avcodec/avs3_parser: set has_b_frames properly

has_b_frames should be output_reorder_delay field in AVS3 sequence
header and larger than 1. The parser implementation doesn't parse
that field. Decoder can set has_b_frames properly, so use FFMAX
here to avoid resetting has_b_frames from output_reorder_delay to 1.
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>

libavcodec/avs3_parser.c

@@ -71,7 +71,7 @@ static void parse_avs3_nal_units(AVCodecParserContext *s, const uint8_t *buf,
     if (buf[0] == 0x0 && buf[1] == 0x0 && buf[2] == 0x1) {
         if (buf[3] == AVS3_SEQ_START_CODE) {
             GetBitContext gb;
-            int profile, ratecode;
+            int profile, ratecode, low_delay;
 
             init_get_bits8(&gb, buf + 4, buf_size - 4);
 
@@ -114,7 +114,8 @@ static void parse_avs3_nal_units(AVCodecParserContext *s, const uint8_t *buf,
             //            bitrate_high(12)
             skip_bits(&gb, 32);
 
-            avctx->has_b_frames = !get_bits(&gb, 1);
+            low_delay = get_bits(&gb, 1);
+            avctx->has_b_frames = FFMAX(avctx->has_b_frames, !low_delay);
 
             avctx->framerate.num = avctx->time_base.den = ff_avs3_frame_rate_tab[ratecode].num;
             avctx->framerate.den = avctx->time_base.num = ff_avs3_frame_rate_tab[ratecode].den;