summaryrefslogtreecommitdiff
path: root/sys/androidmedia/gstamcaudiodec.c
diff options
context:
space:
mode:
authorSebastian Dröge <sebastian.droege@collabora.co.uk>2012-10-18 16:41:07 +0200
committerSebastian Dröge <sebastian.droege@collabora.co.uk>2012-10-25 14:05:48 +0200
commit079c68e4de878390304da28330c149508448fa6e (patch)
treea455dac4255b3590c194328db68b58d98e469da4 /sys/androidmedia/gstamcaudiodec.c
parent36680b1190b3ddbe733e526e870ce9e904b129c4 (diff)
downloadgstreamer-plugins-bad-079c68e4de878390304da28330c149508448fa6e.tar.gz
androidmedia: Port to 1.0
Diffstat (limited to 'sys/androidmedia/gstamcaudiodec.c')
-rw-r--r--sys/androidmedia/gstamcaudiodec.c230
1 files changed, 130 insertions, 100 deletions
diff --git a/sys/androidmedia/gstamcaudiodec.c b/sys/androidmedia/gstamcaudiodec.c
index 31a8bf5f2..886556ae2 100644
--- a/sys/androidmedia/gstamcaudiodec.c
+++ b/sys/androidmedia/gstamcaudiodec.c
@@ -28,7 +28,7 @@
#endif
#include <gst/gst.h>
-#include <gst/audio/multichannel.h>
+#include <gst/audio/audio.h>
#include <string.h>
#ifdef HAVE_ORC
@@ -69,11 +69,12 @@ enum
/* class initialization */
-#define DEBUG_INIT(bla) \
+#define DEBUG_INIT \
GST_DEBUG_CATEGORY_INIT (gst_amc_audio_dec_debug_category, "amcaudiodec", 0, \
"Android MediaCodec audio decoder");
+#define parent_class gst_amc_audio_dec_parent_class
-GST_BOILERPLATE_FULL (GstAmcAudioDec, gst_amc_audio_dec, GstAudioDecoder,
+G_DEFINE_TYPE_WITH_CODE (GstAmcAudioDec, gst_amc_audio_dec,
GST_TYPE_AUDIO_DECODER, DEBUG_INIT);
static GstCaps *
@@ -95,21 +96,21 @@ create_sink_caps (const GstAmcCodecInfo * codec_info)
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"parsed", G_TYPE_BOOLEAN, TRUE, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else if (strcmp (type->mime, "audio/3gpp") == 0) {
GstStructure *tmp;
tmp = gst_structure_new ("audio/AMR",
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else if (strcmp (type->mime, "audio/amr-wb") == 0) {
GstStructure *tmp;
tmp = gst_structure_new ("audio/AMR-WB",
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else if (strcmp (type->mime, "audio/mp4a-latm") == 0) {
gint j;
GstStructure *tmp, *tmp2;
@@ -147,13 +148,13 @@ create_sink_caps (const GstAmcCodecInfo * codec_info)
tmp2 = gst_structure_copy (tmp);
gst_structure_set (tmp2, "profile", G_TYPE_STRING, profile, NULL);
- gst_caps_merge_structure (ret, tmp2);
+ ret = gst_caps_merge_structure (ret, tmp2);
have_profile = TRUE;
}
if (!have_profile) {
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else {
gst_structure_free (tmp);
}
@@ -163,21 +164,21 @@ create_sink_caps (const GstAmcCodecInfo * codec_info)
tmp = gst_structure_new ("audio/x-alaw",
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else if (strcmp (type->mime, "audio/g711-mlaw") == 0) {
GstStructure *tmp;
tmp = gst_structure_new ("audio/x-mulaw",
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else if (strcmp (type->mime, "audio/vorbis") == 0) {
GstStructure *tmp;
tmp = gst_structure_new ("audio/x-vorbis",
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else if (strcmp (type->mime, "audio/flac") == 0) {
GstStructure *tmp;
@@ -185,7 +186,7 @@ create_sink_caps (const GstAmcCodecInfo * codec_info)
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"framed", G_TYPE_BOOLEAN, TRUE, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else if (strcmp (type->mime, "audio/mpeg-L2") == 0) {
GstStructure *tmp;
@@ -195,7 +196,7 @@ create_sink_caps (const GstAmcCodecInfo * codec_info)
"rate", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"channels", GST_TYPE_INT_RANGE, 1, G_MAXINT,
"parsed", G_TYPE_BOOLEAN, TRUE, NULL);
- gst_caps_merge_structure (ret, tmp);
+ ret = gst_caps_merge_structure (ret, tmp);
} else {
GST_WARNING ("Unsupported mimetype '%s'", type->mime);
}
@@ -264,22 +265,38 @@ create_src_caps (const GstAmcCodecInfo * codec_info)
}
static void
-gst_amc_audio_dec_base_init (gpointer g_class)
+gst_amc_audio_dec_class_init (GstAmcAudioDecClass * klass)
{
- GstElementClass *element_class = GST_ELEMENT_CLASS (g_class);
- GstAmcAudioDecClass *audiodec_class = GST_AMC_AUDIO_DEC_CLASS (g_class);
+ GObjectClass *gobject_class = G_OBJECT_CLASS (klass);
+ GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
+ GstAudioDecoderClass *audiodec_class = GST_AUDIO_DECODER_CLASS (klass);
+ GstAmcAudioDecClass *amcaudiodec_class = GST_AMC_AUDIO_DEC_CLASS (klass);
const GstAmcCodecInfo *codec_info;
GstPadTemplate *templ;
GstCaps *caps;
gchar *longname;
+ gobject_class->finalize = gst_amc_audio_dec_finalize;
+
+ element_class->change_state =
+ GST_DEBUG_FUNCPTR (gst_amc_audio_dec_change_state);
+
+ audiodec_class->start = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_start);
+ audiodec_class->stop = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_stop);
+ audiodec_class->open = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_open);
+ audiodec_class->close = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_close);
+ audiodec_class->flush = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_flush);
+ audiodec_class->set_format = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_set_format);
+ audiodec_class->handle_frame =
+ GST_DEBUG_FUNCPTR (gst_amc_audio_dec_handle_frame);
+
codec_info =
- g_type_get_qdata (G_TYPE_FROM_CLASS (g_class), gst_amc_codec_info_quark);
+ g_type_get_qdata (G_TYPE_FROM_CLASS (klass), gst_amc_codec_info_quark);
/* This happens for the base class and abstract subclasses */
if (!codec_info)
return;
- audiodec_class->codec_info = codec_info;
+ amcaudiodec_class->codec_info = codec_info;
/* Add pad templates */
caps = create_sink_caps (codec_info);
@@ -293,7 +310,7 @@ gst_amc_audio_dec_base_init (gpointer g_class)
gst_object_unref (templ);
longname = g_strdup_printf ("Android MediaCodec %s", codec_info->name);
- gst_element_class_set_details_simple (element_class,
+ gst_element_class_set_metadata (element_class,
codec_info->name,
"Codec/Decoder/Audio",
longname, "Sebastian Dröge <sebastian.droege@collabora.co.uk>");
@@ -301,31 +318,7 @@ gst_amc_audio_dec_base_init (gpointer g_class)
}
static void
-gst_amc_audio_dec_class_init (GstAmcAudioDecClass * klass)
-{
- GObjectClass *gobject_class = G_OBJECT_CLASS (klass);
- GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
- GstAudioDecoderClass *audiodec_class = GST_AUDIO_DECODER_CLASS (klass);
-
- gobject_class->finalize = gst_amc_audio_dec_finalize;
-
- element_class->change_state =
- GST_DEBUG_FUNCPTR (gst_amc_audio_dec_change_state);
-
- audiodec_class->start = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_start);
- audiodec_class->stop = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_stop);
-#if 0
- audiodec_class->open = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_open);
- audiodec_class->close = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_close);
-#endif
- audiodec_class->flush = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_flush);
- audiodec_class->set_format = GST_DEBUG_FUNCPTR (gst_amc_audio_dec_set_format);
- audiodec_class->handle_frame =
- GST_DEBUG_FUNCPTR (gst_amc_audio_dec_handle_frame);
-}
-
-static void
-gst_amc_audio_dec_init (GstAmcAudioDec * self, GstAmcAudioDecClass * klass)
+gst_amc_audio_dec_init (GstAmcAudioDec * self)
{
gst_audio_decoder_set_needs_format (GST_AUDIO_DECODER (self), TRUE);
gst_audio_decoder_set_drainable (GST_AUDIO_DECODER (self), TRUE);
@@ -400,8 +393,6 @@ gst_amc_audio_dec_change_state (GstElement * element, GstStateChange transition)
self->downstream_flow_ret = GST_FLOW_OK;
self->draining = FALSE;
self->started = FALSE;
- if (!gst_amc_audio_dec_open (GST_AUDIO_DECODER (self)))
- return GST_STATE_CHANGE_FAILURE;
break;
case GST_STATE_CHANGE_PAUSED_TO_PLAYING:
break;
@@ -429,9 +420,7 @@ gst_amc_audio_dec_change_state (GstElement * element, GstStateChange transition)
case GST_STATE_CHANGE_PLAYING_TO_PAUSED:
break;
case GST_STATE_CHANGE_PAUSED_TO_READY:
- if (!gst_amc_audio_dec_close (GST_AUDIO_DECODER (self)))
- return GST_STATE_CHANGE_FAILURE;
- self->downstream_flow_ret = GST_FLOW_WRONG_STATE;
+ self->downstream_flow_ret = GST_FLOW_FLUSHING;
self->started = FALSE;
break;
case GST_STATE_CHANGE_READY_TO_NULL:
@@ -449,6 +438,7 @@ gst_amc_audio_dec_set_src_caps (GstAmcAudioDec * self, GstAmcFormat * format)
GstCaps *caps;
gint rate, channels;
guint32 channel_mask = 0;
+ GstAudioChannelPosition to[64];
if (!gst_amc_format_get_int (format, "sample-rate", &rate) ||
!gst_amc_format_get_int (format, "channel-count", &channels)) {
@@ -465,26 +455,23 @@ gst_amc_audio_dec_set_src_caps (GstAmcAudioDec * self, GstAmcFormat * format)
if (gst_amc_format_contains_key (format, "channel-mask"))
gst_amc_format_get_int (format, "channel-mask", (gint *) & channel_mask);
- if (self->positions)
- g_free (self->positions);
- self->positions =
- gst_amc_audio_channel_mask_to_positions (channel_mask, channels);
+ gst_amc_audio_channel_mask_to_positions (channel_mask, channels,
+ self->positions);
+ memcpy (to, self->positions, sizeof (to));
+ gst_audio_channel_positions_to_valid_order (to, channels);
+ self->needs_reorder =
+ (memcmp (self->positions, to,
+ sizeof (GstAudioChannelPosition) * channels) != 0);
+ if (self->needs_reorder)
+ gst_audio_get_channel_reorder_map (channels, self->positions, to,
+ self->reorder_map);
+ gst_audio_info_init (&self->info);
+ gst_audio_info_set_format (&self->info, GST_AUDIO_FORMAT_S16, rate, channels,
+ to);
- caps = gst_caps_new_simple ("audio/x-raw-int",
- "rate", G_TYPE_INT, rate,
- "channels", G_TYPE_INT, channels,
- "width", G_TYPE_INT, 16,
- "depth", G_TYPE_INT, 16,
- "signed", G_TYPE_BOOLEAN, TRUE,
- "endianness", G_TYPE_INT, G_BYTE_ORDER, NULL);
+ caps = gst_audio_info_to_caps (&self->info);
- if (self->positions)
- gst_audio_set_channel_positions (gst_caps_get_structure (caps, 0),
- self->positions);
-
- self->channels = channels;
- self->rate = rate;
gst_pad_set_caps (GST_AUDIO_DECODER_SRC_PAD (self), caps);
gst_caps_unref (caps);
@@ -592,6 +579,7 @@ retry:
GstAmcAudioDecClass *klass = GST_AMC_AUDIO_DEC_GET_CLASS (self);
GstBuffer *outbuf;
GstAmcBuffer *buf;
+ GstMapInfo minfo;
/* This sometimes happens at EOS or if the input is not properly framed,
* let's handle it gracefully by allocating a new buffer for the current
@@ -610,13 +598,33 @@ retry:
}
}
- outbuf = gst_buffer_try_new_and_alloc (buffer_info.size);
+ outbuf =
+ gst_audio_decoder_allocate_output_buffer (GST_AUDIO_DECODER (self),
+ buffer_info.size);
if (!outbuf)
goto failed_allocate;
+ gst_buffer_map (outbuf, &minfo, GST_MAP_WRITE);
buf = &self->output_buffers[idx];
- orc_memcpy (GST_BUFFER_DATA (outbuf), buf->data + buffer_info.offset,
- buffer_info.size);
+ if (self->needs_reorder) {
+ gint i, n_samples, c, n_channels;
+ gint *reorder_map = self->reorder_map;
+ gint16 *dest, *source;
+
+ dest = (gint16 *) minfo.data;
+ source = (gint16 *) (buf->data + buffer_info.offset);
+ n_samples = buffer_info.size / self->info.bpf;
+ n_channels = self->info.channels;
+
+ for (i = 0; i < n_samples; i++) {
+ for (c = 0; c < n_channels; c++) {
+ dest[i * n_channels + reorder_map[c]] = source[i * n_channels + c];
+ }
+ }
+ } else {
+ orc_memcpy (minfo.data, buf->data + buffer_info.offset, buffer_info.size);
+ }
+ gst_buffer_unmap (outbuf, &minfo);
/* FIXME: We should get one decoded input frame here for
* every buffer. If this is not the case somewhere, we will
@@ -630,7 +638,7 @@ done:
if (!gst_amc_codec_release_output_buffer (self->codec, idx))
goto failed_release;
- if (is_eos || flow_ret == GST_FLOW_UNEXPECTED) {
+ if (is_eos || flow_ret == GST_FLOW_EOS) {
GST_AUDIO_DECODER_STREAM_UNLOCK (self);
g_mutex_lock (self->drain_lock);
if (self->draining) {
@@ -639,7 +647,7 @@ done:
g_cond_broadcast (self->drain_cond);
} else if (flow_ret == GST_FLOW_OK) {
GST_DEBUG_OBJECT (self, "Component signalled EOS");
- flow_ret = GST_FLOW_UNEXPECTED;
+ flow_ret = GST_FLOW_EOS;
}
g_mutex_unlock (self->drain_lock);
GST_AUDIO_DECODER_STREAM_LOCK (self);
@@ -702,20 +710,19 @@ flushing:
{
GST_DEBUG_OBJECT (self, "Flushing -- stopping task");
gst_pad_pause_task (GST_AUDIO_DECODER_SRC_PAD (self));
- self->downstream_flow_ret = GST_FLOW_WRONG_STATE;
+ self->downstream_flow_ret = GST_FLOW_FLUSHING;
GST_AUDIO_DECODER_STREAM_UNLOCK (self);
return;
}
flow_error:
{
- if (flow_ret == GST_FLOW_UNEXPECTED) {
+ if (flow_ret == GST_FLOW_EOS) {
GST_DEBUG_OBJECT (self, "EOS");
gst_pad_push_event (GST_AUDIO_DECODER_SRC_PAD (self),
gst_event_new_eos ());
gst_pad_pause_task (GST_AUDIO_DECODER_SRC_PAD (self));
- } else
- if (flow_ret == GST_FLOW_NOT_LINKED || flow_ret < GST_FLOW_UNEXPECTED) {
+ } else if (flow_ret == GST_FLOW_NOT_LINKED || flow_ret < GST_FLOW_EOS) {
GST_ELEMENT_ERROR (self, STREAM, FAILED,
("Internal data stream error."), ("stream stopped, reason %s",
gst_flow_get_name (flow_ret)));
@@ -786,20 +793,19 @@ gst_amc_audio_dec_stop (GstAudioDecoder * decoder)
}
gst_pad_stop_task (GST_AUDIO_DECODER_SRC_PAD (decoder));
- g_free (self->positions);
- self->positions = NULL;
+ memset (self->positions, 0, sizeof (self->positions));
- g_list_foreach (self->codec_datas, (GFunc) gst_buffer_unref, NULL);
+ g_list_foreach (self->codec_datas, (GFunc) g_free, NULL);
g_list_free (self->codec_datas);
self->codec_datas = NULL;
- self->downstream_flow_ret = GST_FLOW_WRONG_STATE;
+ self->downstream_flow_ret = GST_FLOW_FLUSHING;
self->eos = FALSE;
g_mutex_lock (self->drain_lock);
self->draining = FALSE;
g_cond_broadcast (self->drain_cond);
g_mutex_unlock (self->drain_lock);
- gst_buffer_replace (&self->codec_data, NULL);
+
GST_DEBUG_OBJECT (self, "Stopped decoder");
return TRUE;
}
@@ -883,10 +889,14 @@ gst_amc_audio_dec_set_format (GstAudioDecoder * decoder, GstCaps * caps)
if (gst_structure_has_field (s, "codec_data")) {
const GValue *h = gst_structure_get_value (s, "codec_data");
GstBuffer *codec_data = gst_value_get_buffer (h);
-
- self->codec_datas =
- g_list_prepend (self->codec_datas, gst_buffer_ref (codec_data));
- gst_amc_format_set_buffer (format, "csd-0", codec_data);
+ GstMapInfo minfo;
+ guint8 *data;
+
+ gst_buffer_map (codec_data, &minfo, GST_MAP_READ);
+ data = g_memdup (minfo.data, minfo.size);
+ self->codec_datas = g_list_prepend (self->codec_datas, data);
+ gst_amc_format_set_buffer (format, "csd-0", data, minfo.size);
+ gst_buffer_unmap (codec_data, &minfo);
} else if (gst_structure_has_field (s, "streamheader")) {
const GValue *sh = gst_structure_get_value (s, "streamheader");
gint nsheaders = gst_value_array_get_size (sh);
@@ -894,13 +904,17 @@ gst_amc_audio_dec_set_format (GstAudioDecoder * decoder, GstCaps * caps)
const GValue *h;
gint i, j;
gchar *fname;
+ GstMapInfo minfo;
+ guint8 *data;
for (i = 0, j = 0; i < nsheaders; i++) {
h = gst_value_array_get_value (sh, i);
buf = gst_value_get_buffer (h);
if (strcmp (mime, "audio/vorbis") == 0) {
- guint8 header_type = GST_BUFFER_DATA (buf)[0];
+ guint8 header_type;
+
+ gst_buffer_extract (buf, 0, &header_type, 1);
/* Only use the identification and setup packets */
if (header_type != 0x01 && header_type != 0x05)
@@ -908,9 +922,11 @@ gst_amc_audio_dec_set_format (GstAudioDecoder * decoder, GstCaps * caps)
}
fname = g_strdup_printf ("csd-%d", j);
- self->codec_datas =
- g_list_prepend (self->codec_datas, gst_buffer_ref (buf));
- gst_amc_format_set_buffer (format, fname, buf);
+ gst_buffer_map (buf, &minfo, GST_MAP_READ);
+ data = g_memdup (minfo.data, minfo.size);
+ self->codec_datas = g_list_prepend (self->codec_datas, data);
+ gst_amc_format_set_buffer (format, fname, data, minfo.size);
+ gst_buffer_unmap (buf, &minfo);
g_free (fname);
j++;
}
@@ -949,7 +965,7 @@ gst_amc_audio_dec_set_format (GstAudioDecoder * decoder, GstCaps * caps)
self->flushing = FALSE;
self->downstream_flow_ret = GST_FLOW_OK;
gst_pad_start_task (GST_AUDIO_DECODER_SRC_PAD (self),
- (GstTaskFunction) gst_amc_audio_dec_loop, decoder);
+ (GstTaskFunction) gst_amc_audio_dec_loop, decoder, NULL);
return TRUE;
}
@@ -985,7 +1001,7 @@ gst_amc_audio_dec_flush (GstAudioDecoder * decoder, gboolean hard)
self->eos = FALSE;
self->downstream_flow_ret = GST_FLOW_OK;
gst_pad_start_task (GST_AUDIO_DECODER_SRC_PAD (self),
- (GstTaskFunction) gst_amc_audio_dec_loop, decoder);
+ (GstTaskFunction) gst_amc_audio_dec_loop, decoder, NULL);
GST_DEBUG_OBJECT (self, "Reset decoder");
}
@@ -999,6 +1015,9 @@ gst_amc_audio_dec_handle_frame (GstAudioDecoder * decoder, GstBuffer * inbuf)
GstAmcBufferInfo buffer_info;
guint offset = 0;
GstClockTime timestamp, duration, timestamp_offset = 0;
+ GstMapInfo minfo;
+
+ memset (&minfo, 0, sizeof (minfo));
self = GST_AMC_AUDIO_DEC (decoder);
@@ -1021,7 +1040,7 @@ gst_amc_audio_dec_handle_frame (GstAudioDecoder * decoder, GstBuffer * inbuf)
GST_WARNING_OBJECT (self, "Got frame after EOS");
if (inbuf)
gst_buffer_unref (inbuf);
- return GST_FLOW_UNEXPECTED;
+ return GST_FLOW_EOS;
}
if (self->flushing)
@@ -1033,10 +1052,12 @@ gst_amc_audio_dec_handle_frame (GstAudioDecoder * decoder, GstBuffer * inbuf)
if (!inbuf)
return gst_amc_audio_dec_drain (self);
- timestamp = GST_BUFFER_TIMESTAMP (inbuf);
+ timestamp = GST_BUFFER_PTS (inbuf);
duration = GST_BUFFER_DURATION (inbuf);
- while (offset < GST_BUFFER_SIZE (inbuf)) {
+ gst_buffer_map (inbuf, &minfo, GST_MAP_READ);
+
+ while (offset < minfo.size) {
/* Make sure to release the base class stream lock, otherwise
* _loop() can't call _finish_frame() and we might block forever
* because no input buffers are released */
@@ -1085,15 +1106,14 @@ gst_amc_audio_dec_handle_frame (GstAudioDecoder * decoder, GstBuffer * inbuf)
memset (&buffer_info, 0, sizeof (buffer_info));
buffer_info.offset = 0;
- buffer_info.size = MIN (GST_BUFFER_SIZE (inbuf) - offset, buf->size);
+ buffer_info.size = MIN (minfo.size - offset, buf->size);
- orc_memcpy (buf->data, GST_BUFFER_DATA (inbuf) + offset, buffer_info.size);
+ orc_memcpy (buf->data, minfo.data + offset, buffer_info.size);
/* Interpolate timestamps if we're passing the buffer
* in multiple chunks */
if (offset != 0 && duration != GST_CLOCK_TIME_NONE) {
- timestamp_offset =
- gst_util_uint64_scale (offset, duration, GST_BUFFER_SIZE (inbuf));
+ timestamp_offset = gst_util_uint64_scale (offset, duration, minfo.size);
}
if (timestamp != GST_CLOCK_TIME_NONE) {
@@ -1117,7 +1137,7 @@ gst_amc_audio_dec_handle_frame (GstAudioDecoder * decoder, GstBuffer * inbuf)
if (!gst_amc_codec_queue_input_buffer (self->codec, idx, &buffer_info))
goto queue_error;
}
-
+ gst_buffer_unmap (inbuf, &minfo);
gst_buffer_unref (inbuf);
return self->downstream_flow_ret;
@@ -1126,6 +1146,8 @@ downstream_error:
{
GST_ERROR_OBJECT (self, "Downstream returned %s",
gst_flow_get_name (self->downstream_flow_ret));
+ if (minfo.data)
+ gst_buffer_unmap (inbuf, &minfo);
if (inbuf)
gst_buffer_unref (inbuf);
return self->downstream_flow_ret;
@@ -1134,6 +1156,8 @@ invalid_buffer_index:
{
GST_ELEMENT_ERROR (self, LIBRARY, FAILED, (NULL),
("Invalid input buffer index %d of %d", idx, self->n_input_buffers));
+ if (minfo.data)
+ gst_buffer_unmap (inbuf, &minfo);
if (inbuf)
gst_buffer_unref (inbuf);
return GST_FLOW_ERROR;
@@ -1142,6 +1166,8 @@ dequeue_error:
{
GST_ELEMENT_ERROR (self, LIBRARY, FAILED, (NULL),
("Failed to dequeue input buffer"));
+ if (minfo.data)
+ gst_buffer_unmap (inbuf, &minfo);
if (inbuf)
gst_buffer_unref (inbuf);
return GST_FLOW_ERROR;
@@ -1150,16 +1176,20 @@ queue_error:
{
GST_ELEMENT_ERROR (self, LIBRARY, FAILED, (NULL),
("Failed to queue input buffer"));
+ if (minfo.data)
+ gst_buffer_unmap (inbuf, &minfo);
if (inbuf)
gst_buffer_unref (inbuf);
return GST_FLOW_ERROR;
}
flushing:
{
- GST_DEBUG_OBJECT (self, "Flushing -- returning WRONG_STATE");
+ GST_DEBUG_OBJECT (self, "Flushing -- returning FLUSHING");
+ if (minfo.data)
+ gst_buffer_unmap (inbuf, &minfo);
if (inbuf)
gst_buffer_unref (inbuf);
- return GST_FLOW_WRONG_STATE;
+ return GST_FLOW_FLUSHING;
}
}