lavf: move the packet keyframe setting code.

compute_pkt_fields() is for unreliable estimates or guessing. The
keyframe information from the parser is (at least in theory) reliable,
so it should be used even when the other guessing is disabled with the
AVFMT_FLAG_NOFILLIN flag.

Therefore, move setting the packet keyframe flag based on parser
information from compute_pkt_fields() to read_frame_internal().

libavformat/utils.c

@@ -984,14 +984,6 @@ static void compute_pkt_fields(AVFormatContext *s, AVStream *st,
     /* update flags */
     if(is_intra_only(st->codec))
         pkt->flags |= AV_PKT_FLAG_KEY;
-    else if (pc) {
-        pkt->flags = 0;
-        /* keyframe computation */
-        if (pc->key_frame == 1)
-            pkt->flags |= AV_PKT_FLAG_KEY;
-        else if (pc->key_frame == -1 && pc->pict_type == AV_PICTURE_TYPE_I)
-            pkt->flags |= AV_PKT_FLAG_KEY;
-    }
     if (pc)
         pkt->convergence_duration = pc->convergence_duration;
 }
@@ -1053,6 +1045,10 @@ static int read_frame_internal(AVFormatContext *s, AVPacket *pkt)
                     pkt->pts = st->parser->pts;
                     pkt->dts = st->parser->dts;
                     pkt->pos = st->parser->pos;
+                    if (st->parser->key_frame == 1 ||
+                        (st->parser->key_frame == -1 &&
+                         st->parser->pict_type == AV_PICTURE_TYPE_I))
+                        pkt->flags |= AV_PKT_FLAG_KEY;
                     if(pkt->data == st->cur_pkt.data && pkt->size == st->cur_pkt.size){
                         s->cur_st = NULL;
                         pkt->destruct= st->cur_pkt.destruct;