avconv.c

/*
 * avconv main
 * Copyright (c) 2000-2011 The libav developers.
 *
 * This file is part of Libav.
 *
 * Libav is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Lesser General Public
 * License as published by the Free Software Foundation; either
 * version 2.1 of the License, or (at your option) any later version.
 *
 * Libav is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Lesser General Public License for more details.
 *
 * You should have received a copy of the GNU Lesser General Public
 * License along with Libav; if not, write to the Free Software
 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
 */

#include "config.h"
#include <ctype.h>
#include <string.h>
#include <math.h>
#include <stdlib.h>
#include <errno.h>
#include <signal.h>
#include <limits.h>
#include "libavformat/avformat.h"
#include "libavdevice/avdevice.h"
#include "libswscale/swscale.h"
#include "libavresample/avresample.h"
#include "libavutil/opt.h"
#include "libavutil/audioconvert.h"
#include "libavutil/parseutils.h"
#include "libavutil/samplefmt.h"
#include "libavutil/colorspace.h"
#include "libavutil/fifo.h"
#include "libavutil/intreadwrite.h"
#include "libavutil/dict.h"
#include "libavutil/mathematics.h"
#include "libavutil/pixdesc.h"
#include "libavutil/avstring.h"
#include "libavutil/libm.h"
#include "libavutil/imgutils.h"
#include "libavutil/time.h"
#include "libavformat/os_support.h"

# include "libavfilter/avfilter.h"
# include "libavfilter/avfiltergraph.h"
# include "libavfilter/buffersrc.h"
# include "libavfilter/buffersink.h"

#if HAVE_SYS_RESOURCE_H
#include <sys/types.h>
#include <sys/resource.h>
#elif HAVE_GETPROCESSTIMES
#include <windows.h>
#endif
#if HAVE_GETPROCESSMEMORYINFO
#include <windows.h>
#include <psapi.h>
#endif

#if HAVE_SYS_SELECT_H
#include <sys/select.h>
#endif

#if HAVE_PTHREADS
#include <pthread.h>
#endif

#include <time.h>

#include "avconv.h"
#include "cmdutils.h"

#include "libavutil/avassert.h"

const char program_name[] = "avconv";
const int program_birth_year = 2000;

static FILE *vstats_file;

static int64_t video_size = 0;
static int64_t audio_size = 0;
static int64_t extra_size = 0;
static int nb_frames_dup = 0;
static int nb_frames_drop = 0;



#if HAVE_PTHREADS
/* signal to input threads that they should exit; set by the main thread */
static int transcoding_finished;
#endif

#define DEFAULT_PASS_LOGFILENAME_PREFIX "av2pass"

InputStream **input_streams = NULL;
int        nb_input_streams = 0;
InputFile   **input_files   = NULL;
int        nb_input_files   = 0;

OutputStream **output_streams = NULL;
int         nb_output_streams = 0;
OutputFile   **output_files   = NULL;
int         nb_output_files   = 0;

FilterGraph **filtergraphs;
int        nb_filtergraphs;

static void term_exit(void)
{
    av_log(NULL, AV_LOG_QUIET, "");
}

static volatile int received_sigterm = 0;
static volatile int received_nb_signals = 0;

static void
sigterm_handler(int sig)
{
    received_sigterm = sig;
    received_nb_signals++;
    term_exit();
}

static void term_init(void)
{
    signal(SIGINT , sigterm_handler); /* Interrupt (ANSI).    */
    signal(SIGTERM, sigterm_handler); /* Termination (ANSI).  */
#ifdef SIGXCPU
    signal(SIGXCPU, sigterm_handler);
#endif
}

static int decode_interrupt_cb(void *ctx)
{
    return received_nb_signals > 1;
}

const AVIOInterruptCB int_cb = { decode_interrupt_cb, NULL };

static void exit_program(void)
{
    int i, j;

    for (i = 0; i < nb_filtergraphs; i++) {
        avfilter_graph_free(&filtergraphs[i]->graph);
        for (j = 0; j < filtergraphs[i]->nb_inputs; j++) {
            av_freep(&filtergraphs[i]->inputs[j]->name);
            av_freep(&filtergraphs[i]->inputs[j]);
        }
        av_freep(&filtergraphs[i]->inputs);
        for (j = 0; j < filtergraphs[i]->nb_outputs; j++) {
            av_freep(&filtergraphs[i]->outputs[j]->name);
            av_freep(&filtergraphs[i]->outputs[j]);
        }
        av_freep(&filtergraphs[i]->outputs);
        av_freep(&filtergraphs[i]);
    }
    av_freep(&filtergraphs);

    /* close files */
    for (i = 0; i < nb_output_files; i++) {
        AVFormatContext *s = output_files[i]->ctx;
        if (!(s->oformat->flags & AVFMT_NOFILE) && s->pb)
            avio_close(s->pb);
        avformat_free_context(s);
        av_dict_free(&output_files[i]->opts);
        av_freep(&output_files[i]);
    }
    for (i = 0; i < nb_output_streams; i++) {
        AVBitStreamFilterContext *bsfc = output_streams[i]->bitstream_filters;
        while (bsfc) {
            AVBitStreamFilterContext *next = bsfc->next;
            av_bitstream_filter_close(bsfc);
            bsfc = next;
        }
        output_streams[i]->bitstream_filters = NULL;
        avcodec_free_frame(&output_streams[i]->filtered_frame);

        av_freep(&output_streams[i]->forced_keyframes);
        av_freep(&output_streams[i]->avfilter);
        av_freep(&output_streams[i]->logfile_prefix);
        av_freep(&output_streams[i]);
    }
    for (i = 0; i < nb_input_files; i++) {
        avformat_close_input(&input_files[i]->ctx);
        av_freep(&input_files[i]);
    }
    for (i = 0; i < nb_input_streams; i++) {
        avcodec_free_frame(&input_streams[i]->decoded_frame);
        av_dict_free(&input_streams[i]->opts);
        free_buffer_pool(&input_streams[i]->buffer_pool);
        av_freep(&input_streams[i]->filters);
        av_freep(&input_streams[i]);
    }

    if (vstats_file)
        fclose(vstats_file);
    av_free(vstats_filename);

    av_freep(&input_streams);
    av_freep(&input_files);
    av_freep(&output_streams);
    av_freep(&output_files);

    uninit_opts();

    avfilter_uninit();
    avformat_network_deinit();

    if (received_sigterm) {
        av_log(NULL, AV_LOG_INFO, "Received signal %d: terminating.\n",
               (int) received_sigterm);
        exit (255);
    }
}

void assert_avoptions(AVDictionary *m)
{
    AVDictionaryEntry *t;
    if ((t = av_dict_get(m, "", NULL, AV_DICT_IGNORE_SUFFIX))) {
        av_log(NULL, AV_LOG_FATAL, "Option %s not found.\n", t->key);
        exit(1);
    }
}

static void assert_codec_experimental(AVCodecContext *c, int encoder)
{
    const char *codec_string = encoder ? "encoder" : "decoder";
    AVCodec *codec;
    if (c->codec->capabilities & CODEC_CAP_EXPERIMENTAL &&
        c->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
        av_log(NULL, AV_LOG_FATAL, "%s '%s' is experimental and might produce bad "
                "results.\nAdd '-strict experimental' if you want to use it.\n",
                codec_string, c->codec->name);
        codec = encoder ? avcodec_find_encoder(c->codec->id) : avcodec_find_decoder(c->codec->id);
        if (!(codec->capabilities & CODEC_CAP_EXPERIMENTAL))
            av_log(NULL, AV_LOG_FATAL, "Or use the non experimental %s '%s'.\n",
                   codec_string, codec->name);
        exit(1);
    }
}

/**
 * Update the requested input sample format based on the output sample format.
 * This is currently only used to request float output from decoders which
 * support multiple sample formats, one of which is AV_SAMPLE_FMT_FLT.
 * Ideally this will be removed in the future when decoders do not do format
 * conversion and only output in their native format.
 */
static void update_sample_fmt(AVCodecContext *dec, AVCodec *dec_codec,
                              AVCodecContext *enc)
{
    /* if sample formats match or a decoder sample format has already been
       requested, just return */
    if (enc->sample_fmt == dec->sample_fmt ||
        dec->request_sample_fmt > AV_SAMPLE_FMT_NONE)
        return;

    /* if decoder supports more than one output format */
    if (dec_codec && dec_codec->sample_fmts &&
        dec_codec->sample_fmts[0] != AV_SAMPLE_FMT_NONE &&
        dec_codec->sample_fmts[1] != AV_SAMPLE_FMT_NONE) {
        const enum AVSampleFormat *p;
        int min_dec = INT_MAX, min_inc = INT_MAX;
        enum AVSampleFormat dec_fmt = AV_SAMPLE_FMT_NONE;
        enum AVSampleFormat inc_fmt = AV_SAMPLE_FMT_NONE;

        /* find a matching sample format in the encoder */
        for (p = dec_codec->sample_fmts; *p != AV_SAMPLE_FMT_NONE; p++) {
            if (*p == enc->sample_fmt) {
                dec->request_sample_fmt = *p;
                return;
            } else {
                enum AVSampleFormat dfmt = av_get_packed_sample_fmt(*p);
                enum AVSampleFormat efmt = av_get_packed_sample_fmt(enc->sample_fmt);
                int fmt_diff = 32 * abs(dfmt - efmt);
                if (av_sample_fmt_is_planar(*p) !=
                    av_sample_fmt_is_planar(enc->sample_fmt))
                    fmt_diff++;
                if (dfmt == efmt) {
                    min_inc = fmt_diff;
                    inc_fmt = *p;
                } else if (dfmt > efmt) {
                    if (fmt_diff < min_inc) {
                        min_inc = fmt_diff;
                        inc_fmt = *p;
                    }
                } else {
                    if (fmt_diff < min_dec) {
                        min_dec = fmt_diff;
                        dec_fmt = *p;
                    }
                }
            }
        }

        /* if none match, provide the one that matches quality closest */
        dec->request_sample_fmt = min_inc != INT_MAX ? inc_fmt : dec_fmt;
    }
}

static void write_frame(AVFormatContext *s, AVPacket *pkt, OutputStream *ost)
{
    AVBitStreamFilterContext *bsfc = ost->bitstream_filters;
    AVCodecContext          *avctx = ost->st->codec;
    int ret;

    /*
     * Audio encoders may split the packets --  #frames in != #packets out.
     * But there is no reordering, so we can limit the number of output packets
     * by simply dropping them here.
     * Counting encoded video frames needs to be done separately because of
     * reordering, see do_video_out()
     */
    if (!(avctx->codec_type == AVMEDIA_TYPE_VIDEO && avctx->codec)) {
        if (ost->frame_number >= ost->max_frames) {
            av_free_packet(pkt);
            return;
        }
        ost->frame_number++;
    }

    while (bsfc) {
        AVPacket new_pkt = *pkt;
        int a = av_bitstream_filter_filter(bsfc, avctx, NULL,
                                           &new_pkt.data, &new_pkt.size,
                                           pkt->data, pkt->size,
                                           pkt->flags & AV_PKT_FLAG_KEY);
        if (a > 0) {
            av_free_packet(pkt);
            new_pkt.destruct = av_destruct_packet;
        } else if (a < 0) {
            av_log(NULL, AV_LOG_ERROR, "%s failed for stream %d, codec %s",
                   bsfc->filter->name, pkt->stream_index,
                   avctx->codec ? avctx->codec->name : "copy");
            print_error("", a);
            if (exit_on_error)
                exit(1);
        }
        *pkt = new_pkt;

        bsfc = bsfc->next;
    }

    pkt->stream_index = ost->index;
    ret = av_interleaved_write_frame(s, pkt);
    if (ret < 0) {
        print_error("av_interleaved_write_frame()", ret);
        exit(1);
    }
}

static int check_recording_time(OutputStream *ost)
{
    OutputFile *of = output_files[ost->file_index];

    if (of->recording_time != INT64_MAX &&
        av_compare_ts(ost->sync_opts - ost->first_pts, ost->st->codec->time_base, of->recording_time,
                      AV_TIME_BASE_Q) >= 0) {
        ost->finished = 1;
        return 0;
    }
    return 1;
}

static void do_audio_out(AVFormatContext *s, OutputStream *ost,
                         AVFrame *frame)
{
    AVCodecContext *enc = ost->st->codec;
    AVPacket pkt;
    int got_packet = 0;

    av_init_packet(&pkt);
    pkt.data = NULL;
    pkt.size = 0;

    if (!check_recording_time(ost))
        return;

    if (frame->pts == AV_NOPTS_VALUE || audio_sync_method < 0)
        frame->pts = ost->sync_opts;
    ost->sync_opts = frame->pts + frame->nb_samples;

    if (avcodec_encode_audio2(enc, &pkt, frame, &got_packet) < 0) {
        av_log(NULL, AV_LOG_FATAL, "Audio encoding failed\n");
        exit(1);
    }

    if (got_packet) {
        if (pkt.pts != AV_NOPTS_VALUE)
            pkt.pts      = av_rescale_q(pkt.pts,      enc->time_base, ost->st->time_base);
        if (pkt.dts != AV_NOPTS_VALUE)
            pkt.dts      = av_rescale_q(pkt.dts,      enc->time_base, ost->st->time_base);
        if (pkt.duration > 0)
            pkt.duration = av_rescale_q(pkt.duration, enc->time_base, ost->st->time_base);

        write_frame(s, &pkt, ost);

        audio_size += pkt.size;
    }
}

static void pre_process_video_frame(InputStream *ist, AVPicture *picture, void **bufp)
{
    AVCodecContext *dec;
    AVPicture *picture2;
    AVPicture picture_tmp;
    uint8_t *buf = 0;

    dec = ist->st->codec;

    /* deinterlace : must be done before any resize */
    if (do_deinterlace) {
        int size;

        /* create temporary picture */
        size = avpicture_get_size(dec->pix_fmt, dec->width, dec->height);
        buf  = av_malloc(size);
        if (!buf)
            return;

        picture2 = &picture_tmp;
        avpicture_fill(picture2, buf, dec->pix_fmt, dec->width, dec->height);

        if (avpicture_deinterlace(picture2, picture,
                                 dec->pix_fmt, dec->width, dec->height) < 0) {
            /* if error, do not deinterlace */
            av_log(NULL, AV_LOG_WARNING, "Deinterlacing failed\n");
            av_free(buf);
            buf = NULL;
            picture2 = picture;
        }
    } else {
        picture2 = picture;
    }

    if (picture != picture2)
        *picture = *picture2;
    *bufp = buf;
}

static void do_subtitle_out(AVFormatContext *s,
                            OutputStream *ost,
                            InputStream *ist,
                            AVSubtitle *sub,
                            int64_t pts)
{
    static uint8_t *subtitle_out = NULL;
    int subtitle_out_max_size = 1024 * 1024;
    int subtitle_out_size, nb, i;
    AVCodecContext *enc;
    AVPacket pkt;

    if (pts == AV_NOPTS_VALUE) {
        av_log(NULL, AV_LOG_ERROR, "Subtitle packets must have a pts\n");
        if (exit_on_error)
            exit(1);
        return;
    }

    enc = ost->st->codec;

    if (!subtitle_out) {
        subtitle_out = av_malloc(subtitle_out_max_size);
    }

    /* Note: DVB subtitle need one packet to draw them and one other
       packet to clear them */
    /* XXX: signal it in the codec context ? */
    if (enc->codec_id == AV_CODEC_ID_DVB_SUBTITLE)
        nb = 2;
    else
        nb = 1;

    for (i = 0; i < nb; i++) {
        ost->sync_opts = av_rescale_q(pts, ist->st->time_base, enc->time_base);
        if (!check_recording_time(ost))
            return;

        sub->pts = av_rescale_q(pts, ist->st->time_base, AV_TIME_BASE_Q);
        // start_display_time is required to be 0
        sub->pts               += av_rescale_q(sub->start_display_time, (AVRational){ 1, 1000 }, AV_TIME_BASE_Q);
        sub->end_display_time  -= sub->start_display_time;
        sub->start_display_time = 0;
        subtitle_out_size = avcodec_encode_subtitle(enc, subtitle_out,
                                                    subtitle_out_max_size, sub);
        if (subtitle_out_size < 0) {
            av_log(NULL, AV_LOG_FATAL, "Subtitle encoding failed\n");
            exit(1);
        }

        av_init_packet(&pkt);
        pkt.data = subtitle_out;
        pkt.size = subtitle_out_size;
        pkt.pts  = av_rescale_q(sub->pts, AV_TIME_BASE_Q, ost->st->time_base);
        if (enc->codec_id == AV_CODEC_ID_DVB_SUBTITLE) {
            /* XXX: the pts correction is handled here. Maybe handling
               it in the codec would be better */
            if (i == 0)
                pkt.pts += 90 * sub->start_display_time;
            else
                pkt.pts += 90 * sub->end_display_time;
        }
        write_frame(s, &pkt, ost);
    }
}

static void do_video_out(AVFormatContext *s,
                         OutputStream *ost,
                         AVFrame *in_picture,
                         int *frame_size)
{
    int ret, format_video_sync;
    AVPacket pkt;
    AVCodecContext *enc = ost->st->codec;

    *frame_size = 0;

    format_video_sync = video_sync_method;
    if (format_video_sync == VSYNC_AUTO)
        format_video_sync = (s->oformat->flags & AVFMT_NOTIMESTAMPS) ? VSYNC_PASSTHROUGH :
                            (s->oformat->flags & AVFMT_VARIABLE_FPS) ? VSYNC_VFR : VSYNC_CFR;
    if (format_video_sync != VSYNC_PASSTHROUGH &&
        ost->frame_number &&
        in_picture->pts != AV_NOPTS_VALUE &&
        in_picture->pts < ost->sync_opts) {
        nb_frames_drop++;
        av_log(NULL, AV_LOG_VERBOSE, "*** drop!\n");
        return;
    }

    if (in_picture->pts == AV_NOPTS_VALUE)
        in_picture->pts = ost->sync_opts;
    ost->sync_opts = in_picture->pts;


    if (!ost->frame_number)
        ost->first_pts = in_picture->pts;

    av_init_packet(&pkt);
    pkt.data = NULL;
    pkt.size = 0;

    if (!check_recording_time(ost) ||
        ost->frame_number >= ost->max_frames)
        return;

    if (s->oformat->flags & AVFMT_RAWPICTURE &&
        enc->codec->id == AV_CODEC_ID_RAWVIDEO) {
        /* raw pictures are written as AVPicture structure to
           avoid any copies. We support temporarily the older
           method. */
        enc->coded_frame->interlaced_frame = in_picture->interlaced_frame;
        enc->coded_frame->top_field_first  = in_picture->top_field_first;
        pkt.data   = (uint8_t *)in_picture;
        pkt.size   =  sizeof(AVPicture);
        pkt.pts    = av_rescale_q(in_picture->pts, enc->time_base, ost->st->time_base);
        pkt.flags |= AV_PKT_FLAG_KEY;

        write_frame(s, &pkt, ost);
    } else {
        int got_packet;
        AVFrame big_picture;

        big_picture = *in_picture;
        /* better than nothing: use input picture interlaced
           settings */
        big_picture.interlaced_frame = in_picture->interlaced_frame;
        if (ost->st->codec->flags & (CODEC_FLAG_INTERLACED_DCT|CODEC_FLAG_INTERLACED_ME)) {
            if (ost->top_field_first == -1)
                big_picture.top_field_first = in_picture->top_field_first;
            else
                big_picture.top_field_first = !!ost->top_field_first;
        }

        big_picture.quality = ost->st->codec->global_quality;
        if (!enc->me_threshold)
            big_picture.pict_type = 0;
        if (ost->forced_kf_index < ost->forced_kf_count &&
            big_picture.pts >= ost->forced_kf_pts[ost->forced_kf_index]) {
            big_picture.pict_type = AV_PICTURE_TYPE_I;
            ost->forced_kf_index++;
        }
        ret = avcodec_encode_video2(enc, &pkt, &big_picture, &got_packet);
        if (ret < 0) {
            av_log(NULL, AV_LOG_FATAL, "Video encoding failed\n");
            exit(1);
        }

        if (got_packet) {
            if (pkt.pts != AV_NOPTS_VALUE)
                pkt.pts = av_rescale_q(pkt.pts, enc->time_base, ost->st->time_base);
            if (pkt.dts != AV_NOPTS_VALUE)
                pkt.dts = av_rescale_q(pkt.dts, enc->time_base, ost->st->time_base);

            write_frame(s, &pkt, ost);
            *frame_size = pkt.size;
            video_size += pkt.size;

            /* if two pass, output log */
            if (ost->logfile && enc->stats_out) {
                fprintf(ost->logfile, "%s", enc->stats_out);
            }
        }
    }
    ost->sync_opts++;
    /*
     * For video, number of frames in == number of packets out.
     * But there may be reordering, so we can't throw away frames on encoder
     * flush, we need to limit them here, before they go into encoder.
     */
    ost->frame_number++;
}

static double psnr(double d)
{
    return -10.0 * log(d) / log(10.0);
}

static void do_video_stats(AVFormatContext *os, OutputStream *ost,
                           int frame_size)
{
    AVCodecContext *enc;
    int frame_number;
    double ti1, bitrate, avg_bitrate;

    /* this is executed just the first time do_video_stats is called */
    if (!vstats_file) {
        vstats_file = fopen(vstats_filename, "w");
        if (!vstats_file) {
            perror("fopen");
            exit(1);
        }
    }

    enc = ost->st->codec;
    if (enc->codec_type == AVMEDIA_TYPE_VIDEO) {
        frame_number = ost->frame_number;
        fprintf(vstats_file, "frame= %5d q= %2.1f ", frame_number, enc->coded_frame->quality / (float)FF_QP2LAMBDA);
        if (enc->flags&CODEC_FLAG_PSNR)
            fprintf(vstats_file, "PSNR= %6.2f ", psnr(enc->coded_frame->error[0] / (enc->width * enc->height * 255.0 * 255.0)));

        fprintf(vstats_file,"f_size= %6d ", frame_size);
        /* compute pts value */
        ti1 = ost->sync_opts * av_q2d(enc->time_base);
        if (ti1 < 0.01)
            ti1 = 0.01;

        bitrate     = (frame_size * 8) / av_q2d(enc->time_base) / 1000.0;
        avg_bitrate = (double)(video_size * 8) / ti1 / 1000.0;
        fprintf(vstats_file, "s_size= %8.0fkB time= %0.3f br= %7.1fkbits/s avg_br= %7.1fkbits/s ",
               (double)video_size / 1024, ti1, bitrate, avg_bitrate);
        fprintf(vstats_file, "type= %c\n", av_get_picture_type_char(enc->coded_frame->pict_type));
    }
}

/**
 * Read one frame for lavfi output for ost and encode it.
 */
static int poll_filter(OutputStream *ost)
{
    OutputFile    *of = output_files[ost->file_index];
    AVFilterBufferRef *picref;
    AVFrame *filtered_frame = NULL;
    int frame_size, ret;

    if (!ost->filtered_frame && !(ost->filtered_frame = avcodec_alloc_frame())) {
        return AVERROR(ENOMEM);
    } else
        avcodec_get_frame_defaults(ost->filtered_frame);
    filtered_frame = ost->filtered_frame;

    if (ost->enc->type == AVMEDIA_TYPE_AUDIO &&
        !(ost->enc->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE))
        ret = av_buffersink_read_samples(ost->filter->filter, &picref,
                                         ost->st->codec->frame_size);
    else
        ret = av_buffersink_read(ost->filter->filter, &picref);

    if (ret < 0)
        return ret;

    avfilter_copy_buf_props(filtered_frame, picref);
    if (picref->pts != AV_NOPTS_VALUE) {
        filtered_frame->pts = av_rescale_q(picref->pts,
                                           ost->filter->filter->inputs[0]->time_base,
                                           ost->st->codec->time_base) -
                              av_rescale_q(of->start_time,
                                           AV_TIME_BASE_Q,
                                           ost->st->codec->time_base);

        if (of->start_time && filtered_frame->pts < 0) {
            avfilter_unref_buffer(picref);
            return 0;
        }
    }

    switch (ost->filter->filter->inputs[0]->type) {
    case AVMEDIA_TYPE_VIDEO:
        if (!ost->frame_aspect_ratio)
            ost->st->codec->sample_aspect_ratio = picref->video->pixel_aspect;

        do_video_out(of->ctx, ost, filtered_frame, &frame_size);
        if (vstats_filename && frame_size)
            do_video_stats(of->ctx, ost, frame_size);
        break;
    case AVMEDIA_TYPE_AUDIO:
        do_audio_out(of->ctx, ost, filtered_frame);
        break;
    default:
        // TODO support subtitle filters
        av_assert0(0);
    }

    avfilter_unref_buffer(picref);

    return 0;
}

/**
 * Read as many frames from possible from lavfi and encode them.
 *
 * Always read from the active stream with the lowest timestamp. If no frames
 * are available for it then return EAGAIN and wait for more input. This way we
 * can use lavfi sources that generate unlimited amount of frames without memory
 * usage exploding.
 */
static int poll_filters(void)
{
    int i, j, ret = 0;

    while (ret >= 0 && !received_sigterm) {
        OutputStream *ost = NULL;
        int64_t min_pts = INT64_MAX;

        /* choose output stream with the lowest timestamp */
        for (i = 0; i < nb_output_streams; i++) {
            int64_t pts = output_streams[i]->sync_opts;

            if (!output_streams[i]->filter || output_streams[i]->finished)
                continue;

            pts = av_rescale_q(pts, output_streams[i]->st->codec->time_base,
                               AV_TIME_BASE_Q);
            if (pts < min_pts) {
                min_pts = pts;
                ost = output_streams[i];
            }
        }

        if (!ost)
            break;

        ret = poll_filter(ost);

        if (ret == AVERROR_EOF) {
            OutputFile *of = output_files[ost->file_index];

            ost->finished = 1;

            if (of->shortest) {
                for (j = 0; j < of->ctx->nb_streams; j++)
                    output_streams[of->ost_index + j]->finished = 1;
            }

            ret = 0;
        } else if (ret == AVERROR(EAGAIN))
            return 0;
    }

    return ret;
}

static void print_report(int is_last_report, int64_t timer_start)
{
    char buf[1024];
    OutputStream *ost;
    AVFormatContext *oc;
    int64_t total_size;
    AVCodecContext *enc;
    int frame_number, vid, i;
    double bitrate, ti1, pts;
    static int64_t last_time = -1;
    static int qp_histogram[52];

    if (!print_stats && !is_last_report)
        return;

    if (!is_last_report) {
        int64_t cur_time;
        /* display the report every 0.5 seconds */
        cur_time = av_gettime();
        if (last_time == -1) {
            last_time = cur_time;
            return;
        }
        if ((cur_time - last_time) < 500000)
            return;
        last_time = cur_time;
    }


    oc = output_files[0]->ctx;

    total_size = avio_size(oc->pb);
    if (total_size < 0) // FIXME improve avio_size() so it works with non seekable output too
        total_size = avio_tell(oc->pb);

    buf[0] = '\0';
    ti1 = 1e10;
    vid = 0;
    for (i = 0; i < nb_output_streams; i++) {
        float q = -1;
        ost = output_streams[i];
        enc = ost->st->codec;
        if (!ost->stream_copy && enc->coded_frame)
            q = enc->coded_frame->quality / (float)FF_QP2LAMBDA;
        if (vid && enc->codec_type == AVMEDIA_TYPE_VIDEO) {
            snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "q=%2.1f ", q);
        }
        if (!vid && enc->codec_type == AVMEDIA_TYPE_VIDEO) {
            float t = (av_gettime() - timer_start) / 1000000.0;

            frame_number = ost->frame_number;
            snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "frame=%5d fps=%3d q=%3.1f ",
                     frame_number, (t > 1) ? (int)(frame_number / t + 0.5) : 0, q);
            if (is_last_report)
                snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "L");
            if (qp_hist) {
                int j;
                int qp = lrintf(q);
                if (qp >= 0 && qp < FF_ARRAY_ELEMS(qp_histogram))
                    qp_histogram[qp]++;
                for (j = 0; j < 32; j++)
                    snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "%X", (int)lrintf(log2(qp_histogram[j] + 1)));
            }
            if (enc->flags&CODEC_FLAG_PSNR) {
                int j;
                double error, error_sum = 0;
                double scale, scale_sum = 0;
                char type[3] = { 'Y','U','V' };
                snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "PSNR=");
                for (j = 0; j < 3; j++) {
                    if (is_last_report) {
                        error = enc->error[j];
                        scale = enc->width * enc->height * 255.0 * 255.0 * frame_number;
                    } else {
                        error = enc->coded_frame->error[j];
                        scale = enc->width * enc->height * 255.0 * 255.0;
                    }
                    if (j)
                        scale /= 4;
                    error_sum += error;
                    scale_sum += scale;
                    snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "%c:%2.2f ", type[j], psnr(error / scale));
                }
                snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "*:%2.2f ", psnr(error_sum / scale_sum));
            }
            vid = 1;
        }
        /* compute min output value */
        pts = (double)ost->st->pts.val * av_q2d(ost->st->time_base);
        if ((pts < ti1) && (pts > 0))
            ti1 = pts;
    }
    if (ti1 < 0.01)
        ti1 = 0.01;

    bitrate = (double)(total_size * 8) / ti1 / 1000.0;

    snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf),
            "size=%8.0fkB time=%0.2f bitrate=%6.1fkbits/s",
            (double)total_size / 1024, ti1, bitrate);

    if (nb_frames_dup || nb_frames_drop)
        snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), " dup=%d drop=%d",
                nb_frames_dup, nb_frames_drop);

    av_log(NULL, AV_LOG_INFO, "%s    \r", buf);

    fflush(stderr);

    if (is_last_report) {
        int64_t raw= audio_size + video_size + extra_size;
        av_log(NULL, AV_LOG_INFO, "\n");
        av_log(NULL, AV_LOG_INFO, "video:%1.0fkB audio:%1.0fkB global headers:%1.0fkB muxing overhead %f%%\n",
               video_size / 1024.0,
               audio_size / 1024.0,
               extra_size / 1024.0,
               100.0 * (total_size - raw) / raw
        );
    }
}

static void flush_encoders(void)
{
    int i, ret;

    for (i = 0; i < nb_output_streams; i++) {
        OutputStream   *ost = output_streams[i];
        AVCodecContext *enc = ost->st->codec;
        AVFormatContext *os = output_files[ost->file_index]->ctx;
        int stop_encoding = 0;

        if (!ost->encoding_needed)
            continue;

        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 == AV_CODEC_ID_RAWVIDEO)
            continue;

        for (;;) {
            int (*encode)(AVCodecContext*, AVPacket*, const AVFrame*, int*) = NULL;
            const char *desc;
            int64_t *size;

            switch (ost->st->codec->codec_type) {
            case AVMEDIA_TYPE_AUDIO:
                encode = avcodec_encode_audio2;
                desc   = "Audio";
                size   = &audio_size;
                break;
            case AVMEDIA_TYPE_VIDEO:
                encode = avcodec_encode_video2;
                desc   = "Video";
                size   = &video_size;
                break;
            default:
                stop_encoding = 1;
            }

            if (encode) {
                AVPacket pkt;
                int got_packet;
                av_init_packet(&pkt);
                pkt.data = NULL;
                pkt.size = 0;

                ret = encode(enc, &pkt, NULL, &got_packet);
                if (ret < 0) {
                    av_log(NULL, AV_LOG_FATAL, "%s encoding failed\n", desc);
                    exit(1);
                }
                *size += ret;
                if (ost->logfile && enc->stats_out) {
                    fprintf(ost->logfile, "%s", enc->stats_out);
                }
                if (!got_packet) {
                    stop_encoding = 1;
                    break;
                }
                if (pkt.pts != AV_NOPTS_VALUE)
                    pkt.pts = av_rescale_q(pkt.pts, enc->time_base, ost->st->time_base);
                if (pkt.dts != AV_NOPTS_VALUE)
                    pkt.dts = av_rescale_q(pkt.dts, enc->time_base, ost->st->time_base);
                write_frame(os, &pkt, ost);
            }

            if (stop_encoding)
                break;
        }
    }
}

/*
 * 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  = input_files[ist->file_index]->ist_index + ist->st->index;

    if (ost->source_index != ist_index)
        return 0;

    if (of->start_time && ist->last_dts < of->start_time)
        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);
    AVPacket opkt;

    av_init_packet(&opkt);

    if ((!ost->frame_number && !(pkt->flags & AV_PKT_FLAG_KEY)) &&
        !ost->copy_initial_nonkeyframes)
        return;

    if (of->recording_time != INT64_MAX &&
        ist->last_dts >= of->recording_time + of->start_time) {
        ost->finished = 1;
        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++;
    }

    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->last_dts, 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 != AV_CODEC_ID_H264
       && ost->st->codec->codec_id != AV_CODEC_ID_MPEG1VIDEO
       && ost->st->codec->codec_id != AV_CODEC_ID_MPEG2VIDEO
       && ost->st->codec->codec_id != AV_CODEC_ID_VC1