Skip to content

Commit 4fc769e

Browse files
marconemarcone
committed
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
1 parent b411579 commit 4fc769e

File tree

3 files changed

+59
-107
lines changed

3 files changed

+59
-107
lines changed

‎media/libstagefright/MP3Extractor.cpp‎

Lines changed: 7 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -317,6 +317,13 @@ MP3Extractor::MP3Extractor(
317317
mSeeker = VBRISeeker::CreateFromSource(mDataSource, post_id3_pos);
318318
}
319319

320+
if (mSeeker != NULL) {
321+
// While it is safe to send the XING/VBRI frame to the decoder, this will
322+
// result in an extra 1152 samples being output. The real first frame to
323+
// decode is after the XING/VBRI frame, so skip there.
324+
mFirstFramePos += frame_size;
325+
}
326+
320327
int64_t durationUs;
321328

322329
if (mSeeker == NULL || !mSeeker->getDuration(&durationUs)) {

‎media/libstagefright/VBRISeeker.cpp‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -92,7 +92,7 @@ sp<VBRISeeker> VBRISeeker::CreateFromSource(
9292
}
9393

9494
sp<VBRISeeker> seeker = new VBRISeeker;
95-
seeker->mBasePos = post_id3_pos;
95+
seeker->mBasePos = post_id3_pos + frameSize;
9696
seeker->mDurationUs = durationUs;
9797

9898
off64_t offset = post_id3_pos;

‎media/libstagefright/XINGSeeker.cpp‎

Lines changed: 51 additions & 106 deletions
Original file line numberDiff line numberDiff line change
@@ -15,35 +15,13 @@
1515
*/
1616

1717
#include "include/XINGSeeker.h"
18+
#include "include/avc_utils.h"
1819

1920
#include <media/stagefright/DataSource.h>
2021
#include <media/stagefright/Utils.h>
2122

2223
namespace android {
2324

24-
static bool parse_xing_header(
25-
const sp<DataSource> &source, off64_t first_frame_pos,
26-
int32_t *frame_number = NULL, int32_t *byte_number = NULL,
27-
unsigned char *table_of_contents = NULL, bool *toc_is_valid = NULL,
28-
int32_t *quality_indicator = NULL, int64_t *duration = NULL);
29-
30-
// static
31-
sp<XINGSeeker> XINGSeeker::CreateFromSource(
32-
const sp<DataSource> &source, off64_t first_frame_pos) {
33-
sp<XINGSeeker> seeker = new XINGSeeker;
34-
35-
seeker->mFirstFramePos = first_frame_pos;
36-
37-
if (!parse_xing_header(
38-
source, first_frame_pos,
39-
NULL, &seeker->mSizeBytes, seeker->mTOC, &seeker->mTOCValid,
40-
NULL, &seeker->mDurationUs)) {
41-
return NULL;
42-
}
43-
44-
return seeker;
45-
}
46-
4725
XINGSeeker::XINGSeeker()
4826
: mDurationUs(-1),
4927
mSizeBytes(0) {
@@ -91,134 +69,101 @@ bool XINGSeeker::getOffsetForTime(int64_t *timeUs, off64_t *pos) {
9169
return true;
9270
}
9371

94-
static bool parse_xing_header(
95-
const sp<DataSource> &source, off64_t first_frame_pos,
96-
int32_t *frame_number, int32_t *byte_number,
97-
unsigned char *table_of_contents, bool *toc_valid,
98-
int32_t *quality_indicator,
99-
int64_t *duration) {
100-
if (frame_number) {
101-
*frame_number = 0;
102-
}
103-
if (byte_number) {
104-
*byte_number = 0;
105-
}
106-
if (toc_valid) {
107-
*toc_valid = false;
108-
}
109-
if (quality_indicator) {
110-
*quality_indicator = 0;
111-
}
112-
if (duration) {
113-
*duration = 0;
114-
}
72+
// static
73+
sp<XINGSeeker> XINGSeeker::CreateFromSource(
74+
const sp<DataSource> &source, off64_t first_frame_pos) {
75+
sp<XINGSeeker> seeker = new XINGSeeker;
76+
77+
seeker->mFirstFramePos = first_frame_pos;
78+
79+
ALOGI("xingseeker first frame pos: %lld", first_frame_pos);
80+
81+
seeker->mSizeBytes = 0;
82+
seeker->mTOCValid = false;
83+
seeker->mDurationUs = 0;
11584

11685
uint8_t buffer[4];
11786
int offset = first_frame_pos;
11887
if (source->readAt(offset, &buffer, 4) < 4) { // get header
119-
return false;
88+
return NULL;
12089
}
12190
offset += 4;
12291

123-
uint8_t id, layer, sr_index, mode;
124-
layer = (buffer[1] >> 1) & 3;
125-
id = (buffer[1] >> 3) & 3;
126-
sr_index = (buffer[2] >> 2) & 3;
127-
mode = (buffer[3] >> 6) & 3;
128-
if (layer == 0) {
129-
return false;
130-
}
131-
if (id == 1) {
132-
return false;
133-
}
134-
if (sr_index == 3) {
135-
return false;
92+
int header = U32_AT(buffer);;
93+
size_t xingframesize = 0;
94+
int sampling_rate = 0;
95+
int num_channels;
96+
int samples_per_frame = 0;
97+
if (!GetMPEGAudioFrameSize(header, &xingframesize, &sampling_rate, &num_channels,
98+
NULL, &samples_per_frame)) {
99+
return NULL;
136100
}
101+
seeker->mFirstFramePos += xingframesize;
102+
103+
uint8_t version = (buffer[1] >> 3) & 3;
104+
137105
// determine offset of XING header
138-
if(id&1) { // mpeg1
139-
if (mode != 3) offset += 32;
106+
if(version & 1) { // mpeg1
107+
if (num_channels != 1) offset += 32;
140108
else offset += 17;
141-
} else { // mpeg2
142-
if (mode != 3) offset += 17;
109+
} else { // mpeg 2 or 2.5
110+
if (num_channels != 1) offset += 17;
143111
else offset += 9;
144112
}
145113

146114
if (source->readAt(offset, &buffer, 4) < 4) { // XING header ID
147-
return false;
115+
return NULL;
148116
}
149117
offset += 4;
150118
// Check XING ID
151119
if ((buffer[0] != 'X') || (buffer[1] != 'i')
152120
|| (buffer[2] != 'n') || (buffer[3] != 'g')) {
153121
if ((buffer[0] != 'I') || (buffer[1] != 'n')
154122
|| (buffer[2] != 'f') || (buffer[3] != 'o')) {
155-
return false;
123+
return NULL;
156124
}
157125
}
158126

159127
if (source->readAt(offset, &buffer, 4) < 4) { // flags
160-
return false;
128+
return NULL;
161129
}
162130
offset += 4;
163131
uint32_t flags = U32_AT(buffer);
164132

165133
if (flags & 0x0001) { // Frames field is present
166134
if (source->readAt(offset, buffer, 4) < 4) {
167-
return false;
168-
}
169-
if (frame_number) {
170-
*frame_number = U32_AT(buffer);
171-
}
172-
int32_t frame = U32_AT(buffer);
173-
// Samples per Frame: 1. index = MPEG Version ID, 2. index = Layer
174-
const int samplesPerFrames[2][3] =
175-
{
176-
{ 384, 1152, 576 }, // MPEG 2, 2.5: layer1, layer2, layer3
177-
{ 384, 1152, 1152 }, // MPEG 1: layer1, layer2, layer3
178-
};
179-
// sampling rates in hertz: 1. index = MPEG Version ID, 2. index = sampling rate index
180-
const int samplingRates[4][3] =
181-
{
182-
{ 11025, 12000, 8000, }, // MPEG 2.5
183-
{ 0, 0, 0, }, // reserved
184-
{ 22050, 24000, 16000, }, // MPEG 2
185-
{ 44100, 48000, 32000, } // MPEG 1
186-
};
187-
if (duration) {
188-
*duration = (int64_t)frame * samplesPerFrames[id&1][3-layer] * 1000000LL
189-
/ samplingRates[id][sr_index];
135+
return NULL;
190136
}
137+
int32_t frames = U32_AT(buffer);
138+
seeker->mDurationUs = (int64_t)frames * samples_per_frame * 1000000LL / sampling_rate;
191139
offset += 4;
192140
}
193141
if (flags & 0x0002) { // Bytes field is present
194-
if (byte_number) {
195-
if (source->readAt(offset, buffer, 4) < 4) {
196-
return false;
197-
}
198-
*byte_number = U32_AT(buffer);
142+
if (source->readAt(offset, buffer, 4) < 4) {
143+
return NULL;
199144
}
145+
seeker->mSizeBytes = U32_AT(buffer);
200146
offset += 4;
201147
}
202148
if (flags & 0x0004) { // TOC field is present
203-
if (table_of_contents) {
204-
if (source->readAt(offset + 1, table_of_contents, 99) < 99) {
205-
return false;
206-
}
207-
if (toc_valid) {
208-
*toc_valid = true;
209-
}
149+
if (source->readAt(offset + 1, seeker->mTOC, 99) < 99) {
150+
return NULL;
210151
}
152+
seeker->mTOCValid = true;
211153
offset += 100;
212154
}
155+
156+
#if 0
213157
if (flags & 0x0008) { // Quality indicator field is present
214-
if (quality_indicator) {
215-
if (source->readAt(offset, buffer, 4) < 4) {
216-
return false;
217-
}
218-
*quality_indicator = U32_AT(buffer);
158+
if (source->readAt(offset, buffer, 4) < 4) {
159+
return NULL;
219160
}
161+
// do something with the quality indicator
162+
offset += 4;
220163
}
221-
return true;
164+
#endif
165+
166+
return seeker;
222167
}
223168

224169
} // namespace android

0 commit comments

Comments
 (0)