Merge "Support AAC recording" into kraken

3 files changed, 166 insertions, 25 deletions

diff --git a/media/libstagefright/AudioSource.cpp b/media/libstagefright/AudioSource.cpp
index edabaf9..abd8abc 100644
--- a/media/libstagefright/AudioSource.cpp
+++ b/media/libstagefright/AudioSource.cpp
@@ -14,6 +14,10 @@
  * limitations under the License.
  */
 
+//#define LOG_NDEBUG 0
+#define LOG_TAG "AudioSource"
+#include <utils/Log.h>
+
 #include <media/stagefright/AudioSource.h>
 
 #include <media/AudioRecord.h>
diff --git a/media/libstagefright/MPEG4Writer.cpp b/media/libstagefright/MPEG4Writer.cpp
index 5ff2abe..e6336e7 100644
--- a/media/libstagefright/MPEG4Writer.cpp
+++ b/media/libstagefright/MPEG4Writer.cpp
@@ -14,6 +14,10 @@
  * limitations under the License.
  */
 
+//#define LOG_NDEBUG 0
+#define LOG_TAG "MPEG4Writer"
+#include <utils/Log.h>
+
 #include <arpa/inet.h>
 
 #include <ctype.h>
@@ -58,6 +62,8 @@ private:
     };
     List<SampleInfo> mSampleInfos;
 
+    List<int32_t> mStssTableEntries;
+
     void *mCodecSpecificData;
     size_t mCodecSpecificDataSize;
     bool mGotAllCodecSpecificData;
@@ -522,7 +528,8 @@ void MPEG4Writer::Track::threadEntry() {
     sp<MetaData> meta = mSource->getFormat();
     const char *mime;
     meta->findCString(kKeyMIMEType, &mime);
-    bool is_mpeg4 = !strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_MPEG4);
+    bool is_mpeg4 = !strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_MPEG4) ||
+                    !strcasecmp(mime, MEDIA_MIMETYPE_AUDIO_AAC);
     bool is_avc = !strcasecmp(mime, MEDIA_MIMETYPE_VIDEO_AVC);
     int32_t count = 0;
 
@@ -668,6 +675,9 @@ void MPEG4Writer::Track::threadEntry() {
 
         info.offset = offset;
 
+
+        bool is_audio = !strncasecmp(mime, "audio/", 6);
+
         int64_t timestampUs;
         CHECK(buffer->meta_data()->findInt64(kKeyTime, &timestampUs));
 
@@ -680,6 +690,12 @@ void MPEG4Writer::Track::threadEntry() {
 
         mSampleInfos.push_back(info);
 
+        int32_t isSync = false;
+        buffer->meta_data()->findInt32(kKeyIsSyncFrame, &isSync);
+        if (isSync) {
+            mStssTableEntries.push_back(mSampleInfos.size());
+        }
+        // Our timestamp is in ms.
         buffer->release();
         buffer = NULL;
     }
@@ -735,8 +751,8 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
             success = success && mMeta->findInt32(kKeyHeight, &height);
             CHECK(success);
 
-            mOwner->writeInt32(width);
-            mOwner->writeInt32(height);
+            mOwner->writeInt32(width << 16);   // 32-bit fixed-point value
+            mOwner->writeInt32(height << 16);  // 32-bit fixed-point value
         }
       mOwner->endBox();  // tkhd
 
@@ -754,26 +770,15 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
 
         mOwner->beginBox("hdlr");
           mOwner->writeInt32(0);             // version=0, flags=0
-          mOwner->writeInt32(0);             // predefined
-          mOwner->writeFourcc(is_audio ? "soun" : "vide");
+          mOwner->writeInt32(0);             // component type: should be mhlr
+          mOwner->writeFourcc(is_audio ? "soun" : "vide");  // component subtype
           mOwner->writeInt32(0);             // reserved
           mOwner->writeInt32(0);             // reserved
           mOwner->writeInt32(0);             // reserved
-          mOwner->writeCString("");          // name
+          mOwner->writeCString("SoundHandler");          // name
         mOwner->endBox();
 
         mOwner->beginBox("minf");
-
-          mOwner->beginBox("dinf");
-            mOwner->beginBox("dref");
-              mOwner->writeInt32(0);  // version=0, flags=0
-              mOwner->writeInt32(1);
-              mOwner->beginBox("url ");
-                mOwner->writeInt32(1);  // version=0, flags=1
-              mOwner->endBox();  // url
-            mOwner->endBox();  // dref
-          mOwner->endBox();  // dinf
-
           if (is_audio) {
               mOwner->beginBox("smhd");
               mOwner->writeInt32(0);           // version=0, flags=0
@@ -789,7 +794,18 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
               mOwner->writeInt16(0);
               mOwner->endBox();
           }
-        mOwner->endBox();  // minf
+
+          mOwner->beginBox("dinf");
+            mOwner->beginBox("dref");
+              mOwner->writeInt32(0);  // version=0, flags=0
+              mOwner->writeInt32(1);
+              mOwner->beginBox("url ");
+                mOwner->writeInt32(1);  // version=0, flags=1
+              mOwner->endBox();  // url
+            mOwner->endBox();  // dref
+          mOwner->endBox();  // dinf
+
+       mOwner->endBox();  // minf
 
         mOwner->beginBox("stbl");
 
@@ -802,6 +818,8 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
                     fourcc = "samr";
                 } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AMR_WB, mime)) {
                     fourcc = "sawb";
+                } else if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime)) {
+                    fourcc = "mp4a";
                 } else {
                     LOGE("Unknown mime type '%s'.", mime);
                     CHECK(!"should not be here, unknown mime type.");
@@ -810,10 +828,12 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
                 mOwner->beginBox(fourcc);          // audio format
                   mOwner->writeInt32(0);           // reserved
                   mOwner->writeInt16(0);           // reserved
-                  mOwner->writeInt16(0);           // data ref index
+                  mOwner->writeInt16(0x1);         // data ref index
                   mOwner->writeInt32(0);           // reserved
                   mOwner->writeInt32(0);           // reserved
-                  mOwner->writeInt16(2);           // channel count
+                  int32_t nChannels;
+                  CHECK_EQ(true, mMeta->findInt32(kKeyChannelCount, &nChannels));
+                  mOwner->writeInt16(nChannels);   // channel count
                   mOwner->writeInt16(16);          // sample size
                   mOwner->writeInt16(0);           // predefined
                   mOwner->writeInt16(0);           // reserved
@@ -823,6 +843,38 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
                   CHECK(success);
 
                   mOwner->writeInt32(samplerate << 16);
+                  if (!strcasecmp(MEDIA_MIMETYPE_AUDIO_AAC, mime)) {
+                    mOwner->beginBox("esds");
+
+                        mOwner->writeInt32(0);     // version=0, flags=0
+                        mOwner->writeInt8(0x03);   // ES_DescrTag
+                        mOwner->writeInt8(23 + mCodecSpecificDataSize);
+                        mOwner->writeInt16(0x0000);// ES_ID
+                        mOwner->writeInt8(0x00);
+
+                        mOwner->writeInt8(0x04);   // DecoderConfigDescrTag
+                        mOwner->writeInt8(15 + mCodecSpecificDataSize);
+                        mOwner->writeInt8(0x40);   // objectTypeIndication ISO/IEC 14492-2
+                        mOwner->writeInt8(0x15);   // streamType AudioStream
+
+                        mOwner->writeInt16(0x03);  // XXX
+                        mOwner->writeInt8(0x00);   // buffer size 24-bit
+                        mOwner->writeInt32(96000); // max bit rate
+                        mOwner->writeInt32(96000); // avg bit rate
+
+                        mOwner->writeInt8(0x05);   // DecoderSpecificInfoTag
+                        mOwner->writeInt8(mCodecSpecificDataSize);
+                        mOwner->write(mCodecSpecificData, mCodecSpecificDataSize);
+
+                        static const uint8_t kData2[] = {
+                            0x06,  // SLConfigDescriptorTag
+                            0x01,
+                            0x02
+                        };
+                        mOwner->write(kData2, sizeof(kData2));
+
+                    mOwner->endBox();  // esds
+                  }
                 mOwner->endBox();
             } else {
                 if (!strcasecmp(MEDIA_MIMETYPE_VIDEO_MPEG4, mime)) {
@@ -883,7 +935,7 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
                             0x00, 0x03, 0xe8, 0x00
                         };
                         mOwner->write(kData, sizeof(kData));
-                        
+
                         mOwner->writeInt8(0x05);  // DecoderSpecificInfoTag
 
                         mOwner->writeInt8(mCodecSpecificDataSize);
@@ -943,6 +995,17 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
 
           mOwner->endBox();  // stts
 
+          if (!is_audio) {
+            mOwner->beginBox("stss");
+              mOwner->writeInt32(0);  // version=0, flags=0
+              mOwner->writeInt32(mStssTableEntries.size());  // number of sync frames
+              for (List<int32_t>::iterator it = mStssTableEntries.begin();
+                   it != mStssTableEntries.end(); ++it) {
+                  mOwner->writeInt32(*it);
+              }
+            mOwner->endBox();  // stss
+          }
+
           mOwner->beginBox("stsz");
             mOwner->writeInt32(0);  // version=0, flags=0
             mOwner->writeInt32(0);  // default sample size
@@ -969,7 +1032,7 @@ void MPEG4Writer::Track::writeTrackHeader(int32_t trackID) {
             mOwner->writeInt32(0);  // version=0, flags=0
             mOwner->writeInt32(mSampleInfos.size());
             for (List<SampleInfo>::iterator it = mSampleInfos.begin();
-                 it != mSampleInfos.end(); ++it, ++n) {
+                 it != mSampleInfos.end(); ++it) {
                 mOwner->writeInt64((*it).offset);
             }
           mOwner->endBox();  // co64
diff --git a/media/libstagefright/OMXCodec.cpp b/media/libstagefright/OMXCodec.cpp
index 41be9ac..b7d6d42 100644
--- a/media/libstagefright/OMXCodec.cpp
+++ b/media/libstagefright/OMXCodec.cpp
@@ -520,6 +520,7 @@ status_t OMXCodec::configureCodec(const sp<MetaData> &meta) {
 
         setAACFormat(numChannels, sampleRate);
     }
+
     if (!strncasecmp(mMIME, "video/", 6)) {
         int32_t width, height;
         bool success = meta->findInt32(kKeyWidth, &width);
@@ -567,7 +568,8 @@ status_t OMXCodec::configureCodec(const sp<MetaData> &meta) {
     }
 
     if (!strcmp(mComponentName, "OMX.TI.AMR.encode")
-        || !strcmp(mComponentName, "OMX.TI.WBAMR.encode")) {
+        || !strcmp(mComponentName, "OMX.TI.WBAMR.encode")
+        || !strcmp(mComponentName, "OMX.TI.AAC.encode")) {
         setMinBufferSize(kPortIndexOutput, 8192);  // XXX
     }
 
@@ -708,7 +710,7 @@ void OMXCodec::setVideoInputFormat(
 
     OMX_COLOR_FORMATTYPE colorFormat = OMX_COLOR_FormatYUV420SemiPlanar;
     if (!strcasecmp("OMX.TI.Video.encoder", mComponentName)) {
-        colorFormat = OMX_COLOR_FormatYUV420Planar;
+        colorFormat = OMX_COLOR_FormatYCbYCr;
     }
 
 
@@ -2127,11 +2129,24 @@ void OMXCodec::setState(State newState) {
 
 void OMXCodec::setRawAudioFormat(
         OMX_U32 portIndex, int32_t sampleRate, int32_t numChannels) {
+
+    // port definition
+    OMX_PARAM_PORTDEFINITIONTYPE def;
+    InitOMXParams(&def);
+    def.nPortIndex = portIndex;
+    status_t err = mOMX->getParameter(
+            mNode, OMX_IndexParamPortDefinition, &def, sizeof(def));
+    CHECK_EQ(err, OK);
+    def.format.audio.eEncoding = OMX_AUDIO_CodingPCM;
+    CHECK_EQ(mOMX->setParameter(mNode, OMX_IndexParamPortDefinition,
+            &def, sizeof(def)), OK);
+
+    // pcm param
     OMX_AUDIO_PARAM_PCMMODETYPE pcmParams;
     InitOMXParams(&pcmParams);
     pcmParams.nPortIndex = portIndex;
 
-    status_t err = mOMX->getParameter(
+    err = mOMX->getParameter(
             mNode, OMX_IndexParamAudioPcm, &pcmParams, sizeof(pcmParams));
 
     CHECK_EQ(err, OK);
@@ -2171,6 +2186,8 @@ void OMXCodec::setAMRFormat(bool isWAMR) {
     CHECK_EQ(err, OK);
 
     def.eAMRFrameFormat = OMX_AUDIO_AMRFrameFormatFSF;
+
+    // XXX: Select bandmode based on bit rate
     def.eAMRBandMode =
         isWAMR ? OMX_AUDIO_AMRBandModeWB0 : OMX_AUDIO_AMRBandModeNB0;
 
@@ -2191,8 +2208,60 @@ void OMXCodec::setAMRFormat(bool isWAMR) {
 }
 
 void OMXCodec::setAACFormat(int32_t numChannels, int32_t sampleRate) {
+    CHECK(numChannels == 1 || numChannels == 2);
     if (mIsEncoder) {
+        //////////////// input port ////////////////////
         setRawAudioFormat(kPortIndexInput, sampleRate, numChannels);
+
+        //////////////// output port ////////////////////
+        // format
+        OMX_AUDIO_PARAM_PORTFORMATTYPE format;
+        format.nPortIndex = kPortIndexOutput;
+        format.nIndex = 0;
+        status_t err = OMX_ErrorNone;
+        while (OMX_ErrorNone == err) {
+            CHECK_EQ(mOMX->getParameter(mNode, OMX_IndexParamAudioPortFormat,
+                    &format, sizeof(format)), OK);
+            if (format.eEncoding == OMX_AUDIO_CodingAAC) {
+                break;
+            }
+            format.nIndex++;
+        }
+        CHECK_EQ(OK, err);
+        CHECK_EQ(mOMX->setParameter(mNode, OMX_IndexParamAudioPortFormat,
+                &format, sizeof(format)), OK);
+
+        // port definition
+        OMX_PARAM_PORTDEFINITIONTYPE def;
+        InitOMXParams(&def);
+        def.nPortIndex = kPortIndexOutput;
+        CHECK_EQ(mOMX->getParameter(mNode, OMX_IndexParamPortDefinition,
+                &def, sizeof(def)), OK);
+        def.format.audio.bFlagErrorConcealment = OMX_TRUE;
+        def.format.audio.eEncoding = OMX_AUDIO_CodingAAC;
+        CHECK_EQ(mOMX->setParameter(mNode, OMX_IndexParamPortDefinition,
+                &def, sizeof(def)), OK);
+
+        // profile
+        OMX_AUDIO_PARAM_AACPROFILETYPE profile;
+        InitOMXParams(&profile);
+        profile.nPortIndex = kPortIndexOutput;
+        CHECK_EQ(mOMX->getParameter(mNode, OMX_IndexParamAudioAac,
+                &profile, sizeof(profile)), OK);
+        profile.nChannels = numChannels;
+        profile.eChannelMode = (numChannels == 1?
+                OMX_AUDIO_ChannelModeMono: OMX_AUDIO_ChannelModeStereo);
+        profile.nSampleRate = sampleRate;
+        profile.nBitRate = 96000;   // XXX
+        profile.nAudioBandWidth = 0;
+        profile.nFrameLength = 0;
+        profile.nAACtools = OMX_AUDIO_AACToolAll;
+        profile.nAACERtools = OMX_AUDIO_AACERNone;
+        profile.eAACProfile = OMX_AUDIO_AACObjectLC;
+        profile.eAACStreamFormat = OMX_AUDIO_AACStreamFormatMP4FF;
+        CHECK_EQ(mOMX->setParameter(mNode, OMX_IndexParamAudioAac,
+                &profile, sizeof(profile)), OK);
+
     } else {
         OMX_AUDIO_PARAM_AACPROFILETYPE profile;
         InitOMXParams(&profile);
@@ -2961,6 +3030,11 @@ void OMXCodec::initOutputFormat(const sp<MetaData> &inputFormat) {
             } else if (audio_def->eEncoding == OMX_AUDIO_CodingAAC) {
                 mOutputFormat->setCString(
                         kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_AAC);
+                int32_t numChannels, sampleRate;
+                inputFormat->findInt32(kKeyChannelCount, &numChannels);
+                inputFormat->findInt32(kKeySampleRate, &sampleRate);
+                mOutputFormat->setInt32(kKeyChannelCount, numChannels);
+                mOutputFormat->setInt32(kKeySampleRate, sampleRate);
             } else {
                 CHECK(!"Should not be here. Unknown audio encoding.");
             }