From: Vincent Penquerc'h <vincent.penquerch@collabora.co.uk>
Date: Wed, 30 Apr 2014 14:30:45 +0000 (+0100)
Subject: avaudenc: avoid using wrong number of samples
X-Git-Tag: 1.19.3~499^2~565
X-Git-Url: http://review.tizen.org/git/?a=commitdiff_plain;h=17e2e9acd9b9be3cc3637a60507892ca4a97bd0a;p=platform%2Fupstream%2Fgstreamer.git

avaudenc: avoid using wrong number of samples

If audio_in is NULL, we'll send a NULL frame to libav, to flush
the codec. In that case, we won't know how many samples the codec
will have used, so we use -1 (for don't know) when letting the
base class know about the buffer.

Coverity 1195177
---

diff --git a/ext/libav/gstavaudenc.c b/ext/libav/gstavaudenc.c
index fcb176d..a9aa88a 100644
--- a/ext/libav/gstavaudenc.c
+++ b/ext/libav/gstavaudenc.c
@@ -553,7 +553,7 @@ gst_ffmpegaudenc_encode_audio (GstFFMpegAudEnc * ffmpegaudenc,
         pkt, gst_ffmpegaudenc_free_avpacket);
 
     codec = ffmpegaudenc->context->codec;
-    if ((codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE)) {
+    if ((codec->capabilities & CODEC_CAP_VARIABLE_FRAME_SIZE) || !audio_in) {
       ret = gst_audio_encoder_finish_frame (enc, outbuf, -1);
     } else {
       ret = gst_audio_encoder_finish_frame (enc, outbuf, frame.nb_samples);