/* -*- Mode: C++; tab-width: 2; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim:set ts=2 sw=2 sts=2 et cindent: */
/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

#include <string.h>

#include "mozilla/DebugOnly.h"
#include "mozilla/StandardInteger.h"

#include "nsDebug.h"
#include "OggCodecState.h"
#include "OggDecoder.h"
#include "nsTraceRefcnt.h"
#include "VideoUtils.h"
#include "MediaDecoderReader.h"
#include <algorithm>

namespace mozilla {

#ifdef PR_LOGGING
extern PRLogModuleInfo* gMediaDecoderLog;
#define LOG(type, msg) PR_LOG(gMediaDecoderLog, type, msg)
#else
#define LOG(type, msg)
#endif

// Reads a little-endian encoded unsigned 32bit integer at p.
static uint32_t LEUint32(const unsigned char* p)
{
  return p[0] +
        (p[1] << 8) +
        (p[2] << 16) +
        (p[3] << 24);
}

// Reads a little-endian encoded 64bit integer at p.
static int64_t LEInt64(const unsigned char* p)
{
  uint32_t lo = LEUint32(p);
  uint32_t hi = LEUint32(p + 4);
  return static_cast<int64_t>(lo) | (static_cast<int64_t>(hi) << 32);
}

// Reads a little-endian encoded unsigned 16bit integer at p.
static uint16_t LEUint16(const unsigned char* p)
{
  return p[0] + (p[1] << 8);
}

// Reads a little-endian encoded signed 16bit integer at p.
static int16_t LEInt16(const unsigned char* p)
{
  return static_cast<int16_t>(LEUint16(p));
}

/** Decoder base class for Ogg-encapsulated streams. */
OggCodecState*
OggCodecState::Create(ogg_page* aPage)
{
  NS_ASSERTION(ogg_page_bos(aPage), "Only call on BOS page!");
  nsAutoPtr<OggCodecState> codecState;
  if (aPage->body_len > 6 && memcmp(aPage->body+1, "theora", 6) == 0) {
    codecState = new TheoraState(aPage);
  } else if (aPage->body_len > 6 && memcmp(aPage->body+1, "vorbis", 6) == 0) {
    codecState = new VorbisState(aPage);
#ifdef MOZ_OPUS
  } else if (aPage->body_len > 8 && memcmp(aPage->body, "OpusHead", 8) == 0) {
    codecState = new OpusState(aPage);
#endif
  } else if (aPage->body_len > 8 && memcmp(aPage->body, "fishead\0", 8) == 0) {
    codecState = new SkeletonState(aPage);
  } else {
    codecState = new OggCodecState(aPage, false);
  }
  return codecState->OggCodecState::Init() ? codecState.forget() : nullptr;
}

OggCodecState::OggCodecState(ogg_page* aBosPage, bool aActive) :
  mPacketCount(0),
  mSerial(ogg_page_serialno(aBosPage)),
  mActive(aActive),
  mDoneReadingHeaders(!aActive)
{
  MOZ_COUNT_CTOR(OggCodecState);
  memset(&mState, 0, sizeof(ogg_stream_state));
}

OggCodecState::~OggCodecState() {
  MOZ_COUNT_DTOR(OggCodecState);
  Reset();
#ifdef DEBUG
  int ret =
#endif
  ogg_stream_clear(&mState);
  NS_ASSERTION(ret == 0, "ogg_stream_clear failed");
}

nsresult OggCodecState::Reset() {
  if (ogg_stream_reset(&mState) != 0) {
    return NS_ERROR_FAILURE;
  }
  mPackets.Erase();
  ClearUnstamped();
  return NS_OK;
}

void OggCodecState::ClearUnstamped()
{
  for (uint32_t i = 0; i < mUnstamped.Length(); ++i) {
    OggCodecState::ReleasePacket(mUnstamped[i]);
  }
  mUnstamped.Clear();
}

bool OggCodecState::Init() {
  int ret = ogg_stream_init(&mState, mSerial);
  return ret == 0;
}

bool OggCodecState::IsValidVorbisTagName(nsCString& aName)
{
  // Tag names must consist of ASCII 0x20 through 0x7D,
  // excluding 0x3D '=' which is the separator.
  uint32_t length = aName.Length();
  const char* data = aName.Data();
  for (uint32_t i = 0; i < length; i++) {
    if (data[i] < 0x20 || data[i] > 0x7D || data[i] == '=') {
      return false;
    }
  }
  return true;
}

bool OggCodecState::AddVorbisComment(MetadataTags* aTags,
                                       const char* aComment,
                                       uint32_t aLength)
{
  const char* div = (const char*)memchr(aComment, '=', aLength);
  if (!div) {
    LOG(PR_LOG_DEBUG, ("Skipping comment: no separator"));
    return false;
  }
  nsCString key = nsCString(aComment, div-aComment);
  if (!IsValidVorbisTagName(key)) {
    LOG(PR_LOG_DEBUG, ("Skipping comment: invalid tag name"));
    return false;
  }
  uint32_t valueLength = aLength - (div-aComment);
  nsCString value = nsCString(div + 1, valueLength);
  if (!IsUTF8(value)) {
    LOG(PR_LOG_DEBUG, ("Skipping comment: invalid UTF-8 in value"));
    return false;
  }
  aTags->Put(key, value);
  return true;
}

void VorbisState::RecordVorbisPacketSamples(ogg_packet* aPacket,
                                              long aSamples)
{
#ifdef VALIDATE_VORBIS_SAMPLE_CALCULATION
  mVorbisPacketSamples[aPacket] = aSamples;
#endif
}

void VorbisState::ValidateVorbisPacketSamples(ogg_packet* aPacket,
                                                long aSamples)
{
#ifdef VALIDATE_VORBIS_SAMPLE_CALCULATION
  NS_ASSERTION(mVorbisPacketSamples[aPacket] == aSamples,
    "Decoded samples for Vorbis packet don't match expected!");
  mVorbisPacketSamples.erase(aPacket);
#endif
}

void VorbisState::AssertHasRecordedPacketSamples(ogg_packet* aPacket)
{
#ifdef VALIDATE_VORBIS_SAMPLE_CALCULATION
  NS_ASSERTION(mVorbisPacketSamples.count(aPacket) == 1,
    "Must have recorded packet samples");
#endif
}

static ogg_packet* Clone(ogg_packet* aPacket) {
  ogg_packet* p = new ogg_packet();
  memcpy(p, aPacket, sizeof(ogg_packet));
  p->packet = new unsigned char[p->bytes];
  memcpy(p->packet, aPacket->packet, p->bytes);
  return p;
}

void OggCodecState::ReleasePacket(ogg_packet* aPacket) {
  if (aPacket)
    delete [] aPacket->packet;
  delete aPacket;
}

void OggPacketQueue::Append(ogg_packet* aPacket) {
  nsDeque::Push(aPacket);
}

ogg_packet* OggCodecState::PacketOut() {
  if (mPackets.IsEmpty()) {
    return nullptr;
  }
  return mPackets.PopFront();
}

nsresult OggCodecState::PageIn(ogg_page* aPage) {
  if (!mActive)
    return NS_OK;
  NS_ASSERTION(static_cast<uint32_t>(ogg_page_serialno(aPage)) == mSerial,
               "Page must be for this stream!");
  if (ogg_stream_pagein(&mState, aPage) == -1)
    return NS_ERROR_FAILURE;
  int r;
  do {
    ogg_packet packet;
    r = ogg_stream_packetout(&mState, &packet);
    if (r == 1) {
      mPackets.Append(Clone(&packet));
    }
  } while (r != 0);
  if (ogg_stream_check(&mState)) {
    NS_WARNING("Unrecoverable error in ogg_stream_packetout");
    return NS_ERROR_FAILURE;
  }
  return NS_OK;
}

nsresult OggCodecState::PacketOutUntilGranulepos(bool& aFoundGranulepos) {
  int r;
  aFoundGranulepos = false;
  // Extract packets from the sync state until either no more packets
  // come out, or we get a data packet with non -1 granulepos.
  do {
    ogg_packet packet;
    r = ogg_stream_packetout(&mState, &packet);
    if (r == 1) {
      ogg_packet* clone = Clone(&packet);
      if (IsHeader(&packet)) {
        // Header packets go straight into the packet queue.
        mPackets.Append(clone);
      } else {
        // We buffer data packets until we encounter a granulepos. We'll
        // then use the granulepos to figure out the granulepos of the
        // preceeding packets.
        mUnstamped.AppendElement(clone);
        aFoundGranulepos = packet.granulepos > 0;
      }
    }
  } while (r != 0 && !aFoundGranulepos);
  if (ogg_stream_check(&mState)) {
    NS_WARNING("Unrecoverable error in ogg_stream_packetout");
    return NS_ERROR_FAILURE;
  }
  return NS_OK;
}

TheoraState::TheoraState(ogg_page* aBosPage) :
  OggCodecState(aBosPage, true),
  mSetup(0),
  mCtx(0),
  mPixelAspectRatio(0)
{
  MOZ_COUNT_CTOR(TheoraState);
  th_info_init(&mInfo);
  th_comment_init(&mComment);
}

TheoraState::~TheoraState() {
  MOZ_COUNT_DTOR(TheoraState);
  th_setup_free(mSetup);
  th_decode_free(mCtx);
  th_comment_clear(&mComment);
  th_info_clear(&mInfo);
}

bool TheoraState::Init() {
  if (!mActive)
    return false;

  int64_t n = mInfo.aspect_numerator;
  int64_t d = mInfo.aspect_denominator;

  mPixelAspectRatio = (n == 0 || d == 0) ?
    1.0f : static_cast<float>(n) / static_cast<float>(d);

  // Ensure the frame and picture regions aren't larger than our prescribed
  // maximum, or zero sized.
  nsIntSize frame(mInfo.frame_width, mInfo.frame_height);
  nsIntRect picture(mInfo.pic_x, mInfo.pic_y, mInfo.pic_width, mInfo.pic_height);
  if (!VideoInfo::ValidateVideoRegion(frame, picture, frame)) {
    return mActive = false;
  }

  mCtx = th_decode_alloc(&mInfo, mSetup);
  if (mCtx == NULL) {
    return mActive = false;
  }

  return true;
}

bool
TheoraState::DecodeHeader(ogg_packet* aPacket)
{
  nsAutoRef<ogg_packet> autoRelease(aPacket);
  mPacketCount++;
  int ret = th_decode_headerin(&mInfo,
                               &mComment,
                               &mSetup,
                               aPacket);
 
  // We must determine when we've read the last header packet.
  // th_decode_headerin() does not tell us when it's read the last header, so
  // we must keep track of the headers externally.
  //
  // There are 3 header packets, the Identification, Comment, and Setup
  // headers, which must be in that order. If they're out of order, the file
  // is invalid. If we've successfully read a header, and it's the setup
  // header, then we're done reading headers. The first byte of each packet
  // determines it's type as follows:
  //    0x80 -> Identification header
  //    0x81 -> Comment header
  //    0x82 -> Setup header
  // See http://www.theora.org/doc/Theora.pdf Chapter 6, "Bitstream Headers",
  // for more details of the Ogg/Theora containment scheme.
  bool isSetupHeader = aPacket->bytes > 0 && aPacket->packet[0] == 0x82;
  if (ret < 0 || mPacketCount > 3) {
    // We've received an error, or the first three packets weren't valid
    // header packets. Assume bad input.
    // Our caller will deactivate the bitstream.
    return false;
  } else if (ret > 0 && isSetupHeader && mPacketCount == 3) {
    // Successfully read the three header packets.
    mDoneReadingHeaders = true;
  }
  return true;
}

int64_t
TheoraState::Time(int64_t granulepos) {
  if (!mActive) {
    return -1;
  }
  return TheoraState::Time(&mInfo, granulepos);
}

bool
TheoraState::IsHeader(ogg_packet* aPacket) {
  return th_packet_isheader(aPacket);
}

# define TH_VERSION_CHECK(_info,_maj,_min,_sub) \
 (((_info)->version_major>(_maj)||(_info)->version_major==(_maj))&& \
 (((_info)->version_minor>(_min)||(_info)->version_minor==(_min))&& \
 (_info)->version_subminor>=(_sub)))

int64_t TheoraState::Time(th_info* aInfo, int64_t aGranulepos)
{
  if (aGranulepos < 0 || aInfo->fps_numerator == 0) {
    return -1;
  }
  // Implementation of th_granule_frame inlined here to operate
  // on the th_info structure instead of the theora_state.
  int shift = aInfo->keyframe_granule_shift; 
  ogg_int64_t iframe = aGranulepos >> shift;
  ogg_int64_t pframe = aGranulepos - (iframe << shift);
  int64_t frameno = iframe + pframe - TH_VERSION_CHECK(aInfo, 3, 2, 1);
  CheckedInt64 t = ((CheckedInt64(frameno) + 1) * USECS_PER_S) * aInfo->fps_denominator;
  if (!t.isValid())
    return -1;
  t /= aInfo->fps_numerator;
  return t.isValid() ? t.value() : -1;
}

int64_t TheoraState::StartTime(int64_t granulepos) {
  if (granulepos < 0 || !mActive || mInfo.fps_numerator == 0) {
    return -1;
  }
  CheckedInt64 t = (CheckedInt64(th_granule_frame(mCtx, granulepos)) * USECS_PER_S) * mInfo.fps_denominator;
  if (!t.isValid())
    return -1;
  return t.value() / mInfo.fps_numerator;
}

int64_t
TheoraState::MaxKeyframeOffset()
{
  // Determine the maximum time in microseconds by which a key frame could
  // offset for the theora bitstream. Theora granulepos encode time as:
  // ((key_frame_number << granule_shift) + frame_offset).
  // Therefore the maximum possible time by which any frame could be offset
  // from a keyframe is the duration of (1 << granule_shift) - 1) frames.
  int64_t frameDuration;
  
  // Max number of frames keyframe could possibly be offset.
  int64_t keyframeDiff = (1 << mInfo.keyframe_granule_shift) - 1;

  // Length of frame in usecs.
  frameDuration = (mInfo.fps_denominator * USECS_PER_S) / mInfo.fps_numerator;

  // Total time in usecs keyframe can be offset from any given frame.
  return frameDuration * keyframeDiff;
}

nsresult
TheoraState::PageIn(ogg_page* aPage)
{
  if (!mActive)
    return NS_OK;
  NS_ASSERTION(static_cast<uint32_t>(ogg_page_serialno(aPage)) == mSerial,
               "Page must be for this stream!");
  if (ogg_stream_pagein(&mState, aPage) == -1)
    return NS_ERROR_FAILURE;
  bool foundGp;
  nsresult res = PacketOutUntilGranulepos(foundGp);
  if (NS_FAILED(res))
    return res;
  if (foundGp && mDoneReadingHeaders) {
    // We've found a packet with a granulepos, and we've loaded our metadata
    // and initialized our decoder. Determine granulepos of buffered packets.
    ReconstructTheoraGranulepos();
    for (uint32_t i = 0; i < mUnstamped.Length(); ++i) {
      ogg_packet* packet = mUnstamped[i];
#ifdef DEBUG
      NS_ASSERTION(!IsHeader(packet), "Don't try to recover header packet gp");
      NS_ASSERTION(packet->granulepos != -1, "Packet must have gp by now");
#endif
      mPackets.Append(packet);
    }
    mUnstamped.Clear();
  }
  return NS_OK;
}

// Returns 1 if the Theora info struct is decoding a media of Theora
// version (maj,min,sub) or later, otherwise returns 0.
int
TheoraVersion(th_info* info,
              unsigned char maj,
              unsigned char min,
              unsigned char sub)
{
  ogg_uint32_t ver = (maj << 16) + (min << 8) + sub;
  ogg_uint32_t th_ver = (info->version_major << 16) +
                        (info->version_minor << 8) +
                        info->version_subminor;
  return (th_ver >= ver) ? 1 : 0;
}

void TheoraState::ReconstructTheoraGranulepos()
{
  if (mUnstamped.Length() == 0) {
    return;
  }
  ogg_int64_t lastGranulepos = mUnstamped[mUnstamped.Length() - 1]->granulepos;
  NS_ASSERTION(lastGranulepos != -1, "Must know last granulepos");

  // Reconstruct the granulepos (and thus timestamps) of the decoded
  // frames. Granulepos are stored as ((keyframe<<shift)+offset). We
  // know the granulepos of the last frame in the list, so we can infer
  // the granulepos of the intermediate frames using their frame numbers.
  ogg_int64_t shift = mInfo.keyframe_granule_shift;
  ogg_int64_t version_3_2_1 = TheoraVersion(&mInfo,3,2,1);
  ogg_int64_t lastFrame = th_granule_frame(mCtx,
                                           lastGranulepos) + version_3_2_1;
  ogg_int64_t firstFrame = lastFrame - mUnstamped.Length() + 1;

  // Until we encounter a keyframe, we'll assume that the "keyframe"
  // segment of the granulepos is the first frame, or if that causes
  // the "offset" segment to overflow, we assume the required
  // keyframe is maximumally offset. Until we encounter a keyframe
  // the granulepos will probably be wrong, but we can't decode the
  // frame anyway (since we don't have its keyframe) so it doesn't really
  // matter.
  ogg_int64_t keyframe = lastGranulepos >> shift;

  // The lastFrame, firstFrame, keyframe variables, as well as the frame
  // variable in the loop below, store the frame number for Theora
  // version >= 3.2.1 streams, and store the frame index for Theora
  // version < 3.2.1 streams.
  for (uint32_t i = 0; i < mUnstamped.Length() - 1; ++i) {
    ogg_int64_t frame = firstFrame + i;
    ogg_int64_t granulepos;
    ogg_packet* packet = mUnstamped[i];
    bool isKeyframe = th_packet_iskeyframe(packet) == 1;

    if (isKeyframe) {
      granulepos = frame << shift;
      keyframe = frame;
    } else if (frame >= keyframe &&
                frame - keyframe < ((ogg_int64_t)1 << shift))
    {
      // (frame - keyframe) won't overflow the "offset" segment of the
      // granulepos, so it's safe to calculate the granulepos.
      granulepos = (keyframe << shift) + (frame - keyframe);
    } else {
      // (frame - keyframeno) will overflow the "offset" segment of the
      // granulepos, so we take "keyframe" to be the max possible offset
      // frame instead.
      ogg_int64_t k = std::max(frame - (((ogg_int64_t)1 << shift) - 1), version_3_2_1);
      granulepos = (k << shift) + (frame - k);
    }
    // Theora 3.2.1+ granulepos store frame number [1..N], so granulepos
    // should be > 0.
    // Theora 3.2.0 granulepos store the frame index [0..(N-1)], so
    // granulepos should be >= 0. 
    NS_ASSERTION(granulepos >= version_3_2_1,
                  "Invalid granulepos for Theora version");

    // Check that the frame's granule number is one more than the
    // previous frame's.
    NS_ASSERTION(i == 0 ||
                 th_granule_frame(mCtx, granulepos) ==
                 th_granule_frame(mCtx, mUnstamped[i-1]->granulepos) + 1,
                 "Granulepos calculation is incorrect!");

    packet->granulepos = granulepos;
  }

  // Check that the second to last frame's granule number is one less than
  // the last frame's (the known granule number). If not our granulepos
  // recovery missed a beat.
  NS_ASSERTION(mUnstamped.Length() < 2 ||
    th_granule_frame(mCtx, mUnstamped[mUnstamped.Length()-2]->granulepos) + 1 ==
    th_granule_frame(mCtx, lastGranulepos),
    "Granulepos recovery should catch up with packet->granulepos!");
}

nsresult VorbisState::Reset()
{
  nsresult res = NS_OK;
  if (mActive && vorbis_synthesis_restart(&mDsp) != 0) {
    res = NS_ERROR_FAILURE;
  }
  if (NS_FAILED(OggCodecState::Reset())) {
    return NS_ERROR_FAILURE;
  }

  mGranulepos = 0;
  mPrevVorbisBlockSize = 0;

  return res;
}

VorbisState::VorbisState(ogg_page* aBosPage) :
  OggCodecState(aBosPage, true),
  mPrevVorbisBlockSize(0),
  mGranulepos(0)
{
  MOZ_COUNT_CTOR(VorbisState);
  vorbis_info_init(&mInfo);
  vorbis_comment_init(&mComment);
  memset(&mDsp, 0, sizeof(vorbis_dsp_state));
  memset(&mBlock, 0, sizeof(vorbis_block));
}

VorbisState::~VorbisState() {
  MOZ_COUNT_DTOR(VorbisState);
  Reset();
  vorbis_block_clear(&mBlock);
  vorbis_dsp_clear(&mDsp);
  vorbis_info_clear(&mInfo);
  vorbis_comment_clear(&mComment);
}

bool VorbisState::DecodeHeader(ogg_packet* aPacket) {
  nsAutoRef<ogg_packet> autoRelease(aPacket);
  mPacketCount++;
  int ret = vorbis_synthesis_headerin(&mInfo,
                                      &mComment,
                                      aPacket);
  // We must determine when we've read the last header packet.
  // vorbis_synthesis_headerin() does not tell us when it's read the last
  // header, so we must keep track of the headers externally.
  //
  // There are 3 header packets, the Identification, Comment, and Setup
  // headers, which must be in that order. If they're out of order, the file
  // is invalid. If we've successfully read a header, and it's the setup
  // header, then we're done reading headers. The first byte of each packet
  // determines it's type as follows:
  //    0x1 -> Identification header
  //    0x3 -> Comment header
  //    0x5 -> Setup header
  // For more details of the Vorbis/Ogg containment scheme, see the Vorbis I
  // Specification, Chapter 4, Codec Setup and Packet Decode:
  // http://www.xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-580004

  bool isSetupHeader = aPacket->bytes > 0 && aPacket->packet[0] == 0x5;

  if (ret < 0 || mPacketCount > 3) {
    // We've received an error, or the first three packets weren't valid
    // header packets. Assume bad input. Our caller will deactivate the
    // bitstream.
    return false;
  } else if (ret == 0 && isSetupHeader && mPacketCount == 3) {
    // Successfully read the three header packets.
    // The bitstream remains active.
    mDoneReadingHeaders = true;
  }
  return true;
}

bool VorbisState::Init()
{
  if (!mActive)
    return false;

  int ret = vorbis_synthesis_init(&mDsp, &mInfo);
  if (ret != 0) {
    NS_WARNING("vorbis_synthesis_init() failed initializing vorbis bitstream");
    return mActive = false;
  }
  ret = vorbis_block_init(&mDsp, &mBlock);
  if (ret != 0) {
    NS_WARNING("vorbis_block_init() failed initializing vorbis bitstream");
    if (mActive) {
      vorbis_dsp_clear(&mDsp);
    }
    return mActive = false;
  }
  return true;
}

int64_t VorbisState::Time(int64_t granulepos)
{
  if (!mActive) {
    return -1;
  }

  return VorbisState::Time(&mInfo, granulepos);
}

int64_t VorbisState::Time(vorbis_info* aInfo, int64_t aGranulepos)
{
  if (aGranulepos == -1 || aInfo->rate == 0) {
    return -1;
  }
  CheckedInt64 t = CheckedInt64(aGranulepos) * USECS_PER_S;
  if (!t.isValid())
    t = 0;
  return t.value() / aInfo->rate;
}

bool
VorbisState::IsHeader(ogg_packet* aPacket)
{
  // The first byte in each Vorbis header packet is either 0x01, 0x03, or 0x05,
  // i.e. the first bit is odd. Audio data packets have their first bit as 0x0.
  // Any packet with its first bit set cannot be a data packet, it's a
  // (possibly invalid) header packet.
  // See: http://xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-610004.2.1
  return aPacket->bytes > 0 ? (aPacket->packet[0] & 0x1) : false;
}

MetadataTags*
VorbisState::GetTags()
{
  MetadataTags* tags;
  NS_ASSERTION(mComment.user_comments, "no vorbis comment strings!");
  NS_ASSERTION(mComment.comment_lengths, "no vorbis comment lengths!");
  tags = new MetadataTags;
  tags->Init();
  for (int i = 0; i < mComment.comments; i++) {
    AddVorbisComment(tags, mComment.user_comments[i],
                     mComment.comment_lengths[i]);
  }
  return tags;
}

nsresult
VorbisState::PageIn(ogg_page* aPage)
{
  if (!mActive)
    return NS_OK;
  NS_ASSERTION(static_cast<uint32_t>(ogg_page_serialno(aPage)) == mSerial,
               "Page must be for this stream!");
  if (ogg_stream_pagein(&mState, aPage) == -1)
    return NS_ERROR_FAILURE;
  bool foundGp;
  nsresult res = PacketOutUntilGranulepos(foundGp);
  if (NS_FAILED(res))
    return res;
  if (foundGp && mDoneReadingHeaders) {
    // We've found a packet with a granulepos, and we've loaded our metadata
    // and initialized our decoder. Determine granulepos of buffered packets.
    ReconstructVorbisGranulepos();
    for (uint32_t i = 0; i < mUnstamped.Length(); ++i) {
      ogg_packet* packet = mUnstamped[i];
      AssertHasRecordedPacketSamples(packet);
      NS_ASSERTION(!IsHeader(packet), "Don't try to recover header packet gp");
      NS_ASSERTION(packet->granulepos != -1, "Packet must have gp by now");
      mPackets.Append(packet);
    }
    mUnstamped.Clear();
  }
  return NS_OK;
}

nsresult VorbisState::ReconstructVorbisGranulepos()
{
  // The number of samples in a Vorbis packet is:
  // window_blocksize(previous_packet)/4+window_blocksize(current_packet)/4
  // See: http://xiph.org/vorbis/doc/Vorbis_I_spec.html#x1-230001.3.2
  // So we maintain mPrevVorbisBlockSize, the block size of the last packet
  // encountered. We also maintain mGranulepos, which is the granulepos of
  // the last encountered packet. This enables us to give granulepos to
  // packets when the last packet in mUnstamped doesn't have a granulepos
  // (for example if the stream was truncated).
  //
  // We validate our prediction of the number of samples decoded when
  // VALIDATE_VORBIS_SAMPLE_CALCULATION is defined by recording the predicted
  // number of samples, and verifing we extract that many when decoding
  // each packet.

  NS_ASSERTION(mUnstamped.Length() > 0, "Length must be > 0");
  ogg_packet* last = mUnstamped[mUnstamped.Length()-1];
  NS_ASSERTION(last->e_o_s || last->granulepos >= 0,
    "Must know last granulepos!");
  if (mUnstamped.Length() == 1) {
    ogg_packet* packet = mUnstamped[0];
    long blockSize = vorbis_packet_blocksize(&mInfo, packet);
    if (blockSize < 0) {
      // On failure vorbis_packet_blocksize returns < 0. If we've got
      // a bad packet, we just assume that decode will have to skip this
      // packet, i.e. assume 0 samples are decodable from this packet.
      blockSize = 0;
      mPrevVorbisBlockSize = 0;
    }
    long samples = mPrevVorbisBlockSize / 4 + blockSize / 4;
    mPrevVorbisBlockSize = blockSize;
    if (packet->granulepos == -1) {
      packet->granulepos = mGranulepos + samples;
    }

    // Account for a partial last frame
    if (packet->e_o_s && packet->granulepos >= mGranulepos) {
       samples = packet->granulepos - mGranulepos;
    }
 
    mGranulepos = packet->granulepos;
    RecordVorbisPacketSamples(packet, samples);
    return NS_OK;
  }

  bool unknownGranulepos = last->granulepos == -1;
  int totalSamples = 0;
  for (int32_t i = mUnstamped.Length() - 1; i > 0; i--) {
    ogg_packet* packet = mUnstamped[i];
    ogg_packet* prev = mUnstamped[i-1];
    ogg_int64_t granulepos = packet->granulepos;
    NS_ASSERTION(granulepos != -1, "Must know granulepos!");
    long prevBlockSize = vorbis_packet_blocksize(&mInfo, prev);
    long blockSize = vorbis_packet_blocksize(&mInfo, packet);

    if (blockSize < 0 || prevBlockSize < 0) {
      // On failure vorbis_packet_blocksize returns < 0. If we've got
      // a bad packet, we just assume that decode will have to skip this
      // packet, i.e. assume 0 samples are decodable from this packet.
      blockSize = 0;
      prevBlockSize = 0;
    }

    long samples = prevBlockSize / 4 + blockSize / 4;
    totalSamples += samples;
    prev->granulepos = granulepos - samples;
    RecordVorbisPacketSamples(packet, samples);
  }

  if (unknownGranulepos) {
    for (uint32_t i = 0; i < mUnstamped.Length(); i++) {
      ogg_packet* packet = mUnstamped[i];
      packet->granulepos += mGranulepos + totalSamples + 1;
    }
  }

  ogg_packet* first = mUnstamped[0];
  long blockSize = vorbis_packet_blocksize(&mInfo, first);
  if (blockSize < 0) {
    mPrevVorbisBlockSize = 0;
    blockSize = 0;
  }

  long samples = (mPrevVorbisBlockSize == 0) ? 0 :
                  mPrevVorbisBlockSize / 4 + blockSize / 4;
  int64_t start = first->granulepos - samples;
  RecordVorbisPacketSamples(first, samples);

  if (last->e_o_s && start < mGranulepos) {
    // We've calculated that there are more samples in this page than its
    // granulepos claims, and it's the last page in the stream. This is legal,
    // and we will need to prune the trailing samples when we come to decode it.
    // We must correct the timestamps so that they follow the last Vorbis page's
    // samples.
    int64_t pruned = mGranulepos - start;
    for (uint32_t i = 0; i < mUnstamped.Length() - 1; i++) {
      mUnstamped[i]->granulepos += pruned;
    }
#ifdef VALIDATE_VORBIS_SAMPLE_CALCULATION
    mVorbisPacketSamples[last] -= pruned;
#endif
  }

  mPrevVorbisBlockSize = vorbis_packet_blocksize(&mInfo, last);
  mPrevVorbisBlockSize = std::max(static_cast<long>(0), mPrevVorbisBlockSize);
  mGranulepos = last->granulepos;

  return NS_OK;
}

#ifdef MOZ_OPUS
OpusState::OpusState(ogg_page* aBosPage) :
  OggCodecState(aBosPage, true),
  mRate(0),
  mNominalRate(0),
  mChannels(0),
  mPreSkip(0),
#ifdef MOZ_SAMPLE_TYPE_FLOAT32
  mGain(1.0f),
#else
  mGain_Q16(65536),
#endif
  mChannelMapping(0),
  mStreams(0),
  mCoupledStreams(0),
  mDecoder(NULL),
  mSkip(0),
  mPrevPacketGranulepos(0),
  mPrevPageGranulepos(0)
{
  MOZ_COUNT_CTOR(OpusState);
}

OpusState::~OpusState() {
  MOZ_COUNT_DTOR(OpusState);
  Reset();

  if (mDecoder) {
    opus_multistream_decoder_destroy(mDecoder);
    mDecoder = NULL;
  }
}

nsresult OpusState::Reset()
{
  return Reset(false);
}

nsresult OpusState::Reset(bool aStart)
{
  nsresult res = NS_OK;

  if (mActive && mDecoder) {
    // Reset the decoder.
    opus_multistream_decoder_ctl(mDecoder, OPUS_RESET_STATE);
    // Let the seek logic handle pre-roll if we're not seeking to the start.
    mSkip = aStart ? mPreSkip : 0;
    // This lets us distinguish the first page being the last page vs. just
    // not having processed the previous page when we encounter the last page.
    mPrevPageGranulepos = aStart ? 0 : -1;
    mPrevPacketGranulepos = aStart ? 0 : -1;
  }

  // Clear queued data.
  if (NS_FAILED(OggCodecState::Reset())) {
    return NS_ERROR_FAILURE;
  }

  LOG(PR_LOG_DEBUG, ("Opus decoder reset, to skip %d", mSkip));

  return res;
}

bool OpusState::Init(void)
{
  if (!mActive)
    return false;

  int error;

  NS_ASSERTION(mDecoder == NULL, "leaking OpusDecoder");

  mDecoder = opus_multistream_decoder_create(mRate,
                                             mChannels,
                                             mStreams,
                                             mCoupledStreams,
                                             mMappingTable,
                                             &error);

  mSkip = mPreSkip;

  LOG(PR_LOG_DEBUG, ("Opus decoder init, to skip %d", mSkip));

  return error == OPUS_OK;
}

bool OpusState::DecodeHeader(ogg_packet* aPacket)
{
  nsAutoRef<ogg_packet> autoRelease(aPacket);
  switch(mPacketCount++) {
    // Parse the id header.
    case 0: {
      if (aPacket->bytes < 19 || memcmp(aPacket->packet, "OpusHead", 8)) {
        LOG(PR_LOG_DEBUG, ("Invalid Opus file: unrecognized header"));
        return false;
      }

      mRate = 48000; // The Opus decoder runs at 48 kHz regardless.

      int version = aPacket->packet[8];
      // Accept file format versions 0.x.
      if ((version & 0xf0) != 0) {
        LOG(PR_LOG_DEBUG, ("Rejecting unknown Opus file version %d", version));
        return false;
      }

      mChannels = aPacket->packet[9];
      if (mChannels<1) {
        LOG(PR_LOG_DEBUG, ("Invalid Opus file: Number of channels %d", mChannels));
        return false;
      }
#ifndef MOZ_SAMPLE_TYPE_FLOAT32
      // Downmixing more than 2 channels it is not supported for integer
      // output samples. It is only supported for float output.
      if (mChannels>2)
        return false;
#endif
      mPreSkip = LEUint16(aPacket->packet + 10);
      mNominalRate = LEUint32(aPacket->packet + 12);
      double gain_dB = LEInt16(aPacket->packet + 16) / 256.0;
#ifdef MOZ_SAMPLE_TYPE_FLOAT32
      mGain = static_cast<float>(pow(10,0.05*gain_dB));
#else
      mGain_Q16 = static_cast<int32_t>(std::min(65536*pow(10,0.05*gain_dB)+0.5,
                                              static_cast<double>(INT32_MAX)));
#endif
      mChannelMapping = aPacket->packet[18];

      if (mChannelMapping == 0) {
        // Mapping family 0 only allows two channels
        if (mChannels>2) {
          LOG(PR_LOG_DEBUG, ("Invalid Opus file: too many channels (%d) for"
                             " mapping family 0.", mChannels));
          return false;
        }
        mStreams = 1;
        mCoupledStreams = mChannels - 1;
        mMappingTable[0] = 0;
        mMappingTable[1] = 1;
      } else if (mChannelMapping == 1) {
        // Currently only up to 8 channels are defined for mapping family 1
        if (mChannels>8) {
          LOG(PR_LOG_DEBUG, ("Invalid Opus file: too many channels (%d) for"
                             " mapping family 1.", mChannels));
          return false;
        }
        if (aPacket->bytes>20+mChannels) {
          mStreams = aPacket->packet[19];
          mCoupledStreams = aPacket->packet[20];
          int i;
          for (i=0; i<mChannels; i++)
            mMappingTable[i] = aPacket->packet[21+i];
        } else {
          LOG(PR_LOG_DEBUG, ("Invalid Opus file: channel mapping %d,"
                             " but no channel mapping table", mChannelMapping));
          return false;
        }
      } else {
        LOG(PR_LOG_DEBUG, ("Invalid Opus file: unsupported channel mapping "
                           "family %d", mChannelMapping));
        return false;
      }
      if (mStreams < 1) {
        LOG(PR_LOG_DEBUG, ("Invalid Opus file: no streams"));
        return false;
      }
      if (mCoupledStreams > mStreams) {
        LOG(PR_LOG_DEBUG, ("Invalid Opus file: more coupled streams (%d) than "
                           "total streams (%d)", mCoupledStreams, mStreams));
        return false;
      }

#ifdef DEBUG
      LOG(PR_LOG_DEBUG, ("Opus stream header:"));
      LOG(PR_LOG_DEBUG, (" channels: %d", mChannels));
      LOG(PR_LOG_DEBUG, ("  preskip: %d", mPreSkip));
      LOG(PR_LOG_DEBUG, (" original: %d Hz", mNominalRate));
      LOG(PR_LOG_DEBUG, ("     gain: %.2f dB", gain_dB));
      LOG(PR_LOG_DEBUG, ("Channel Mapping:"));
      LOG(PR_LOG_DEBUG, ("   family: %d", mChannelMapping));
      LOG(PR_LOG_DEBUG, ("  streams: %d", mStreams));
#endif
    }
    break;

    // Parse the metadata header.
    case 1: {
      if (aPacket->bytes < 16 || memcmp(aPacket->packet, "OpusTags", 8))
        return false;

      // Copy out the raw comment lines, but only do basic validation
      // checks against the string packing: too little data, too many
      // comments, or comments that are too long. Rejecting these cases
      // helps reduce the propagation of broken files.
      // We do not ensure they are valid UTF-8 here, nor do we validate
      // the required ASCII_TAG=value format of the user comments.
      const unsigned char* buf = aPacket->packet + 8;
      uint32_t bytes = aPacket->bytes - 8;
      uint32_t len;
      // Read the vendor string.
      len = LEUint32(buf);
      buf += 4;
      bytes -= 4;
      if (len > bytes)
        return false;
      mVendorString = nsCString(reinterpret_cast<const char*>(buf), len);
      buf += len;
      bytes -= len;
      // Read the user comments.
      if (bytes < 4)
        return false;
      uint32_t ncomments = LEUint32(buf);
      buf += 4;
      bytes -= 4;
      // If there are so many comments even their length fields
      // won't fit in the packet, stop reading now.
      if (ncomments > (bytes>>2))
        return false;
      uint32_t i;
      for (i = 0; i < ncomments; i++) {
        if (bytes < 4)
          return false;
        len = LEUint32(buf);
        buf += 4;
        bytes -= 4;
        if (len > bytes)
          return false;
        mTags.AppendElement(nsCString(reinterpret_cast<const char*>(buf), len));
        buf += len;
        bytes -= len;
      }

#ifdef DEBUG
      LOG(PR_LOG_DEBUG, ("Opus metadata header:"));
      LOG(PR_LOG_DEBUG, ("  vendor: %s", mVendorString.get()));
      for (uint32_t i = 0; i < mTags.Length(); i++) {
        LOG(PR_LOG_DEBUG, (" %s", mTags[i].get()));
      }
#endif
    }
    break;

    // We made it to the first data packet (which includes reconstructing
    // timestamps for it in PageIn). Success!
    default: {
      mDoneReadingHeaders = true;
      // Put it back on the queue so we can decode it.
      mPackets.PushFront(autoRelease.disown());
    }
    break;
  }
  return true;
}

/* Construct and return a tags hashmap from our internal array */
MetadataTags* OpusState::GetTags()
{
  MetadataTags* tags;

  tags = new MetadataTags;
  tags->Init();
  for (uint32_t i = 0; i < mTags.Length(); i++) {
    AddVorbisComment(tags, mTags[i].Data(), mTags[i].Length());
  }

  return tags;
}

/* Return the timestamp (in microseconds) equivalent to a granulepos. */
int64_t OpusState::Time(int64_t aGranulepos)
{
  if (!mActive)
    return -1;

  return Time(mPreSkip, aGranulepos);
}

int64_t OpusState::Time(int aPreSkip, int64_t aGranulepos)
{
  if (aGranulepos < 0)
    return -1;

  // Ogg Opus always runs at a granule rate of 48 kHz.
  CheckedInt64 t = CheckedInt64(aGranulepos - aPreSkip) * USECS_PER_S;
  return t.isValid() ? t.value() / 48000 : -1;
}

bool OpusState::IsHeader(ogg_packet* aPacket)
{
  return aPacket->bytes >= 16 &&
         (!memcmp(aPacket->packet, "OpusHead", 8) ||
          !memcmp(aPacket->packet, "OpusTags", 8));
}

nsresult OpusState::PageIn(ogg_page* aPage)
{
  if (!mActive)
    return NS_OK;
  NS_ASSERTION(static_cast<uint32_t>(ogg_page_serialno(aPage)) == mSerial,
               "Page must be for this stream!");
  if (ogg_stream_pagein(&mState, aPage) == -1)
    return NS_ERROR_FAILURE;

  bool haveGranulepos;
  nsresult rv = PacketOutUntilGranulepos(haveGranulepos);
  if (NS_FAILED(rv) || !haveGranulepos || mPacketCount < 2)
    return rv;
  if(!ReconstructOpusGranulepos())
    return NS_ERROR_FAILURE;
  for (uint32_t i = 0; i < mUnstamped.Length(); i++) {
    ogg_packet* packet = mUnstamped[i];
    NS_ASSERTION(!IsHeader(packet), "Don't try to play a header packet");
    NS_ASSERTION(packet->granulepos != -1, "Packet should have a granulepos");
    mPackets.Append(packet);
  }
  mUnstamped.Clear();
  return NS_OK;
}

// Helper method to return the change in granule position due to an Opus packet
// (as distinct from the number of samples in the packet, which depends on the
// decoder rate). It should work with a multistream Opus file, and continue to
// work should we ever allow the decoder to decode at a rate other than 48 kHz.
// It even works before we've created the actual Opus decoder.
static int GetOpusDeltaGP(ogg_packet* packet)
{
  int nframes;
  nframes = opus_packet_get_nb_frames(packet->packet, packet->bytes);
  if (nframes > 0) {
    return nframes*opus_packet_get_samples_per_frame(packet->packet, 48000);
  }
  NS_WARNING("Invalid Opus packet.");
  return nframes;
}

bool OpusState::ReconstructOpusGranulepos(void)
{
  NS_ASSERTION(mUnstamped.Length() > 0, "Must have unstamped packets");
  ogg_packet* last = mUnstamped[mUnstamped.Length()-1];
  NS_ASSERTION(last->e_o_s || last->granulepos > 0,
      "Must know last granulepos!");
  int64_t gp;
  // If this is the last page, and we've seen at least one previous page (or
  // this is the first page)...
  if (last->e_o_s) {
    if (mPrevPageGranulepos != -1) {
      // If this file only has one page and the final granule position is
      // smaller than the pre-skip amount, we MUST reject the stream.
      if (!mDoneReadingHeaders && last->granulepos < mPreSkip)
        return false;
      int64_t last_gp = last->granulepos;
      gp = mPrevPageGranulepos;
      // Loop through the packets forwards, adding the current packet's
      // duration to the previous granulepos to get the value for the
      // current packet.
      for (uint32_t i = 0; i < mUnstamped.Length() - 1; ++i) {
        ogg_packet* packet = mUnstamped[i];
        int offset = GetOpusDeltaGP(packet);
        // Check for error (negative offset) and overflow.
        if (offset >= 0 && gp <= INT64_MAX - offset) {
          gp += offset;
          if (gp >= last_gp) {
            NS_WARNING("Opus end trimming removed more than a full packet.");
            // We were asked to remove a full packet's worth of data or more.
            // Encoders SHOULD NOT produce streams like this, but we'll handle
            // it for them anyway.
            gp = last_gp;
            for (uint32_t j = i+1; j < mUnstamped.Length(); ++j) {
              OggCodecState::ReleasePacket(mUnstamped[j]);
            }
            mUnstamped.RemoveElementsAt(i+1, mUnstamped.Length() - (i+1));
            last = packet;
            last->e_o_s = 1;
          }
        }
        packet->granulepos = gp;
      }
      mPrevPageGranulepos = last_gp;
      return true;
    } else {
      NS_WARNING("No previous granule position to use for Opus end trimming.");
      // If we don't have a previous granule position, fall through.
      // We simply won't trim any samples from the end.
      // TODO: Are we guaranteed to have seen a previous page if there is one?
    }
  }

  gp = last->granulepos;
  // Loop through the packets backwards, subtracting the next
  // packet's duration from its granulepos to get the value
  // for the current packet.
  for (uint32_t i = mUnstamped.Length() - 1; i > 0; i--) {
    int offset = GetOpusDeltaGP(mUnstamped[i]);
    // Check for error (negative offset) and overflow.
    if (offset >= 0) {
      if (offset <= gp) {
        gp -= offset;
      } else {
        // If the granule position of the first data page is smaller than the
        // number of decodable audio samples on that page, then we MUST reject
        // the stream.
        if (!mDoneReadingHeaders)
          return false;
        // It's too late to reject the stream.
        // If we get here, this almost certainly means the file has screwed-up
        // timestamps somewhere after the first page.
        NS_WARNING("Clamping negative Opus granulepos to zero.");
        gp = 0;
      }
    }
    mUnstamped[i - 1]->granulepos = gp;
  }

  // Check to make sure the first granule position is at least as large as the
  // total number of samples decodable from the first page with completed
  // packets. This requires looking at the duration of the first packet, too.
  // We MUST reject such streams.
  if (!mDoneReadingHeaders && GetOpusDeltaGP(mUnstamped[0]) > gp)
    return false;
  mPrevPageGranulepos = last->granulepos;
  return true;
}
#endif /* MOZ_OPUS */

SkeletonState::SkeletonState(ogg_page* aBosPage) :
  OggCodecState(aBosPage, true),
  mVersion(0),
  mPresentationTime(0),
  mLength(0)
{
  MOZ_COUNT_CTOR(SkeletonState);
}
 
SkeletonState::~SkeletonState()
{
  MOZ_COUNT_DTOR(SkeletonState);
}

// Support for Ogg Skeleton 4.0, as per specification at:
// http://wiki.xiph.org/Ogg_Skeleton_4

// Minimum length in bytes of a Skeleton header packet.
static const long SKELETON_MIN_HEADER_LEN = 28;
static const long SKELETON_4_0_MIN_HEADER_LEN = 80;

// Minimum length in bytes of a Skeleton 4.0 index packet.
static const long SKELETON_4_0_MIN_INDEX_LEN = 42;

// Minimum possible size of a compressed index keypoint.
static const size_t MIN_KEY_POINT_SIZE = 2;

// Byte offset of the major and minor version numbers in the
// Ogg Skeleton 4.0 header packet.
static const size_t SKELETON_VERSION_MAJOR_OFFSET = 8;
static const size_t SKELETON_VERSION_MINOR_OFFSET = 10;

// Byte-offsets of the presentation time numerator and denominator
static const size_t SKELETON_PRESENTATION_TIME_NUMERATOR_OFFSET = 12;
static const size_t SKELETON_PRESENTATION_TIME_DENOMINATOR_OFFSET = 20;

// Byte-offsets of the length of file field in the Skeleton 4.0 header packet.
static const size_t SKELETON_FILE_LENGTH_OFFSET = 64;

// Byte-offsets of the fields in the Skeleton index packet.
static const size_t INDEX_SERIALNO_OFFSET = 6;
static const size_t INDEX_NUM_KEYPOINTS_OFFSET = 10;
static const size_t INDEX_TIME_DENOM_OFFSET = 18;
static const size_t INDEX_FIRST_NUMER_OFFSET = 26;
static const size_t INDEX_LAST_NUMER_OFFSET = 34;
static const size_t INDEX_KEYPOINT_OFFSET = 42;

static bool IsSkeletonBOS(ogg_packet* aPacket)
{
  return aPacket->bytes >= SKELETON_MIN_HEADER_LEN && 
         memcmp(reinterpret_cast<char*>(aPacket->packet), "fishead", 8) == 0;
}

static bool IsSkeletonIndex(ogg_packet* aPacket)
{
  return aPacket->bytes >= SKELETON_4_0_MIN_INDEX_LEN &&
         memcmp(reinterpret_cast<char*>(aPacket->packet), "index", 5) == 0;
}

// Reads a variable length encoded integer at p. Will not read
// past aLimit. Returns pointer to character after end of integer.
static const unsigned char* ReadVariableLengthInt(const unsigned char* p,
                                                  const unsigned char* aLimit,
                                                  int64_t& n)
{
  int shift = 0;
  int64_t byte = 0;
  n = 0;
  while (p < aLimit &&
         (byte & 0x80) != 0x80 &&
         shift < 57)
  {
    byte = static_cast<int64_t>(*p);
    n |= ((byte & 0x7f) << shift);
    shift += 7;
    p++;
  }
  return p;
}

bool SkeletonState::DecodeIndex(ogg_packet* aPacket)
{
  NS_ASSERTION(aPacket->bytes >= SKELETON_4_0_MIN_INDEX_LEN,
               "Index must be at least minimum size");
  if (!mActive) {
    return false;
  }

  uint32_t serialno = LEUint32(aPacket->packet + INDEX_SERIALNO_OFFSET);
  int64_t numKeyPoints = LEInt64(aPacket->packet + INDEX_NUM_KEYPOINTS_OFFSET);

  int64_t endTime = 0, startTime = 0;
  const unsigned char* p = aPacket->packet;

  int64_t timeDenom = LEInt64(aPacket->packet + INDEX_TIME_DENOM_OFFSET);
  if (timeDenom == 0) {
    LOG(PR_LOG_DEBUG, ("Ogg Skeleton Index packet for stream %u has 0 "
                       "timestamp denominator.", serialno));
    return (mActive = false);
  }

  // Extract the start time.
  CheckedInt64 t = CheckedInt64(LEInt64(p + INDEX_FIRST_NUMER_OFFSET)) * USECS_PER_S;
  if (!t.isValid()) {
    return (mActive = false);
  } else {
    startTime = t.value() / timeDenom;
  }

  // Extract the end time.
  t = LEInt64(p + INDEX_LAST_NUMER_OFFSET) * USECS_PER_S;
  if (!t.isValid()) {
    return (mActive = false);
  } else {
    endTime = t.value() / timeDenom;
  }

  // Check the numKeyPoints value read, ensure we're not going to run out of
  // memory while trying to decode the index packet.
  CheckedInt64 minPacketSize = (CheckedInt64(numKeyPoints) * MIN_KEY_POINT_SIZE) + INDEX_KEYPOINT_OFFSET;
  if (!minPacketSize.isValid())
  {
    return (mActive = false);
  }
  
  int64_t sizeofIndex = aPacket->bytes - INDEX_KEYPOINT_OFFSET;
  int64_t maxNumKeyPoints = sizeofIndex / MIN_KEY_POINT_SIZE;
  if (aPacket->bytes < minPacketSize.value() ||
      numKeyPoints > maxNumKeyPoints || 
      numKeyPoints < 0)
  {
    // Packet size is less than the theoretical minimum size, or the packet is
    // claiming to store more keypoints than it's capable of storing. This means
    // that the numKeyPoints field is too large or small for the packet to
    // possibly contain as many packets as it claims to, so the numKeyPoints
    // field is possibly malicious. Don't try decoding this index, we may run
    // out of memory.
    LOG(PR_LOG_DEBUG, ("Possibly malicious number of key points reported "
                       "(%lld) in index packet for stream %u.",
                       numKeyPoints,
                       serialno));
    return (mActive = false);
  }

  nsAutoPtr<nsKeyFrameIndex> keyPoints(new nsKeyFrameIndex(startTime, endTime));
  
  p = aPacket->packet + INDEX_KEYPOINT_OFFSET;
  const unsigned char* limit = aPacket->packet + aPacket->bytes;
  int64_t numKeyPointsRead = 0;
  CheckedInt64 offset = 0;
  CheckedInt64 time = 0;
  while (p < limit &&
         numKeyPointsRead < numKeyPoints)
  {
    int64_t delta = 0;
    p = ReadVariableLengthInt(p, limit, delta);
    offset += delta;
    if (p == limit ||
        !offset.isValid() ||
        offset.value() > mLength ||
        offset.value() < 0)
    {
      return (mActive = false);
    }
    p = ReadVariableLengthInt(p, limit, delta);
    time += delta;
    if (!time.isValid() ||
        time.value() > endTime ||
        time.value() < startTime)
    {
      return (mActive = false);
    }
    CheckedInt64 timeUsecs = time * USECS_PER_S;
    if (!timeUsecs.isValid())
      return mActive = false;
    timeUsecs /= timeDenom;
    keyPoints->Add(offset.value(), timeUsecs.value());
    numKeyPointsRead++;
  }

  int32_t keyPointsRead = keyPoints->Length();
  if (keyPointsRead > 0) {
    mIndex.Put(serialno, keyPoints.forget());
  }

  LOG(PR_LOG_DEBUG, ("Loaded %d keypoints for Skeleton on stream %u",
                     keyPointsRead, serialno));
  return true;
}

nsresult SkeletonState::IndexedSeekTargetForTrack(uint32_t aSerialno,
                                                    int64_t aTarget,
                                                    nsKeyPoint& aResult)
{
  nsKeyFrameIndex* index = nullptr;
  mIndex.Get(aSerialno, &index);

  if (!index ||
      index->Length() == 0 ||
      aTarget < index->mStartTime ||
      aTarget > index->mEndTime)
  {
    return NS_ERROR_FAILURE;
  }

  // Binary search to find the last key point with time less than target.
  int start = 0;
  int end = index->Length() - 1;
  while (end > start) {
    int mid = start + ((end - start + 1) >> 1);
    if (index->Get(mid).mTime == aTarget) {
       start = mid;
       break;
    } else if (index->Get(mid).mTime < aTarget) {
      start = mid;
    } else {
      end = mid - 1;
    }
  }

  aResult = index->Get(start);
  NS_ASSERTION(aResult.mTime <= aTarget, "Result should have time <= target");
  return NS_OK;
}

nsresult SkeletonState::IndexedSeekTarget(int64_t aTarget,
                                            nsTArray<uint32_t>& aTracks,
                                            nsSeekTarget& aResult)
{
  if (!mActive || mVersion < SKELETON_VERSION(4,0)) {
    return NS_ERROR_FAILURE;
  }
  // Loop over all requested tracks' indexes, and get the keypoint for that
  // seek target. Record the keypoint with the lowest offset, this will be
  // our seek result. User must seek to the one with lowest offset to ensure we
  // pass "keyframes" on all tracks when we decode forwards to the seek target.
  nsSeekTarget r;
  for (uint32_t i=0; i<aTracks.Length(); i++) {
    nsKeyPoint k;
    if (NS_SUCCEEDED(IndexedSeekTargetForTrack(aTracks[i], aTarget, k)) &&
        k.mOffset < r.mKeyPoint.mOffset)
    {
      r.mKeyPoint = k;
      r.mSerial = aTracks[i];
    }
  }
  if (r.IsNull()) {
    return NS_ERROR_FAILURE;
  }
  LOG(PR_LOG_DEBUG, ("Indexed seek target for time %lld is offset %lld",
                     aTarget, r.mKeyPoint.mOffset));
  aResult = r;
  return NS_OK;
}

nsresult SkeletonState::GetDuration(const nsTArray<uint32_t>& aTracks,
                                      int64_t& aDuration)
{
  if (!mActive ||
      mVersion < SKELETON_VERSION(4,0) ||
      !HasIndex() ||
      aTracks.Length() == 0)
  {
    return NS_ERROR_FAILURE;
  }
  int64_t endTime = INT64_MIN;
  int64_t startTime = INT64_MAX;
  for (uint32_t i=0; i<aTracks.Length(); i++) {
    nsKeyFrameIndex* index = nullptr;
    mIndex.Get(aTracks[i], &index);
    if (!index) {
      // Can't get the timestamps for one of the required tracks, fail.
      return NS_ERROR_FAILURE;
    }
    if (index->mEndTime > endTime) {
      endTime = index->mEndTime;
    }
    if (index->mStartTime < startTime) {
      startTime = index->mStartTime;
    }
  }
  NS_ASSERTION(endTime > startTime, "Duration must be positive");
  CheckedInt64 duration = CheckedInt64(endTime) - startTime;
  aDuration = duration.isValid() ? duration.value() : 0;
  return duration.isValid() ? NS_OK : NS_ERROR_FAILURE;
}

bool SkeletonState::DecodeHeader(ogg_packet* aPacket)
{
  nsAutoRef<ogg_packet> autoRelease(aPacket);
  if (IsSkeletonBOS(aPacket)) {
    uint16_t verMajor = LEUint16(aPacket->packet + SKELETON_VERSION_MAJOR_OFFSET);
    uint16_t verMinor = LEUint16(aPacket->packet + SKELETON_VERSION_MINOR_OFFSET);

    // Read the presentation time. We read this before the version check as the
    // presentation time exists in all versions.
    int64_t n = LEInt64(aPacket->packet + SKELETON_PRESENTATION_TIME_NUMERATOR_OFFSET);
    int64_t d = LEInt64(aPacket->packet + SKELETON_PRESENTATION_TIME_DENOMINATOR_OFFSET);
    mPresentationTime = d == 0 ? 0 : (static_cast<float>(n) / static_cast<float>(d)) * USECS_PER_S;

    mVersion = SKELETON_VERSION(verMajor, verMinor);
    // We can only care to parse Skeleton version 4.0+.
    if (mVersion < SKELETON_VERSION(4,0) ||
        mVersion >= SKELETON_VERSION(5,0) ||
        aPacket->bytes < SKELETON_4_0_MIN_HEADER_LEN)
      return false;

    // Extract the segment length.
    mLength = LEInt64(aPacket->packet + SKELETON_FILE_LENGTH_OFFSET);

    LOG(PR_LOG_DEBUG, ("Skeleton segment length: %lld", mLength));

    // Initialize the serialno-to-index map.
    mIndex.Init();
    return true;
  } else if (IsSkeletonIndex(aPacket) && mVersion >= SKELETON_VERSION(4,0)) {
    return DecodeIndex(aPacket);
  } else if (aPacket->e_o_s) {
    mDoneReadingHeaders = true;
    return true;
  }
  return false;
}


} // namespace mozilla