Update to current webrtc library

This is from the upstream library commit id
3326535126e435f1ba647885ce43a8f0f3d317eb, corresponding to Chromium
88.0.4290.1.
This commit is contained in:
Arun Raghavan
2020-10-12 18:08:02 -04:00
parent b1b02581d3
commit bcec8b0b21
859 changed files with 76187 additions and 49580 deletions

View File

@ -12,28 +12,42 @@
// https://ccrma.stanford.edu/courses/422/projects/WaveFormat/ and
// http://www-mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/WAVE.html
#include "webrtc/common_audio/wav_header.h"
#include "common_audio/wav_header.h"
#include <algorithm>
#include <cstring>
#include <limits>
#include <string>
#include "webrtc/base/checks.h"
#include "webrtc/common_audio/include/audio_util.h"
#include "rtc_base/checks.h"
#include "rtc_base/logging.h"
#include "rtc_base/sanitizer.h"
#include "rtc_base/system/arch.h"
namespace webrtc {
namespace {
#ifndef WEBRTC_ARCH_LITTLE_ENDIAN
#error "Code not working properly for big endian platforms."
#endif
#pragma pack(2)
struct ChunkHeader {
uint32_t ID;
uint32_t Size;
};
static_assert(sizeof(ChunkHeader) == 8, "ChunkHeader size");
#pragma pack(2)
struct RiffHeader {
ChunkHeader header;
uint32_t Format;
};
static_assert(sizeof(RiffHeader) == sizeof(ChunkHeader) + 4, "RiffHeader size");
// We can't nest this definition in WavHeader, because VS2013 gives an error
// on sizeof(WavHeader::fmt): "error C2070: 'unknown': illegal sizeof operand".
struct FmtSubchunk {
#pragma pack(2)
struct FmtPcmSubchunk {
ChunkHeader header;
uint16_t AudioFormat;
uint16_t NumChannels;
@ -42,37 +56,247 @@ struct FmtSubchunk {
uint16_t BlockAlign;
uint16_t BitsPerSample;
};
static_assert(sizeof(FmtSubchunk) == 24, "FmtSubchunk size");
const uint32_t kFmtSubchunkSize = sizeof(FmtSubchunk) - sizeof(ChunkHeader);
static_assert(sizeof(FmtPcmSubchunk) == 24, "FmtPcmSubchunk size");
const uint32_t kFmtPcmSubchunkSize =
sizeof(FmtPcmSubchunk) - sizeof(ChunkHeader);
struct WavHeader {
struct {
ChunkHeader header;
uint32_t Format;
} riff;
FmtSubchunk fmt;
// Pack struct to avoid additional padding bytes.
#pragma pack(2)
struct FmtIeeeFloatSubchunk {
ChunkHeader header;
uint16_t AudioFormat;
uint16_t NumChannels;
uint32_t SampleRate;
uint32_t ByteRate;
uint16_t BlockAlign;
uint16_t BitsPerSample;
uint16_t ExtensionSize;
};
static_assert(sizeof(FmtIeeeFloatSubchunk) == 26, "FmtIeeeFloatSubchunk size");
const uint32_t kFmtIeeeFloatSubchunkSize =
sizeof(FmtIeeeFloatSubchunk) - sizeof(ChunkHeader);
// Simple PCM wav header. It does not include chunks that are not essential to
// read audio samples.
#pragma pack(2)
struct WavHeaderPcm {
WavHeaderPcm(const WavHeaderPcm&) = default;
WavHeaderPcm& operator=(const WavHeaderPcm&) = default;
RiffHeader riff;
FmtPcmSubchunk fmt;
struct {
ChunkHeader header;
} data;
};
static_assert(sizeof(WavHeader) == kWavHeaderSize, "no padding in header");
static_assert(sizeof(WavHeaderPcm) == kPcmWavHeaderSize,
"no padding in header");
} // namespace
// IEEE Float Wav header, includes extra chunks necessary for proper non-PCM
// WAV implementation.
#pragma pack(2)
struct WavHeaderIeeeFloat {
WavHeaderIeeeFloat(const WavHeaderIeeeFloat&) = default;
WavHeaderIeeeFloat& operator=(const WavHeaderIeeeFloat&) = default;
RiffHeader riff;
FmtIeeeFloatSubchunk fmt;
struct {
ChunkHeader header;
uint32_t SampleLength;
} fact;
struct {
ChunkHeader header;
} data;
};
static_assert(sizeof(WavHeaderIeeeFloat) == kIeeeFloatWavHeaderSize,
"no padding in header");
bool CheckWavParameters(int num_channels,
uint32_t PackFourCC(char a, char b, char c, char d) {
uint32_t packed_value =
static_cast<uint32_t>(a) | static_cast<uint32_t>(b) << 8 |
static_cast<uint32_t>(c) << 16 | static_cast<uint32_t>(d) << 24;
return packed_value;
}
std::string ReadFourCC(uint32_t x) {
return std::string(reinterpret_cast<char*>(&x), 4);
}
uint16_t MapWavFormatToHeaderField(WavFormat format) {
switch (format) {
case WavFormat::kWavFormatPcm:
return 1;
case WavFormat::kWavFormatIeeeFloat:
return 3;
case WavFormat::kWavFormatALaw:
return 6;
case WavFormat::kWavFormatMuLaw:
return 7;
}
RTC_CHECK(false);
}
WavFormat MapHeaderFieldToWavFormat(uint16_t format_header_value) {
if (format_header_value == 1) {
return WavFormat::kWavFormatPcm;
}
if (format_header_value == 3) {
return WavFormat::kWavFormatIeeeFloat;
}
RTC_CHECK(false) << "Unsupported WAV format";
}
uint32_t RiffChunkSize(size_t bytes_in_payload, size_t header_size) {
return static_cast<uint32_t>(bytes_in_payload + header_size -
sizeof(ChunkHeader));
}
uint32_t ByteRate(size_t num_channels,
int sample_rate,
size_t bytes_per_sample) {
return static_cast<uint32_t>(num_channels * sample_rate * bytes_per_sample);
}
uint16_t BlockAlign(size_t num_channels, size_t bytes_per_sample) {
return static_cast<uint16_t>(num_channels * bytes_per_sample);
}
// Finds a chunk having the sought ID. If found, then |readable| points to the
// first byte of the sought chunk data. If not found, the end of the file is
// reached.
bool FindWaveChunk(ChunkHeader* chunk_header,
WavHeaderReader* readable,
const std::string sought_chunk_id) {
RTC_DCHECK_EQ(sought_chunk_id.size(), 4);
while (true) {
if (readable->Read(chunk_header, sizeof(*chunk_header)) !=
sizeof(*chunk_header))
return false; // EOF.
if (ReadFourCC(chunk_header->ID) == sought_chunk_id)
return true; // Sought chunk found.
// Ignore current chunk by skipping its payload.
if (!readable->SeekForward(chunk_header->Size))
return false; // EOF or error.
}
}
bool ReadFmtChunkData(FmtPcmSubchunk* fmt_subchunk, WavHeaderReader* readable) {
// Reads "fmt " chunk payload.
if (readable->Read(&(fmt_subchunk->AudioFormat), kFmtPcmSubchunkSize) !=
kFmtPcmSubchunkSize)
return false;
const uint32_t fmt_size = fmt_subchunk->header.Size;
if (fmt_size != kFmtPcmSubchunkSize) {
// There is an optional two-byte extension field permitted to be present
// with PCM, but which must be zero.
int16_t ext_size;
if (kFmtPcmSubchunkSize + sizeof(ext_size) != fmt_size)
return false;
if (readable->Read(&ext_size, sizeof(ext_size)) != sizeof(ext_size))
return false;
if (ext_size != 0)
return false;
}
return true;
}
void WritePcmWavHeader(size_t num_channels,
int sample_rate,
size_t bytes_per_sample,
size_t num_samples,
uint8_t* buf,
size_t* header_size) {
RTC_CHECK(buf);
RTC_CHECK(header_size);
*header_size = kPcmWavHeaderSize;
auto header = rtc::MsanUninitialized<WavHeaderPcm>({});
const size_t bytes_in_payload = bytes_per_sample * num_samples;
header.riff.header.ID = PackFourCC('R', 'I', 'F', 'F');
header.riff.header.Size = RiffChunkSize(bytes_in_payload, *header_size);
header.riff.Format = PackFourCC('W', 'A', 'V', 'E');
header.fmt.header.ID = PackFourCC('f', 'm', 't', ' ');
header.fmt.header.Size = kFmtPcmSubchunkSize;
header.fmt.AudioFormat = MapWavFormatToHeaderField(WavFormat::kWavFormatPcm);
header.fmt.NumChannels = static_cast<uint16_t>(num_channels);
header.fmt.SampleRate = sample_rate;
header.fmt.ByteRate = ByteRate(num_channels, sample_rate, bytes_per_sample);
header.fmt.BlockAlign = BlockAlign(num_channels, bytes_per_sample);
header.fmt.BitsPerSample = static_cast<uint16_t>(8 * bytes_per_sample);
header.data.header.ID = PackFourCC('d', 'a', 't', 'a');
header.data.header.Size = static_cast<uint32_t>(bytes_in_payload);
// Do an extra copy rather than writing everything to buf directly, since buf
// might not be correctly aligned.
memcpy(buf, &header, *header_size);
}
void WriteIeeeFloatWavHeader(size_t num_channels,
int sample_rate,
size_t bytes_per_sample,
size_t num_samples,
uint8_t* buf,
size_t* header_size) {
RTC_CHECK(buf);
RTC_CHECK(header_size);
*header_size = kIeeeFloatWavHeaderSize;
auto header = rtc::MsanUninitialized<WavHeaderIeeeFloat>({});
const size_t bytes_in_payload = bytes_per_sample * num_samples;
header.riff.header.ID = PackFourCC('R', 'I', 'F', 'F');
header.riff.header.Size = RiffChunkSize(bytes_in_payload, *header_size);
header.riff.Format = PackFourCC('W', 'A', 'V', 'E');
header.fmt.header.ID = PackFourCC('f', 'm', 't', ' ');
header.fmt.header.Size = kFmtIeeeFloatSubchunkSize;
header.fmt.AudioFormat =
MapWavFormatToHeaderField(WavFormat::kWavFormatIeeeFloat);
header.fmt.NumChannels = static_cast<uint16_t>(num_channels);
header.fmt.SampleRate = sample_rate;
header.fmt.ByteRate = ByteRate(num_channels, sample_rate, bytes_per_sample);
header.fmt.BlockAlign = BlockAlign(num_channels, bytes_per_sample);
header.fmt.BitsPerSample = static_cast<uint16_t>(8 * bytes_per_sample);
header.fmt.ExtensionSize = 0;
header.fact.header.ID = PackFourCC('f', 'a', 'c', 't');
header.fact.header.Size = 4;
header.fact.SampleLength = static_cast<uint32_t>(num_channels * num_samples);
header.data.header.ID = PackFourCC('d', 'a', 't', 'a');
header.data.header.Size = static_cast<uint32_t>(bytes_in_payload);
// Do an extra copy rather than writing everything to buf directly, since buf
// might not be correctly aligned.
memcpy(buf, &header, *header_size);
}
// Returns the number of bytes per sample for the format.
size_t GetFormatBytesPerSample(WavFormat format) {
switch (format) {
case WavFormat::kWavFormatPcm:
// Other values may be OK, but for now we're conservative.
return 2;
case WavFormat::kWavFormatALaw:
case WavFormat::kWavFormatMuLaw:
return 1;
case WavFormat::kWavFormatIeeeFloat:
return 4;
default:
RTC_CHECK(false);
return 2;
}
}
bool CheckWavParameters(size_t num_channels,
int sample_rate,
WavFormat format,
int bytes_per_sample,
uint32_t num_samples) {
size_t bytes_per_sample,
size_t num_samples) {
// num_channels, sample_rate, and bytes_per_sample must be positive, must fit
// in their respective fields, and their product must fit in the 32-bit
// ByteRate field.
if (num_channels <= 0 || sample_rate <= 0 || bytes_per_sample <= 0)
if (num_channels == 0 || sample_rate <= 0 || bytes_per_sample == 0)
return false;
if (static_cast<uint64_t>(sample_rate) > std::numeric_limits<uint32_t>::max())
return false;
if (static_cast<uint64_t>(num_channels) >
std::numeric_limits<uint16_t>::max())
if (num_channels > std::numeric_limits<uint16_t>::max())
return false;
if (static_cast<uint64_t>(bytes_per_sample) * 8 >
std::numeric_limits<uint16_t>::max())
@ -83,26 +307,29 @@ bool CheckWavParameters(int num_channels,
// format and bytes_per_sample must agree.
switch (format) {
case kWavFormatPcm:
case WavFormat::kWavFormatPcm:
// Other values may be OK, but for now we're conservative:
if (bytes_per_sample != 1 && bytes_per_sample != 2)
return false;
break;
case kWavFormatALaw:
case kWavFormatMuLaw:
case WavFormat::kWavFormatALaw:
case WavFormat::kWavFormatMuLaw:
if (bytes_per_sample != 1)
return false;
break;
case WavFormat::kWavFormatIeeeFloat:
if (bytes_per_sample != 4)
return false;
break;
default:
return false;
}
// The number of bytes in the file, not counting the first ChunkHeader, must
// be less than 2^32; otherwise, the ChunkSize field overflows.
const uint32_t max_samples =
(std::numeric_limits<uint32_t>::max()
- (kWavHeaderSize - sizeof(ChunkHeader))) /
bytes_per_sample;
const size_t header_size = kPcmWavHeaderSize - sizeof(ChunkHeader);
const size_t max_samples =
(std::numeric_limits<uint32_t>::max() - header_size) / bytes_per_sample;
if (num_samples > max_samples)
return false;
@ -113,130 +340,102 @@ bool CheckWavParameters(int num_channels,
return true;
}
#ifdef WEBRTC_ARCH_LITTLE_ENDIAN
static inline void WriteLE16(uint16_t* f, uint16_t x) { *f = x; }
static inline void WriteLE32(uint32_t* f, uint32_t x) { *f = x; }
static inline void WriteFourCC(uint32_t* f, char a, char b, char c, char d) {
*f = static_cast<uint32_t>(a)
| static_cast<uint32_t>(b) << 8
| static_cast<uint32_t>(c) << 16
| static_cast<uint32_t>(d) << 24;
} // namespace
bool CheckWavParameters(size_t num_channels,
int sample_rate,
WavFormat format,
size_t num_samples) {
return CheckWavParameters(num_channels, sample_rate, format,
GetFormatBytesPerSample(format), num_samples);
}
static inline uint16_t ReadLE16(uint16_t x) { return x; }
static inline uint32_t ReadLE32(uint32_t x) { return x; }
static inline std::string ReadFourCC(uint32_t x) {
return std::string(reinterpret_cast<char*>(&x), 4);
}
#else
#error "Write be-to-le conversion functions"
#endif
static inline uint32_t RiffChunkSize(uint32_t bytes_in_payload) {
return bytes_in_payload + kWavHeaderSize - sizeof(ChunkHeader);
}
static inline uint32_t ByteRate(int num_channels, int sample_rate,
int bytes_per_sample) {
return static_cast<uint32_t>(num_channels) * sample_rate * bytes_per_sample;
}
static inline uint16_t BlockAlign(int num_channels, int bytes_per_sample) {
return num_channels * bytes_per_sample;
}
void WriteWavHeader(uint8_t* buf,
int num_channels,
void WriteWavHeader(size_t num_channels,
int sample_rate,
WavFormat format,
int bytes_per_sample,
uint32_t num_samples) {
size_t num_samples,
uint8_t* buf,
size_t* header_size) {
RTC_CHECK(buf);
RTC_CHECK(header_size);
const size_t bytes_per_sample = GetFormatBytesPerSample(format);
RTC_CHECK(CheckWavParameters(num_channels, sample_rate, format,
bytes_per_sample, num_samples));
WavHeader header;
const uint32_t bytes_in_payload = bytes_per_sample * num_samples;
WriteFourCC(&header.riff.header.ID, 'R', 'I', 'F', 'F');
WriteLE32(&header.riff.header.Size, RiffChunkSize(bytes_in_payload));
WriteFourCC(&header.riff.Format, 'W', 'A', 'V', 'E');
WriteFourCC(&header.fmt.header.ID, 'f', 'm', 't', ' ');
WriteLE32(&header.fmt.header.Size, kFmtSubchunkSize);
WriteLE16(&header.fmt.AudioFormat, format);
WriteLE16(&header.fmt.NumChannels, num_channels);
WriteLE32(&header.fmt.SampleRate, sample_rate);
WriteLE32(&header.fmt.ByteRate, ByteRate(num_channels, sample_rate,
bytes_per_sample));
WriteLE16(&header.fmt.BlockAlign, BlockAlign(num_channels, bytes_per_sample));
WriteLE16(&header.fmt.BitsPerSample, 8 * bytes_per_sample);
WriteFourCC(&header.data.header.ID, 'd', 'a', 't', 'a');
WriteLE32(&header.data.header.Size, bytes_in_payload);
// Do an extra copy rather than writing everything to buf directly, since buf
// might not be correctly aligned.
memcpy(buf, &header, kWavHeaderSize);
if (format == WavFormat::kWavFormatPcm) {
WritePcmWavHeader(num_channels, sample_rate, bytes_per_sample, num_samples,
buf, header_size);
} else {
RTC_CHECK_EQ(format, WavFormat::kWavFormatIeeeFloat);
WriteIeeeFloatWavHeader(num_channels, sample_rate, bytes_per_sample,
num_samples, buf, header_size);
}
}
bool ReadWavHeader(ReadableWav* readable,
int* num_channels,
bool ReadWavHeader(WavHeaderReader* readable,
size_t* num_channels,
int* sample_rate,
WavFormat* format,
int* bytes_per_sample,
uint32_t* num_samples) {
WavHeader header;
if (readable->Read(&header, kWavHeaderSize - sizeof(header.data)) !=
kWavHeaderSize - sizeof(header.data))
return false;
size_t* bytes_per_sample,
size_t* num_samples,
int64_t* data_start_pos) {
// Read using the PCM header, even though it might be float Wav file
auto header = rtc::MsanUninitialized<WavHeaderPcm>({});
const uint32_t fmt_size = ReadLE32(header.fmt.header.Size);
if (fmt_size != kFmtSubchunkSize) {
// There is an optional two-byte extension field permitted to be present
// with PCM, but which must be zero.
int16_t ext_size;
if (kFmtSubchunkSize + sizeof(ext_size) != fmt_size)
return false;
if (readable->Read(&ext_size, sizeof(ext_size)) != sizeof(ext_size))
return false;
if (ext_size != 0)
return false;
}
if (readable->Read(&header.data, sizeof(header.data)) != sizeof(header.data))
// Read RIFF chunk.
if (readable->Read(&header.riff, sizeof(header.riff)) != sizeof(header.riff))
return false;
// Parse needed fields.
*format = static_cast<WavFormat>(ReadLE16(header.fmt.AudioFormat));
*num_channels = ReadLE16(header.fmt.NumChannels);
*sample_rate = ReadLE32(header.fmt.SampleRate);
*bytes_per_sample = ReadLE16(header.fmt.BitsPerSample) / 8;
const uint32_t bytes_in_payload = ReadLE32(header.data.header.Size);
if (*bytes_per_sample <= 0)
return false;
*num_samples = bytes_in_payload / *bytes_per_sample;
// Sanity check remaining fields.
if (ReadFourCC(header.riff.header.ID) != "RIFF")
return false;
if (ReadFourCC(header.riff.Format) != "WAVE")
return false;
if (ReadFourCC(header.fmt.header.ID) != "fmt ")
return false;
if (ReadFourCC(header.data.header.ID) != "data")
return false;
if (ReadLE32(header.riff.header.Size) < RiffChunkSize(bytes_in_payload))
// Find "fmt " and "data" chunks. While the official Wave file specification
// does not put requirements on the chunks order, it is uncommon to find the
// "data" chunk before the "fmt " one. The code below fails if this is not the
// case.
if (!FindWaveChunk(&header.fmt.header, readable, "fmt ")) {
RTC_LOG(LS_ERROR) << "Cannot find 'fmt ' chunk.";
return false;
if (ReadLE32(header.fmt.ByteRate) !=
}
if (!ReadFmtChunkData(&header.fmt, readable)) {
RTC_LOG(LS_ERROR) << "Cannot read 'fmt ' chunk.";
return false;
}
if (!FindWaveChunk(&header.data.header, readable, "data")) {
RTC_LOG(LS_ERROR) << "Cannot find 'data' chunk.";
return false;
}
// Parse needed fields.
*format = MapHeaderFieldToWavFormat(header.fmt.AudioFormat);
*num_channels = header.fmt.NumChannels;
*sample_rate = header.fmt.SampleRate;
*bytes_per_sample = header.fmt.BitsPerSample / 8;
const size_t bytes_in_payload = header.data.header.Size;
if (*bytes_per_sample == 0)
return false;
*num_samples = bytes_in_payload / *bytes_per_sample;
const size_t header_size = *format == WavFormat::kWavFormatPcm
? kPcmWavHeaderSize
: kIeeeFloatWavHeaderSize;
if (header.riff.header.Size < RiffChunkSize(bytes_in_payload, header_size))
return false;
if (header.fmt.ByteRate !=
ByteRate(*num_channels, *sample_rate, *bytes_per_sample))
return false;
if (ReadLE16(header.fmt.BlockAlign) !=
BlockAlign(*num_channels, *bytes_per_sample))
if (header.fmt.BlockAlign != BlockAlign(*num_channels, *bytes_per_sample))
return false;
return CheckWavParameters(*num_channels, *sample_rate, *format,
*bytes_per_sample, *num_samples);
if (!CheckWavParameters(*num_channels, *sample_rate, *format,
*bytes_per_sample, *num_samples)) {
return false;
}
*data_start_pos = readable->GetPosition();
return true;
}
} // namespace webrtc