mmaldec: fix problems with flush logic

Don't try to do a blocking wait for MMAL output if we haven't even sent
a single real packet, but only flush packets. Obviously we can't expect
to get anything back.

Additionally, don't send a flush packet to MMAL in the same case. It
appears the MMAL decoder will sometimes hang in mmal_vc_port_disable()
(called from ffmmal_close_decoder()), waiting for a reply from the GPU
which never arrives. Either MMAL disallows sending flush packets without
preceding real data, or it's a MMAL bug.

libavcodec/mmaldec.c

@@ -445,8 +445,6 @@ static int ffmmal_add_packet(AVCodecContext *avctx, AVPacket *avpkt)
     uint8_t *start;
     int ret = 0;
 
-    ctx->packets_sent++;
-
     if (avpkt->size) {
         if (ctx->bsfc) {
             uint8_t *tmp_data;
@@ -472,6 +470,14 @@ static int ffmmal_add_packet(AVCodecContext *avctx, AVPacket *avpkt)
         }
         size = buf->size;
         data = buf->data;
+        ctx->packets_sent++;
+    } else {
+        if (!ctx->packets_sent) {
+            // Short-cut the flush logic to avoid upsetting MMAL.
+            ctx->eos_sent = 1;
+            ctx->eos_received = 1;
+            goto done;
+        }
     }
 
     start = data;
@@ -641,7 +647,8 @@ static int ffmmal_read_frame(AVCodecContext *avctx, AVFrame *frame, int *got_fra
         // excessive buffering.
         // We also wait if we sent eos, but didn't receive it yet (think of decoding
         // stream with a very low number of frames).
-        if (ctx->frames_output || ctx->packets_sent > MAX_DELAYED_FRAMES || ctx->eos_sent) {
+        if (ctx->frames_output || ctx->packets_sent > MAX_DELAYED_FRAMES ||
+            (ctx->packets_sent && ctx->eos_sent)) {
             // MMAL will ignore broken input packets, which means the frame we
             // expect here may never arrive. Dealing with this correctly is
             // complicated, so here's a hack to avoid that it freezes forever