Merge remote-tracking branch 'qatar/master'
authorMichael Niedermayer <michaelni@gmx.at>
Tue, 6 Dec 2011 00:37:27 +0000 (01:37 +0100)
committerMichael Niedermayer <michaelni@gmx.at>
Tue, 6 Dec 2011 00:37:27 +0000 (01:37 +0100)
* qatar/master:
  mov: Don't av_malloc(0).
  avconv: only allocate 1 AVFrame per input stream
  avconv: fix memleaks due to not freeing the AVFrame for audio
  h264-fate: remove -strict 1 except where necessary (mr4/5-tandberg).
  misc Doxygen markup improvements
  doxygen: eliminate Qt-style doxygen syntax
  g722: Add a regression test for muxing/demuxing in wav
  g722: Change bits per sample to 4
  g722dec: Signal skipping the lower bits via AVOptions instead of bits_per_coded_sample
  api-example: update to use avcodec_decode_audio4()
  avplay: use avcodec_decode_audio4()
  avplay: use a separate buffer for playing silence
  avformat: use avcodec_decode_audio4() in avformat_find_stream_info()
  avconv: use avcodec_decode_audio4() instead of avcodec_decode_audio3()
  mov: Allow empty stts atom.
  doc: document preferred Doxygen syntax and make patcheck detect it

Conflicts:
avconv.c
ffplay.c
libavcodec/mlpdec.c
libavcodec/version.h
libavformat/mov.c
tests/codec-regression.sh
tests/fate/h264.mak

Merged-by: Michael Niedermayer <michaelni@gmx.at>
31 files changed:
1  2 
avconv.c
cmdutils.h
doc/developer.texi
doc/examples/decoding_encoding.c
ffmpeg.c
ffplay.c
libavcodec/amrnbdec.c
libavcodec/cinepak.c
libavcodec/eamad.c
libavcodec/ivi_common.h
libavcodec/lsp.c
libavcodec/mlpdec.c
libavcodec/qcelpdata.h
libavcodec/rtjpeg.c
libavcodec/utils.c
libavcodec/version.h
libavdevice/pulse.c
libavdevice/x11grab.c
libavformat/avformat.h
libavformat/matroskadec.c
libavformat/mov.c
libavformat/nuv.c
libavformat/oggdec.c
libavformat/utils.c
libavutil/lzo.c
libavutil/lzo.h
libpostproc/postprocess.c
tests/codec-regression.sh
tests/fate/h264.mak
tests/ref/acodec/g722
tools/patcheck

diff --cc avconv.c
+++ b/avconv.c
@@@ -1820,10 -1751,10 +1828,9 @@@ static int transcode_video(InputStream 
      quality = same_quant ? decoded_frame->quality : 0;
      if (!*got_output) {
          /* no picture yet */
-         av_freep(&decoded_frame);
          return ret;
      }
 -    ist->next_pts = ist->pts = guess_correct_pts(&ist->pts_ctx, decoded_frame->pkt_pts,
 -                                                 decoded_frame->pkt_dts);
 +    ist->next_pts = ist->pts = decoded_frame->best_effort_timestamp;
      if (pkt->duration)
          ist->next_pts += av_rescale_q(pkt->duration, ist->st->time_base, AV_TIME_BASE_Q);
      else if (ist->st->codec->time_base.num != 0) {
  
  #if CONFIG_AVFILTER
          if (ost->input_video_filter) {
 -            AVRational sar;
 -            if (ist->st->sample_aspect_ratio.num)
 -                sar = ist->st->sample_aspect_ratio;
 -            else
 -                sar = ist->st->codec->sample_aspect_ratio;
 -            av_vsrc_buffer_add_frame(ost->input_video_filter, decoded_frame, ist->pts, sar);
 +            if (!decoded_frame->sample_aspect_ratio.num)
 +                decoded_frame->sample_aspect_ratio = ist->st->sample_aspect_ratio;
 +            decoded_frame->pts = ist->pts;
 +
 +            av_vsrc_buffer_add_frame(ost->input_video_filter, decoded_frame, AV_VSRC_BUF_FLAG_OVERWRITE);
-             if (!(filtered_frame = avcodec_alloc_frame())) {
-                 ret = AVERROR(ENOMEM);
-                 goto fail;
-             }
+             if (!ist->filtered_frame && !(ist->filtered_frame = avcodec_alloc_frame())) {
+                 av_free(buffer_to_free);
+                 return AVERROR(ENOMEM);
+             } else
+                 avcodec_get_frame_defaults(ist->filtered_frame);
+             filtered_frame = ist->filtered_frame;
              frame_available = avfilter_poll_frame(ost->output_video_filter->inputs[0]);
          }
          while (frame_available) {
diff --cc cmdutils.h
Simple merge
Simple merge
   * format handling
   */
  
 -#include <stdlib.h>
 -#include <stdio.h>
 -#include <string.h>
 -
 -#ifdef HAVE_AV_CONFIG_H
 -#undef HAVE_AV_CONFIG_H
 -#endif
 -
 +#include "libavutil/imgutils.h"
 +#include "libavutil/opt.h"
  #include "libavcodec/avcodec.h"
  #include "libavutil/mathematics.h"
+ #include "libavutil/samplefmt.h"
  
  #define INBUF_SIZE 4096
  #define AUDIO_INBUF_SIZE 20480
diff --cc ffmpeg.c
+++ b/ffmpeg.c
@@@ -168,38 -230,15 +168,39 @@@ static uint8_t *audio_buf
  static uint8_t *audio_out;
  static unsigned int allocated_audio_out_size, allocated_audio_buf_size;
  
- static void *samples;
 -static short *samples;
 -
 -static AVBitStreamFilterContext *video_bitstream_filters=NULL;
 -static AVBitStreamFilterContext *audio_bitstream_filters=NULL;
 -static AVBitStreamFilterContext *subtitle_bitstream_filters=NULL;
 +static uint8_t *input_tmp= NULL;
  
  #define DEFAULT_PASS_LOGFILENAME_PREFIX "ffmpeg2pass"
  
 -struct InputStream;
 +typedef struct InputStream {
 +    int file_index;
 +    AVStream *st;
 +    int discard;             /* true if stream data should be discarded */
 +    int decoding_needed;     /* true if the packets must be decoded in 'raw_fifo' */
 +    AVCodec *dec;
++    AVFrame *decoded_frame;
++    AVFrame *filtered_frame;
 +
 +    int64_t       start;     /* time when read started */
 +    int64_t       next_pts;  /* synthetic pts for cases where pkt.pts
 +                                is not defined */
 +    int64_t       pts;       /* current pts */
 +    double ts_scale;
 +    int is_start;            /* is 1 at the start and after a discontinuity */
 +    int showed_multi_packet_warning;
 +    AVDictionary *opts;
 +} InputStream;
 +
 +typedef struct InputFile {
 +    AVFormatContext *ctx;
 +    int eof_reached;      /* true if eof reached */
 +    int ist_index;        /* index of first stream in input_streams */
 +    int buffer_size;      /* current total buffer size */
 +    int64_t ts_offset;
 +    int nb_streams;       /* number of stream that ffmpeg is aware of; may be different
 +                             from ctx.nb_streams if new streams appear during av_read_frame() */
 +    int rate_emu;
 +} InputFile;
  
  typedef struct OutputStream {
      int file_index;          /* file index */
@@@ -658,8 -462,11 +659,11 @@@ void av_noreturn exit_program(int ret
      for(i=0;i<nb_input_files;i++) {
          av_close_input_file(input_files[i].ctx);
      }
--    for (i = 0; i < nb_input_streams; i++)
++    for (i = 0; i < nb_input_streams; i++) {
++        av_freep(&input_streams[i].decoded_frame);
++        av_freep(&input_streams[i].filtered_frame);
          av_dict_free(&input_streams[i].opts);
 -
 -    av_free(intra_matrix);
 -    av_free(inter_matrix);
++    }
  
      if (vstats_file)
          fclose(vstats_file);
      av_free(audio_buf);
      av_free(audio_out);
      allocated_audio_buf_size= allocated_audio_out_size= 0;
--    av_free(samples);
  
  #if CONFIG_AVFILTER
      avfilter_uninit();
@@@ -830,22 -751,16 +833,19 @@@ static void write_frame(AVFormatContex
      }
  }
  
 -#define MAX_AUDIO_PACKET_SIZE (128 * 1024)
 +static void generate_silence(uint8_t* buf, enum AVSampleFormat sample_fmt, size_t size)
 +{
 +    int fill_char = 0x00;
 +    if (sample_fmt == AV_SAMPLE_FMT_U8)
 +        fill_char = 0x80;
 +    memset(buf, fill_char, size);
 +}
  
--static void do_audio_out(AVFormatContext *s,
--                         OutputStream *ost,
--                         InputStream *ist,
--                         unsigned char *buf, int size)
++static void do_audio_out(AVFormatContext *s, OutputStream *ost,
++                         InputStream *ist, AVFrame *decoded_frame)
  {
      uint8_t *buftmp;
      int64_t audio_out_size, audio_buf_size;
--    int64_t allocated_for_size= size;
  
      int size_out, frame_bytes, ret, resample_changed;
      AVCodecContext *enc= ost->st->codec;
      int osize = av_get_bytes_per_sample(enc->sample_fmt);
      int isize = av_get_bytes_per_sample(dec->sample_fmt);
      const int coded_bps = av_get_bits_per_sample(enc->codec->id);
++    uint8_t *buf = decoded_frame->data[0];
++    int size     = decoded_frame->nb_samples * dec->channels * isize;
++    int64_t allocated_for_size = size;
  
  need_realloc:
      audio_buf_size= (allocated_for_size + isize*dec->channels - 1) / (isize*dec->channels);
@@@ -1550,536 -1441,462 +1553,540 @@@ static void print_report(OutputFile *ou
      }
  }
  
 -static void generate_silence(uint8_t* buf, enum AVSampleFormat sample_fmt, size_t size)
 +static void flush_encoders(OutputStream *ost_table, int nb_ostreams)
  {
 -    int fill_char = 0x00;
 -    if (sample_fmt == AV_SAMPLE_FMT_U8)
 -        fill_char = 0x80;
 -    memset(buf, fill_char, size);
 -}
 +    int i, ret;
  
 -/* pkt = NULL means EOF (needed to flush decoder buffers) */
 -static int output_packet(InputStream *ist, int ist_index,
 -                         OutputStream **ost_table, int nb_ostreams,
 -                         const AVPacket *pkt)
 -{
 -    AVFormatContext *os;
 -    OutputStream *ost;
 -    int ret, i;
 -    int got_output;
 -    AVFrame picture;
 -    void *buffer_to_free = NULL;
 -    static unsigned int samples_size= 0;
 -    AVSubtitle subtitle, *subtitle_to_free;
 -    int64_t pkt_pts = AV_NOPTS_VALUE;
 -#if CONFIG_AVFILTER
 -    int frame_available;
 -#endif
 -    float quality;
 -
 -    AVPacket avpkt;
 -    int bps = av_get_bytes_per_sample(ist->st->codec->sample_fmt);
 -
 -    if(ist->next_pts == AV_NOPTS_VALUE)
 -        ist->next_pts= ist->pts;
 +    for (i = 0; i < nb_ostreams; i++) {
 +        OutputStream   *ost = &ost_table[i];
 +        AVCodecContext *enc = ost->st->codec;
 +        AVFormatContext *os = output_files[ost->file_index].ctx;
  
 -    if (pkt == NULL) {
 -        /* EOF handling */
 -        av_init_packet(&avpkt);
 -        avpkt.data = NULL;
 -        avpkt.size = 0;
 -        goto handle_eof;
 -    } else {
 -        avpkt = *pkt;
 -    }
 +        if (!ost->encoding_needed)
 +            continue;
  
 -    if(pkt->dts != AV_NOPTS_VALUE)
 -        ist->next_pts = ist->pts = av_rescale_q(pkt->dts, ist->st->time_base, AV_TIME_BASE_Q);
 -    if(pkt->pts != AV_NOPTS_VALUE)
 -        pkt_pts = av_rescale_q(pkt->pts, ist->st->time_base, AV_TIME_BASE_Q);
 +        if (ost->st->codec->codec_type == AVMEDIA_TYPE_AUDIO && enc->frame_size <=1)
 +            continue;
 +        if (ost->st->codec->codec_type == AVMEDIA_TYPE_VIDEO && (os->oformat->flags & AVFMT_RAWPICTURE) && enc->codec->id == CODEC_ID_RAWVIDEO)
 +            continue;
  
 -    //while we have more to decode or while the decoder did output something on EOF
 -    while (avpkt.size > 0 || (!pkt && got_output)) {
 -        uint8_t *data_buf, *decoded_data_buf;
 -        int data_size, decoded_data_size;
 -    handle_eof:
 -        ist->pts= ist->next_pts;
 +        for(;;) {
 +            AVPacket pkt;
 +            int fifo_bytes;
 +            av_init_packet(&pkt);
 +            pkt.stream_index= ost->index;
  
 -        if(avpkt.size && avpkt.size != pkt->size &&
 -           ((!ist->showed_multi_packet_warning && verbose>0) || verbose>1)){
 -            fprintf(stderr, "Multiple frames in a packet from stream %d\n", pkt->stream_index);
 -            ist->showed_multi_packet_warning=1;
 -        }
 +            switch (ost->st->codec->codec_type) {
 +            case AVMEDIA_TYPE_AUDIO:
 +                fifo_bytes = av_fifo_size(ost->fifo);
 +                ret = 0;
 +                /* encode any samples remaining in fifo */
 +                if (fifo_bytes > 0) {
 +                    int osize = av_get_bytes_per_sample(enc->sample_fmt);
 +                    int fs_tmp = enc->frame_size;
 +
 +                    av_fifo_generic_read(ost->fifo, audio_buf, fifo_bytes, NULL);
 +                    if (enc->codec->capabilities & CODEC_CAP_SMALL_LAST_FRAME) {
 +                        enc->frame_size = fifo_bytes / (osize * enc->channels);
 +                    } else { /* pad */
 +                        int frame_bytes = enc->frame_size*osize*enc->channels;
 +                        if (allocated_audio_buf_size < frame_bytes)
 +                            exit_program(1);
 +                        generate_silence(audio_buf+fifo_bytes, enc->sample_fmt, frame_bytes - fifo_bytes);
 +                    }
  
 -        /* decode the packet if needed */
 -        decoded_data_buf = NULL; /* fail safe */
 -        decoded_data_size= 0;
 -        data_buf  = avpkt.data;
 -        data_size = avpkt.size;
 -        subtitle_to_free = NULL;
 -        if (ist->decoding_needed) {
 -            switch(ist->st->codec->codec_type) {
 -            case AVMEDIA_TYPE_AUDIO:{
 -                if(pkt && samples_size < FFMAX(pkt->size*sizeof(*samples), AVCODEC_MAX_AUDIO_FRAME_SIZE)) {
 -                    samples_size = FFMAX(pkt->size*sizeof(*samples), AVCODEC_MAX_AUDIO_FRAME_SIZE);
 -                    av_free(samples);
 -                    samples= av_malloc(samples_size);
 +                    ret = avcodec_encode_audio(enc, bit_buffer, bit_buffer_size, (short *)audio_buf);
 +                    pkt.duration = av_rescale((int64_t)enc->frame_size*ost->st->time_base.den,
 +                                              ost->st->time_base.num, enc->sample_rate);
 +                    enc->frame_size = fs_tmp;
                  }
 -                decoded_data_size= samples_size;
 -                    /* XXX: could avoid copy if PCM 16 bits with same
 -                       endianness as CPU */
 -                ret = avcodec_decode_audio3(ist->st->codec, samples, &decoded_data_size,
 -                                            &avpkt);
 -                if (ret < 0)
 -                    return ret;
 -                avpkt.data += ret;
 -                avpkt.size -= ret;
 -                data_size   = ret;
 -                got_output  = decoded_data_size > 0;
 -                /* Some bug in mpeg audio decoder gives */
 -                /* decoded_data_size < 0, it seems they are overflows */
 -                if (!got_output) {
 -                    /* no audio frame */
 -                    continue;
 +                if (ret <= 0) {
 +                    ret = avcodec_encode_audio(enc, bit_buffer, bit_buffer_size, NULL);
                  }
 -                decoded_data_buf = (uint8_t *)samples;
 -                ist->next_pts += ((int64_t)AV_TIME_BASE/bps * decoded_data_size) /
 -                    (ist->st->codec->sample_rate * ist->st->codec->channels);
 -                break;}
 -            case AVMEDIA_TYPE_VIDEO:
 -                    decoded_data_size = (ist->st->codec->width * ist->st->codec->height * 3) / 2;
 -                    /* XXX: allocate picture correctly */
 -                    avcodec_get_frame_defaults(&picture);
 -                    avpkt.pts = pkt_pts;
 -                    avpkt.dts = ist->pts;
 -                    pkt_pts = AV_NOPTS_VALUE;
 -
 -                    ret = avcodec_decode_video2(ist->st->codec,
 -                                                &picture, &got_output, &avpkt);
 -                    quality = same_quality ? picture.quality : 0;
 -                    if (ret < 0)
 -                        return ret;
 -                    if (!got_output) {
 -                        /* no picture yet */
 -                        goto discard_packet;
 -                    }
 -                    ist->next_pts = ist->pts = guess_correct_pts(&ist->pts_ctx, picture.pkt_pts, picture.pkt_dts);
 -                    if (ist->st->codec->time_base.num != 0) {
 -                        int ticks= ist->st->parser ? ist->st->parser->repeat_pict+1 : ist->st->codec->ticks_per_frame;
 -                        ist->next_pts += ((int64_t)AV_TIME_BASE *
 -                                          ist->st->codec->time_base.num * ticks) /
 -                            ist->st->codec->time_base.den;
 -                    }
 -                    avpkt.size = 0;
 -                    buffer_to_free = NULL;
 -                    pre_process_video_frame(ist, (AVPicture *)&picture, &buffer_to_free);
 -                    break;
 -            case AVMEDIA_TYPE_SUBTITLE:
 -                ret = avcodec_decode_subtitle2(ist->st->codec,
 -                                               &subtitle, &got_output, &avpkt);
 -                if (ret < 0)
 -                    return ret;
 -                if (!got_output) {
 -                    goto discard_packet;
 +                if (ret < 0) {
 +                    av_log(NULL, AV_LOG_FATAL, "Audio encoding failed\n");
 +                    exit_program(1);
                  }
 -                subtitle_to_free = &subtitle;
 -                avpkt.size = 0;
 -                break;
 -            default:
 -                return -1;
 -            }
 -        } else {
 -            switch(ist->st->codec->codec_type) {
 -            case AVMEDIA_TYPE_AUDIO:
 -                ist->next_pts += ((int64_t)AV_TIME_BASE * ist->st->codec->frame_size) /
 -                    ist->st->codec->sample_rate;
 +                audio_size += ret;
 +                pkt.flags |= AV_PKT_FLAG_KEY;
                  break;
              case AVMEDIA_TYPE_VIDEO:
 -                if (ist->st->codec->time_base.num != 0) {
 -                    int ticks= ist->st->parser ? ist->st->parser->repeat_pict+1 : ist->st->codec->ticks_per_frame;
 -                    ist->next_pts += ((int64_t)AV_TIME_BASE *
 -                                      ist->st->codec->time_base.num * ticks) /
 -                        ist->st->codec->time_base.den;
 +                ret = avcodec_encode_video(enc, bit_buffer, bit_buffer_size, NULL);
 +                if (ret < 0) {
 +                    av_log(NULL, AV_LOG_FATAL, "Video encoding failed\n");
 +                    exit_program(1);
 +                }
 +                video_size += ret;
 +                if(enc->coded_frame && enc->coded_frame->key_frame)
 +                    pkt.flags |= AV_PKT_FLAG_KEY;
 +                if (ost->logfile && enc->stats_out) {
 +                    fprintf(ost->logfile, "%s", enc->stats_out);
                  }
                  break;
 +            default:
 +                ret=-1;
              }
 -            ret = avpkt.size;
 -            avpkt.size = 0;
 +
 +            if (ret <= 0)
 +                break;
 +            pkt.data = bit_buffer;
 +            pkt.size = ret;
 +            if (enc->coded_frame && enc->coded_frame->pts != AV_NOPTS_VALUE)
 +                pkt.pts= av_rescale_q(enc->coded_frame->pts, enc->time_base, ost->st->time_base);
 +            write_frame(os, &pkt, ost->st->codec, ost->bitstream_filters);
          }
 +    }
 +}
  
 -#if CONFIG_AVFILTER
 -        if (ist->st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
 -            for (i = 0; i < nb_ostreams; i++) {
 -                ost = ost_table[i];
 -                if (ost->input_video_filter && ost->source_index == ist_index) {
 -                    AVRational sar;
 -                    if (ist->st->sample_aspect_ratio.num)
 -                        sar = ist->st->sample_aspect_ratio;
 -                    else
 -                        sar = ist->st->codec->sample_aspect_ratio;
 -                    // add it to be filtered
 -                    av_vsrc_buffer_add_frame(ost->input_video_filter, &picture,
 -                                             ist->pts,
 -                                             sar);
 -                }
 +/*
 + * Check whether a packet from ist should be written into ost at this time
 + */
 +static int check_output_constraints(InputStream *ist, OutputStream *ost)
 +{
 +    OutputFile *of = &output_files[ost->file_index];
 +    int ist_index  = ist - input_streams;
 +
 +    if (ost->source_index != ist_index)
 +        return 0;
 +
 +    if (of->start_time && ist->pts < of->start_time)
 +        return 0;
 +
 +    if (of->recording_time != INT64_MAX &&
 +        av_compare_ts(ist->pts, AV_TIME_BASE_Q, of->recording_time + of->start_time,
 +                      (AVRational){1, 1000000}) >= 0) {
 +        ost->is_past_recording_time = 1;
 +        return 0;
 +    }
 +
 +    return 1;
 +}
 +
 +static void do_streamcopy(InputStream *ist, OutputStream *ost, const AVPacket *pkt)
 +{
 +    OutputFile *of = &output_files[ost->file_index];
 +    int64_t ost_tb_start_time = av_rescale_q(of->start_time, AV_TIME_BASE_Q, ost->st->time_base);
 +    AVPicture pict;
 +    AVPacket opkt;
 +
 +    av_init_packet(&opkt);
 +
 +    if ((!ost->frame_number && !(pkt->flags & AV_PKT_FLAG_KEY)) &&
 +        !ost->copy_initial_nonkeyframes)
 +        return;
 +
 +    /* force the input stream PTS */
 +    if (ost->st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
 +        audio_size += pkt->size;
 +    else if (ost->st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
 +        video_size += pkt->size;
 +        ost->sync_opts++;
 +    }
 +
 +    opkt.stream_index = ost->index;
 +    if (pkt->pts != AV_NOPTS_VALUE)
 +        opkt.pts = av_rescale_q(pkt->pts, ist->st->time_base, ost->st->time_base) - ost_tb_start_time;
 +    else
 +        opkt.pts = AV_NOPTS_VALUE;
 +
 +    if (pkt->dts == AV_NOPTS_VALUE)
 +        opkt.dts = av_rescale_q(ist->pts, AV_TIME_BASE_Q, ost->st->time_base);
 +    else
 +        opkt.dts = av_rescale_q(pkt->dts, ist->st->time_base, ost->st->time_base);
 +    opkt.dts -= ost_tb_start_time;
 +
 +    opkt.duration = av_rescale_q(pkt->duration, ist->st->time_base, ost->st->time_base);
 +    opkt.flags    = pkt->flags;
 +
 +    //FIXME remove the following 2 lines they shall be replaced by the bitstream filters
 +    if(   ost->st->codec->codec_id != CODEC_ID_H264
 +       && ost->st->codec->codec_id != CODEC_ID_MPEG1VIDEO
 +       && ost->st->codec->codec_id != CODEC_ID_MPEG2VIDEO
 +       ) {
 +        if (av_parser_change(ist->st->parser, ost->st->codec, &opkt.data, &opkt.size, pkt->data, pkt->size, pkt->flags & AV_PKT_FLAG_KEY))
 +            opkt.destruct = av_destruct_packet;
 +    } else {
 +        opkt.data = pkt->data;
 +        opkt.size = pkt->size;
 +    }
 +    if (of->ctx->oformat->flags & AVFMT_RAWPICTURE) {
 +        /* store AVPicture in AVPacket, as expected by the output format */
 +        avpicture_fill(&pict, opkt.data, ost->st->codec->pix_fmt, ost->st->codec->width, ost->st->codec->height);
 +        opkt.data = (uint8_t *)&pict;
 +        opkt.size = sizeof(AVPicture);
 +        opkt.flags |= AV_PKT_FLAG_KEY;
 +    }
 +
 +    write_frame(of->ctx, &opkt, ost->st->codec, ost->bitstream_filters);
 +    ost->st->codec->frame_number++;
 +    ost->frame_number++;
 +    av_free_packet(&opkt);
 +}
 +
 +static void rate_emu_sleep(InputStream *ist)
 +{
 +    if (input_files[ist->file_index].rate_emu) {
 +        int64_t pts = av_rescale(ist->pts, 1000000, AV_TIME_BASE);
 +        int64_t now = av_gettime() - ist->start;
 +        if (pts > now)
 +            usleep(pts - now);
 +    }
 +}
 +
 +static int transcode_audio(InputStream *ist, AVPacket *pkt, int *got_output)
 +{
-     static unsigned int samples_size = 0;
++    AVFrame *decoded_frame;
++    AVCodecContext *avctx = ist->st->codec;
 +    int bps = av_get_bytes_per_sample(ist->st->codec->sample_fmt);
-     uint8_t *decoded_data_buf  = NULL;
-     int      decoded_data_size = 0;
 +    int i, ret;
 +
-     if (pkt && samples_size < FFMAX(pkt->size * bps, AVCODEC_MAX_AUDIO_FRAME_SIZE)) {
-         av_free(samples);
-         samples_size = FFMAX(pkt->size * bps, AVCODEC_MAX_AUDIO_FRAME_SIZE);
-         samples      = av_malloc(samples_size);
-     }
-     decoded_data_size = samples_size;
++    if (!ist->decoded_frame && !(ist->decoded_frame = avcodec_alloc_frame()))
++        return AVERROR(ENOMEM);
++    else
++        avcodec_get_frame_defaults(ist->decoded_frame);
++    decoded_frame = ist->decoded_frame;
 +
-     ret = avcodec_decode_audio3(ist->st->codec, samples, &decoded_data_size,
-                                 pkt);
-     if (ret < 0)
++    ret = avcodec_decode_audio4(avctx, decoded_frame, got_output, pkt);
++    if (ret < 0) {
 +        return ret;
-     *got_output  = decoded_data_size > 0;
++    }
 +
-     /* Some bug in mpeg audio decoder gives */
-     /* decoded_data_size < 0, it seems they are overflows */
 +    if (!*got_output) {
 +        /* no audio frame */
 +        return ret;
 +    }
 +
-     decoded_data_buf = (uint8_t *)samples;
-     ist->next_pts   += ((int64_t)AV_TIME_BASE/bps * decoded_data_size) /
-                        (ist->st->codec->sample_rate * ist->st->codec->channels);
++    /* if the decoder provides a pts, use it instead of the last packet pts.
++       the decoder could be delaying output by a packet or more. */
++    if (decoded_frame->pts != AV_NOPTS_VALUE)
++        ist->next_pts = decoded_frame->pts;
++
++    /* increment next_pts to use for the case where the input stream does not
++       have timestamps or there are multiple frames in the packet */
++    ist->next_pts += ((int64_t)AV_TIME_BASE * decoded_frame->nb_samples) /
++                     avctx->sample_rate;
 +
 +    // preprocess audio (volume)
 +    if (audio_volume != 256) {
-         switch (ist->st->codec->sample_fmt) {
++        int decoded_data_size = decoded_frame->nb_samples * avctx->channels * bps;
++        void *samples = decoded_frame->data[0];
++        switch (avctx->sample_fmt) {
 +        case AV_SAMPLE_FMT_U8:
 +        {
 +            uint8_t *volp = samples;
 +            for (i = 0; i < (decoded_data_size / sizeof(*volp)); i++) {
 +                int v = (((*volp - 128) * audio_volume + 128) >> 8) + 128;
 +                *volp++ = av_clip_uint8(v);
              }
 +            break;
          }
 -#endif
 -
 -        // preprocess audio (volume)
 -        if (ist->st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
 -            if (audio_volume != 256) {
 -                short *volp;
 -                volp = samples;
 -                for(i=0;i<(decoded_data_size / sizeof(short));i++) {
 -                    int v = ((*volp) * audio_volume + 128) >> 8;
 -                    if (v < -32768) v = -32768;
 -                    if (v >  32767) v = 32767;
 -                    *volp++ = v;
 -                }
 +        case AV_SAMPLE_FMT_S16:
 +        {
 +            int16_t *volp = samples;
 +            for (i = 0; i < (decoded_data_size / sizeof(*volp)); i++) {
 +                int v = ((*volp) * audio_volume + 128) >> 8;
 +                *volp++ = av_clip_int16(v);
              }
 +            break;
          }
 -
 -        /* frame rate emulation */
 -        if (rate_emu) {
 -            int64_t pts = av_rescale(ist->pts, 1000000, AV_TIME_BASE);
 -            int64_t now = av_gettime() - ist->start;
 -            if (pts > now)
 -                usleep(pts - now);
 +        case AV_SAMPLE_FMT_S32:
 +        {
 +            int32_t *volp = samples;
 +            for (i = 0; i < (decoded_data_size / sizeof(*volp)); i++) {
 +                int64_t v = (((int64_t)*volp * audio_volume + 128) >> 8);
 +                *volp++ = av_clipl_int32(v);
 +            }
 +            break;
          }
 -        /* if output time reached then transcode raw format,
 -           encode packets and output them */
 -        if (start_time == 0 || ist->pts >= start_time)
 -            for(i=0;i<nb_ostreams;i++) {
 -                int frame_size;
 -
 -                ost = ost_table[i];
 -                if (ost->source_index == ist_index) {
 -#if CONFIG_AVFILTER
 -                frame_available = ist->st->codec->codec_type != AVMEDIA_TYPE_VIDEO ||
 -                    !ost->output_video_filter || avfilter_poll_frame(ost->output_video_filter->inputs[0]);
 -                while (frame_available) {
 -                    AVRational ist_pts_tb;
 -                    if (ist->st->codec->codec_type == AVMEDIA_TYPE_VIDEO && ost->output_video_filter)
 -                        get_filtered_video_frame(ost->output_video_filter, &picture, &ost->picref, &ist_pts_tb);
 -                    if (ost->picref)
 -                        ist->pts = av_rescale_q(ost->picref->pts, ist_pts_tb, AV_TIME_BASE_Q);
 -#endif
 -                    os = output_files[ost->file_index];
 -
 -                    /* set the input output pts pairs */
 -                    //ost->sync_ipts = (double)(ist->pts + input_files[ist->file_index].ts_offset - start_time)/ AV_TIME_BASE;
 -
 -                    if (ost->encoding_needed) {
 -                        av_assert0(ist->decoding_needed);
 -                        switch(ost->st->codec->codec_type) {
 -                        case AVMEDIA_TYPE_AUDIO:
 -                            do_audio_out(os, ost, ist, decoded_data_buf, decoded_data_size);
 -                            break;
 -                        case AVMEDIA_TYPE_VIDEO:
 +        case AV_SAMPLE_FMT_FLT:
 +        {
 +            float *volp = samples;
 +            float scale = audio_volume / 256.f;
 +            for (i = 0; i < (decoded_data_size / sizeof(*volp)); i++) {
 +                *volp++ *= scale;
 +            }
 +            break;
 +        }
 +        case AV_SAMPLE_FMT_DBL:
 +        {
 +            double *volp = samples;
 +            double scale = audio_volume / 256.;
 +            for (i = 0; i < (decoded_data_size / sizeof(*volp)); i++) {
 +                *volp++ *= scale;
 +            }
 +            break;
 +        }
 +        default:
 +            av_log(NULL, AV_LOG_FATAL,
 +                   "Audio volume adjustment on sample format %s is not supported.\n",
 +                   av_get_sample_fmt_name(ist->st->codec->sample_fmt));
 +            exit_program(1);
 +        }
 +    }
 +
 +    rate_emu_sleep(ist);
 +
 +    for (i = 0; i < nb_output_streams; i++) {
 +        OutputStream *ost = &output_streams[i];
 +
 +        if (!check_output_constraints(ist, ost) || !ost->encoding_needed)
 +            continue;
-         do_audio_out(output_files[ost->file_index].ctx, ost, ist,
-                      decoded_data_buf, decoded_data_size);
++        do_audio_out(output_files[ost->file_index].ctx, ost, ist, decoded_frame);
 +    }
++
 +    return ret;
 +}
 +
 +static int transcode_video(InputStream *ist, AVPacket *pkt, int *got_output, int64_t *pkt_pts, int64_t *pkt_dts)
 +{
 +    AVFrame *decoded_frame, *filtered_frame = NULL;
 +    void *buffer_to_free = NULL;
 +    int i, ret = 0;
 +    float quality = 0;
  #if CONFIG_AVFILTER
 -                            if (ost->picref->video && !ost->frame_aspect_ratio)
 -                                ost->st->codec->sample_aspect_ratio = ost->picref->video->pixel_aspect;
 +    int frame_available = 1;
  #endif
 -                            do_video_out(os, ost, ist, &picture, &frame_size,
 -                                         same_quality ? quality : ost->st->codec->global_quality);
 -                            if (vstats_filename && frame_size)
 -                                do_video_stats(os, ost, frame_size);
 -                            break;
 -                        case AVMEDIA_TYPE_SUBTITLE:
 -                            do_subtitle_out(os, ost, ist, &subtitle,
 -                                            pkt->pts);
 -                            break;
 -                        default:
 -                            abort();
 -                        }
 -                    } else {
 -                        AVFrame avframe; //FIXME/XXX remove this
 -                        AVPacket opkt;
 -                        int64_t ost_tb_start_time= av_rescale_q(start_time, AV_TIME_BASE_Q, ost->st->time_base);
 +    int duration=0;
 +    int64_t *best_effort_timestamp;
 +    AVRational *frame_sample_aspect;
  
-     if (!(decoded_frame = avcodec_alloc_frame()))
 -                        av_init_packet(&opkt);
++    if (!ist->decoded_frame && !(ist->decoded_frame = avcodec_alloc_frame()))
 +        return AVERROR(ENOMEM);
++    else
++        avcodec_get_frame_defaults(ist->decoded_frame);
++    decoded_frame = ist->decoded_frame;
 +    pkt->pts  = *pkt_pts;
 +    pkt->dts  = *pkt_dts;
 +    *pkt_pts  = AV_NOPTS_VALUE;
 +
 +    if (pkt->duration) {
 +        duration = av_rescale_q(pkt->duration, ist->st->time_base, AV_TIME_BASE_Q);
 +    } else if(ist->st->codec->time_base.num != 0) {
 +        int ticks= ist->st->parser ? ist->st->parser->repeat_pict+1 : ist->st->codec->ticks_per_frame;
 +        duration = ((int64_t)AV_TIME_BASE *
 +                          ist->st->codec->time_base.num * ticks) /
 +                          ist->st->codec->time_base.den;
 +    }
  
 -                        if ((!ost->frame_number && !(pkt->flags & AV_PKT_FLAG_KEY)) && !copy_initial_nonkeyframes)
 -#if !CONFIG_AVFILTER
 -                            continue;
 -#else
 -                            goto cont;
 -#endif
 +    if(*pkt_dts != AV_NOPTS_VALUE && duration) {
 +        *pkt_dts += duration;
 +    }else
 +        *pkt_dts = AV_NOPTS_VALUE;
 +
 +    ret = avcodec_decode_video2(ist->st->codec,
 +                                decoded_frame, got_output, pkt);
 +    if (ret < 0)
-         goto fail;
++        return ret;
  
 -                        /* no reencoding needed : output the packet directly */
 -                        /* force the input stream PTS */
 +    quality = same_quant ? decoded_frame->quality : 0;
 +    if (!*got_output) {
 +        /* no picture yet */
-         av_freep(&decoded_frame);
 +        return ret;
 +    }
  
 -                        avcodec_get_frame_defaults(&avframe);
 -                        ost->st->codec->coded_frame= &avframe;
 -                        avframe.key_frame = pkt->flags & AV_PKT_FLAG_KEY;
 +    best_effort_timestamp= av_opt_ptr(avcodec_get_frame_class(), decoded_frame, "best_effort_timestamp");
 +    if(*best_effort_timestamp != AV_NOPTS_VALUE)
 +        ist->next_pts = ist->pts = *best_effort_timestamp;
  
 -                        if(ost->st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
 -                            audio_size += data_size;
 -                        else if (ost->st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
 -                            video_size += data_size;
 -                            ost->sync_opts++;
 -                        }
 +    ist->next_pts += duration;
 +    pkt->size = 0;
  
 -                        opkt.stream_index= ost->index;
 -                        if(pkt->pts != AV_NOPTS_VALUE)
 -                            opkt.pts= av_rescale_q(pkt->pts, ist->st->time_base, ost->st->time_base) - ost_tb_start_time;
 -                        else
 -                            opkt.pts= AV_NOPTS_VALUE;
 -
 -                        if (pkt->dts == AV_NOPTS_VALUE)
 -                            opkt.dts = av_rescale_q(ist->pts, AV_TIME_BASE_Q, ost->st->time_base);
 -                        else
 -                            opkt.dts = av_rescale_q(pkt->dts, ist->st->time_base, ost->st->time_base);
 -                        opkt.dts -= ost_tb_start_time;
 -
 -                        opkt.duration = av_rescale_q(pkt->duration, ist->st->time_base, ost->st->time_base);
 -                        opkt.flags= pkt->flags;
 -
 -                        //FIXME remove the following 2 lines they shall be replaced by the bitstream filters
 -                        if(   ost->st->codec->codec_id != CODEC_ID_H264
 -                           && ost->st->codec->codec_id != CODEC_ID_MPEG1VIDEO
 -                           && ost->st->codec->codec_id != CODEC_ID_MPEG2VIDEO
 -                           ) {
 -                            if(av_parser_change(ist->st->parser, ost->st->codec, &opkt.data, &opkt.size, data_buf, data_size, pkt->flags & AV_PKT_FLAG_KEY))
 -                                opkt.destruct= av_destruct_packet;
 -                        } else {
 -                            opkt.data = data_buf;
 -                            opkt.size = data_size;
 -                        }
 +    pre_process_video_frame(ist, (AVPicture *)decoded_frame, &buffer_to_free);
  
 -                        write_frame(os, &opkt, ost->st->codec, ost->bitstream_filters);
 -                        ost->st->codec->frame_number++;
 -                        ost->frame_number++;
 -                        av_free_packet(&opkt);
 -                    }
  #if CONFIG_AVFILTER
 -                    cont:
 -                    frame_available = (ist->st->codec->codec_type == AVMEDIA_TYPE_VIDEO) &&
 -                                       ost->output_video_filter && avfilter_poll_frame(ost->output_video_filter->inputs[0]);
 -                    if (ost->picref)
 -                        avfilter_unref_buffer(ost->picref);
 -                }
 -#endif
 -                }
 -            }
 -
 -        av_free(buffer_to_free);
 -        /* XXX: allocate the subtitles in the codec ? */
 -        if (subtitle_to_free) {
 -            avsubtitle_free(subtitle_to_free);
 -            subtitle_to_free = NULL;
 +    frame_sample_aspect= av_opt_ptr(avcodec_get_frame_class(), decoded_frame, "sample_aspect_ratio");
 +    for(i=0;i<nb_output_streams;i++) {
 +        OutputStream *ost = ost = &output_streams[i];
 +        if(check_output_constraints(ist, ost)){
 +            if (!frame_sample_aspect->num)
 +                *frame_sample_aspect = ist->st->sample_aspect_ratio;
 +            decoded_frame->pts = ist->pts;
 +
 +            av_vsrc_buffer_add_frame(ost->input_video_filter, decoded_frame, AV_VSRC_BUF_FLAG_OVERWRITE);
          }
      }
 - discard_packet:
 -    if (pkt == NULL) {
 -        /* EOF handling */
 +#endif
  
 -        for(i=0;i<nb_ostreams;i++) {
 -            ost = ost_table[i];
 -            if (ost->source_index == ist_index) {
 -                AVCodecContext *enc= ost->st->codec;
 -                os = output_files[ost->file_index];
 -
 -                if(ost->st->codec->codec_type == AVMEDIA_TYPE_AUDIO && enc->frame_size <=1)
 -                    continue;
 -                if(ost->st->codec->codec_type == AVMEDIA_TYPE_VIDEO && (os->oformat->flags & AVFMT_RAWPICTURE))
 -                    continue;
 -
 -                if (ost->encoding_needed) {
 -                    for(;;) {
 -                        AVPacket pkt;
 -                        int fifo_bytes;
 -                        av_init_packet(&pkt);
 -                        pkt.stream_index= ost->index;
 -
 -                        switch(ost->st->codec->codec_type) {
 -                        case AVMEDIA_TYPE_AUDIO:
 -                            fifo_bytes = av_fifo_size(ost->fifo);
 -                            ret = 0;
 -                            /* encode any samples remaining in fifo */
 -                            if (fifo_bytes > 0) {
 -                                int osize = av_get_bytes_per_sample(enc->sample_fmt);
 -                                int fs_tmp = enc->frame_size;
 -
 -                                av_fifo_generic_read(ost->fifo, audio_buf, fifo_bytes, NULL);
 -                                if (enc->codec->capabilities & CODEC_CAP_SMALL_LAST_FRAME) {
 -                                    enc->frame_size = fifo_bytes / (osize * enc->channels);
 -                                } else { /* pad */
 -                                    int frame_bytes = enc->frame_size*osize*enc->channels;
 -                                    if (allocated_audio_buf_size < frame_bytes)
 -                                        exit_program(1);
 -                                    generate_silence(audio_buf+fifo_bytes, enc->sample_fmt, frame_bytes - fifo_bytes);
 -                                }
 -
 -                                ret = avcodec_encode_audio(enc, bit_buffer, bit_buffer_size, (short *)audio_buf);
 -                                pkt.duration = av_rescale((int64_t)enc->frame_size*ost->st->time_base.den,
 -                                                          ost->st->time_base.num, enc->sample_rate);
 -                                enc->frame_size = fs_tmp;
 -                            }
 -                            if(ret <= 0) {
 -                                ret = avcodec_encode_audio(enc, bit_buffer, bit_buffer_size, NULL);
 -                            }
 -                            if (ret < 0) {
 -                                fprintf(stderr, "Audio encoding failed\n");
 -                                exit_program(1);
 -                            }
 -                            audio_size += ret;
 -                            pkt.flags |= AV_PKT_FLAG_KEY;
 -                            break;
 -                        case AVMEDIA_TYPE_VIDEO:
 -                            ret = avcodec_encode_video(enc, bit_buffer, bit_buffer_size, NULL);
 -                            if (ret < 0) {
 -                                fprintf(stderr, "Video encoding failed\n");
 -                                exit_program(1);
 -                            }
 -                            video_size += ret;
 -                            if(enc->coded_frame && enc->coded_frame->key_frame)
 -                                pkt.flags |= AV_PKT_FLAG_KEY;
 -                            if (ost->logfile && enc->stats_out) {
 -                                fprintf(ost->logfile, "%s", enc->stats_out);
 -                            }
 -                            break;
 -                        default:
 -                            ret=-1;
 -                        }
 +    rate_emu_sleep(ist);
  
 -                        if(ret<=0)
 -                            break;
 -                        pkt.data= bit_buffer;
 -                        pkt.size= ret;
 -                        if(enc->coded_frame && enc->coded_frame->pts != AV_NOPTS_VALUE)
 -                            pkt.pts= av_rescale_q(enc->coded_frame->pts, enc->time_base, ost->st->time_base);
 -                        write_frame(os, &pkt, ost->st->codec, ost->bitstream_filters);
 -                    }
 +    for (i = 0; i < nb_output_streams; i++) {
 +        OutputStream *ost = &output_streams[i];
 +        int frame_size;
 +
 +        if (!check_output_constraints(ist, ost) || !ost->encoding_needed)
 +            continue;
 +
 +#if CONFIG_AVFILTER
 +        if (ost->input_video_filter) {
 +            frame_available = avfilter_poll_frame(ost->output_video_filter->inputs[0]);
 +        }
 +        while (frame_available) {
 +            if (ost->output_video_filter) {
 +                AVRational ist_pts_tb = ost->output_video_filter->inputs[0]->time_base;
 +                if (av_buffersink_get_buffer_ref(ost->output_video_filter, &ost->picref, 0) < 0)
 +                    goto cont;
-                 if (!filtered_frame && !(filtered_frame = avcodec_alloc_frame())) {
-                     ret = AVERROR(ENOMEM);
-                     goto fail;
-                 }
++                if (!ist->filtered_frame && !(ist->filtered_frame = avcodec_alloc_frame())) {
++                    av_free(buffer_to_free);
++                    return AVERROR(ENOMEM);
++                } else
++                    avcodec_get_frame_defaults(ist->filtered_frame);
++                filtered_frame = ist->filtered_frame;
 +                *filtered_frame= *decoded_frame; //for me_threshold
 +                if (ost->picref) {
 +                    avfilter_fill_frame_from_video_buffer_ref(filtered_frame, ost->picref);
 +                    ist->pts = av_rescale_q(ost->picref->pts, ist_pts_tb, AV_TIME_BASE_Q);
                  }
              }
 +            if (ost->picref->video && !ost->frame_aspect_ratio)
 +                ost->st->codec->sample_aspect_ratio = ost->picref->video->sample_aspect_ratio;
 +#else
 +            filtered_frame = decoded_frame;
 +#endif
 +
 +            do_video_out(output_files[ost->file_index].ctx, ost, ist, filtered_frame, &frame_size,
 +                         same_quant ? quality : ost->st->codec->global_quality);
 +            if (vstats_filename && frame_size)
 +                do_video_stats(output_files[ost->file_index].ctx, ost, frame_size);
 +#if CONFIG_AVFILTER
 +            cont:
 +            frame_available = ost->output_video_filter && avfilter_poll_frame(ost->output_video_filter->inputs[0]);
 +            avfilter_unref_buffer(ost->picref);
          }
-         av_freep(&filtered_frame);
 +#endif
      }
  
- fail:
 -    return 0;
 +    av_free(buffer_to_free);
-     av_freep(&decoded_frame);
 +    return ret;
  }
  
 -static void print_sdp(AVFormatContext **avc, int n)
 +static int transcode_subtitles(InputStream *ist, AVPacket *pkt, int *got_output)
  {
 -    char sdp[2048];
 +    AVSubtitle subtitle;
 +    int i, ret = avcodec_decode_subtitle2(ist->st->codec,
 +                                          &subtitle, got_output, pkt);
 +    if (ret < 0)
 +        return ret;
 +    if (!*got_output)
 +        return ret;
  
 -    av_sdp_create(avc, n, sdp, sizeof(sdp));
 -    printf("SDP:\n%s\n", sdp);
 -    fflush(stdout);
 +    rate_emu_sleep(ist);
 +
 +    for (i = 0; i < nb_output_streams; i++) {
 +        OutputStream *ost = &output_streams[i];
 +
 +        if (!check_output_constraints(ist, ost) || !ost->encoding_needed)
 +            continue;
 +
 +        do_subtitle_out(output_files[ost->file_index].ctx, ost, ist, &subtitle, pkt->pts);
 +    }
 +
 +    avsubtitle_free(&subtitle);
 +    return ret;
  }
  
 -static int copy_chapters(int infile, int outfile)
 +/* pkt = NULL means EOF (needed to flush decoder buffers) */
 +static int output_packet(InputStream *ist,
 +                         OutputStream *ost_table, int nb_ostreams,
 +                         const AVPacket *pkt)
  {
 -    AVFormatContext *is = input_files[infile].ctx;
 -    AVFormatContext *os = output_files[outfile];
 -    int i;
 +    int ret = 0, i;
 +    int got_output;
 +    int64_t pkt_dts = AV_NOPTS_VALUE;
 +    int64_t pkt_pts = AV_NOPTS_VALUE;
  
 -    for (i = 0; i < is->nb_chapters; i++) {
 -        AVChapter *in_ch = is->chapters[i], *out_ch;
 -        int64_t ts_off   = av_rescale_q(start_time - input_files[infile].ts_offset,
 -                                      AV_TIME_BASE_Q, in_ch->time_base);
 -        int64_t rt       = (recording_time == INT64_MAX) ? INT64_MAX :
 -                           av_rescale_q(recording_time, AV_TIME_BASE_Q, in_ch->time_base);
 +    AVPacket avpkt;
  
 +    if (ist->next_pts == AV_NOPTS_VALUE)
 +        ist->next_pts = ist->pts;
  
 -        if (in_ch->end < ts_off)
 -            continue;
 -        if (rt != INT64_MAX && in_ch->start > rt + ts_off)
 +    if (pkt == NULL) {
 +        /* EOF handling */
 +        av_init_packet(&avpkt);
 +        avpkt.data = NULL;
 +        avpkt.size = 0;
 +        goto handle_eof;
 +    } else {
 +        avpkt = *pkt;
 +    }
 +
 +    if(pkt->dts != AV_NOPTS_VALUE){
 +        if(ist->st->codec->codec_type != AVMEDIA_TYPE_VIDEO || !ist->decoding_needed)
 +            ist->next_pts = ist->pts = av_rescale_q(pkt->dts, ist->st->time_base, AV_TIME_BASE_Q);
 +        pkt_dts = av_rescale_q(pkt->dts, ist->st->time_base, AV_TIME_BASE_Q);
 +    }
 +    if(pkt->pts != AV_NOPTS_VALUE)
 +        pkt_pts = av_rescale_q(pkt->pts, ist->st->time_base, AV_TIME_BASE_Q);
 +
 +    //while we have more to decode or while the decoder did output something on EOF
 +    while (ist->decoding_needed && (avpkt.size > 0 || (!pkt && got_output))) {
 +    handle_eof:
 +
 +        ist->pts = ist->next_pts;
 +
 +        if (avpkt.size && avpkt.size != pkt->size) {
 +            av_log(NULL, ist->showed_multi_packet_warning ? AV_LOG_VERBOSE : AV_LOG_WARNING,
 +                   "Multiple frames in a packet from stream %d\n", pkt->stream_index);
 +            ist->showed_multi_packet_warning = 1;
 +        }
 +
 +        switch(ist->st->codec->codec_type) {
 +        case AVMEDIA_TYPE_AUDIO:
 +            ret = transcode_audio    (ist, &avpkt, &got_output);
 +            break;
 +        case AVMEDIA_TYPE_VIDEO:
 +            ret = transcode_video    (ist, &avpkt, &got_output, &pkt_pts, &pkt_dts);
              break;
 +        case AVMEDIA_TYPE_SUBTITLE:
 +            ret = transcode_subtitles(ist, &avpkt, &got_output);
 +            break;
 +        default:
 +            return -1;
 +        }
  
 -        out_ch = av_mallocz(sizeof(AVChapter));
 -        if (!out_ch)
 -            return AVERROR(ENOMEM);
 +        if (ret < 0)
 +            return ret;
 +        // touch data and size only if not EOF
 +        if (pkt) {
 +            if(ist->st->codec->codec_type != AVMEDIA_TYPE_AUDIO)
 +                ret = avpkt.size;
 +            avpkt.data += ret;
 +            avpkt.size -= ret;
 +        }
 +        if (!got_output) {
 +            continue;
 +        }
 +    }
  
 -        out_ch->id        = in_ch->id;
 -        out_ch->time_base = in_ch->time_base;
 -        out_ch->start     = FFMAX(0,  in_ch->start - ts_off);
 -        out_ch->end       = FFMIN(rt, in_ch->end   - ts_off);
 +    /* handle stream copy */
 +    if (!ist->decoding_needed) {
 +        rate_emu_sleep(ist);
 +        ist->pts = ist->next_pts;
 +        switch (ist->st->codec->codec_type) {
 +        case AVMEDIA_TYPE_AUDIO:
 +            ist->next_pts += ((int64_t)AV_TIME_BASE * ist->st->codec->frame_size) /
 +                             ist->st->codec->sample_rate;
 +            break;
 +        case AVMEDIA_TYPE_VIDEO:
 +            if (pkt->duration) {
 +                ist->next_pts += av_rescale_q(pkt->duration, ist->st->time_base, AV_TIME_BASE_Q);
 +            } else if(ist->st->codec->time_base.num != 0) {
 +                int ticks= ist->st->parser ? ist->st->parser->repeat_pict+1 : ist->st->codec->ticks_per_frame;
 +                ist->next_pts += ((int64_t)AV_TIME_BASE *
 +                                  ist->st->codec->time_base.num * ticks) /
 +                                  ist->st->codec->time_base.den;
 +            }
 +            break;
 +        }
 +    }
 +    for (i = 0; pkt && i < nb_ostreams; i++) {
 +        OutputStream *ost = &ost_table[i];
  
 -        if (metadata_chapters_autocopy)
 -            av_dict_copy(&out_ch->metadata, in_ch->metadata, 0);
 +        if (!check_output_constraints(ist, ost) || ost->encoding_needed)
 +            continue;
  
 -        os->nb_chapters++;
 -        os->chapters = av_realloc(os->chapters, sizeof(AVChapter)*os->nb_chapters);
 -        if (!os->chapters)
 -            return AVERROR(ENOMEM);
 -        os->chapters[os->nb_chapters - 1] = out_ch;
 +        do_streamcopy(ist, ost, pkt);
      }
 +
      return 0;
  }
  
diff --cc ffplay.c
+++ b/ffplay.c
@@@ -151,33 -153,18 +151,33 @@@ typedef struct VideoState 
      AVStream *audio_st;
      PacketQueue audioq;
      int audio_hw_buf_size;
-     /* samples output by the codec. we reserve more space for avsync
-        compensation, resampling and format conversion */
-     DECLARE_ALIGNED(16,uint8_t,audio_buf1)[AVCODEC_MAX_AUDIO_FRAME_SIZE * 4];
 +    DECLARE_ALIGNED(16,uint8_t,audio_buf2)[AVCODEC_MAX_AUDIO_FRAME_SIZE * 4];
+     uint8_t silence_buf[SDL_AUDIO_BUFFER_SIZE];
      uint8_t *audio_buf;
+     uint8_t *audio_buf1;
      unsigned int audio_buf_size; /* in bytes */
      int audio_buf_index; /* in bytes */
 +    int audio_write_buf_size;
      AVPacket audio_pkt_temp;
      AVPacket audio_pkt;
      enum AVSampleFormat audio_src_fmt;
 -    AVAudioConvert *reformat_ctx;
 +    enum AVSampleFormat audio_tgt_fmt;
 +    int audio_src_channels;
 +    int audio_tgt_channels;
 +    int64_t audio_src_channel_layout;
 +    int64_t audio_tgt_channel_layout;
 +    int audio_src_freq;
 +    int audio_tgt_freq;
 +    struct SwrContext *swr_ctx;
 +    double audio_current_pts;
 +    double audio_current_pts_drift;
 +    int frame_drops_early;
 +    int frame_drops_late;
+     AVFrame *frame;
  
 -    int show_audio; /* if true, display audio samples */
 +    enum ShowMode {
 +        SHOW_MODE_NONE = -1, SHOW_MODE_VIDEO = 0, SHOW_MODE_WAVES, SHOW_MODE_RDFT, SHOW_MODE_NB
 +    } show_mode;
      int16_t sample_array[SAMPLE_ARRAY_SIZE];
      int sample_array_index;
      int last_i_start;
@@@ -1998,8 -1964,8 +1998,8 @@@ static int synchronize_audio(VideoStat
                      max_size = ((nb_samples * (100 + SAMPLE_CORRECTION_PERCENT_MAX)) / 100) * n;
                      if (wanted_size < min_size)
                          wanted_size = min_size;
-                     else if (wanted_size > FFMIN3(max_size, sizeof(is->audio_buf1), sizeof(is->audio_buf2)))
-                         wanted_size = FFMIN3(max_size, sizeof(is->audio_buf1), sizeof(is->audio_buf2));
 -                    else if (wanted_size > max_size)
 -                        wanted_size = max_size;
++                    else if (wanted_size > FFMIN3(max_size, samples_size, sizeof(is->audio_buf2)))
++                        wanted_size = FFMIN3(max_size, samples_size, sizeof(is->audio_buf2));
  
                      /* add or remove samples to correction the synchro */
                      if (wanted_size < samples_size) {
@@@ -2042,8 -2008,7 +2042,8 @@@ static int audio_decode_frame(VideoStat
      AVPacket *pkt_temp = &is->audio_pkt_temp;
      AVPacket *pkt = &is->audio_pkt;
      AVCodecContext *dec= is->audio_st->codec;
 -    int n, len1, data_size, got_frame;
 +    int len1, len2, data_size, resampled_data_size;
-     int64_t dec_channel_layout;
++    int64_t dec_channel_layout, got_frame;
      double pts;
      int new_packet = 0;
      int flush_complete = 0;
                      flush_complete = 1;
                  continue;
              }
+             data_size = av_samples_get_buffer_size(NULL, dec->channels,
+                                                    is->frame->nb_samples,
+                                                    dec->sample_fmt, 1);
  
 -            if (dec->sample_fmt != is->audio_src_fmt) {
 -                if (is->reformat_ctx)
 -                    av_audio_convert_free(is->reformat_ctx);
 -                is->reformat_ctx= av_audio_convert_alloc(AV_SAMPLE_FMT_S16, 1,
 -                                                         dec->sample_fmt, 1, NULL, 0);
 -                if (!is->reformat_ctx) {
 -                    fprintf(stderr, "Cannot convert %s sample format to %s sample format\n",
 +            dec_channel_layout = (dec->channel_layout && dec->channels == av_get_channel_layout_nb_channels(dec->channel_layout)) ? dec->channel_layout : av_get_default_channel_layout(dec->channels);
 +
 +            if (dec->sample_fmt != is->audio_src_fmt || dec_channel_layout != is->audio_src_channel_layout || dec->sample_rate != is->audio_src_freq) {
 +                if (is->swr_ctx)
 +                    swr_free(&is->swr_ctx);
 +                is->swr_ctx = swr_alloc_set_opts(NULL,
 +                                                 is->audio_tgt_channel_layout, is->audio_tgt_fmt, is->audio_tgt_freq,
 +                                                 dec_channel_layout,           dec->sample_fmt,   dec->sample_rate,
 +                                                 0, NULL);
 +                if (!is->swr_ctx || swr_init(is->swr_ctx) < 0) {
 +                    fprintf(stderr, "Cannot create sample rate converter for conversion of %d Hz %s %d channels to %d Hz %s %d channels!\n",
 +                        dec->sample_rate,
                          av_get_sample_fmt_name(dec->sample_fmt),
 -                        av_get_sample_fmt_name(AV_SAMPLE_FMT_S16));
 -                        break;
 +                        dec->channels,
 +                        is->audio_tgt_freq,
 +                        av_get_sample_fmt_name(is->audio_tgt_fmt),
 +                        is->audio_tgt_channels);
 +                    break;
                  }
 -                is->audio_src_fmt= dec->sample_fmt;
 +                is->audio_src_channel_layout = dec_channel_layout;
 +                is->audio_src_channels = dec->channels;
 +                is->audio_src_freq = dec->sample_rate;
 +                is->audio_src_fmt = dec->sample_fmt;
              }
  
 -            if (is->reformat_ctx) {
 -                const void *ibuf[6]= { is->frame->data[0] };
 -                void *obuf[6];
 -                int istride[6]= {av_get_bytes_per_sample(dec->sample_fmt)};
 -                int ostride[6]= {2};
 -                int len= data_size/istride[0];
 -                obuf[0] = av_realloc(is->audio_buf1, FFALIGN(len * ostride[0], 32));
 -                if (!obuf[0]) {
 -                    return AVERROR(ENOMEM);
 -                }
 -                is->audio_buf1 = obuf[0];
 -                if (av_audio_convert(is->reformat_ctx, obuf, ostride, ibuf, istride, len)<0) {
 -                    printf("av_audio_convert() failed\n");
 +            resampled_data_size = data_size;
 +            if (is->swr_ctx) {
-                 const uint8_t *in[] = {is->audio_buf1};
++                const uint8_t *in[] = { is->frame->data[0] };
 +                uint8_t *out[] = {is->audio_buf2};
 +                len2 = swr_convert(is->swr_ctx, out, sizeof(is->audio_buf2) / is->audio_tgt_channels / av_get_bytes_per_sample(is->audio_tgt_fmt),
 +                                                in, data_size / dec->channels / av_get_bytes_per_sample(dec->sample_fmt));
 +                if (len2 < 0) {
 +                    fprintf(stderr, "audio_resample() failed\n");
                      break;
                  }
 -                is->audio_buf = is->audio_buf1;
 -                /* FIXME: existing code assume that data_size equals framesize*channels*2
 -                          remove this legacy cruft */
 -                data_size= len*2;
 -            }else{
 +                if (len2 == sizeof(is->audio_buf2) / is->audio_tgt_channels / av_get_bytes_per_sample(is->audio_tgt_fmt)) {
 +                    fprintf(stderr, "warning: audio buffer is probably too small\n");
 +                    swr_init(is->swr_ctx);
 +                }
 +                is->audio_buf = is->audio_buf2;
 +                resampled_data_size = len2 * is->audio_tgt_channels * av_get_bytes_per_sample(is->audio_tgt_fmt);
 +            } else {
-                 is->audio_buf= is->audio_buf1;
+                 is->audio_buf = is->frame->data[0];
              }
  
              /* if no pts, then compute it */
@@@ -2178,11 -2138,10 +2180,10 @@@ static void sdl_audio_callback(void *op
             audio_size = audio_decode_frame(is, &pts);
             if (audio_size < 0) {
                  /* if error, just output silence */
-                is->audio_buf = is->audio_buf1;
-                is->audio_buf_size = 256 * is->audio_tgt_channels * av_get_bytes_per_sample(is->audio_tgt_fmt);
-                memset(is->audio_buf, 0, is->audio_buf_size);
+                is->audio_buf      = is->silence_buf;
+                is->audio_buf_size = sizeof(is->silence_buf);
             } else {
 -               if (is->show_audio)
 +               if (is->show_mode != SHOW_MODE_VIDEO)
                     update_sample_display(is, (int16_t *)is->audio_buf, audio_size);
                 audio_size = synchronize_audio(is, (int16_t *)is->audio_buf, audio_size,
                                                pts);
@@@ -2353,9 -2279,13 +2354,12 @@@ static void stream_component_close(Vide
          SDL_CloseAudio();
  
          packet_queue_end(&is->audioq);
 +        if (is->swr_ctx)
 +            swr_free(&is->swr_ctx);
          av_free_packet(&is->audio_pkt);
 -        if (is->reformat_ctx)
 -            av_audio_convert_free(is->reformat_ctx);
 -        is->reformat_ctx = NULL;
+         av_freep(&is->audio_buf1);
+         is->audio_buf = NULL;
+         av_freep(&is->frame);
  
          if (is->rdft) {
              av_rdft_end(is->rdft);
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
@@@ -122,24 -122,21 +122,24 @@@ typedef struct MLPDecodeContext 
      AVCodecContext *avctx;
      AVFrame     frame;
  
-     //! Current access unit being read has a major sync.
+     /// Current access unit being read has a major sync.
      int         is_major_sync_unit;
  
-     //! Set if a valid major sync block has been read. Otherwise no decoding is possible.
+     /// Set if a valid major sync block has been read. Otherwise no decoding is possible.
      uint8_t     params_valid;
  
-     //! Number of substreams contained within this stream.
+     /// Number of substreams contained within this stream.
      uint8_t     num_substreams;
  
-     //! Index of the last substream to decode - further substreams are skipped.
+     /// Index of the last substream to decode - further substreams are skipped.
      uint8_t     max_decoded_substream;
  
-     //! Stream needs channel reordering to comply with FFmpeg's channel order
++    /// Stream needs channel reordering to comply with FFmpeg's channel order
 +    uint8_t     needs_reordering;
 +
-     //! number of PCM samples contained in each frame
+     /// number of PCM samples contained in each frame
      int         access_unit_size;
-     //! next power of two above the number of samples in each frame
+     /// next power of two above the number of samples in each frame
      int         access_unit_size_pow2;
  
      SubStream   substream[MAX_SUBSTREAMS];
Simple merge
Simple merge
Simple merge
@@@ -21,7 -21,7 +21,7 @@@
  #define AVCODEC_VERSION_H
  
  #define LIBAVCODEC_VERSION_MAJOR 53
- #define LIBAVCODEC_VERSION_MINOR 40
 -#define LIBAVCODEC_VERSION_MINOR 26
++#define LIBAVCODEC_VERSION_MINOR 41
  #define LIBAVCODEC_VERSION_MICRO  0
  
  #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
Simple merge
@@@ -27,8 -26,9 +26,8 @@@
      WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
      OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
      DEALINGS IN THE SOFTWARE.
**/
 */
  
 -
  #include <stdio.h>
  #include "oggdec.h"
  #include "avformat.h"
Simple merge
diff --cc libavutil/lzo.c
Simple merge
diff --cc libavutil/lzo.h
Simple merge
Simple merge
@@@ -347,13 -291,13 +347,18 @@@ do_audio_encoding ac3.rm "-vn -acodec a
  #$tiny_psnr $pcm_dst $pcm_ref 2 1024
  fi
  
 +if [ -n "$do_g723_1" ] ; then
 +do_audio_encoding g723_1.tco "-b:a 6.3k -ac 1 -ar 8000 -acodec g723_1"
 +do_audio_decoding
 +fi
 +
+ if [ -n "$do_g722" ] ; then
+ do_audio_encoding g722.wav "-b 64k -ac 1 -ar 16000 -acodec g722"
+ do_audio_decoding
+ fi
  if [ -n "$do_g726" ] ; then
 -do_audio_encoding g726.wav "-b 32k -ac 1 -ar 8000 -acodec g726"
 +do_audio_encoding g726.wav "-b:a 32k -ac 1 -ar 8000 -acodec g726"
  do_audio_decoding
  fi
  
@@@ -330,15 -315,8 +330,15 @@@ fate-h264-conformance-frext-pph10i4_pan
  fate-h264-conformance-frext-pph10i5_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH10I5_Panasonic_A.264 -pix_fmt yuv420p10le
  fate-h264-conformance-frext-pph10i6_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH10I6_Panasonic_A.264 -pix_fmt yuv420p10le
  fate-h264-conformance-frext-pph10i7_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH10I7_Panasonic_A.264 -pix_fmt yuv420p10le
- fate-h264-conformance-hcbp2_hhi_a: CMD = framecrc  -vsync 0 -strict 1 -i $(SAMPLES)/h264-conformance/HCBP2_HHI_A.264
- fate-h264-conformance-hcmp1_hhi_a: CMD = framecrc  -vsync 0 -strict 1 -i $(SAMPLES)/h264-conformance/HCMP1_HHI_A.264
 +fate-h264-conformance-frext-pph422i1_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH422I1_Panasonic_A.264 -pix_fmt yuv422p10le
 +fate-h264-conformance-frext-pph422i2_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH422I2_Panasonic_A.264 -pix_fmt yuv422p10le
 +fate-h264-conformance-frext-pph422i3_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH422I3_Panasonic_A.264 -pix_fmt yuv422p10le
 +fate-h264-conformance-frext-pph422i4_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH422I4_Panasonic_A.264 -pix_fmt yuv422p10le
 +fate-h264-conformance-frext-pph422i5_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH422I5_Panasonic_A.264 -pix_fmt yuv422p10le
 +fate-h264-conformance-frext-pph422i6_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH422I6_Panasonic_A.264 -pix_fmt yuv422p10le
 +fate-h264-conformance-frext-pph422i7_panasonic_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/FRext/PPH422I7_Panasonic_A.264 -pix_fmt yuv422p10le
+ fate-h264-conformance-hcbp2_hhi_a: CMD = framecrc  -vsync 0 -i $(SAMPLES)/h264-conformance/HCBP2_HHI_A.264
+ fate-h264-conformance-hcmp1_hhi_a: CMD = framecrc  -vsync 0 -i $(SAMPLES)/h264-conformance/HCMP1_HHI_A.264
  fate-h264-conformance-ls_sva_d: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/LS_SVA_D.264
  fate-h264-conformance-midr_mw_d: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/MIDR_MW_D.264
  fate-h264-conformance-mps_mw_a: CMD = framecrc -vsync 0 -i $(SAMPLES)/h264-conformance/MPS_MW_A.264
index 0000000,a1fc72a..1ca02e2
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,4 +1,4 @@@
 -b380355e0360b4e50ee78f33fd60a0f5 *./tests/data/acodec/g722.wav
++156f63e3391b95020ae882dbae6eccf3 *./tests/data/acodec/g722.wav
+ 47991 ./tests/data/acodec/g722.wav
 -82fdd5bb059336e0550de7ba5947c5bb *./tests/data/g722.acodec.out.wav
 -stddev: 8860.44 PSNR: 17.38 MAXDIFF:33814 bytes:   191732/  1058400
++8f65de513acc08b37a488d6a802b4f00 *./tests/data/g722.acodec.out.wav
++stddev: 8860.50 PSNR: 17.38 MAXDIFF:33814 bytes:   191732/  1058400
diff --cc tools/patcheck
Simple merge