ffmpeg: do packet ts rescaling in write_packet()

This will be useful in the following commit, after which the muxer
timebase is not always available when encoding.

This merges Libav commit 3e265ca5. It was previously skipped.

There are some changes with how/when the mux_timebase field is set,
because the Libav approach often causes a too imprecise time base
to be set. This is hard, because the muxer's write_header function
can readjust the timebase, at which point we might already have
encoded packets buffered. (It might be better to buffer them after
the encoder, instead of after all the timestamp handling logic
before muxing.)

The two FATE tests change because the output time base is raised
for subtitles. (Needed to avoid certain rounding issues in other
cases.)

Includes a minor merge fix by Mark Thompson, and

    avconv: Move rescale to stream timebase before monotonisation

also by Mark Thompson <sw@jkqxz.net>.
Signed-off-by: wm4 <nfxjfg@googlemail.com>

ffmpeg.c

@@ -713,10 +713,12 @@ static void write_packet(OutputFile *of, AVPacket *pkt, OutputStream *ost)
             if (pkt->duration > 0)
                 av_log(NULL, AV_LOG_WARNING, "Overriding packet duration by frame rate, this should not happen\n");
             pkt->duration = av_rescale_q(1, av_inv_q(ost->frame_rate),
-                                         ost->st->time_base);
+                                         ost->mux_timebase);
         }
     }
 
+    av_packet_rescale_ts(pkt, ost->mux_timebase, ost->st->time_base);
+
     if (!(s->oformat->flags & AVFMT_NOTIMESTAMPS)) {
         if (pkt->dts != AV_NOPTS_VALUE &&
             pkt->pts != AV_NOPTS_VALUE &&
@@ -907,13 +909,13 @@ static void do_audio_out(OutputFile *of, OutputStream *ost,
 
         update_benchmark("encode_audio %d.%d", ost->file_index, ost->index);
 
-        av_packet_rescale_ts(&pkt, enc->time_base, ost->st->time_base);
+        av_packet_rescale_ts(&pkt, enc->time_base, ost->mux_timebase);
 
         if (debug_ts) {
             av_log(NULL, AV_LOG_INFO, "encoder -> type:audio "
                    "pkt_pts:%s pkt_pts_time:%s pkt_dts:%s pkt_dts_time:%s\n",
-                   av_ts2str(pkt.pts), av_ts2timestr(pkt.pts, &ost->st->time_base),
-                   av_ts2str(pkt.dts), av_ts2timestr(pkt.dts, &ost->st->time_base));
+                   av_ts2str(pkt.pts), av_ts2timestr(pkt.pts, &enc->time_base),
+                   av_ts2str(pkt.dts), av_ts2timestr(pkt.dts, &enc->time_base));
         }
 
         output_packet(of, &pkt, ost);
@@ -993,15 +995,15 @@ static void do_subtitle_out(OutputFile *of,
         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);
-        pkt.duration = av_rescale_q(sub->end_display_time, (AVRational){ 1, 1000 }, ost->st->time_base);
+        pkt.pts  = av_rescale_q(sub->pts, AV_TIME_BASE_Q, ost->mux_timebase);
+        pkt.duration = av_rescale_q(sub->end_display_time, (AVRational){ 1, 1000 }, ost->mux_timebase);
         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;
+                pkt.pts += av_rescale_q(sub->start_display_time, (AVRational){ 1, 1000 }, ost->mux_timebase);
             else
-                pkt.pts += 90 * sub->end_display_time;
+                pkt.pts += av_rescale_q(sub->end_display_time, (AVRational){ 1, 1000 }, ost->mux_timebase);
         }
         pkt.dts = pkt.pts;
         output_packet(of, &pkt, ost);
@@ -1187,7 +1189,7 @@ static void do_video_out(OutputFile *of,
             mux_par->field_order = AV_FIELD_PROGRESSIVE;
         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.pts    = av_rescale_q(in_picture->pts, enc->time_base, ost->mux_timebase);
         pkt.flags |= AV_PKT_FLAG_KEY;
 
         output_packet(of, &pkt, ost);
@@ -1283,13 +1285,13 @@ static void do_video_out(OutputFile *of,
             if (pkt.pts == AV_NOPTS_VALUE && !(enc->codec->capabilities & AV_CODEC_CAP_DELAY))
                 pkt.pts = ost->sync_opts;
 
-            av_packet_rescale_ts(&pkt, enc->time_base, ost->st->time_base);
+            av_packet_rescale_ts(&pkt, enc->time_base, ost->mux_timebase);
 
             if (debug_ts) {
                 av_log(NULL, AV_LOG_INFO, "encoder -> type:video "
                     "pkt_pts:%s pkt_pts_time:%s pkt_dts:%s pkt_dts_time:%s\n",
-                    av_ts2str(pkt.pts), av_ts2timestr(pkt.pts, &ost->st->time_base),
-                    av_ts2str(pkt.dts), av_ts2timestr(pkt.dts, &ost->st->time_base));
+                    av_ts2str(pkt.pts), av_ts2timestr(pkt.pts, &ost->mux_timebase),
+                    av_ts2str(pkt.dts), av_ts2timestr(pkt.dts, &ost->mux_timebase));
             }
 
             frame_size = pkt.size;
@@ -1862,7 +1864,7 @@ static void flush_encoders(void)
                     av_packet_unref(&pkt);
                     continue;
                 }
-                av_packet_rescale_ts(&pkt, enc->time_base, ost->st->time_base);
+                av_packet_rescale_ts(&pkt, enc->time_base, ost->mux_timebase);
                 pkt_size = pkt.size;
                 output_packet(of, &pkt, ost);
                 if (ost->enc_ctx->codec_type == AVMEDIA_TYPE_VIDEO && vstats_filename) {
@@ -1897,7 +1899,7 @@ static void do_streamcopy(InputStream *ist, OutputStream *ost, const AVPacket *p
     OutputFile *of = output_files[ost->file_index];
     InputFile   *f = input_files [ist->file_index];
     int64_t start_time = (of->start_time == AV_NOPTS_VALUE) ? 0 : of->start_time;
-    int64_t ost_tb_start_time = av_rescale_q(start_time, AV_TIME_BASE_Q, ost->st->time_base);
+    int64_t ost_tb_start_time = av_rescale_q(start_time, AV_TIME_BASE_Q, ost->mux_timebase);
     AVPicture pict;
     AVPacket opkt;
 
@@ -1938,14 +1940,14 @@ static void do_streamcopy(InputStream *ist, OutputStream *ost, const AVPacket *p
         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;
+        opkt.pts = av_rescale_q(pkt->pts, ist->st->time_base, ost->mux_timebase) - ost_tb_start_time;
     else
         opkt.pts = AV_NOPTS_VALUE;
 
     if (pkt->dts == AV_NOPTS_VALUE)
-        opkt.dts = av_rescale_q(ist->dts, AV_TIME_BASE_Q, ost->st->time_base);
+        opkt.dts = av_rescale_q(ist->dts, AV_TIME_BASE_Q, ost->mux_timebase);
     else
-        opkt.dts = av_rescale_q(pkt->dts, ist->st->time_base, ost->st->time_base);
+        opkt.dts = av_rescale_q(pkt->dts, ist->st->time_base, ost->mux_timebase);
     opkt.dts -= ost_tb_start_time;
 
     if (ost->st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO && pkt->dts != AV_NOPTS_VALUE) {
@@ -1954,10 +1956,11 @@ static void do_streamcopy(InputStream *ist, OutputStream *ost, const AVPacket *p
             duration = ist->dec_ctx->frame_size;
         opkt.dts = opkt.pts = av_rescale_delta(ist->st->time_base, pkt->dts,
                                                (AVRational){1, ist->dec_ctx->sample_rate}, duration, &ist->filter_in_rescale_delta_last,
-                                               ost->st->time_base) - ost_tb_start_time;
+                                               ost->mux_timebase) - ost_tb_start_time;
     }
 
-    opkt.duration = av_rescale_q(pkt->duration, ist->st->time_base, ost->st->time_base);
+    opkt.duration = av_rescale_q(pkt->duration, ist->st->time_base, ost->mux_timebase);
+
     opkt.flags    = pkt->flags;
     // FIXME remove the following 2 lines they shall be replaced by the bitstream filters
     if (  ost->st->codecpar->codec_id != AV_CODEC_ID_H264
@@ -2822,6 +2825,10 @@ static int check_init_output_file(OutputFile *of, int file_index)
     for (i = 0; i < of->ctx->nb_streams; i++) {
         OutputStream *ost = output_streams[of->ost_index + i];
 
+        /* try to improve muxing time_base (only possible if nothing has been written yet) */
+        if (!av_fifo_size(ost->muxing_queue))
+            ost->mux_timebase = ost->st->time_base;
+
         while (av_fifo_size(ost->muxing_queue)) {
             AVPacket pkt;
             av_fifo_generic_read(ost->muxing_queue, &pkt, sizeof(pkt), NULL);
@@ -2981,6 +2988,8 @@ static int init_output_stream_streamcopy(OutputStream *ost)
         break;
     }
 
+    ost->mux_timebase = ist->st->time_base;
+
     return 0;
 }
 
@@ -3236,7 +3245,7 @@ static int init_output_stream_encode(OutputStream *ost)
         }
         break;
     case AVMEDIA_TYPE_SUBTITLE:
-        enc_ctx->time_base = (AVRational){1, 1000};
+        enc_ctx->time_base = AV_TIME_BASE_Q;
         if (!enc_ctx->width) {
             enc_ctx->width     = input_streams[ost->source_index]->st->codecpar->width;
             enc_ctx->height    = input_streams[ost->source_index]->st->codecpar->height;
@@ -3249,6 +3258,8 @@ static int init_output_stream_encode(OutputStream *ost)
         break;
     }
 
+    ost->mux_timebase = enc_ctx->time_base;
+
     return 0;
 }
 

ffmpeg.h

@@ -446,6 +446,8 @@ typedef struct OutputStream {
     int64_t first_pts;
     /* dts of the last packet sent to the muxer */
     int64_t last_mux_dts;
+    // the timebase of the packets sent to the muxer
+    AVRational mux_timebase;
 
     int                    nb_bitstream_filters;
     uint8_t                  *bsf_extradata_updated;

tests/ref/fate/binsub-movtextenc

-ef264064c522389d0cf267c4d6235561
+af6a8f38d7c11d9af7823cc44554d2ad