(buffer
);
372 EXPECT_EQ(AudioDecoder::kDecodeError
, last_decode_status());
375 TEST_P(OpusAudioDecoderBehavioralTest
, InitializeWithNoCodecDelay
) {
376 ASSERT_EQ(GetParam().decoder_type
, OPUS
);
377 AudioDecoderConfig decoder_config
;
378 decoder_config
.Initialize(kCodecOpus
,
380 CHANNEL_LAYOUT_STEREO
,
383 arraysize(kOpusExtraData
),
385 base::TimeDelta::FromMilliseconds(80),
387 InitializeDecoder(decoder_config
);
390 TEST_P(OpusAudioDecoderBehavioralTest
, InitializeWithBadCodecDelay
) {
391 ASSERT_EQ(GetParam().decoder_type
, OPUS
);
392 AudioDecoderConfig decoder_config
;
393 decoder_config
.Initialize(
396 CHANNEL_LAYOUT_STEREO
,
399 arraysize(kOpusExtraData
),
401 base::TimeDelta::FromMilliseconds(80),
402 // Use a different codec delay than in the extradata.
404 InitializeDecoderWithResult(decoder_config
, false);
407 TEST_P(FFmpegAudioDecoderBehavioralTest
, InitializeWithBadConfig
) {
408 const AudioDecoderConfig
decoder_config(kCodecVorbis
,
410 CHANNEL_LAYOUT_STEREO
,
411 // Invalid sample rate of zero.
416 InitializeDecoderWithResult(decoder_config
, false);
419 #if defined(OPUS_FIXED_POINT)
420 const DecodedBufferExpectations kSfxOpusExpectations
[] = {
421 {0, 13500, "-2.70,-1.41,-0.78,-1.27,-2.56,-3.73,"},
422 {13500, 20000, "5.48,5.93,6.05,5.83,5.54,5.46,"},
423 {33500, 20000, "-3.44,-3.34,-3.57,-4.11,-4.74,-5.13,"},
426 const DecodedBufferExpectations kSfxOpusExpectations
[] = {
427 {0, 13500, "-2.70,-1.41,-0.78,-1.27,-2.56,-3.73,"},
428 {13500, 20000, "5.48,5.93,6.04,5.83,5.54,5.45,"},
429 {33500, 20000, "-3.45,-3.35,-3.57,-4.12,-4.74,-5.14,"},
433 const DecodedBufferExpectations kBearOpusExpectations
[] = {
434 {500, 3500, "-0.26,0.87,1.36,0.84,-0.30,-1.22,"},
435 {4000, 10000, "0.09,0.23,0.21,0.03,-0.17,-0.24,"},
436 {14000, 10000, "0.10,0.24,0.23,0.04,-0.14,-0.23,"},
439 const DecoderTestData kOpusTests
[] = {
440 {OPUS
, kCodecOpus
, "sfx-opus.ogg", kSfxOpusExpectations
, -312, 48000,
441 CHANNEL_LAYOUT_MONO
},
442 {OPUS
, kCodecOpus
, "bear-opus.ogg", kBearOpusExpectations
, 24, 48000,
443 CHANNEL_LAYOUT_STEREO
},
446 // Dummy data for behavioral tests.
447 const DecoderTestData kOpusBehavioralTest
[] = {
448 {OPUS
, kUnknownAudioCodec
, "", NULL
, 0, 0, CHANNEL_LAYOUT_NONE
},
451 INSTANTIATE_TEST_CASE_P(OpusAudioDecoderTest
,
453 testing::ValuesIn(kOpusTests
));
454 INSTANTIATE_TEST_CASE_P(OpusAudioDecoderBehavioralTest
,
455 OpusAudioDecoderBehavioralTest
,
456 testing::ValuesIn(kOpusBehavioralTest
));
458 #if defined(USE_PROPRIETARY_CODECS)
459 const DecodedBufferExpectations kSfxMp3Expectations
[] = {
460 {0, 1065, "2.81,3.99,4.53,4.10,3.08,2.46,"},
461 {1065, 26122, "-3.81,-4.14,-3.90,-3.36,-3.03,-3.23,"},
462 {27188, 26122, "4.24,3.95,4.22,4.78,5.13,4.93,"},
465 const DecodedBufferExpectations kSfxAdtsExpectations
[] = {
466 {0, 23219, "-1.90,-1.53,-0.15,1.28,1.23,-0.33,"},
467 {23219, 23219, "0.54,0.88,2.19,3.54,3.24,1.63,"},
468 {46439, 23219, "1.42,1.69,2.95,4.23,4.02,2.36,"},
472 #if defined(OS_CHROMEOS)
473 const DecodedBufferExpectations kSfxFlacExpectations
[] = {
474 {0, 104489, "-2.42,-1.12,0.71,1.70,1.09,-0.68,"},
475 {104489, 104489, "-1.99,-0.67,1.18,2.19,1.60,-0.16,"},
476 {208979, 79433, "2.84,2.70,3.23,4.06,4.59,4.44,"},
480 const DecodedBufferExpectations kSfxWaveExpectations
[] = {
481 {0, 23219, "-1.23,-0.87,0.47,1.85,1.88,0.29,"},
482 {23219, 23219, "0.75,1.10,2.43,3.78,3.53,1.93,"},
483 {46439, 23219, "1.27,1.56,2.83,4.13,3.87,2.23,"},
486 const DecodedBufferExpectations kFourChannelWaveExpectations
[] = {
487 {0, 11609, "-1.68,1.68,0.89,-3.45,1.52,1.15,"},
488 {11609, 11609, "43.26,9.06,18.27,35.98,19.45,7.46,"},
489 {23219, 11609, "36.37,9.45,16.04,27.67,18.81,10.15,"},
492 const DecodedBufferExpectations kSfxOggExpectations
[] = {
493 {0, 13061, "-0.33,1.25,2.86,3.26,2.09,0.14,"},
494 {13061, 23219, "-2.79,-2.42,-1.06,0.33,0.93,-0.64,"},
495 {36281, 23219, "-1.19,-0.80,0.57,1.97,2.08,0.51,"},
498 const DecodedBufferExpectations kBearOgvExpectations
[] = {
499 {0, 13061, "-1.25,0.10,2.11,2.29,1.50,-0.68,"},
500 {13061, 23219, "-1.80,-1.41,-0.13,1.30,1.65,0.01,"},
501 {36281, 23219, "-1.43,-1.25,0.11,1.29,1.86,0.14,"},
504 const DecoderTestData kFFmpegTests
[] = {
505 #if defined(USE_PROPRIETARY_CODECS)
506 {FFMPEG
, kCodecMP3
, "sfx.mp3", kSfxMp3Expectations
, 0, 44100,
507 CHANNEL_LAYOUT_MONO
},
508 {FFMPEG
, kCodecAAC
, "sfx.adts", kSfxAdtsExpectations
, 0, 44100,
509 CHANNEL_LAYOUT_MONO
},
511 #if defined(OS_CHROMEOS)
512 {FFMPEG
, kCodecFLAC
, "sfx.flac", kSfxFlacExpectations
, 0, 44100,
513 CHANNEL_LAYOUT_MONO
},
515 {FFMPEG
, kCodecPCM
, "sfx_f32le.wav", kSfxWaveExpectations
, 0, 44100,
516 CHANNEL_LAYOUT_MONO
},
517 {FFMPEG
, kCodecPCM
, "4ch.wav", kFourChannelWaveExpectations
, 0, 44100,
518 CHANNEL_LAYOUT_QUAD
},
519 {FFMPEG
, kCodecVorbis
, "sfx.ogg", kSfxOggExpectations
, 0, 44100,
520 CHANNEL_LAYOUT_MONO
},
521 // Note: bear.ogv is incorrectly muxed such that valid samples are given
522 // negative timestamps, this marks them for discard per the ogg vorbis spec.
523 {FFMPEG
, kCodecVorbis
, "bear.ogv", kBearOgvExpectations
, -704, 44100,
524 CHANNEL_LAYOUT_STEREO
},
527 // Dummy data for behavioral tests.
528 const DecoderTestData kFFmpegBehavioralTest
[] = {
529 {FFMPEG
, kUnknownAudioCodec
, "", NULL
, 0, 0, CHANNEL_LAYOUT_NONE
},
532 INSTANTIATE_TEST_CASE_P(FFmpegAudioDecoderTest
,
534 testing::ValuesIn(kFFmpegTests
));
535 INSTANTIATE_TEST_CASE_P(FFmpegAudioDecoderBehavioralTest
,
536 FFmpegAudioDecoderBehavioralTest
,
537 testing::ValuesIn(kFFmpegBehavioralTest
));