Merge to M43: "Disable splices for overlap of less than a millisecond."

There appear to be frequent cases where YT is overlapping appends by a
few hundred microseconds which are not actually overlapped.  I believe
this to be a limitation of the muxed timestamps, i.e. we frequently see
things with only millisecond granularity.

To avoid glitches and pops in these cases, disable splice frames unless
we have at least a millisecond of data available.

BUG=484414
TEST=new unittest. No glitches on YT or Play Music.
TBR=chcunningham,sandersd

Review URL: https://codereview.chromium.org/1134533004

Cr-Commit-Position: refs/heads/master@{#329063}
(cherry picked from commit 62a5bb3d60f9863666194da9446d467c69fe3280)

Review URL: https://codereview.chromium.org/1138113002

Cr-Commit-Position: refs/branch-heads/2357@{#356}
Cr-Branched-From: 59d4494849b405682265ed5d3f5164573b9a939b-refs/heads/master@{#323860}

2 files changed, 37 insertions, 5 deletions

diff --git a/media/filters/source_buffer_stream.cc b/media/filters/source_buffer_stream.cc
index fb699db..3d81409 100644
--- a/media/filters/source_buffer_stream.cc
+++ b/media/filters/source_buffer_stream.cc
@@ -1536,14 +1536,16 @@ void SourceBufferStream::GenerateSpliceFrame(const BufferQueue& new_buffers) {
     }
   }
 
-  // Don't generate splice frames which represent less than two frames, since we
-  // need at least that much to generate a crossfade.  Per the spec, make this
-  // check using the sample rate of the overlapping buffers.
+  // Don't generate splice frames which represent less than a millisecond (which
+  // is frequently the extent of timestamp resolution for poorly encoded media)
+  // or less than two frames (need at least two to crossfade).
   const base::TimeDelta splice_duration =
       pre_splice_buffers.back()->timestamp() +
       pre_splice_buffers.back()->duration() - splice_timestamp;
-  const base::TimeDelta minimum_splice_duration = base::TimeDelta::FromSecondsD(
-      2.0 / audio_configs_[append_config_index_].samples_per_second());
+  const base::TimeDelta minimum_splice_duration = std::max(
+      base::TimeDelta::FromMilliseconds(1),
+      base::TimeDelta::FromSecondsD(
+          2.0 / audio_configs_[append_config_index_].samples_per_second()));
   if (splice_duration < minimum_splice_duration) {
     DVLOG(1) << "Can't generate splice: not enough samples for crossfade; have "
              << splice_duration.InMicroseconds() << " us, but need "
@@ -1551,6 +1553,9 @@ void SourceBufferStream::GenerateSpliceFrame(const BufferQueue& new_buffers) {
     return;
   }
 
+  DVLOG(1) << "Generating splice frame @ " << new_buffers.front()->timestamp()
+           << ", splice duration: " << splice_duration.InMicroseconds()
+           << " us";
   new_buffers.front()->ConvertToSpliceBuffer(pre_splice_buffers);
 }
 
diff --git a/media/filters/source_buffer_stream_unittest.cc b/media/filters/source_buffer_stream_unittest.cc
index 87509c2..f3914b4 100644
--- a/media/filters/source_buffer_stream_unittest.cc
+++ b/media/filters/source_buffer_stream_unittest.cc
@@ -3909,6 +3909,33 @@ TEST_F(SourceBufferStreamTest, Audio_SpliceFrame_NoTinySplices) {
   CheckNoNextBuffer();
 }
 
+TEST_F(SourceBufferStreamTest, Audio_SpliceFrame_NoMillisecondSplices) {
+  video_config_ = TestVideoConfig::Invalid();
+  audio_config_.Initialize(kCodecVorbis, kSampleFormatPlanarF32,
+                           CHANNEL_LAYOUT_STEREO, 4000, NULL, 0, false, false,
+                           base::TimeDelta(), 0);
+  stream_.reset(new SourceBufferStream(audio_config_, log_cb(), true));
+  // Equivalent to 0.5ms per frame.
+  SetStreamInfo(2000, 2000);
+  Seek(0);
+
+  // Append four buffers with a 0.5ms duration each.
+  NewSegmentAppend(0, 4);
+  CheckExpectedRangesByTimestamp("{ [0,2) }");
+
+  // Overlap the range [0, 2) with [1.25, 2); this results in an overlap of
+  // 0.75ms between the ranges.
+  NewSegmentAppend_OffsetFirstBuffer(2, 2,
+                                     base::TimeDelta::FromMillisecondsD(0.25));
+  CheckExpectedRangesByTimestamp("{ [0,2) }");
+
+  // A splice frame should not be generated (indicated by the lack of a config
+  // change in the expected buffer string) since it requires at least 1ms of
+  // data to crossfade.
+  CheckExpectedBuffers("0K 0K 1K 1K 1K");
+  CheckNoNextBuffer();
+}
+
 TEST_F(SourceBufferStreamTest, Audio_SpliceFrame_Preroll) {
   SetAudioStream();
   Seek(0);