Skip XING/VBRI frame when decoding
Don't send the XING/VBRI frame to the decoder, as it will result in extra data being output. Also clean up the XING seeker code a bit. Change-Id: I714a94180edae9bedc1bc0c96bf382fe6f928ac6
This commit is contained in:
@@ -317,6 +317,13 @@ MP3Extractor::MP3Extractor(
|
||||
mSeeker = VBRISeeker::CreateFromSource(mDataSource, post_id3_pos);
|
||||
}
|
||||
|
||||
if (mSeeker != NULL) {
|
||||
// While it is safe to send the XING/VBRI frame to the decoder, this will
|
||||
// result in an extra 1152 samples being output. The real first frame to
|
||||
// decode is after the XING/VBRI frame, so skip there.
|
||||
mFirstFramePos += frame_size;
|
||||
}
|
||||
|
||||
int64_t durationUs;
|
||||
|
||||
if (mSeeker == NULL || !mSeeker->getDuration(&durationUs)) {
|
||||
|
||||
@@ -92,7 +92,7 @@ sp<VBRISeeker> VBRISeeker::CreateFromSource(
|
||||
}
|
||||
|
||||
sp<VBRISeeker> seeker = new VBRISeeker;
|
||||
seeker->mBasePos = post_id3_pos;
|
||||
seeker->mBasePos = post_id3_pos + frameSize;
|
||||
seeker->mDurationUs = durationUs;
|
||||
|
||||
off64_t offset = post_id3_pos;
|
||||
|
||||
@@ -15,35 +15,13 @@
|
||||
*/
|
||||
|
||||
#include "include/XINGSeeker.h"
|
||||
#include "include/avc_utils.h"
|
||||
|
||||
#include <media/stagefright/DataSource.h>
|
||||
#include <media/stagefright/Utils.h>
|
||||
|
||||
namespace android {
|
||||
|
||||
static bool parse_xing_header(
|
||||
const sp<DataSource> &source, off64_t first_frame_pos,
|
||||
int32_t *frame_number = NULL, int32_t *byte_number = NULL,
|
||||
unsigned char *table_of_contents = NULL, bool *toc_is_valid = NULL,
|
||||
int32_t *quality_indicator = NULL, int64_t *duration = NULL);
|
||||
|
||||
// static
|
||||
sp<XINGSeeker> XINGSeeker::CreateFromSource(
|
||||
const sp<DataSource> &source, off64_t first_frame_pos) {
|
||||
sp<XINGSeeker> seeker = new XINGSeeker;
|
||||
|
||||
seeker->mFirstFramePos = first_frame_pos;
|
||||
|
||||
if (!parse_xing_header(
|
||||
source, first_frame_pos,
|
||||
NULL, &seeker->mSizeBytes, seeker->mTOC, &seeker->mTOCValid,
|
||||
NULL, &seeker->mDurationUs)) {
|
||||
return NULL;
|
||||
}
|
||||
|
||||
return seeker;
|
||||
}
|
||||
|
||||
XINGSeeker::XINGSeeker()
|
||||
: mDurationUs(-1),
|
||||
mSizeBytes(0) {
|
||||
@@ -91,60 +69,50 @@ bool XINGSeeker::getOffsetForTime(int64_t *timeUs, off64_t *pos) {
|
||||
return true;
|
||||
}
|
||||
|
||||
static bool parse_xing_header(
|
||||
const sp<DataSource> &source, off64_t first_frame_pos,
|
||||
int32_t *frame_number, int32_t *byte_number,
|
||||
unsigned char *table_of_contents, bool *toc_valid,
|
||||
int32_t *quality_indicator,
|
||||
int64_t *duration) {
|
||||
if (frame_number) {
|
||||
*frame_number = 0;
|
||||
}
|
||||
if (byte_number) {
|
||||
*byte_number = 0;
|
||||
}
|
||||
if (toc_valid) {
|
||||
*toc_valid = false;
|
||||
}
|
||||
if (quality_indicator) {
|
||||
*quality_indicator = 0;
|
||||
}
|
||||
if (duration) {
|
||||
*duration = 0;
|
||||
}
|
||||
// static
|
||||
sp<XINGSeeker> XINGSeeker::CreateFromSource(
|
||||
const sp<DataSource> &source, off64_t first_frame_pos) {
|
||||
sp<XINGSeeker> seeker = new XINGSeeker;
|
||||
|
||||
seeker->mFirstFramePos = first_frame_pos;
|
||||
|
||||
ALOGI("xingseeker first frame pos: %lld", first_frame_pos);
|
||||
|
||||
seeker->mSizeBytes = 0;
|
||||
seeker->mTOCValid = false;
|
||||
seeker->mDurationUs = 0;
|
||||
|
||||
uint8_t buffer[4];
|
||||
int offset = first_frame_pos;
|
||||
if (source->readAt(offset, &buffer, 4) < 4) { // get header
|
||||
return false;
|
||||
return NULL;
|
||||
}
|
||||
offset += 4;
|
||||
|
||||
uint8_t id, layer, sr_index, mode;
|
||||
layer = (buffer[1] >> 1) & 3;
|
||||
id = (buffer[1] >> 3) & 3;
|
||||
sr_index = (buffer[2] >> 2) & 3;
|
||||
mode = (buffer[3] >> 6) & 3;
|
||||
if (layer == 0) {
|
||||
return false;
|
||||
}
|
||||
if (id == 1) {
|
||||
return false;
|
||||
}
|
||||
if (sr_index == 3) {
|
||||
return false;
|
||||
int header = U32_AT(buffer);;
|
||||
size_t xingframesize = 0;
|
||||
int sampling_rate = 0;
|
||||
int num_channels;
|
||||
int samples_per_frame = 0;
|
||||
if (!GetMPEGAudioFrameSize(header, &xingframesize, &sampling_rate, &num_channels,
|
||||
NULL, &samples_per_frame)) {
|
||||
return NULL;
|
||||
}
|
||||
seeker->mFirstFramePos += xingframesize;
|
||||
|
||||
uint8_t version = (buffer[1] >> 3) & 3;
|
||||
|
||||
// determine offset of XING header
|
||||
if(id&1) { // mpeg1
|
||||
if (mode != 3) offset += 32;
|
||||
if(version & 1) { // mpeg1
|
||||
if (num_channels != 1) offset += 32;
|
||||
else offset += 17;
|
||||
} else { // mpeg2
|
||||
if (mode != 3) offset += 17;
|
||||
} else { // mpeg 2 or 2.5
|
||||
if (num_channels != 1) offset += 17;
|
||||
else offset += 9;
|
||||
}
|
||||
|
||||
if (source->readAt(offset, &buffer, 4) < 4) { // XING header ID
|
||||
return false;
|
||||
return NULL;
|
||||
}
|
||||
offset += 4;
|
||||
// Check XING ID
|
||||
@@ -152,73 +120,50 @@ static bool parse_xing_header(
|
||||
|| (buffer[2] != 'n') || (buffer[3] != 'g')) {
|
||||
if ((buffer[0] != 'I') || (buffer[1] != 'n')
|
||||
|| (buffer[2] != 'f') || (buffer[3] != 'o')) {
|
||||
return false;
|
||||
return NULL;
|
||||
}
|
||||
}
|
||||
|
||||
if (source->readAt(offset, &buffer, 4) < 4) { // flags
|
||||
return false;
|
||||
return NULL;
|
||||
}
|
||||
offset += 4;
|
||||
uint32_t flags = U32_AT(buffer);
|
||||
|
||||
if (flags & 0x0001) { // Frames field is present
|
||||
if (source->readAt(offset, buffer, 4) < 4) {
|
||||
return false;
|
||||
}
|
||||
if (frame_number) {
|
||||
*frame_number = U32_AT(buffer);
|
||||
}
|
||||
int32_t frame = U32_AT(buffer);
|
||||
// Samples per Frame: 1. index = MPEG Version ID, 2. index = Layer
|
||||
const int samplesPerFrames[2][3] =
|
||||
{
|
||||
{ 384, 1152, 576 }, // MPEG 2, 2.5: layer1, layer2, layer3
|
||||
{ 384, 1152, 1152 }, // MPEG 1: layer1, layer2, layer3
|
||||
};
|
||||
// sampling rates in hertz: 1. index = MPEG Version ID, 2. index = sampling rate index
|
||||
const int samplingRates[4][3] =
|
||||
{
|
||||
{ 11025, 12000, 8000, }, // MPEG 2.5
|
||||
{ 0, 0, 0, }, // reserved
|
||||
{ 22050, 24000, 16000, }, // MPEG 2
|
||||
{ 44100, 48000, 32000, } // MPEG 1
|
||||
};
|
||||
if (duration) {
|
||||
*duration = (int64_t)frame * samplesPerFrames[id&1][3-layer] * 1000000LL
|
||||
/ samplingRates[id][sr_index];
|
||||
return NULL;
|
||||
}
|
||||
int32_t frames = U32_AT(buffer);
|
||||
seeker->mDurationUs = (int64_t)frames * samples_per_frame * 1000000LL / sampling_rate;
|
||||
offset += 4;
|
||||
}
|
||||
if (flags & 0x0002) { // Bytes field is present
|
||||
if (byte_number) {
|
||||
if (source->readAt(offset, buffer, 4) < 4) {
|
||||
return false;
|
||||
}
|
||||
*byte_number = U32_AT(buffer);
|
||||
if (source->readAt(offset, buffer, 4) < 4) {
|
||||
return NULL;
|
||||
}
|
||||
seeker->mSizeBytes = U32_AT(buffer);
|
||||
offset += 4;
|
||||
}
|
||||
if (flags & 0x0004) { // TOC field is present
|
||||
if (table_of_contents) {
|
||||
if (source->readAt(offset + 1, table_of_contents, 99) < 99) {
|
||||
return false;
|
||||
}
|
||||
if (toc_valid) {
|
||||
*toc_valid = true;
|
||||
}
|
||||
if (source->readAt(offset + 1, seeker->mTOC, 99) < 99) {
|
||||
return NULL;
|
||||
}
|
||||
seeker->mTOCValid = true;
|
||||
offset += 100;
|
||||
}
|
||||
|
||||
#if 0
|
||||
if (flags & 0x0008) { // Quality indicator field is present
|
||||
if (quality_indicator) {
|
||||
if (source->readAt(offset, buffer, 4) < 4) {
|
||||
return false;
|
||||
}
|
||||
*quality_indicator = U32_AT(buffer);
|
||||
if (source->readAt(offset, buffer, 4) < 4) {
|
||||
return NULL;
|
||||
}
|
||||
// do something with the quality indicator
|
||||
offset += 4;
|
||||
}
|
||||
return true;
|
||||
#endif
|
||||
|
||||
return seeker;
|
||||
}
|
||||
|
||||
} // namespace android
|
||||
|
||||
Reference in New Issue
Block a user