1
0
mirror of https://git.ffmpeg.org/ffmpeg.git synced 2024-12-23 07:42:51 +00:00

vmdaudio: remove unnecessary fields from VmdAudioContext and use the corresponding AVCodecContext fields instead.

Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
This commit is contained in:
Justin Ruggles 2011-02-23 13:11:12 -05:00 committed by Ronald S. Bultje
parent 1e86d685e0
commit 2ec7862db8

View File

@ -421,9 +421,6 @@ static av_cold int vmdvideo_decode_end(AVCodecContext *avctx)
typedef struct VmdAudioContext {
AVCodecContext *avctx;
int out_bps;
int channels;
int bits;
int block_align;
int predictors[2];
} VmdAudioContext;
@ -448,14 +445,13 @@ static av_cold int vmdaudio_decode_init(AVCodecContext *avctx)
VmdAudioContext *s = avctx->priv_data;
s->avctx = avctx;
s->channels = avctx->channels;
s->bits = avctx->bits_per_coded_sample;
s->block_align = avctx->block_align;
avctx->sample_fmt = AV_SAMPLE_FMT_S16;
s->out_bps = av_get_bits_per_sample_fmt(avctx->sample_fmt) >> 3;
av_log(s->avctx, AV_LOG_DEBUG, "%d channels, %d bits/sample, block align = %d, sample rate = %d\n",
s->channels, s->bits, s->block_align, avctx->sample_rate);
av_log(avctx, AV_LOG_DEBUG, "%d channels, %d bits/sample, "
"block align = %d, sample rate = %d\n",
avctx->channels, avctx->bits_per_coded_sample, avctx->block_align,
avctx->sample_rate);
return 0;
}
@ -482,14 +478,14 @@ static int vmdaudio_loadsound(VmdAudioContext *s, unsigned char *data,
const uint8_t *buf, int silent_chunks, int data_size)
{
int i;
int silent_size = s->block_align * silent_chunks * s->out_bps;
int silent_size = s->avctx->block_align * silent_chunks * s->out_bps;
if (silent_chunks) {
memset(data, 0, silent_size);
data += silent_size;
}
if (s->bits == 16)
vmdaudio_decode_audio(s, data, buf, data_size, s->channels == 2);
if (s->avctx->bits_per_coded_sample == 16)
vmdaudio_decode_audio(s, data, buf, data_size, s->avctx->channels == 2);
else {
/* copy the data but convert it to signed */
for (i = 0; i < data_size; i++){
@ -537,7 +533,7 @@ static int vmdaudio_decode_frame(AVCodecContext *avctx,
}
/* ensure output buffer is large enough */
if (*data_size < (s->block_align*silent_chunks + buf_size) * s->out_bps)
if (*data_size < (avctx->block_align*silent_chunks + buf_size) * s->out_bps)
return -1;
*data_size = vmdaudio_loadsound(s, output_samples, buf, silent_chunks, buf_size);