avconv: compute next_pts from input packet duration when possible.

avconv.c

@@ -1747,7 +1747,9 @@ static int transcode_video(InputStream *ist, AVPacket *pkt, int *got_output, int
     }
     ist->next_pts = ist->pts = guess_correct_pts(&ist->pts_ctx, decoded_frame->pkt_pts,
                                                  decoded_frame->pkt_dts);
-    if (ist->st->codec->time_base.num != 0) {
+    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 *