video lavd: skip leading AUD NALU for H.264

H264_mf encoder (Windows) produces single byte AUD unit prior to SPS
NALU. Because only first NALU was evaluated to assess whether to pass
the frame further, it took 10 frames (+ additional 10 for probe) before
decoding started before it surrendered the checks.

src/rtp/rtpdec_h264.c

@@ -356,6 +356,8 @@ get_nalu_name(int type)
         return "H264 SEI";
     case NAL_SPS:
         return "H264 SPS";
+    case NAL_AUD:
+        return "H264 AUD";
     case NAL_HEVC_VPS:
         return "HEVC VPS";
     default:

src/rtp/rtpdec_h264.h

@@ -54,6 +54,7 @@ enum {
         NAL_IDR = 5,
         NAL_SEI = 6,
         NAL_SPS = 7,
+        NAL_AUD = 9,
         NAL_MAX = 23,
         // HEVC
         NAL_HEVC_VPS = 32,

src/video_decompress/libavcodec.c

@@ -806,16 +806,22 @@ static _Bool check_first_sps_vps(struct state_libavcodec_decompress *s, unsigned
                 s->sps_vps_found = 1;
                 return 1;
         }
-        const unsigned char *first_nal = rtpenc_h264_get_next_nal(src, src_len, NULL);
+
-        if (!first_nal) {
-                return 0;
-        }
         int nalu_type = 0;
-        if (s->desc.color_spec == H264) {
+        const unsigned char *nal = src;
-                nalu_type = NALU_HDR_GET_TYPE(first_nal[0]);
+        do {
-        } else {
+                nal =
-                nalu_type = first_nal[0] >> 1;
+                    rtpenc_h264_get_next_nal(nal, src_len - (nal - src), NULL);
-        }
+                if (!nal) {
+                        return 0;
+                }
+                if (s->desc.color_spec == H264) {
+                        nalu_type = NALU_HDR_GET_TYPE(nal[0]);
+                } else {
+                        nalu_type = nal[0] >> 1;
+                }
+                debug_msg("Received %s NALU.", get_nalu_name(nalu_type));
+        } while (nalu_type == NAL_AUD);
 
         switch (nalu_type) {
         case NAL_SPS: