avformat: add wve demuxer
[ffmpeg.git] / ffmpeg.c
index 8107b3c..36a68fb 100644 (file)
--- a/ffmpeg.c
+++ b/ffmpeg.c
 #include <limits.h>
 #include <stdint.h>
 
-#if HAVE_ISATTY
 #if HAVE_IO_H
 #include <io.h>
 #endif
 #if HAVE_UNISTD_H
 #include <unistd.h>
 #endif
-#endif
 
 #include "libavformat/avformat.h"
 #include "libavdevice/avdevice.h"
@@ -64,7 +62,6 @@
 #include "libavcodec/mathops.h"
 #include "libavformat/os_support.h"
 
-# include "libavfilter/avcodec.h"
 # include "libavfilter/avfilter.h"
 # include "libavfilter/buffersrc.h"
 # include "libavfilter/buffersink.h"
@@ -172,8 +169,8 @@ static int sub2video_get_blank_frame(InputStream *ist)
     AVFrame *frame = ist->sub2video.frame;
 
     av_frame_unref(frame);
-    ist->sub2video.frame->width  = ist->sub2video.w;
-    ist->sub2video.frame->height = ist->sub2video.h;
+    ist->sub2video.frame->width  = ist->dec_ctx->width  ? ist->dec_ctx->width  : ist->sub2video.w;
+    ist->sub2video.frame->height = ist->dec_ctx->height ? ist->dec_ctx->height : ist->sub2video.h;
     ist->sub2video.frame->format = AV_PIX_FMT_RGB32;
     if ((ret = av_frame_get_buffer(frame, 32)) < 0)
         return ret;
@@ -193,7 +190,9 @@ static void sub2video_copy_rect(uint8_t *dst, int dst_linesize, int w, int h,
         return;
     }
     if (r->x < 0 || r->x + r->w > w || r->y < 0 || r->y + r->h > h) {
-        av_log(NULL, AV_LOG_WARNING, "sub2video: rectangle overflowing\n");
+        av_log(NULL, AV_LOG_WARNING, "sub2video: rectangle (%d %d %d %d) overflowing %d %d\n",
+            r->x, r->y, r->w, r->h, w, h
+        );
         return;
     }
 
@@ -225,7 +224,6 @@ static void sub2video_push_ref(InputStream *ist, int64_t pts)
 
 static void sub2video_update(InputStream *ist, AVSubtitle *sub)
 {
-    int w = ist->sub2video.w, h = ist->sub2video.h;
     AVFrame *frame = ist->sub2video.frame;
     int8_t *dst;
     int     dst_linesize;
@@ -253,7 +251,7 @@ static void sub2video_update(InputStream *ist, AVSubtitle *sub)
     dst          = frame->data    [0];
     dst_linesize = frame->linesize[0];
     for (i = 0; i < num_rects; i++)
-        sub2video_copy_rect(dst, dst_linesize, w, h, sub->rects[i]);
+        sub2video_copy_rect(dst, dst_linesize, frame->width, frame->height, sub->rects[i]);
     sub2video_push_ref(ist, pts);
     ist->sub2video.end_pts = end_pts;
 }
@@ -294,7 +292,7 @@ static void sub2video_flush(InputStream *ist)
     if (ist->sub2video.end_pts < INT64_MAX)
         sub2video_update(ist, NULL);
     for (i = 0; i < ist->nb_filters; i++)
-        av_buffersrc_add_ref(ist->filters[i]->filter, NULL, 0);
+        av_buffersrc_add_frame(ist->filters[i]->filter, NULL);
 }
 
 /* end of sub2video hack */
@@ -370,11 +368,7 @@ void term_init(void)
 #if HAVE_TERMIOS_H
     if(!run_as_daemon){
         struct termios tty;
-        int istty = 1;
-#if HAVE_ISATTY
-        istty = isatty(0) && isatty(2);
-#endif
-        if (istty && tcgetattr (0, &tty) == 0) {
+        if (tcgetattr (0, &tty) == 0) {
             oldtty = tty;
             restore_tty = 1;
 
@@ -434,10 +428,6 @@ static int read_key(void)
         is_pipe = !GetConsoleMode(input_handle, &dw);
     }
 
-    if (stdin->_cnt > 0) {
-        read(0, &ch, 1);
-        return ch;
-    }
     if (is_pipe) {
         /* When running under a GUI, you will end here. */
         if (!PeekNamedPipe(input_handle, NULL, 0, NULL, &nchars, NULL)) {
@@ -673,6 +663,7 @@ static void write_frame(AVFormatContext *s, AVPacket *pkt, OutputStream *ost)
         uint8_t *sd = av_packet_get_side_data(pkt, AV_PKT_DATA_QUALITY_STATS,
                                               NULL);
         ost->quality = sd ? AV_RL32(sd) : -1;
+        ost->pict_type = sd ? sd[4] : AV_PICTURE_TYPE_NONE;
 
         for (i = 0; i<FF_ARRAY_ELEMS(ost->error); i++) {
             if (sd && i < sd[5])
@@ -695,7 +686,7 @@ static void write_frame(AVFormatContext *s, AVPacket *pkt, OutputStream *ost)
                                            &new_pkt.data, &new_pkt.size,
                                            pkt->data, pkt->size,
                                            pkt->flags & AV_PKT_FLAG_KEY);
-        if(a == 0 && new_pkt.data != pkt->data && new_pkt.destruct) {
+        if(a == 0 && new_pkt.data != pkt->data) {
             uint8_t *t = av_malloc(new_pkt.size + AV_INPUT_BUFFER_PADDING_SIZE); //the new should be a subset of the old so cannot overflow
             if(t) {
                 memcpy(t, new_pkt.data, new_pkt.size);
@@ -1108,6 +1099,7 @@ static void do_video_out(AVFormatContext *s,
 #endif
         return;
 
+#if FF_API_LAVF_FMT_RAWPICTURE
     if (s->oformat->flags & AVFMT_RAWPICTURE &&
         enc->codec->id == AV_CODEC_ID_RAWVIDEO) {
         /* raw pictures are written as AVPicture structure to
@@ -1123,7 +1115,9 @@ static void do_video_out(AVFormatContext *s,
         pkt.flags |= AV_PKT_FLAG_KEY;
 
         write_frame(s, &pkt, ost);
-    } else {
+    } else
+#endif
+    {
         int got_packet, forced_keyframe = 0;
         double pts_time;
 
@@ -1153,7 +1147,7 @@ static void do_video_out(AVFormatContext *s,
             ost->forced_keyframes_expr_const_values[FKF_T] = pts_time;
             res = av_expr_eval(ost->forced_keyframes_pexpr,
                                ost->forced_keyframes_expr_const_values, NULL);
-            av_dlog(NULL, "force_key_frame: n:%f n_forced:%f prev_forced_n:%f t:%f prev_forced_t:%f -> res:%f\n",
+            ff_dlog(NULL, "force_key_frame: n:%f n_forced:%f prev_forced_n:%f t:%f prev_forced_t:%f -> res:%f\n",
                     ost->forced_keyframes_expr_const_values[FKF_N],
                     ost->forced_keyframes_expr_const_values[FKF_N_FORCED],
                     ost->forced_keyframes_expr_const_values[FKF_PREV_FORCED_N],
@@ -1287,7 +1281,7 @@ static void do_video_stats(OutputStream *ost, int frame_size)
         avg_bitrate = (double)(ost->data_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)ost->data_size / 1024, ti1, bitrate, avg_bitrate);
-        fprintf(vstats_file, "type= %c\n", enc->coded_frame ? av_get_picture_type_char(enc->coded_frame->pict_type) : 'I');
+        fprintf(vstats_file, "type= %c\n", av_get_picture_type_char(ost->pict_type));
     }
 }
 
@@ -1532,7 +1526,7 @@ static void print_report(int is_last_report, int64_t timer_start, int64_t cur_ti
     AVCodecContext *enc;
     int frame_number, vid, i;
     double bitrate;
-    int64_t pts = INT64_MIN;
+    int64_t pts = INT64_MIN + 1;
     static int64_t last_time = -1;
     static int qp_histogram[52];
     int hours, mins, secs, us;
@@ -1594,7 +1588,7 @@ static void print_report(int is_last_report, int64_t timer_start, int64_t cur_ti
                     snprintf(buf + strlen(buf), sizeof(buf) - strlen(buf), "%X", (int)lrintf(log2(qp_histogram[j] + 1)));
             }
 
-            if ((enc->flags & AV_CODEC_FLAG_PSNR) && (enc->coded_frame || is_last_report)) {
+            if ((enc->flags & AV_CODEC_FLAG_PSNR) && (ost->pict_type != AV_PICTURE_TYPE_NONE || is_last_report)) {
                 int j;
                 double error, error_sum = 0;
                 double scale, scale_sum = 0;
@@ -1713,8 +1707,10 @@ static void flush_encoders(void)
 
         if (enc->codec_type == AVMEDIA_TYPE_AUDIO && enc->frame_size <= 1)
             continue;
+#if FF_API_LAVF_FMT_RAWPICTURE
         if (enc->codec_type == AVMEDIA_TYPE_VIDEO && (os->oformat->flags & AVFMT_RAWPICTURE) && enc->codec->id == AV_CODEC_ID_RAWVIDEO)
             continue;
+#endif
 
         for (;;) {
             int (*encode)(AVCodecContext*, AVPacket*, const AVFrame*, int*) = NULL;
@@ -1745,7 +1741,9 @@ static void flush_encoders(void)
                 ret = encode(enc, &pkt, NULL, &got_packet);
                 update_benchmark("flush %s %d.%d", desc, ost->file_index, ost->index);
                 if (ret < 0) {
-                    av_log(NULL, AV_LOG_FATAL, "%s encoding failed\n", desc);
+                    av_log(NULL, AV_LOG_FATAL, "%s encoding failed: %s\n",
+                           desc,
+                           av_err2str(ret));
                     exit_program(1);
                 }
                 if (ost->logfile && enc->stats_out) {
@@ -1861,17 +1859,22 @@ static void do_streamcopy(InputStream *ist, OutputStream *ost, const AVPacket *p
 
     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->enc_ctx->codec_id != AV_CODEC_ID_H264
-       && ost->enc_ctx->codec_id != AV_CODEC_ID_MPEG1VIDEO
-       && ost->enc_ctx->codec_id != AV_CODEC_ID_MPEG2VIDEO
-       && ost->enc_ctx->codec_id != AV_CODEC_ID_VC1
+    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
        ) {
-        if (av_parser_change(ost->parser, ost->st->codec,
+        int ret = av_parser_change(ost->parser, ost->st->codec,
                              &opkt.data, &opkt.size,
                              pkt->data, pkt->size,
-                             pkt->flags & AV_PKT_FLAG_KEY)) {
+                             pkt->flags & AV_PKT_FLAG_KEY);
+        if (ret < 0) {
+            av_log(NULL, AV_LOG_FATAL, "av_parser_change failed: %s\n",
+                   av_err2str(ret));
+            exit_program(1);
+        }
+        if (ret) {
             opkt.buf = av_buffer_create(opkt.data, opkt.size, av_buffer_default_free, NULL, 0);
             if (!opkt.buf)
                 exit_program(1);
@@ -1882,13 +1885,22 @@ static void do_streamcopy(InputStream *ist, OutputStream *ost, const AVPacket *p
     }
     av_copy_packet_side_data(&opkt, pkt);
 
-    if (ost->st->codec->codec_type == AVMEDIA_TYPE_VIDEO && (of->ctx->oformat->flags & AVFMT_RAWPICTURE)) {
+#if FF_API_LAVF_FMT_RAWPICTURE
+    if (ost->st->codec->codec_type == AVMEDIA_TYPE_VIDEO &&
+        ost->st->codec->codec_id == AV_CODEC_ID_RAWVIDEO &&
+        (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);
+        int ret = avpicture_fill(&pict, opkt.data, ost->st->codec->pix_fmt, ost->st->codec->width, ost->st->codec->height);
+        if (ret < 0) {
+            av_log(NULL, AV_LOG_FATAL, "avpicture_fill failed: %s\n",
+                   av_err2str(ret));
+            exit_program(1);
+        }
         opkt.data = (uint8_t *)&pict;
         opkt.size = sizeof(AVPicture);
         opkt.flags |= AV_PKT_FLAG_KEY;
     }
+#endif
 
     write_frame(of->ctx, &opkt, ost);
 }
@@ -2019,6 +2031,7 @@ static int decode_audio(InputStream *ist, AVPacket *pkt, int *got_output)
         decoded_frame->pts = av_rescale_delta(decoded_frame_tb, decoded_frame->pts,
                                               (AVRational){1, avctx->sample_rate}, decoded_frame->nb_samples, &ist->filter_in_rescale_delta_last,
                                               (AVRational){1, avctx->sample_rate});
+    ist->nb_samples = decoded_frame->nb_samples;
     for (i = 0; i < ist->nb_filters; i++) {
         if (i < ist->nb_filters - 1) {
             f = ist->filter_frame;
@@ -2066,12 +2079,13 @@ static int decode_video(InputStream *ist, AVPacket *pkt, int *got_output)
         if (ist->dec_ctx->codec_id == AV_CODEC_ID_H264) {
             ist->st->codec->has_b_frames = ist->dec_ctx->has_b_frames;
         } else
-            av_log_ask_for_sample(
-                ist->dec_ctx,
-                "has_b_frames is larger in decoder than demuxer %d > %d ",
-                ist->dec_ctx->has_b_frames,
-                ist->st->codec->has_b_frames
-            );
+            av_log(ist->dec_ctx, AV_LOG_WARNING,
+                   "has_b_frames is larger in decoder than demuxer %d > %d.\n"
+                   "If you want to help, upload a sample "
+                   "of this file to ftp://upload.ffmpeg.org/incoming/ "
+                   "and contact the ffmpeg-devel mailing list. (ffmpeg-devel@ffmpeg.org)",
+                   ist->dec_ctx->has_b_frames,
+                   ist->st->codec->has_b_frames);
     }
 
     if (*got_output || ret<0)
@@ -2247,11 +2261,7 @@ static int send_filter_eof(InputStream *ist)
 {
     int i, ret;
     for (i = 0; i < ist->nb_filters; i++) {
-#if 1
-        ret = av_buffersrc_add_ref(ist->filters[i]->filter, NULL, 0);
-#else
         ret = av_buffersrc_add_frame(ist->filters[i]->filter, NULL);
-#endif
         if (ret < 0)
             return ret;
     }
@@ -2259,7 +2269,7 @@ static int send_filter_eof(InputStream *ist)
 }
 
 /* pkt = NULL means EOF (needed to flush decoder buffers) */
-static int process_input_packet(InputStream *ist, const AVPacket *pkt)
+static int process_input_packet(InputStream *ist, const AVPacket *pkt, int no_eof)
 {
     int ret = 0, i;
     int got_output = 0;
@@ -2368,7 +2378,8 @@ static int process_input_packet(InputStream *ist, const AVPacket *pkt)
     }
 
     /* after flushing, send an EOF on all the filter inputs attached to the stream */
-    if (!pkt && ist->decoding_needed && !got_output) {
+    /* except when looping we need to flush but not to send an EOF */
+    if (!pkt && ist->decoding_needed && !got_output && !no_eof) {
         int ret = send_filter_eof(ist);
         if (ret < 0) {
             av_log(NULL, AV_LOG_FATAL, "Error marking filters as finished\n");
@@ -2588,7 +2599,11 @@ static int init_output_stream(OutputStream *ost, char *error, int error_len)
         }
         if (!av_dict_get(ost->encoder_opts, "threads", NULL, 0))
             av_dict_set(&ost->encoder_opts, "threads", "auto", 0);
-        av_dict_set(&ost->encoder_opts, "side_data_only_packets", "1", 0);
+        if (ost->enc->type == AVMEDIA_TYPE_AUDIO &&
+            !codec->defaults &&
+            !av_dict_get(ost->encoder_opts, "b", NULL, 0) &&
+            !av_dict_get(ost->encoder_opts, "ab", NULL, 0))
+            av_dict_set(&ost->encoder_opts, "b", "128000", 0);
 
         if ((ret = avcodec_open2(ost->enc_ctx, codec, &ost->encoder_opts)) < 0) {
             if (ret == AVERROR_EXPERIMENTAL)
@@ -2894,7 +2909,7 @@ static int transcode_init(void)
                 enc_ctx->time_base = dec_ctx->time_base;
             }
 
-            if (ist && !ost->frame_rate.num)
+            if (!ost->frame_rate.num)
                 ost->frame_rate = ist->framerate;
             if(ost->frame_rate.num)
                 enc_ctx->time_base = av_inv_q(ost->frame_rate);
@@ -3427,9 +3442,17 @@ static int check_keyboard_interaction(int64_t cur_time)
             if(!debug) debug = 1;
             while(debug & (FF_DEBUG_DCT_COEFF|FF_DEBUG_VIS_QP|FF_DEBUG_VIS_MB_TYPE)) //unsupported, would just crash
                 debug += debug;
-        }else
-            if(scanf("%d", &debug)!=1)
+        }else{
+            char buf[32];
+            int k = 0;
+            i = 0;
+            while ((k = read_key()) != '\n' && k != '\r' && i < sizeof(buf)-1)
+                if (k > 0)
+                    buf[i++] = k;
+            buf[i] = 0;
+            if (k <= 0 || sscanf(buf, "%d", &debug)!=1)
                 fprintf(stderr,"error parsing debug value\n");
+        }
         for(i=0;i<nb_input_streams;i++) {
             input_streams[i]->st->codec->debug = debug;
         }
@@ -3592,6 +3615,86 @@ static void reset_eagain(void)
         output_streams[i]->unavailable = 0;
 }
 
+// set duration to max(tmp, duration) in a proper time base and return duration's time_base
+static AVRational duration_max(int64_t tmp, int64_t *duration, AVRational tmp_time_base,
+                                AVRational time_base)
+{
+    int ret;
+
+    if (!*duration) {
+        *duration = tmp;
+        return tmp_time_base;
+    }
+
+    ret = av_compare_ts(*duration, time_base, tmp, tmp_time_base);
+    if (ret < 0) {
+        *duration = tmp;
+        return tmp_time_base;
+    }
+
+    return time_base;
+}
+
+static int seek_to_start(InputFile *ifile, AVFormatContext *is)
+{
+    InputStream *ist;
+    AVCodecContext *avctx;
+    int i, ret, has_audio = 0;
+    int64_t duration = 0;
+
+    ret = av_seek_frame(is, -1, is->start_time, 0);
+    if (ret < 0)
+        return ret;
+
+    for (i = 0; i < ifile->nb_streams; i++) {
+        ist   = input_streams[ifile->ist_index + i];
+        avctx = ist->dec_ctx;
+
+        // flush decoders
+        if (ist->decoding_needed) {
+            process_input_packet(ist, NULL, 1);
+            avcodec_flush_buffers(avctx);
+        }
+
+        /* duration is the length of the last frame in a stream
+         * when audio stream is present we don't care about
+         * last video frame length because it's not defined exactly */
+        if (avctx->codec_type == AVMEDIA_TYPE_AUDIO && ist->nb_samples)
+            has_audio = 1;
+    }
+
+    for (i = 0; i < ifile->nb_streams; i++) {
+        ist   = input_streams[ifile->ist_index + i];
+        avctx = ist->dec_ctx;
+
+        if (has_audio) {
+            if (avctx->codec_type == AVMEDIA_TYPE_AUDIO && ist->nb_samples) {
+                AVRational sample_rate = {1, avctx->sample_rate};
+
+                duration = av_rescale_q(ist->nb_samples, sample_rate, ist->st->time_base);
+            } else
+                continue;
+        } else {
+            if (ist->framerate.num) {
+                duration = av_rescale_q(1, ist->framerate, ist->st->time_base);
+            } else if (ist->st->avg_frame_rate.num) {
+                duration = av_rescale_q(1, ist->st->avg_frame_rate, ist->st->time_base);
+            } else duration = 1;
+        }
+        if (!ifile->duration)
+            ifile->time_base = ist->st->time_base;
+        /* the total duration of the stream, max_pts - min_pts is
+         * the duration of the stream without the last frame */
+        duration += ist->max_pts - ist->min_pts;
+        ifile->time_base = duration_max(duration, &ifile->duration, ist->st->time_base,
+                                        ifile->time_base);
+    }
+
+    ifile->loop--;
+
+    return ret;
+}
+
 /*
  * Return
  * - 0 -- one packet was read and processed
@@ -3606,6 +3709,7 @@ static int process_input(int file_index)
     InputStream *ist;
     AVPacket pkt;
     int ret, i, j;
+    int64_t duration;
 
     is  = ifile->ctx;
     ret = get_input_packet(ifile, &pkt);
@@ -3614,6 +3718,11 @@ static int process_input(int file_index)
         ifile->eagain = 1;
         return ret;
     }
+    if ((ret < 0) && (ifile->loop > 1)) {
+        if ((ret = seek_to_start(ifile, is)) < 0)
+            return ret;
+        ret = get_input_packet(ifile, &pkt);
+    }
     if (ret < 0) {
         if (ret != AVERROR_EOF) {
             print_error(is->filename, ret);
@@ -3624,7 +3733,7 @@ static int process_input(int file_index)
         for (i = 0; i < ifile->nb_streams; i++) {
             ist = input_streams[ifile->ist_index + i];
             if (ist->decoding_needed) {
-                ret = process_input_packet(ist, NULL);
+                ret = process_input_packet(ist, NULL, 0);
                 if (ret>0)
                     return 0;
             }
@@ -3760,6 +3869,16 @@ static int process_input(int file_index)
         }
     }
 
+    duration = av_rescale_q(ifile->duration, ifile->time_base, ist->st->time_base);
+    if (pkt.pts != AV_NOPTS_VALUE) {
+        pkt.pts += duration;
+        ist->max_pts = FFMAX(pkt.pts, ist->max_pts);
+        ist->min_pts = FFMIN(pkt.pts, ist->min_pts);
+    }
+
+    if (pkt.dts != AV_NOPTS_VALUE)
+        pkt.dts += duration;
+
     if ((ist->dec_ctx->codec_type == AVMEDIA_TYPE_VIDEO ||
          ist->dec_ctx->codec_type == AVMEDIA_TYPE_AUDIO) &&
          pkt.dts != AV_NOPTS_VALUE && ist->next_dts != AV_NOPTS_VALUE &&
@@ -3810,7 +3929,7 @@ static int process_input(int file_index)
 
     sub2video_heartbeat(ist, pkt.pts);
 
-    process_input_packet(ist, &pkt);
+    process_input_packet(ist, &pkt, 0);
 
 discard_packet:
     av_free_packet(&pkt);
@@ -3975,7 +4094,7 @@ static int transcode(void)
     for (i = 0; i < nb_input_streams; i++) {
         ist = input_streams[i];
         if (!input_files[ist->file_index]->eof_reached && ist->decoding_needed) {
-            process_input_packet(ist, NULL);
+            process_input_packet(ist, NULL, 0);
         }
     }
     flush_encoders();
@@ -4029,6 +4148,7 @@ static int transcode(void)
                 av_freep(&ost->apad);
                 av_freep(&ost->disposition);
                 av_dict_free(&ost->encoder_opts);
+                av_dict_free(&ost->sws_dict);
                 av_dict_free(&ost->swr_opts);
                 av_dict_free(&ost->resample_opts);
                 av_dict_free(&ost->bsf_args);