Merge pull request #830 from trav90/media-work

Several WebM/VPX fixes

2 files changed, 98 insertions, 61 deletions

diff --git a/dom/media/webm/WebMDemuxer.cpp b/dom/media/webm/WebMDemuxer.cpp
index 20ed71581..b54739c06 100644
--- a/dom/media/webm/WebMDemuxer.cpp
+++ b/dom/media/webm/WebMDemuxer.cpp
@@ -9,6 +9,7 @@
 #include "AbstractMediaDecoder.h"
 #include "MediaResource.h"
 #include "OpusDecoder.h"
+#include "VPXDecoder.h"
 #include "WebMDemuxer.h"
 #include "WebMBufferedParser.h"
 #include "gfx2DGlue.h"
@@ -24,12 +25,9 @@
 #include "mozilla/Sprintf.h"
 
 #include <algorithm>
+#include <numeric>
 #include <stdint.h>
 
-#define VPX_DONT_DEFINE_STDINT_TYPES
-#include "vpx/vp8dx.h"
-#include "vpx/vpx_decoder.h"
-
 #define WEBM_DEBUG(arg, ...) MOZ_LOG(gMediaDemuxerLog, mozilla::LogLevel::Debug, ("WebMDemuxer(%p)::%s: " arg, this, __func__, ##__VA_ARGS__))
 extern mozilla::LazyLogModule gMediaDemuxerLog;
 
@@ -549,7 +547,7 @@ WebMDemuxer::GetTrackCrypto(TrackInfo::TrackType aType, size_t aTrackNumber) {
   return crypto;
 }
 
-bool
+nsresult
 WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSamples)
 {
   if (mIsMediaSource) {
@@ -557,17 +555,18 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
     EnsureUpToDateIndex();
   }
 
-  RefPtr<NesteggPacketHolder> holder(NextPacket(aType));
+  RefPtr<NesteggPacketHolder> holder;
+  nsresult rv = NextPacket(aType, holder);
 
-  if (!holder) {
-    return false;
+  if (NS_FAILED(rv)) {
+    return rv;
   }
 
   int r = 0;
   unsigned int count = 0;
   r = nestegg_packet_count(holder->Packet(), &count);
   if (r == -1) {
-    return false;
+    return NS_ERROR_DOM_MEDIA_DEMUXER_ERR;
   }
   int64_t tstamp = holder->Timestamp();
   int64_t duration = holder->Duration();
@@ -578,7 +577,11 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
   // video frame.
   int64_t next_tstamp = INT64_MIN;
   if (aType == TrackInfo::kAudioTrack) {
-    RefPtr<NesteggPacketHolder> next_holder(NextPacket(aType));
+    RefPtr<NesteggPacketHolder> next_holder;
+    rv = NextPacket(aType, next_holder);
+    if (NS_FAILED(rv) && rv != NS_ERROR_DOM_MEDIA_END_OF_STREAM) {
+      return rv;
+    }
     if (next_holder) {
       next_tstamp = next_holder->Timestamp();
       PushAudioPacket(next_holder);
@@ -593,7 +596,11 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
     }
     mLastAudioFrameTime = Some(tstamp);
   } else if (aType == TrackInfo::kVideoTrack) {
-    RefPtr<NesteggPacketHolder> next_holder(NextPacket(aType));
+    RefPtr<NesteggPacketHolder> next_holder;
+    rv = NextPacket(aType, next_holder);
+    if (NS_FAILED(rv) && rv != NS_ERROR_DOM_MEDIA_END_OF_STREAM) {
+      return rv;
+    }
     if (next_holder) {
       next_tstamp = next_holder->Timestamp();
       PushVideoPacket(next_holder);
@@ -610,7 +617,7 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
   }
 
   if (mIsMediaSource && next_tstamp == INT64_MIN) {
-    return false;
+    return NS_ERROR_DOM_MEDIA_END_OF_STREAM;
   }
 
   int64_t discardPadding = 0;
@@ -626,7 +633,7 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
     r = nestegg_packet_data(holder->Packet(), i, &data, &length);
     if (r == -1) {
       WEBM_DEBUG("nestegg_packet_data failed r=%d", r);
-      return false;
+      return NS_ERROR_DOM_MEDIA_DEMUXER_ERR;
     }
     bool isKeyframe = false;
     if (aType == TrackInfo::kAudioTrack) {
@@ -636,29 +643,31 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
         // Packet is encrypted, can't peek, use packet info
         isKeyframe = nestegg_packet_has_keyframe(holder->Packet()) == NESTEGG_PACKET_HAS_KEYFRAME_TRUE;
       } else {
-        vpx_codec_stream_info_t si;
-        PodZero(&si);
-        si.sz = sizeof(si);
+        auto sample = MakeSpan(data, length);
         switch (mVideoCodec) {
         case NESTEGG_CODEC_VP8:
-          vpx_codec_peek_stream_info(vpx_codec_vp8_dx(), data, length, &si);
+          isKeyframe = VPXDecoder::IsKeyframe(sample, VPXDecoder::Codec::VP8);
           break;
         case NESTEGG_CODEC_VP9:
-          vpx_codec_peek_stream_info(vpx_codec_vp9_dx(), data, length, &si);
+          isKeyframe = VPXDecoder::IsKeyframe(sample, VPXDecoder::Codec::VP9);
           break;
+        default:
+          NS_WARNING("Cannot detect keyframes in unknown WebM video codec");
+          return NS_ERROR_FAILURE;
         }
-        isKeyframe = si.is_kf;
         if (isKeyframe) {
-          // We only look for resolution changes on keyframes for both VP8 and
-          // VP9. Other resolution changes are invalid.
-          if (mLastSeenFrameWidth.isSome() && mLastSeenFrameHeight.isSome() &&
-            (si.w != mLastSeenFrameWidth.value() ||
-              si.h != mLastSeenFrameHeight.value())) {
-            mInfo.mVideo.mDisplay = nsIntSize(si.w, si.h);
+          // For both VP8 and VP9, we only look for resolution changes
+          // on keyframes. Other resolution changes are invalid.
+          auto codec = mVideoCodec == NESTEGG_CODEC_VP8
+                       ? VPXDecoder::Codec::VP8
+                       : VPXDecoder::Codec::VP9;
+          auto dimensions = VPXDecoder::GetFrameSize(sample, codec);
+          if (mLastSeenFrameSize.isSome()
+              && (dimensions != mLastSeenFrameSize.value())) {
+            mInfo.mVideo.mDisplay = dimensions;
             mSharedVideoTrackInfo = new SharedTrackInfo(mInfo.mVideo, ++sStreamSourceID);
           }
-          mLastSeenFrameWidth = Some(si.w);
-          mLastSeenFrameHeight = Some(si.h);
+          mLastSeenFrameSize = Some(dimensions);
         }
       }
     }
@@ -668,7 +677,7 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
     RefPtr<MediaRawData> sample = new MediaRawData(data, length);
     if (length && !sample->Data()) {
       // OOM.
-      return false;
+      return NS_ERROR_OUT_OF_MEMORY;
     }
     sample->mTimecode = tstamp;
     sample->mTime = tstamp;
@@ -721,11 +730,12 @@ WebMDemuxer::GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSampl
     }
     aSamples->Push(sample);
   }
-  return true;
+  return NS_OK;
 }
 
-RefPtr<NesteggPacketHolder>
-WebMDemuxer::NextPacket(TrackInfo::TrackType aType)
+nsresult
+WebMDemuxer::NextPacket(TrackInfo::TrackType aType,
+                        RefPtr<NesteggPacketHolder>& aPacket)
 {
   bool isVideo = aType == TrackInfo::kVideoTrack;
 
@@ -734,56 +744,64 @@ WebMDemuxer::NextPacket(TrackInfo::TrackType aType)
   bool hasType = isVideo ? mHasVideo : mHasAudio;
 
   if (!hasType) {
-    return nullptr;
+    return NS_ERROR_DOM_MEDIA_DEMUXER_ERR;
   }
 
   // The packet queue for the type that we are interested in.
   WebMPacketQueue &packets = isVideo ? mVideoPackets : mAudioPackets;
 
   if (packets.GetSize() > 0) {
-    return packets.PopFront();
+    aPacket = packets.PopFront();
+    return NS_OK;
   }
 
   // Track we are interested in
   uint32_t ourTrack = isVideo ? mVideoTrack : mAudioTrack;
 
   do {
-    RefPtr<NesteggPacketHolder> holder = DemuxPacket(aType);
+    RefPtr<NesteggPacketHolder> holder;
+    nsresult rv = DemuxPacket(aType, holder);
+    if (NS_FAILED(rv)) {
+      return rv;
+    }
     if (!holder) {
-      return nullptr;
+      return NS_ERROR_DOM_MEDIA_DEMUXER_ERR;
     }
 
     if (ourTrack == holder->Track()) {
-      return holder;
+      aPacket = holder;
+      return NS_OK;
     }
   } while (true);
 }
 
-RefPtr<NesteggPacketHolder>
-WebMDemuxer::DemuxPacket(TrackInfo::TrackType aType)
+nsresult
+WebMDemuxer::DemuxPacket(TrackInfo::TrackType aType,
+                         RefPtr<NesteggPacketHolder>& aPacket)
 {
   nestegg_packet* packet;
   int r = nestegg_read_packet(Context(aType), &packet);
   if (r == 0) {
     nestegg_read_reset(Context(aType));
-    return nullptr;
+    return NS_ERROR_DOM_MEDIA_END_OF_STREAM;
   } else if (r < 0) {
-    return nullptr;
+    return NS_ERROR_DOM_MEDIA_DEMUXER_ERR;
   }
 
   unsigned int track = 0;
   r = nestegg_packet_track(packet, &track);
   if (r == -1) {
-    return nullptr;
+    return NS_ERROR_DOM_MEDIA_DEMUXER_ERR;
   }
 
   int64_t offset = Resource(aType).Tell();
   RefPtr<NesteggPacketHolder> holder = new NesteggPacketHolder();
   if (!holder->Init(packet, offset, track, false)) {
-    return nullptr;
+    return NS_ERROR_DOM_MEDIA_DEMUXER_ERR;
   }
 
-  return holder;
+  aPacket = holder;
+  return NS_OK;
 }
 
 void
@@ -943,7 +961,14 @@ WebMTrackDemuxer::Seek(media::TimeUnit aTime)
   media::TimeUnit seekTime = aTime;
   mSamples.Reset();
   mParent->SeekInternal(mType, aTime);
-  mParent->GetNextPacket(mType, &mSamples);
+  nsresult rv = mParent->GetNextPacket(mType, &mSamples);
+  if (NS_FAILED(rv)) {
+    if (rv == NS_ERROR_DOM_MEDIA_END_OF_STREAM) {
+      // Ignore the error for now, the next GetSample will be rejected with EOS.
+      return SeekPromise::CreateAndResolve(media::TimeUnit(), __func__);
+    }
+    return SeekPromise::CreateAndReject(rv, __func__);
+  }
   mNeedKeyframe = true;
 
   // Check what time we actually seeked to.
@@ -956,15 +981,18 @@ WebMTrackDemuxer::Seek(media::TimeUnit aTime)
   return SeekPromise::CreateAndResolve(seekTime, __func__);
 }
 
-RefPtr<MediaRawData>
-WebMTrackDemuxer::NextSample()
+nsresult
+WebMTrackDemuxer::NextSample(RefPtr<MediaRawData>& aData)
 {
-  while (mSamples.GetSize() < 1 && mParent->GetNextPacket(mType, &mSamples)) {
+  nsresult rv;
+  while (mSamples.GetSize() < 1 &&
+         NS_SUCCEEDED((rv = mParent->GetNextPacket(mType, &mSamples)))) {
   }
   if (mSamples.GetSize()) {
-    return mSamples.PopFront();
+    aData = mSamples.PopFront();
+    return NS_OK;
   }
-  return nullptr;
+  return rv;
 }
 
 RefPtr<WebMTrackDemuxer::SamplesPromise>
@@ -973,9 +1001,12 @@ WebMTrackDemuxer::GetSamples(int32_t aNumSamples)
   RefPtr<SamplesHolder> samples = new SamplesHolder;
   MOZ_ASSERT(aNumSamples);
 
+  nsresult rv = NS_ERROR_DOM_MEDIA_END_OF_STREAM;
+
   while (aNumSamples) {
-    RefPtr<MediaRawData> sample(NextSample());
-    if (!sample) {
+    RefPtr<MediaRawData> sample;
+    rv = NextSample(sample);
+    if (NS_FAILED(rv)) {
       break;
     }
     if (mNeedKeyframe && !sample->mKeyframe) {
@@ -987,7 +1018,7 @@ WebMTrackDemuxer::GetSamples(int32_t aNumSamples)
   }
 
   if (samples->mSamples.IsEmpty()) {
-    return SamplesPromise::CreateAndReject(NS_ERROR_DOM_MEDIA_END_OF_STREAM, __func__);
+    return SamplesPromise::CreateAndReject(rv, __func__);
   } else {
     UpdateSamples(samples->mSamples);
     return SamplesPromise::CreateAndResolve(samples, __func__);
@@ -1022,7 +1053,8 @@ WebMTrackDemuxer::SetNextKeyFrameTime()
   }
   // Demux and buffer frames until we find a keyframe.
   RefPtr<MediaRawData> sample;
-  while (!foundKeyframe && (sample = NextSample())) {
+  nsresult rv = NS_OK;
+  while (!foundKeyframe && NS_SUCCEEDED((rv = NextSample(sample)))) {
     if (sample->mKeyframe) {
       frameTime = sample->mTime;
       foundKeyframe = true;
@@ -1104,10 +1136,11 @@ WebMTrackDemuxer::SkipToNextRandomAccessPoint(media::TimeUnit aTimeThreshold)
   uint32_t parsed = 0;
   bool found = false;
   RefPtr<MediaRawData> sample;
+  nsresult rv = NS_OK;
   int64_t sampleTime;
 
   WEBM_DEBUG("TimeThreshold: %f", aTimeThreshold.ToSeconds());
-  while (!found && (sample = NextSample())) {
+  while (!found && NS_SUCCEEDED((rv = NextSample(sample)))) {
     parsed++;
     sampleTime = sample->mTime;
     if (sample->mKeyframe && sampleTime >= aTimeThreshold.ToMicroseconds()) {
@@ -1116,7 +1149,9 @@ WebMTrackDemuxer::SkipToNextRandomAccessPoint(media::TimeUnit aTimeThreshold)
       mSamples.PushFront(sample.forget());
     }
   }
-  SetNextKeyFrameTime();
+  if (NS_SUCCEEDED(rv)) {
+    SetNextKeyFrameTime();
+  }
   if (found) {
     WEBM_DEBUG("next sample: %f (parsed: %d)",
                media::TimeUnit::FromMicroseconds(sampleTime).ToSeconds(),
diff --git a/dom/media/webm/WebMDemuxer.h b/dom/media/webm/WebMDemuxer.h
index 6fff38e7d..09780e8d3 100644
--- a/dom/media/webm/WebMDemuxer.h
+++ b/dom/media/webm/WebMDemuxer.h
@@ -111,7 +111,8 @@ public:
   bool GetOffsetForTime(uint64_t aTime, int64_t* aOffset);
 
   // Demux next WebM packet and append samples to MediaRawDataQueue
-  bool GetNextPacket(TrackInfo::TrackType aType, MediaRawDataQueue *aSamples);
+  nsresult GetNextPacket(TrackInfo::TrackType aType,
+                         MediaRawDataQueue *aSamples);
 
   nsresult Reset(TrackInfo::TrackType aType);
 
@@ -175,11 +176,13 @@ private:
   // Read a packet from the nestegg file. Returns nullptr if all packets for
   // the particular track have been read. Pass TrackInfo::kVideoTrack or
   // TrackInfo::kVideoTrack to indicate the type of the packet we want to read.
-  RefPtr<NesteggPacketHolder> NextPacket(TrackInfo::TrackType aType);
+  nsresult NextPacket(TrackInfo::TrackType aType,
+                      RefPtr<NesteggPacketHolder>& aPacket);
 
   // Internal method that demuxes the next packet from the stream. The caller
   // is responsible for making sure it doesn't get lost.
-  RefPtr<NesteggPacketHolder> DemuxPacket(TrackInfo::TrackType aType);
+  nsresult DemuxPacket(TrackInfo::TrackType aType,
+                       RefPtr<NesteggPacketHolder>& aPacket);
 
   // libnestegg audio and video context for webm container.
   // Access on reader's thread only.
@@ -237,8 +240,7 @@ private:
   int64_t mLastWebMBlockOffset;
   const bool mIsMediaSource;
 
-  Maybe<uint32_t> mLastSeenFrameWidth;
-  Maybe<uint32_t> mLastSeenFrameHeight;
+  Maybe<nsIntSize> mLastSeenFrameSize;
   // This will be populated only if a resolution change occurs, otherwise it
   // will be left as null so the original metadata is used
   RefPtr<SharedTrackInfo> mSharedVideoTrackInfo;
@@ -276,7 +278,7 @@ private:
   ~WebMTrackDemuxer();
   void UpdateSamples(nsTArray<RefPtr<MediaRawData>>& aSamples);
   void SetNextKeyFrameTime();
-  RefPtr<MediaRawData> NextSample ();
+  nsresult NextSample(RefPtr<MediaRawData>& aData);
   RefPtr<WebMDemuxer> mParent;
   TrackInfo::TrackType mType;
   UniquePtr<TrackInfo> mInfo;