Because old thread  now a bit too big for mobile gmail  client.

Yes, I'll try to see if it fails same way on my end. 

code seems to check first 100 frames

2528 static int field_probe(AVFormatContext *fmt_ctx, AVStream *st)

2529 {

2530   AVDictionary *copts = 0;

2531   //av_dict_copy(&copts, opts, 0);

2532   AVCodecID codec_id = st->codecpar->codec_id;

2533 #if LIBAVCODEC_VERSION_INT <= AV_VERSION_INT(58,134,100)

2534   AVCodec *decoder = avcodec_find_decoder(codec_id);

2535 #endif

2536 #if LIBAVCODEC_VERSION_INT >= AV_VERSION_INT(59,16,100)

2537   const AVCodec *decoder = avcodec_find_decoder(codec_id);

2538 #endif

2539   AVCodecContext *ctx = avcodec_alloc_context3(decoder);

2540   if( !ctx ) {

2541     fprintf(stderr,"codec alloc failed\n");

2542     return -1;

2543   }

2544   avcodec_parameters_to_context(ctx, st->codecpar);

2545   if( avcodec_open2(ctx, decoder, &copts) < 0 ) {

2546     fprintf(stderr,"codec open failed\n");

2547     return -1;

2548   }

2549   av_dict_free(&copts);

2550 

2551   AVFrame *ipic = av_frame_alloc();

2552   AVPacket ipkt;

2553   av_init_packet(&ipkt);

2554   int ilaced = -1;

2555   for( int retrys=100; --retrys>=0 && ilaced<0; ) {

2556     av_packet_unref(&ipkt);

2557     int ret = av_read_frame(fmt_ctx, &ipkt);

2558     if( ret == AVERROR_EOF ) break;

2559     if( ret != 0 ) continue;

2560     if( ipkt.stream_index != st->index ) continue;

2561     if( !ipkt.data || !ipkt.size ) continue;

2562     ret = avcodec_send_packet(ctx, &ipkt);

2563     if( ret < 0 ) {

2564       fprintf(stderr, "avcodec_send_packet failed\n");

2565       break;

2566     }

2567     ret = avcodec_receive_frame(ctx, ipic);

2568     if( ret >= 0 ) {

2569       ilaced = ipic->interlaced_frame ? 1 : 0;

2570       break;

2571     }

2572     if( ret != AVERROR(EAGAIN) )

2573       fprintf(stderr, "avcodec_receive_frame failed %d\n", ret);

2574   }

2575   av_packet_unref(&ipkt);

2576   av_frame_free(&ipic);

2577   avcodec_free_context(&ctx);

2578   return ilaced;

may be we can add flags to force interlaced vs progressive to bdwrite ... and/or debug why it fails.