mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2024-11-30 13:41:48 +00:00
Code Issues 1 Projects Releases Wiki Activity
gstreamer/ext/ffmpeg/gstffmpegdec.c
Edward Hervey f3c1a56a60 The proper spelling is 'FFmpeg'.
2009-04-15 22:11:10 +02:00

2714 lines
83 KiB
C
Raw Blame History

/* GStreamer
* Copyright (C) <1999> Erik Walthinsen <omega@cse.ogi.edu>
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <assert.h>
#include <string.h>
#ifdef HAVE_FFMPEG_UNINSTALLED
#include <avcodec.h>
#else
#include <libavcodec/avcodec.h>
#endif
#include <gst/gst.h>
#include "gstffmpeg.h"
#include "gstffmpegcodecmap.h"
/* define to enable alternative buffer refcounting algorithm */
#undef EXTRA_REF
typedef struct _GstFFMpegDec GstFFMpegDec;
struct _GstFFMpegDec
{
GstElement element;
/* We need to keep track of our pads, so we do so here. */
GstPad *srcpad;
GstPad *sinkpad;
/* decoding */
AVCodecContext *context;
AVFrame *picture;
gboolean opened;
union
{
struct
{
gint width, height;
gint clip_width, clip_height;
gint fps_n, fps_d;
gint old_fps_n, old_fps_d;
enum PixelFormat pix_fmt;
} video;
struct
{
gint channels;
gint samplerate;
gint depth;
} audio;
} format;
gboolean waiting_for_key;
gboolean discont;
gboolean clear_ts;
guint64 next_ts;
guint64 in_ts;
gint64 in_offset;
GstClockTime last_out;
gboolean ts_is_dts;
gboolean has_b_frames;
/* parsing */
AVCodecParserContext *pctx;
GstBuffer *pcache;
guint8 *padded;
guint padded_size;
GValue *par; /* pixel aspect ratio of incoming data */
gboolean current_dr; /* if direct rendering is enabled */
gboolean extra_ref; /* keep extra ref around in get/release */
/* some properties */
gint hurry_up;
gint lowres;
gboolean direct_rendering;
gboolean do_padding;
gboolean debug_mv;
gboolean crop;
/* QoS stuff *//* with LOCK */
gdouble proportion;
GstClockTime earliest_time;
/* clipping segment */
GstSegment segment;
gboolean is_realvideo;
/* reverse playback queue */
GList *queued;
};
typedef struct _GstFFMpegDecClass GstFFMpegDecClass;
struct _GstFFMpegDecClass
{
GstElementClass parent_class;
AVCodec *in_plugin;
GstPadTemplate *srctempl, *sinktempl;
};
typedef struct _GstFFMpegDecClassParams GstFFMpegDecClassParams;
struct _GstFFMpegDecClassParams
{
AVCodec *in_plugin;
GstCaps *srccaps, *sinkcaps;
};
#define GST_TYPE_FFMPEGDEC \
(gst_ffmpegdec_get_type())
#define GST_FFMPEGDEC(obj) \
(G_TYPE_CHECK_INSTANCE_CAST((obj),GST_TYPE_FFMPEGDEC,GstFFMpegDec))
#define GST_FFMPEGDEC_CLASS(klass) \
(G_TYPE_CHECK_CLASS_CAST((klass),GST_TYPE_FFMPEGDEC,GstFFMpegDecClass))
#define GST_IS_FFMPEGDEC(obj) \
(G_TYPE_CHECK_INSTANCE_TYPE((obj),GST_TYPE_FFMPEGDEC))
#define GST_IS_FFMPEGDEC_CLASS(klass) \
(G_TYPE_CHECK_CLASS_TYPE((klass),GST_TYPE_FFMPEGDEC))
#define DEFAULT_LOWRES 0
#define DEFAULT_SKIPFRAME 0
#define DEFAULT_DIRECT_RENDERING TRUE
#define DEFAULT_DO_PADDING TRUE
#define DEFAULT_DEBUG_MV FALSE
#define DEFAULT_CROP TRUE
enum
{
PROP_0,
PROP_LOWRES,
PROP_SKIPFRAME,
PROP_DIRECT_RENDERING,
PROP_DO_PADDING,
PROP_DEBUG_MV,
PROP_CROP,
PROP_LAST
};
/* A number of function prototypes are given so we can refer to them later. */
static void gst_ffmpegdec_base_init (GstFFMpegDecClass * klass);
static void gst_ffmpegdec_class_init (GstFFMpegDecClass * klass);
static void gst_ffmpegdec_init (GstFFMpegDec * ffmpegdec);
static void gst_ffmpegdec_finalize (GObject * object);
static gboolean gst_ffmpegdec_query (GstPad * pad, GstQuery * query);
static gboolean gst_ffmpegdec_src_event (GstPad * pad, GstEvent * event);
static gboolean gst_ffmpegdec_setcaps (GstPad * pad, GstCaps * caps);
static gboolean gst_ffmpegdec_sink_event (GstPad * pad, GstEvent * event);
static GstFlowReturn gst_ffmpegdec_chain (GstPad * pad, GstBuffer * buf);
static GstStateChangeReturn gst_ffmpegdec_change_state (GstElement * element,
GstStateChange transition);
static void gst_ffmpegdec_set_property (GObject * object,
guint prop_id, const GValue * value, GParamSpec * pspec);
static void gst_ffmpegdec_get_property (GObject * object,
guint prop_id, GValue * value, GParamSpec * pspec);
static gboolean gst_ffmpegdec_negotiate (GstFFMpegDec * ffmpegdec);
/* some sort of bufferpool handling, but different */
static int gst_ffmpegdec_get_buffer (AVCodecContext * context,
AVFrame * picture);
static void gst_ffmpegdec_release_buffer (AVCodecContext * context,
AVFrame * picture);
#define GST_FFDEC_PARAMS_QDATA g_quark_from_static_string("ffdec-params")
static GstElementClass *parent_class = NULL;
#define GST_FFMPEGDEC_TYPE_LOWRES (gst_ffmpegdec_lowres_get_type())
static GType
gst_ffmpegdec_lowres_get_type (void)
{
static GType ffmpegdec_lowres_type = 0;
if (!ffmpegdec_lowres_type) {
static const GEnumValue ffmpegdec_lowres[] = {
{0, "0", "full"},
{1, "1", "1/2-size"},
{2, "2", "1/4-size"},
{0, NULL, NULL},
};
ffmpegdec_lowres_type =
g_enum_register_static ("GstFFMpegDecLowres", ffmpegdec_lowres);
}
return ffmpegdec_lowres_type;
}
#define GST_FFMPEGDEC_TYPE_SKIPFRAME (gst_ffmpegdec_skipframe_get_type())
static GType
gst_ffmpegdec_skipframe_get_type (void)
{
static GType ffmpegdec_skipframe_type = 0;
if (!ffmpegdec_skipframe_type) {
static const GEnumValue ffmpegdec_skipframe[] = {
{0, "0", "Skip nothing"},
{1, "1", "Skip B-frames"},
{2, "2", "Skip IDCT/Dequantization"},
{5, "5", "Skip everything"},
{0, NULL, NULL},
};
ffmpegdec_skipframe_type =
g_enum_register_static ("GstFFMpegDecSkipFrame", ffmpegdec_skipframe);
}
return ffmpegdec_skipframe_type;
}
static void
gst_ffmpegdec_base_init (GstFFMpegDecClass * klass)
{
GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
GstFFMpegDecClassParams *params;
GstElementDetails details;
GstPadTemplate *sinktempl, *srctempl;
params =
(GstFFMpegDecClassParams *) g_type_get_qdata (G_OBJECT_CLASS_TYPE (klass),
GST_FFDEC_PARAMS_QDATA);
g_assert (params != NULL);
/* construct the element details struct */
details.longname = g_strdup_printf ("FFmpeg %s decoder",
params->in_plugin->long_name);
details.klass = g_strdup_printf ("Codec/Decoder/%s",
(params->in_plugin->type == CODEC_TYPE_VIDEO) ? "Video" : "Audio");
details.description = g_strdup_printf ("FFmpeg %s decoder",
params->in_plugin->name);
details.author = "Wim Taymans <wim.taymans@gmail.com>, "
"Ronald Bultje <rbultje@ronald.bitfreak.net>, "
"Edward Hervey <bilboed@bilboed.com>";
gst_element_class_set_details (element_class, &details);
g_free (details.longname);
g_free (details.klass);
g_free (details.description);
/* pad templates */
sinktempl = gst_pad_template_new ("sink", GST_PAD_SINK,
GST_PAD_ALWAYS, params->sinkcaps);
srctempl = gst_pad_template_new ("src", GST_PAD_SRC,
GST_PAD_ALWAYS, params->srccaps);
gst_element_class_add_pad_template (element_class, srctempl);
gst_element_class_add_pad_template (element_class, sinktempl);
klass->in_plugin = params->in_plugin;
klass->srctempl = srctempl;
klass->sinktempl = sinktempl;
}
static void
gst_ffmpegdec_class_init (GstFFMpegDecClass * klass)
{
GObjectClass *gobject_class = G_OBJECT_CLASS (klass);
GstElementClass *gstelement_class = GST_ELEMENT_CLASS (klass);
parent_class = g_type_class_peek_parent (klass);
gobject_class->finalize = gst_ffmpegdec_finalize;
gobject_class->set_property = gst_ffmpegdec_set_property;
gobject_class->get_property = gst_ffmpegdec_get_property;
if (klass->in_plugin->type == CODEC_TYPE_VIDEO) {
g_object_class_install_property (gobject_class, PROP_SKIPFRAME,
g_param_spec_enum ("skip-frame", "Skip frames",
"Which types of frames to skip during decoding",
GST_FFMPEGDEC_TYPE_SKIPFRAME, 0,
G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
g_object_class_install_property (gobject_class, PROP_LOWRES,
g_param_spec_enum ("lowres", "Low resolution",
"At which resolution to decode images", GST_FFMPEGDEC_TYPE_LOWRES,
0, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
g_object_class_install_property (gobject_class, PROP_DIRECT_RENDERING,
g_param_spec_boolean ("direct-rendering", "Direct Rendering",
"Enable direct rendering", DEFAULT_DIRECT_RENDERING,
G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
g_object_class_install_property (gobject_class, PROP_DO_PADDING,
g_param_spec_boolean ("do-padding", "Do Padding",
"Add 0 padding before decoding data", DEFAULT_DO_PADDING,
G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
g_object_class_install_property (gobject_class, PROP_DEBUG_MV,
g_param_spec_boolean ("debug-mv", "Debug motion vectors",
"Whether ffmpeg should print motion vectors on top of the image",
DEFAULT_DEBUG_MV, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
#if 0
g_object_class_install_property (gobject_class, PROP_CROP,
g_param_spec_boolean ("crop", "Crop",
"Crop images to the display region",
DEFAULT_CROP, G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
#endif
}
gstelement_class->change_state = gst_ffmpegdec_change_state;
}
static void
gst_ffmpegdec_init (GstFFMpegDec * ffmpegdec)
{
GstFFMpegDecClass *oclass;
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
/* setup pads */
ffmpegdec->sinkpad = gst_pad_new_from_template (oclass->sinktempl, "sink");
gst_pad_set_setcaps_function (ffmpegdec->sinkpad,
GST_DEBUG_FUNCPTR (gst_ffmpegdec_setcaps));
gst_pad_set_event_function (ffmpegdec->sinkpad,
GST_DEBUG_FUNCPTR (gst_ffmpegdec_sink_event));
gst_pad_set_chain_function (ffmpegdec->sinkpad,
GST_DEBUG_FUNCPTR (gst_ffmpegdec_chain));
gst_element_add_pad (GST_ELEMENT (ffmpegdec), ffmpegdec->sinkpad);
ffmpegdec->srcpad = gst_pad_new_from_template (oclass->srctempl, "src");
gst_pad_use_fixed_caps (ffmpegdec->srcpad);
gst_pad_set_event_function (ffmpegdec->srcpad,
GST_DEBUG_FUNCPTR (gst_ffmpegdec_src_event));
gst_pad_set_query_function (ffmpegdec->srcpad,
GST_DEBUG_FUNCPTR (gst_ffmpegdec_query));
gst_element_add_pad (GST_ELEMENT (ffmpegdec), ffmpegdec->srcpad);
/* some ffmpeg data */
ffmpegdec->context = avcodec_alloc_context ();
ffmpegdec->picture = avcodec_alloc_frame ();
ffmpegdec->pctx = NULL;
ffmpegdec->pcache = NULL;
ffmpegdec->par = NULL;
ffmpegdec->opened = FALSE;
ffmpegdec->waiting_for_key = TRUE;
ffmpegdec->hurry_up = ffmpegdec->lowres = 0;
ffmpegdec->direct_rendering = DEFAULT_DIRECT_RENDERING;
ffmpegdec->do_padding = DEFAULT_DO_PADDING;
ffmpegdec->debug_mv = DEFAULT_DEBUG_MV;
ffmpegdec->crop = DEFAULT_CROP;
ffmpegdec->format.video.fps_n = -1;
ffmpegdec->format.video.old_fps_n = -1;
gst_segment_init (&ffmpegdec->segment, GST_FORMAT_TIME);
}
static void
gst_ffmpegdec_finalize (GObject * object)
{
GstFFMpegDec *ffmpegdec = (GstFFMpegDec *) object;
if (ffmpegdec->context != NULL) {
av_free (ffmpegdec->context);
ffmpegdec->context = NULL;
}
if (ffmpegdec->picture != NULL) {
av_free (ffmpegdec->picture);
ffmpegdec->picture = NULL;
}
G_OBJECT_CLASS (parent_class)->finalize (object);
}
static gboolean
gst_ffmpegdec_query (GstPad * pad, GstQuery * query)
{
GstFFMpegDec *ffmpegdec;
GstPad *peer;
gboolean res;
ffmpegdec = (GstFFMpegDec *) gst_pad_get_parent (pad);
res = FALSE;
if ((peer = gst_pad_get_peer (ffmpegdec->sinkpad))) {
/* just forward to peer */
res = gst_pad_query (peer, query);
gst_object_unref (peer);
}
#if 0
{
GstFormat bfmt;
bfmt = GST_FORMAT_BYTES;
/* ok, do bitrate calc... */
if ((type != GST_QUERY_POSITION && type != GST_QUERY_TOTAL) ||
*fmt != GST_FORMAT_TIME || ffmpegdec->context->bit_rate == 0 ||
!gst_pad_query (peer, type, &bfmt, value))
return FALSE;
if (ffmpegdec->pcache && type == GST_QUERY_POSITION)
*value -= GST_BUFFER_SIZE (ffmpegdec->pcache);
*value *= GST_SECOND / ffmpegdec->context->bit_rate;
}
#endif
gst_object_unref (ffmpegdec);
return res;
}
static void
gst_ffmpegdec_update_qos (GstFFMpegDec * ffmpegdec, gdouble proportion,
GstClockTime time)
{
GST_OBJECT_LOCK (ffmpegdec);
ffmpegdec->proportion = proportion;
ffmpegdec->earliest_time = time;
GST_OBJECT_UNLOCK (ffmpegdec);
}
static void
gst_ffmpegdec_reset_qos (GstFFMpegDec * ffmpegdec)
{
gst_ffmpegdec_update_qos (ffmpegdec, 0.5, GST_CLOCK_TIME_NONE);
}
static void
gst_ffmpegdec_read_qos (GstFFMpegDec * ffmpegdec, gdouble * proportion,
GstClockTime * time)
{
GST_OBJECT_LOCK (ffmpegdec);
*proportion = ffmpegdec->proportion;
*time = ffmpegdec->earliest_time;
GST_OBJECT_UNLOCK (ffmpegdec);
}
static gboolean
gst_ffmpegdec_src_event (GstPad * pad, GstEvent * event)
{
GstFFMpegDec *ffmpegdec;
gboolean res;
ffmpegdec = (GstFFMpegDec *) gst_pad_get_parent (pad);
switch (GST_EVENT_TYPE (event)) {
case GST_EVENT_QOS:
{
gdouble proportion;
GstClockTimeDiff diff;
GstClockTime timestamp;
gst_event_parse_qos (event, &proportion, &diff, &timestamp);
/* update our QoS values */
gst_ffmpegdec_update_qos (ffmpegdec, proportion, timestamp + diff);
/* forward upstream */
res = gst_pad_push_event (ffmpegdec->sinkpad, event);
break;
}
default:
/* forward upstream */
res = gst_pad_push_event (ffmpegdec->sinkpad, event);
break;
}
gst_object_unref (ffmpegdec);
return res;
}
/* with LOCK */
static void
gst_ffmpegdec_close (GstFFMpegDec * ffmpegdec)
{
if (!ffmpegdec->opened)
return;
GST_LOG_OBJECT (ffmpegdec, "closing ffmpeg codec");
if (ffmpegdec->par) {
g_free (ffmpegdec->par);
ffmpegdec->par = NULL;
}
if (ffmpegdec->context->priv_data)
gst_ffmpeg_avcodec_close (ffmpegdec->context);
ffmpegdec->opened = FALSE;
if (ffmpegdec->context->palctrl) {
av_free (ffmpegdec->context->palctrl);
ffmpegdec->context->palctrl = NULL;
}
if (ffmpegdec->context->extradata) {
av_free (ffmpegdec->context->extradata);
ffmpegdec->context->extradata = NULL;
}
if (ffmpegdec->pctx) {
if (ffmpegdec->pcache) {
gst_buffer_unref (ffmpegdec->pcache);
ffmpegdec->pcache = NULL;
}
av_parser_close (ffmpegdec->pctx);
ffmpegdec->pctx = NULL;
}
ffmpegdec->format.video.fps_n = -1;
ffmpegdec->format.video.old_fps_n = -1;
}
/* with LOCK */
static gboolean
gst_ffmpegdec_open (GstFFMpegDec * ffmpegdec)
{
GstFFMpegDecClass *oclass;
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
if (gst_ffmpeg_avcodec_open (ffmpegdec->context, oclass->in_plugin) < 0)
goto could_not_open;
ffmpegdec->opened = TRUE;
ffmpegdec->is_realvideo = FALSE;
GST_LOG_OBJECT (ffmpegdec, "Opened ffmpeg codec %s, id %d",
oclass->in_plugin->name, oclass->in_plugin->id);
/* open a parser if we can */
switch (oclass->in_plugin->id) {
case CODEC_ID_MPEG4:
case CODEC_ID_MJPEG:
case CODEC_ID_VC1:
GST_LOG_OBJECT (ffmpegdec, "not using parser, blacklisted codec");
ffmpegdec->pctx = NULL;
break;
case CODEC_ID_H264:
/* For H264, only use a parser if there is no context data, if there is,
* we're talking AVC */
if (ffmpegdec->context->extradata_size == 0) {
GST_LOG_OBJECT (ffmpegdec, "H264 with no extradata, creating parser");
ffmpegdec->pctx = av_parser_init (oclass->in_plugin->id);
} else {
GST_LOG_OBJECT (ffmpegdec,
"H264 with extradata implies framed data - not using parser");
ffmpegdec->pctx = NULL;
}
break;
case CODEC_ID_RV10:
case CODEC_ID_RV30:
case CODEC_ID_RV20:
case CODEC_ID_RV40:
ffmpegdec->is_realvideo = TRUE;
break;
default:
ffmpegdec->pctx = av_parser_init (oclass->in_plugin->id);
if (ffmpegdec->pctx)
GST_LOG_OBJECT (ffmpegdec, "Using parser %p", ffmpegdec->pctx);
else
GST_LOG_OBJECT (ffmpegdec, "No parser for codec");
break;
}
switch (oclass->in_plugin->type) {
case CODEC_TYPE_VIDEO:
ffmpegdec->format.video.width = 0;
ffmpegdec->format.video.height = 0;
ffmpegdec->format.video.clip_width = -1;
ffmpegdec->format.video.clip_height = -1;
ffmpegdec->format.video.pix_fmt = PIX_FMT_NB;
break;
case CODEC_TYPE_AUDIO:
ffmpegdec->format.audio.samplerate = 0;
ffmpegdec->format.audio.channels = 0;
ffmpegdec->format.audio.depth = 0;
break;
default:
break;
}
ffmpegdec->last_out = GST_CLOCK_TIME_NONE;
ffmpegdec->next_ts = GST_CLOCK_TIME_NONE;
/* FIXME, reset_qos holds the LOCK */
ffmpegdec->proportion = 0.0;
ffmpegdec->earliest_time = -1;
return TRUE;
/* ERRORS */
could_not_open:
{
gst_ffmpegdec_close (ffmpegdec);
GST_DEBUG_OBJECT (ffmpegdec, "ffdec_%s: Failed to open FFMPEG codec",
oclass->in_plugin->name);
return FALSE;
}
}
static gboolean
gst_ffmpegdec_setcaps (GstPad * pad, GstCaps * caps)
{
GstFFMpegDec *ffmpegdec;
GstFFMpegDecClass *oclass;
GstStructure *structure;
const GValue *par;
const GValue *fps;
gboolean ret = TRUE;
ffmpegdec = (GstFFMpegDec *) (gst_pad_get_parent (pad));
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
GST_DEBUG_OBJECT (pad, "setcaps called");
GST_OBJECT_LOCK (ffmpegdec);
/* stupid check for VC1 */
if ((oclass->in_plugin->id == CODEC_ID_WMV3) ||
(oclass->in_plugin->id == CODEC_ID_VC1))
oclass->in_plugin->id = gst_ffmpeg_caps_to_codecid (caps, NULL);
/* close old session */
gst_ffmpegdec_close (ffmpegdec);
/* set defaults */
avcodec_get_context_defaults (ffmpegdec->context);
/* set buffer functions */
ffmpegdec->context->get_buffer = gst_ffmpegdec_get_buffer;
ffmpegdec->context->release_buffer = gst_ffmpegdec_release_buffer;
ffmpegdec->context->draw_horiz_band = NULL;
/* assume PTS as input, we will adapt when we detect timestamp reordering
* in the output frames. */
ffmpegdec->ts_is_dts = FALSE;
ffmpegdec->has_b_frames = FALSE;
GST_LOG_OBJECT (ffmpegdec, "size %dx%d", ffmpegdec->context->width,
ffmpegdec->context->height);
/* get size and so */
gst_ffmpeg_caps_with_codecid (oclass->in_plugin->id,
oclass->in_plugin->type, caps, ffmpegdec->context);
GST_LOG_OBJECT (ffmpegdec, "size after %dx%d", ffmpegdec->context->width,
ffmpegdec->context->height);
if (!ffmpegdec->context->time_base.den || !ffmpegdec->context->time_base.num) {
GST_DEBUG_OBJECT (ffmpegdec, "forcing 25/1 framerate");
ffmpegdec->context->time_base.num = 1;
ffmpegdec->context->time_base.den = 25;
}
/* get pixel aspect ratio if it's set */
structure = gst_caps_get_structure (caps, 0);
par = gst_structure_get_value (structure, "pixel-aspect-ratio");
if (par) {
GST_DEBUG_OBJECT (ffmpegdec, "sink caps have pixel-aspect-ratio of %d:%d",
gst_value_get_fraction_numerator (par),
gst_value_get_fraction_denominator (par));
/* should be NULL */
if (ffmpegdec->par)
g_free (ffmpegdec->par);
ffmpegdec->par = g_new0 (GValue, 1);
gst_value_init_and_copy (ffmpegdec->par, par);
}
/* get the framerate from incomming caps. fps_n is set to -1 when
* there is no valid framerate */
fps = gst_structure_get_value (structure, "framerate");
if (fps != NULL && GST_VALUE_HOLDS_FRACTION (fps)) {
ffmpegdec->format.video.fps_n = gst_value_get_fraction_numerator (fps);
ffmpegdec->format.video.fps_d = gst_value_get_fraction_denominator (fps);
GST_DEBUG_OBJECT (ffmpegdec, "Using framerate %d/%d from incoming caps",
ffmpegdec->format.video.fps_n, ffmpegdec->format.video.fps_d);
} else {
ffmpegdec->format.video.fps_n = -1;
GST_DEBUG_OBJECT (ffmpegdec, "Using framerate from codec");
}
/* figure out if we can use direct rendering */
ffmpegdec->current_dr = FALSE;
ffmpegdec->extra_ref = FALSE;
if (ffmpegdec->direct_rendering) {
GST_DEBUG_OBJECT (ffmpegdec, "trying to enable direct rendering");
if (oclass->in_plugin->capabilities & CODEC_CAP_DR1) {
if (oclass->in_plugin->id == CODEC_ID_H264) {
GST_DEBUG_OBJECT (ffmpegdec, "disable direct rendering setup for H264");
/* does not work, many stuff reads outside of the planes */
ffmpegdec->current_dr = FALSE;
ffmpegdec->extra_ref = TRUE;
} else {
GST_DEBUG_OBJECT (ffmpegdec, "enabled direct rendering");
ffmpegdec->current_dr = TRUE;
}
} else {
GST_DEBUG_OBJECT (ffmpegdec, "direct rendering not supported");
}
}
if (ffmpegdec->current_dr) {
/* do *not* draw edges when in direct rendering, for some reason it draws
* outside of the memory. */
ffmpegdec->context->flags |= CODEC_FLAG_EMU_EDGE;
}
/* workaround encoder bugs */
ffmpegdec->context->workaround_bugs |= FF_BUG_AUTODETECT;
ffmpegdec->context->error_recognition = 1;
/* for slow cpus */
ffmpegdec->context->lowres = ffmpegdec->lowres;
ffmpegdec->context->hurry_up = ffmpegdec->hurry_up;
/* ffmpeg can draw motion vectors on top of the image (not every decoder
* supports it) */
ffmpegdec->context->debug_mv = ffmpegdec->debug_mv;
/* open codec - we don't select an output pix_fmt yet,
* simply because we don't know! We only get it
* during playback... */
if (!gst_ffmpegdec_open (ffmpegdec))
goto open_failed;
/* clipping region */
gst_structure_get_int (structure, "width",
&ffmpegdec->format.video.clip_width);
gst_structure_get_int (structure, "height",
&ffmpegdec->format.video.clip_height);
/* take into account the lowres property */
if (ffmpegdec->format.video.clip_width != -1)
ffmpegdec->format.video.clip_width >>= ffmpegdec->lowres;
if (ffmpegdec->format.video.clip_height != -1)
ffmpegdec->format.video.clip_height >>= ffmpegdec->lowres;
done:
GST_OBJECT_UNLOCK (ffmpegdec);
gst_object_unref (ffmpegdec);
return ret;
/* ERRORS */
open_failed:
{
GST_DEBUG_OBJECT (ffmpegdec, "Failed to open");
if (ffmpegdec->par) {
g_free (ffmpegdec->par);
ffmpegdec->par = NULL;
}
ret = FALSE;
goto done;
}
}
static GstFlowReturn
alloc_output_buffer (GstFFMpegDec * ffmpegdec, GstBuffer ** outbuf,
gint width, gint height)
{
GstFlowReturn ret;
gint fsize;
ret = GST_FLOW_ERROR;
*outbuf = NULL;
GST_LOG_OBJECT (ffmpegdec, "alloc output buffer");
/* see if we need renegotiation */
if (G_UNLIKELY (!gst_ffmpegdec_negotiate (ffmpegdec)))
goto negotiate_failed;
/* get the size of the gstreamer output buffer given a
* width/height/format */
fsize = gst_ffmpeg_avpicture_get_size (ffmpegdec->context->pix_fmt,
width, height);
if (!ffmpegdec->context->palctrl) {
GST_LOG_OBJECT (ffmpegdec, "calling pad_alloc");
/* no pallete, we can use the buffer size to alloc */
ret = gst_pad_alloc_buffer_and_set_caps (ffmpegdec->srcpad,
GST_BUFFER_OFFSET_NONE, fsize,
GST_PAD_CAPS (ffmpegdec->srcpad), outbuf);
if (G_UNLIKELY (ret != GST_FLOW_OK))
goto alloc_failed;
} else {
GST_LOG_OBJECT (ffmpegdec, "not calling pad_alloc, we have a pallete");
/* for paletted data we can't use pad_alloc_buffer(), because
* fsize contains the size of the palette, so the overall size
* is bigger than ffmpegcolorspace's unit size, which will
* prompt GstBaseTransform to complain endlessly ... */
*outbuf = gst_buffer_new_and_alloc (fsize);
gst_buffer_set_caps (*outbuf, GST_PAD_CAPS (ffmpegdec->srcpad));
ret = GST_FLOW_OK;
}
return ret;
/* special cases */
negotiate_failed:
{
GST_DEBUG_OBJECT (ffmpegdec, "negotiate failed");
return GST_FLOW_NOT_NEGOTIATED;
}
alloc_failed:
{
GST_DEBUG_OBJECT (ffmpegdec, "pad_alloc failed");
return ret;
}
}
static int
gst_ffmpegdec_get_buffer (AVCodecContext * context, AVFrame * picture)
{
GstBuffer *buf = NULL;
GstFFMpegDec *ffmpegdec;
gint width, height;
gint coded_width, coded_height;
gint res;
ffmpegdec = (GstFFMpegDec *) context->opaque;
GST_DEBUG_OBJECT (ffmpegdec, "getting buffer, apply pts %" G_GINT64_FORMAT,
ffmpegdec->in_ts);
/* apply the last timestamp we have seen to this picture, when we get the
* picture back from ffmpeg we can use this to correctly timestamp the output
* buffer */
picture->pts = ffmpegdec->in_ts;
/* make sure we don't free the buffer when it's not ours */
picture->opaque = NULL;
/* take width and height before clipping */
width = context->width;
height = context->height;
coded_width = context->coded_width;
coded_height = context->coded_height;
GST_LOG_OBJECT (ffmpegdec, "dimension %dx%d, coded %dx%d", width, height,
coded_width, coded_height);
if (!ffmpegdec->current_dr) {
GST_LOG_OBJECT (ffmpegdec, "direct rendering disabled, fallback alloc");
res = avcodec_default_get_buffer (context, picture);
GST_LOG_OBJECT (ffmpegdec, "linsize %d %d %d", picture->linesize[0],
picture->linesize[1], picture->linesize[2]);
GST_LOG_OBJECT (ffmpegdec, "data %d %d %d", 0,
picture->data[1] - picture->data[0],
picture->data[2] - picture->data[0]);
return res;
}
switch (context->codec_type) {
case CODEC_TYPE_VIDEO:
/* some ffmpeg video plugins don't see the point in setting codec_type ... */
case CODEC_TYPE_UNKNOWN:
{
GstFlowReturn ret;
gint clip_width, clip_height;
/* take final clipped output size */
if ((clip_width = ffmpegdec->format.video.clip_width) == -1)
clip_width = width;
if ((clip_height = ffmpegdec->format.video.clip_height) == -1)
clip_height = height;
GST_LOG_OBJECT (ffmpegdec, "raw outsize %d/%d", width, height);
/* this is the size ffmpeg needs for the buffer */
avcodec_align_dimensions (context, &width, &height);
GST_LOG_OBJECT (ffmpegdec, "aligned outsize %d/%d, clip %d/%d",
width, height, clip_width, clip_height);
if (width != clip_width || height != clip_height) {
/* We can't alloc if we need to clip the output buffer later */
GST_LOG_OBJECT (ffmpegdec, "we need clipping, fallback alloc");
return avcodec_default_get_buffer (context, picture);
}
/* alloc with aligned dimensions for ffmpeg */
ret = alloc_output_buffer (ffmpegdec, &buf, width, height);
if (G_UNLIKELY (ret != GST_FLOW_OK)) {
/* alloc default buffer when we can't get one from downstream */
GST_LOG_OBJECT (ffmpegdec, "alloc failed, fallback alloc");
return avcodec_default_get_buffer (context, picture);
}
/* copy the right pointers and strides in the picture object */
gst_ffmpeg_avpicture_fill ((AVPicture *) picture,
GST_BUFFER_DATA (buf), context->pix_fmt, width, height);
break;
}
case CODEC_TYPE_AUDIO:
default:
GST_ERROR_OBJECT (ffmpegdec,
"_get_buffer() should never get called for non-video buffers !");
g_assert_not_reached ();
break;
}
/* tell ffmpeg we own this buffer, tranfer the ref we have on the buffer to
* the opaque data. */
picture->type = FF_BUFFER_TYPE_USER;
picture->age = 256 * 256 * 256 * 64;
picture->opaque = buf;
#ifdef EXTRA_REF
if (picture->reference != 0 || ffmpegdec->extra_ref) {
GST_DEBUG_OBJECT (ffmpegdec, "adding extra ref");
gst_buffer_ref (buf);
}
#endif
GST_LOG_OBJECT (ffmpegdec, "returned buffer %p", buf);
return 0;
}
static void
gst_ffmpegdec_release_buffer (AVCodecContext * context, AVFrame * picture)
{
gint i;
GstBuffer *buf;
GstFFMpegDec *ffmpegdec;
ffmpegdec = (GstFFMpegDec *) context->opaque;
/* check if it was our buffer */
if (picture->opaque == NULL) {
GST_DEBUG_OBJECT (ffmpegdec, "default release buffer");
avcodec_default_release_buffer (context, picture);
return;
}
/* we remove the opaque data now */
buf = GST_BUFFER_CAST (picture->opaque);
GST_DEBUG_OBJECT (ffmpegdec, "release buffer %p", buf);
picture->opaque = NULL;
#ifdef EXTRA_REF
if (picture->reference != 0 || ffmpegdec->extra_ref) {
GST_DEBUG_OBJECT (ffmpegdec, "remove extra ref");
gst_buffer_unref (buf);
}
#else
gst_buffer_unref (buf);
#endif
/* zero out the reference in ffmpeg */
for (i = 0; i < 4; i++) {
picture->data[i] = NULL;
picture->linesize[i] = 0;
}
}
static void
gst_ffmpegdec_add_pixel_aspect_ratio (GstFFMpegDec * ffmpegdec,
GstStructure * s)
{
gboolean demuxer_par_set = FALSE;
gboolean decoder_par_set = FALSE;
gint demuxer_num = 1, demuxer_denom = 1;
gint decoder_num = 1, decoder_denom = 1;
GST_OBJECT_LOCK (ffmpegdec);
if (ffmpegdec->par) {
demuxer_num = gst_value_get_fraction_numerator (ffmpegdec->par);
demuxer_denom = gst_value_get_fraction_denominator (ffmpegdec->par);
demuxer_par_set = TRUE;
GST_DEBUG_OBJECT (ffmpegdec, "Demuxer PAR: %d:%d", demuxer_num,
demuxer_denom);
}
if (ffmpegdec->context->sample_aspect_ratio.num &&
ffmpegdec->context->sample_aspect_ratio.den) {
decoder_num = ffmpegdec->context->sample_aspect_ratio.num;
decoder_denom = ffmpegdec->context->sample_aspect_ratio.den;
decoder_par_set = TRUE;
GST_DEBUG_OBJECT (ffmpegdec, "Decoder PAR: %d:%d", decoder_num,
decoder_denom);
}
GST_OBJECT_UNLOCK (ffmpegdec);
if (!demuxer_par_set && !decoder_par_set)
goto no_par;
if (demuxer_par_set && !decoder_par_set)
goto use_demuxer_par;
if (decoder_par_set && !demuxer_par_set)
goto use_decoder_par;
/* Both the demuxer and the decoder provide a PAR. If one of
* the two PARs is 1:1 and the other one is not, use the one
* that is not 1:1. */
if (demuxer_num == demuxer_denom && decoder_num != decoder_denom)
goto use_decoder_par;
if (decoder_num == decoder_denom && demuxer_num != demuxer_denom)
goto use_demuxer_par;
/* Both PARs are non-1:1, so use the PAR provided by the demuxer */
goto use_demuxer_par;
use_decoder_par:
{
GST_DEBUG_OBJECT (ffmpegdec,
"Setting decoder provided pixel-aspect-ratio of %u:%u", decoder_num,
decoder_denom);
gst_structure_set (s, "pixel-aspect-ratio", GST_TYPE_FRACTION, decoder_num,
decoder_denom, NULL);
return;
}
use_demuxer_par:
{
GST_DEBUG_OBJECT (ffmpegdec,
"Setting demuxer provided pixel-aspect-ratio of %u:%u", demuxer_num,
demuxer_denom);
gst_structure_set (s, "pixel-aspect-ratio", GST_TYPE_FRACTION, demuxer_num,
demuxer_denom, NULL);
return;
}
no_par:
{
GST_DEBUG_OBJECT (ffmpegdec,
"Neither demuxer nor codec provide a pixel-aspect-ratio");
return;
}
}
static gboolean
gst_ffmpegdec_negotiate (GstFFMpegDec * ffmpegdec)
{
GstFFMpegDecClass *oclass;
GstCaps *caps;
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
switch (oclass->in_plugin->type) {
case CODEC_TYPE_VIDEO:
if (ffmpegdec->format.video.width == ffmpegdec->context->width &&
ffmpegdec->format.video.height == ffmpegdec->context->height &&
ffmpegdec->format.video.fps_n == ffmpegdec->format.video.old_fps_n &&
ffmpegdec->format.video.fps_d == ffmpegdec->format.video.old_fps_d &&
ffmpegdec->format.video.pix_fmt == ffmpegdec->context->pix_fmt)
return TRUE;
GST_DEBUG_OBJECT (ffmpegdec,
"Renegotiating video from %dx%d@ %d/%d fps to %dx%d@ %d/%d fps",
ffmpegdec->format.video.width, ffmpegdec->format.video.height,
ffmpegdec->format.video.old_fps_n, ffmpegdec->format.video.old_fps_n,
ffmpegdec->context->width, ffmpegdec->context->height,
ffmpegdec->format.video.fps_n, ffmpegdec->format.video.fps_d);
ffmpegdec->format.video.width = ffmpegdec->context->width;
ffmpegdec->format.video.height = ffmpegdec->context->height;
ffmpegdec->format.video.old_fps_n = ffmpegdec->format.video.fps_n;
ffmpegdec->format.video.old_fps_d = ffmpegdec->format.video.fps_d;
ffmpegdec->format.video.pix_fmt = ffmpegdec->context->pix_fmt;
break;
case CODEC_TYPE_AUDIO:
{
gint depth = av_smp_format_depth (ffmpegdec->context->sample_fmt);
if (ffmpegdec->format.audio.samplerate ==
ffmpegdec->context->sample_rate &&
ffmpegdec->format.audio.channels == ffmpegdec->context->channels &&
ffmpegdec->format.audio.depth == depth)
return TRUE;
GST_DEBUG_OBJECT (ffmpegdec,
"Renegotiating audio from %dHz@%dchannels (%d) to %dHz@%dchannels (%d)",
ffmpegdec->format.audio.samplerate, ffmpegdec->format.audio.channels,
ffmpegdec->format.audio.depth,
ffmpegdec->context->sample_rate, ffmpegdec->context->channels, depth);
ffmpegdec->format.audio.samplerate = ffmpegdec->context->sample_rate;
ffmpegdec->format.audio.channels = ffmpegdec->context->channels;
ffmpegdec->format.audio.depth = depth;
}
break;
default:
break;
}
caps = gst_ffmpeg_codectype_to_caps (oclass->in_plugin->type,
ffmpegdec->context, oclass->in_plugin->id, FALSE);
if (caps == NULL)
goto no_caps;
switch (oclass->in_plugin->type) {
case CODEC_TYPE_VIDEO:
{
gint width, height;
width = ffmpegdec->format.video.clip_width;
height = ffmpegdec->format.video.clip_height;
if (width != -1 && height != -1) {
/* overwrite the output size with the dimension of the
* clipping region */
gst_caps_set_simple (caps,
"width", G_TYPE_INT, width, "height", G_TYPE_INT, height, NULL);
}
/* If a demuxer provided a framerate then use it (#313970) */
if (ffmpegdec->format.video.fps_n != -1) {
gst_caps_set_simple (caps, "framerate",
GST_TYPE_FRACTION, ffmpegdec->format.video.fps_n,
ffmpegdec->format.video.fps_d, NULL);
}
gst_ffmpegdec_add_pixel_aspect_ratio (ffmpegdec,
gst_caps_get_structure (caps, 0));
break;
}
case CODEC_TYPE_AUDIO:
{
break;
}
default:
break;
}
if (!gst_pad_set_caps (ffmpegdec->srcpad, caps))
goto caps_failed;
gst_caps_unref (caps);
return TRUE;
/* ERRORS */
no_caps:
{
GST_ELEMENT_ERROR (ffmpegdec, CORE, NEGOTIATION, (NULL),
("could not find caps for codec (%s), unknown type",
oclass->in_plugin->name));
return FALSE;
}
caps_failed:
{
GST_ELEMENT_ERROR (ffmpegdec, CORE, NEGOTIATION, (NULL),
("Could not set caps for ffmpeg decoder (%s), not fixed?",
oclass->in_plugin->name));
gst_caps_unref (caps);
return FALSE;
}
}
/* perform qos calculations before decoding the next frame.
*
* Sets the hurry_up flag and if things are really bad, skips to the next
* keyframe.
*
* Returns TRUE if the frame should be decoded, FALSE if the frame can be dropped
* entirely.
*/
static gboolean
gst_ffmpegdec_do_qos (GstFFMpegDec * ffmpegdec, GstClockTime timestamp,
gboolean * mode_switch)
{
GstClockTimeDiff diff;
gdouble proportion;
GstClockTime qostime, earliest_time;
*mode_switch = FALSE;
/* no timestamp, can't do QoS */
if (G_UNLIKELY (!GST_CLOCK_TIME_IS_VALID (timestamp)))
goto no_qos;
/* get latest QoS observation values */
gst_ffmpegdec_read_qos (ffmpegdec, &proportion, &earliest_time);
/* skip qos if we have no observation (yet) */
if (G_UNLIKELY (!GST_CLOCK_TIME_IS_VALID (earliest_time))) {
/* no hurry_up initialy */
ffmpegdec->context->hurry_up = 0;
goto no_qos;
}
/* qos is done on running time */
qostime = gst_segment_to_running_time (&ffmpegdec->segment, GST_FORMAT_TIME,
timestamp);
/* see how our next timestamp relates to the latest qos timestamp. negative
* values mean we are early, positive values mean we are too late. */
diff = GST_CLOCK_DIFF (qostime, earliest_time);
GST_DEBUG_OBJECT (ffmpegdec, "QOS: qostime %" GST_TIME_FORMAT
", earliest %" GST_TIME_FORMAT, GST_TIME_ARGS (qostime),
GST_TIME_ARGS (earliest_time));
/* if we using less than 40% of the available time, we can try to
* speed up again when we were slow. */
if (proportion < 0.4 && diff < 0) {
goto normal_mode;
} else {
/* if we're more than two seconds late, switch to the next keyframe */
/* FIXME, let the demuxer decide what's the best since we might be dropping
* a lot of frames when the keyframe is far away or we even might not get a new
* keyframe at all.. */
if (diff > ((GstClockTimeDiff) GST_SECOND * 2)
&& !ffmpegdec->waiting_for_key) {
goto skip_to_keyframe;
} else if (diff >= 0) {
/* we're too slow, try to speed up */
if (ffmpegdec->waiting_for_key) {
/* we were waiting for a keyframe, that's ok */
goto skipping;
}
/* switch to hurry_up mode */
goto hurry_up;
}
}
no_qos:
return TRUE;
skipping:
{
return FALSE;
}
normal_mode:
{
if (ffmpegdec->context->hurry_up != 0) {
ffmpegdec->context->hurry_up = 0;
*mode_switch = TRUE;
GST_DEBUG_OBJECT (ffmpegdec, "QOS: normal mode %g < 0.4", proportion);
}
return TRUE;
}
skip_to_keyframe:
{
ffmpegdec->context->hurry_up = 1;
ffmpegdec->waiting_for_key = TRUE;
*mode_switch = TRUE;
GST_DEBUG_OBJECT (ffmpegdec,
"QOS: keyframe, %" G_GINT64_FORMAT " > GST_SECOND/2", diff);
/* we can skip the current frame */
return FALSE;
}
hurry_up:
{
if (ffmpegdec->context->hurry_up != 1) {
ffmpegdec->context->hurry_up = 1;
*mode_switch = TRUE;
GST_DEBUG_OBJECT (ffmpegdec,
"QOS: hurry up, diff %" G_GINT64_FORMAT " >= 0", diff);
}
return TRUE;
}
}
/* returns TRUE if buffer is within segment, else FALSE.
* if Buffer is on segment border, it's timestamp and duration will be clipped */
static gboolean
clip_video_buffer (GstFFMpegDec * dec, GstBuffer * buf, GstClockTime in_ts,
GstClockTime in_dur)
{
gboolean res = TRUE;
gint64 cstart, cstop;
GstClockTime stop;
GST_LOG_OBJECT (dec,
"timestamp:%" GST_TIME_FORMAT " , duration:%" GST_TIME_FORMAT,
GST_TIME_ARGS (in_ts), GST_TIME_ARGS (in_dur));
/* can't clip without TIME segment */
if (G_UNLIKELY (dec->segment.format != GST_FORMAT_TIME))
goto beach;
/* we need a start time */
if (G_UNLIKELY (!GST_CLOCK_TIME_IS_VALID (in_ts)))
goto beach;
/* generate valid stop, if duration unknown, we have unknown stop */
stop =
GST_CLOCK_TIME_IS_VALID (in_dur) ? (in_ts + in_dur) : GST_CLOCK_TIME_NONE;
/* now clip */
res =
gst_segment_clip (&dec->segment, GST_FORMAT_TIME, in_ts, stop, &cstart,
&cstop);
if (G_UNLIKELY (!res))
goto beach;
/* we're pretty sure the duration of this buffer is not till the end of this
* segment (which _clip will assume when the stop is -1) */
if (stop == GST_CLOCK_TIME_NONE)
cstop = GST_CLOCK_TIME_NONE;
/* update timestamp and possibly duration if the clipped stop time is
* valid */
GST_BUFFER_TIMESTAMP (buf) = cstart;
if (GST_CLOCK_TIME_IS_VALID (cstop))
GST_BUFFER_DURATION (buf) = cstop - cstart;
GST_LOG_OBJECT (dec,
"clipped timestamp:%" GST_TIME_FORMAT " , duration:%" GST_TIME_FORMAT,
GST_TIME_ARGS (cstart), GST_TIME_ARGS (GST_BUFFER_DURATION (buf)));
beach:
GST_LOG_OBJECT (dec, "%sdropping", (res ? "not " : ""));
return res;
}
/* figure out if the current picture is a keyframe, return TRUE if that is
* the case. */
static gboolean
check_keyframe (GstFFMpegDec * ffmpegdec)
{
GstFFMpegDecClass *oclass;
gboolean is_itype = FALSE;
gboolean is_reference = FALSE;
gboolean iskeyframe;
/* figure out if we are dealing with a keyframe */
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
/* remember that we have B frames, we need this for the DTS -> PTS conversion
* code */
if (!ffmpegdec->has_b_frames && ffmpegdec->picture->pict_type == FF_B_TYPE) {
GST_DEBUG_OBJECT (ffmpegdec, "we have B frames");
ffmpegdec->has_b_frames = TRUE;
}
is_itype = (ffmpegdec->picture->pict_type == FF_I_TYPE);
is_reference = (ffmpegdec->picture->reference == 1);
iskeyframe = (is_itype || is_reference || ffmpegdec->picture->key_frame)
|| (oclass->in_plugin->id == CODEC_ID_INDEO3)
|| (oclass->in_plugin->id == CODEC_ID_MSZH)
|| (oclass->in_plugin->id == CODEC_ID_ZLIB)
|| (oclass->in_plugin->id == CODEC_ID_VP3)
|| (oclass->in_plugin->id == CODEC_ID_HUFFYUV);
GST_LOG_OBJECT (ffmpegdec,
"current picture: type: %d, is_keyframe:%d, is_itype:%d, is_reference:%d",
ffmpegdec->picture->pict_type, iskeyframe, is_itype, is_reference);
return iskeyframe;
}
/* get an outbuf buffer with the current picture */
static GstFlowReturn
get_output_buffer (GstFFMpegDec * ffmpegdec, GstBuffer ** outbuf)
{
GstFlowReturn ret;
ret = GST_FLOW_OK;
*outbuf = NULL;
if (ffmpegdec->picture->opaque != NULL) {
/* we allocated a picture already for ffmpeg to decode into, let's pick it
* up and use it now. */
*outbuf = (GstBuffer *) ffmpegdec->picture->opaque;
GST_LOG_OBJECT (ffmpegdec, "using opaque buffer %p", *outbuf);
#ifndef EXTRA_REF
gst_buffer_ref (*outbuf);
#endif
} else {
AVPicture pic, *outpic;
gint width, height;
GST_LOG_OBJECT (ffmpegdec, "get output buffer");
/* figure out size of output buffer, this is the clipped output size because
* we will copy the picture into it. */
if ((width = ffmpegdec->format.video.clip_width) == -1)
width = ffmpegdec->context->width;
if ((height = ffmpegdec->format.video.clip_height) == -1)
height = ffmpegdec->context->height;
GST_LOG_OBJECT (ffmpegdec, "clip width %d/height %d", width, height);
ret = alloc_output_buffer (ffmpegdec, outbuf, width, height);
if (G_UNLIKELY (ret != GST_FLOW_OK))
goto alloc_failed;
/* original ffmpeg code does not handle odd sizes correctly.
* This patched up version does */
gst_ffmpeg_avpicture_fill (&pic, GST_BUFFER_DATA (*outbuf),
ffmpegdec->context->pix_fmt, width, height);
outpic = (AVPicture *) ffmpegdec->picture;
GST_LOG_OBJECT (ffmpegdec, "linsize %d %d %d", outpic->linesize[0],
outpic->linesize[1], outpic->linesize[2]);
GST_LOG_OBJECT (ffmpegdec, "data %d %d %d", 0,
outpic->data[1] - outpic->data[0], outpic->data[2] - outpic->data[0]);
av_picture_copy (&pic, outpic, ffmpegdec->context->pix_fmt, width, height);
}
ffmpegdec->picture->pts = -1;
return ret;
/* special cases */
alloc_failed:
{
GST_DEBUG_OBJECT (ffmpegdec, "pad_alloc failed");
return ret;
}
}
static void
clear_queued (GstFFMpegDec * ffmpegdec)
{
g_list_foreach (ffmpegdec->queued, (GFunc) gst_mini_object_unref, NULL);
g_list_free (ffmpegdec->queued);
ffmpegdec->queued = NULL;
}
static GstFlowReturn
flush_queued (GstFFMpegDec * ffmpegdec)
{
GstFlowReturn res = GST_FLOW_OK;
while (ffmpegdec->queued) {
GstBuffer *buf = GST_BUFFER_CAST (ffmpegdec->queued->data);
GST_LOG_OBJECT (ffmpegdec, "pushing buffer %p, offset %"
G_GINT64_FORMAT ", timestamp %"
GST_TIME_FORMAT ", duration %" GST_TIME_FORMAT, buf,
GST_BUFFER_OFFSET (buf),
GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (buf)),
GST_TIME_ARGS (GST_BUFFER_DURATION (buf)));
/* iterate ouput queue an push downstream */
res = gst_pad_push (ffmpegdec->srcpad, buf);
ffmpegdec->queued =
g_list_delete_link (ffmpegdec->queued, ffmpegdec->queued);
}
return res;
}
/* gst_ffmpegdec_[video|audio]_frame:
* ffmpegdec:
* data: pointer to the data to decode
* size: size of data in bytes
* in_timestamp: incoming timestamp.
* in_duration: incoming duration.
* in_offset: incoming offset (frame number).
* outbuf: outgoing buffer. Different from NULL ONLY if it contains decoded data.
* ret: Return flow.
*
* Returns: number of bytes used in decoding. The check for successful decode is
* outbuf being non-NULL.
*/
static gint
gst_ffmpegdec_video_frame (GstFFMpegDec * ffmpegdec,
guint8 * data, guint size,
GstClockTime in_timestamp, GstClockTime in_duration,
gint64 in_offset, GstBuffer ** outbuf, GstFlowReturn * ret)
{
gint len = -1;
gint have_data;
gboolean iskeyframe;
gboolean mode_switch;
gboolean decode;
gint hurry_up = 0;
GstClockTime out_timestamp, out_duration, out_pts;
gint64 out_offset;
*ret = GST_FLOW_OK;
*outbuf = NULL;
ffmpegdec->context->opaque = ffmpegdec;
/* in case we skip frames */
ffmpegdec->picture->pict_type = -1;
/* run QoS code, we don't stop decoding the frame when we are late because
* else we might skip a reference frame */
decode = gst_ffmpegdec_do_qos (ffmpegdec, in_timestamp, &mode_switch);
if (ffmpegdec->is_realvideo && data != NULL) {
gint slice_count;
gint i;
/* setup the slice table for realvideo */
if (ffmpegdec->context->slice_offset == NULL)
ffmpegdec->context->slice_offset = g_malloc (sizeof (guint32) * 1000);
slice_count = (*data++) + 1;
ffmpegdec->context->slice_count = slice_count;
for (i = 0; i < slice_count; i++) {
data += 4;
ffmpegdec->context->slice_offset[i] = GST_READ_UINT32_LE (data);
data += 4;
}
}
if (!decode) {
/* no decoding needed, save previous hurry_up value and brutely skip
* decoding everything */
hurry_up = ffmpegdec->context->hurry_up;
ffmpegdec->context->hurry_up = 2;
}
/* now decode the frame */
len = avcodec_decode_video (ffmpegdec->context,
ffmpegdec->picture, &have_data, data, size);
/* restore previous state */
if (!decode)
ffmpegdec->context->hurry_up = hurry_up;
GST_DEBUG_OBJECT (ffmpegdec, "after decode: len %d, have_data %d",
len, have_data);
/* when we are in hurry_up mode, don't complain when ffmpeg returned
* no data because we told it to skip stuff. */
if (len < 0 && (mode_switch || ffmpegdec->context->hurry_up))
len = 0;
/* no data, we're done */
if (len < 0 || have_data <= 0)
goto beach;
/* get pts */
out_pts = ffmpegdec->picture->pts;
GST_DEBUG_OBJECT (ffmpegdec, "picture: pts %" G_GUINT64_FORMAT, out_pts);
GST_DEBUG_OBJECT (ffmpegdec, "picture: num %d",
ffmpegdec->picture->coded_picture_number);
GST_DEBUG_OBJECT (ffmpegdec, "picture: ref %d",
ffmpegdec->picture->reference);
GST_DEBUG_OBJECT (ffmpegdec, "picture: display %d",
ffmpegdec->picture->display_picture_number);
GST_DEBUG_OBJECT (ffmpegdec, "picture: opaque %p",
ffmpegdec->picture->opaque);
/* check if we are dealing with a keyframe here, this will also check if we
* are dealing with B frames. */
iskeyframe = check_keyframe (ffmpegdec);
/* check that the timestamps go upwards */
if (ffmpegdec->last_out != -1) {
if (ffmpegdec->last_out > out_pts) {
/* timestamps go backwards, this means frames were reordered and we must
* be dealing with DTS as the buffer timestamps */
ffmpegdec->ts_is_dts = TRUE;
GST_DEBUG_OBJECT (ffmpegdec,
"timestamp discont, we have DTS as timestamps");
}
}
if (out_pts == 0 && out_pts == ffmpegdec->last_out) {
GST_LOG_OBJECT (ffmpegdec, "ffmpeg returns 0 timestamps, ignoring");
/* some codecs only output 0 timestamps, when that happens, make us select an
* output timestamp based on the input timestamp. We do this by making the
* ffmpeg timestamp and the interpollated next timestamp invalid. */
out_pts = -1;
ffmpegdec->next_ts = -1;
} else
ffmpegdec->last_out = out_pts;
if (ffmpegdec->ts_is_dts) {
/* we are dealing with DTS as the timestamps, only copy the DTS on the picture
* to the PTS of the output frame if we are dealing with a non-reference
* frame, else we leave the timestamp as -1, which will interpollate from the
* last outputted value. */
if (ffmpegdec->context->has_b_frames && ffmpegdec->has_b_frames &&
ffmpegdec->picture->reference && ffmpegdec->next_ts != -1) {
/* we have b frames and this picture is a reference picture, don't use the
* DTS as the PTS */
GST_DEBUG_OBJECT (ffmpegdec, "DTS as timestamps, interpollate");
out_pts = -1;
}
}
/* when we're waiting for a keyframe, see if we have one or drop the current
* non-keyframe */
if (G_UNLIKELY (ffmpegdec->waiting_for_key)) {
if (G_LIKELY (!iskeyframe))
goto drop_non_keyframe;
/* we have a keyframe, we can stop waiting for one */
ffmpegdec->waiting_for_key = FALSE;
}
/* get a handle to the output buffer */
*ret = get_output_buffer (ffmpegdec, outbuf);
if (G_UNLIKELY (*ret != GST_FLOW_OK))
goto no_output;
/*
* Timestamps:
*
* 1) Copy picture timestamp if valid
* 2) else interpolate from previous output timestamp
* 3) else copy input timestamp
*/
out_timestamp = -1;
if (out_pts != -1) {
/* Get (interpolated) timestamp from FFMPEG */
out_timestamp = (GstClockTime) out_pts;
GST_LOG_OBJECT (ffmpegdec, "using timestamp %" GST_TIME_FORMAT
" returned by ffmpeg", GST_TIME_ARGS (out_timestamp));
}
if (!GST_CLOCK_TIME_IS_VALID (out_timestamp) && ffmpegdec->next_ts != -1) {
out_timestamp = ffmpegdec->next_ts;
GST_LOG_OBJECT (ffmpegdec, "using next timestamp %" GST_TIME_FORMAT,
GST_TIME_ARGS (out_timestamp));
}
if (!GST_CLOCK_TIME_IS_VALID (out_timestamp)) {
out_timestamp = in_timestamp;
GST_LOG_OBJECT (ffmpegdec, "using in timestamp %" GST_TIME_FORMAT,
GST_TIME_ARGS (out_timestamp));
}
GST_BUFFER_TIMESTAMP (*outbuf) = out_timestamp;
/*
* Offset:
*
* 1) Use input offset if valid
*/
if (in_offset == GST_BUFFER_OFFSET_NONE) {
GST_LOG_OBJECT (ffmpegdec, "no valid offset found");
out_offset = GST_BUFFER_OFFSET_NONE;
} else {
GST_LOG_OBJECT (ffmpegdec, "using in_offset %" G_GINT64_FORMAT, in_offset);
out_offset = in_offset;
}
GST_BUFFER_OFFSET (*outbuf) = out_offset;
/*
* Duration:
*
* 1) Copy input duration if valid
* 2) else use input framerate
* 3) else use ffmpeg framerate
*/
out_duration = -1;
if (!GST_CLOCK_TIME_IS_VALID (in_duration)) {
/* if we have an input framerate, use that */
if (ffmpegdec->format.video.fps_n != -1 &&
(ffmpegdec->format.video.fps_n != 1000 &&
ffmpegdec->format.video.fps_d != 1)) {
GST_LOG_OBJECT (ffmpegdec, "using input framerate for duration");
out_duration = gst_util_uint64_scale_int (GST_SECOND,
ffmpegdec->format.video.fps_d, ffmpegdec->format.video.fps_n);
} else {
/* don't try to use the decoder's framerate when it seems a bit abnormal,
* which we assume when den >= 1000... */
if (ffmpegdec->context->time_base.num != 0 &&
(ffmpegdec->context->time_base.den > 0 &&
ffmpegdec->context->time_base.den < 1000)) {
GST_LOG_OBJECT (ffmpegdec, "using decoder's framerate for duration");
out_duration = gst_util_uint64_scale_int (GST_SECOND,
ffmpegdec->context->time_base.num /
ffmpegdec->context->ticks_per_frame,
ffmpegdec->context->time_base.den);
} else {
GST_LOG_OBJECT (ffmpegdec, "no valid duration found");
}
}
} else {
GST_LOG_OBJECT (ffmpegdec, "using in_duration");
out_duration = in_duration;
}
/* Take repeat_pict into account */
if (GST_CLOCK_TIME_IS_VALID (out_duration)) {
out_duration += out_duration * ffmpegdec->picture->repeat_pict / 2;
}
GST_BUFFER_DURATION (*outbuf) = out_duration;
if (out_timestamp != -1 && out_duration != -1)
ffmpegdec->next_ts = out_timestamp + out_duration;
/* palette is not part of raw video frame in gst and the size
* of the outgoing buffer needs to be adjusted accordingly */
if (ffmpegdec->context->palctrl != NULL)
GST_BUFFER_SIZE (*outbuf) -= AVPALETTE_SIZE;
/* now see if we need to clip the buffer against the segment boundaries. */
if (G_UNLIKELY (!clip_video_buffer (ffmpegdec, *outbuf, out_timestamp,
out_duration)))
goto clipped;
/* mark as keyframe or delta unit */
if (!iskeyframe)
GST_BUFFER_FLAG_SET (*outbuf, GST_BUFFER_FLAG_DELTA_UNIT);
beach:
GST_DEBUG_OBJECT (ffmpegdec, "return flow %d, out %p, len %d",
*ret, *outbuf, len);
return len;
/* special cases */
drop_non_keyframe:
{
GST_WARNING_OBJECT (ffmpegdec, "Dropping non-keyframe (seek/init)");
goto beach;
}
no_output:
{
GST_DEBUG_OBJECT (ffmpegdec, "no output buffer");
len = -1;
goto beach;
}
clipped:
{
GST_DEBUG_OBJECT (ffmpegdec, "buffer clipped");
gst_buffer_unref (*outbuf);
*outbuf = NULL;
goto beach;
}
}
/* returns TRUE if buffer is within segment, else FALSE.
* if Buffer is on segment border, it's timestamp and duration will be clipped */
static gboolean
clip_audio_buffer (GstFFMpegDec * dec, GstBuffer * buf, GstClockTime in_ts,
GstClockTime in_dur)
{
GstClockTime stop;
gint64 diff, ctime, cstop;
gboolean res = TRUE;
GST_LOG_OBJECT (dec,
"timestamp:%" GST_TIME_FORMAT ", duration:%" GST_TIME_FORMAT
", size %u", GST_TIME_ARGS (in_ts), GST_TIME_ARGS (in_dur),
GST_BUFFER_SIZE (buf));
/* can't clip without TIME segment */
if (G_UNLIKELY (dec->segment.format != GST_FORMAT_TIME))
goto beach;
/* we need a start time */
if (G_UNLIKELY (!GST_CLOCK_TIME_IS_VALID (in_ts)))
goto beach;
/* trust duration */
stop = in_ts + in_dur;
res = gst_segment_clip (&dec->segment, GST_FORMAT_TIME, in_ts, stop, &ctime,
&cstop);
if (G_UNLIKELY (!res))
goto out_of_segment;
/* see if some clipping happened */
if (G_UNLIKELY ((diff = ctime - in_ts) > 0)) {
/* bring clipped time to bytes */
diff =
gst_util_uint64_scale_int (diff, dec->format.audio.samplerate,
GST_SECOND) * (dec->format.audio.depth * dec->format.audio.channels);
GST_DEBUG_OBJECT (dec, "clipping start to %" GST_TIME_FORMAT " %"
G_GINT64_FORMAT " bytes", GST_TIME_ARGS (ctime), diff);
GST_BUFFER_SIZE (buf) -= diff;
GST_BUFFER_DATA (buf) += diff;
}
if (G_UNLIKELY ((diff = stop - cstop) > 0)) {
/* bring clipped time to bytes */
diff =
gst_util_uint64_scale_int (diff, dec->format.audio.samplerate,
GST_SECOND) * (dec->format.audio.depth * dec->format.audio.channels);
GST_DEBUG_OBJECT (dec, "clipping stop to %" GST_TIME_FORMAT " %"
G_GINT64_FORMAT " bytes", GST_TIME_ARGS (cstop), diff);
GST_BUFFER_SIZE (buf) -= diff;
}
GST_BUFFER_TIMESTAMP (buf) = ctime;
GST_BUFFER_DURATION (buf) = cstop - ctime;
beach:
GST_LOG_OBJECT (dec, "%sdropping", (res ? "not " : ""));
return res;
/* ERRORS */
out_of_segment:
{
GST_LOG_OBJECT (dec, "out of segment");
goto beach;
}
}
static gint
gst_ffmpegdec_audio_frame (GstFFMpegDec * ffmpegdec,
AVCodec * in_plugin, guint8 * data, guint size,
GstClockTime in_timestamp, GstClockTime in_duration,
gint64 in_offset, GstBuffer ** outbuf, GstFlowReturn * ret)
{
gint len = -1;
gint have_data = AVCODEC_MAX_AUDIO_FRAME_SIZE;
GST_DEBUG_OBJECT (ffmpegdec,
"size:%d, offset:%" G_GINT64_FORMAT ", ts:%" GST_TIME_FORMAT ", dur:%"
GST_TIME_FORMAT ", ffmpegdec->next_ts:%" GST_TIME_FORMAT, size,
in_offset, GST_TIME_ARGS (in_timestamp), GST_TIME_ARGS (in_duration),
GST_TIME_ARGS (ffmpegdec->next_ts));
/* outgoing buffer. We use av_malloc() to have properly aligned memory. */
*outbuf = gst_buffer_new ();
GST_BUFFER_DATA (*outbuf) = GST_BUFFER_MALLOCDATA (*outbuf) =
av_malloc (AVCODEC_MAX_AUDIO_FRAME_SIZE);
GST_BUFFER_SIZE (*outbuf) = AVCODEC_MAX_AUDIO_FRAME_SIZE;
GST_BUFFER_FREE_FUNC (*outbuf) = av_free;
len = avcodec_decode_audio2 (ffmpegdec->context,
(int16_t *) GST_BUFFER_DATA (*outbuf), &have_data, data, size);
GST_DEBUG_OBJECT (ffmpegdec,
"Decode audio: len=%d, have_data=%d", len, have_data);
if (len >= 0 && have_data > 0) {
GST_DEBUG_OBJECT (ffmpegdec, "Creating output buffer");
if (!gst_ffmpegdec_negotiate (ffmpegdec)) {
gst_buffer_unref (*outbuf);
*outbuf = NULL;
len = -1;
goto beach;
}
/* Buffer size */
GST_BUFFER_SIZE (*outbuf) = have_data;
/*
* Timestamps:
*
* 1) Copy input timestamp if valid
* 2) else interpolate from previous input timestamp
*/
/* always take timestamps from the input buffer if any */
if (!GST_CLOCK_TIME_IS_VALID (in_timestamp)) {
in_timestamp = ffmpegdec->next_ts;
}
/*
* Duration:
*
* 1) calculate based on number of samples
*/
in_duration = gst_util_uint64_scale (have_data, GST_SECOND,
ffmpegdec->format.audio.depth * ffmpegdec->format.audio.channels *
ffmpegdec->format.audio.samplerate);
GST_DEBUG_OBJECT (ffmpegdec,
"Buffer created. Size:%d , timestamp:%" GST_TIME_FORMAT " , duration:%"
GST_TIME_FORMAT, have_data,
GST_TIME_ARGS (in_timestamp), GST_TIME_ARGS (in_duration));
GST_BUFFER_TIMESTAMP (*outbuf) = in_timestamp;
GST_BUFFER_DURATION (*outbuf) = in_duration;
GST_BUFFER_OFFSET (*outbuf) = in_offset;
/* the next timestamp we'll use when interpolating */
ffmpegdec->next_ts = in_timestamp + in_duration;
/* now see if we need to clip the buffer against the segment boundaries. */
if (G_UNLIKELY (!clip_audio_buffer (ffmpegdec, *outbuf, in_timestamp,
in_duration)))
goto clipped;
} else if (len > 0 && have_data == 0) {
/* cache output, because it may be used for caching (in-place) */
*outbuf = NULL;
} else {
gst_buffer_unref (*outbuf);
*outbuf = NULL;
}
/* If we don't error out after the first failed read with the AAC decoder,
* we must *not* carry on pushing data, else we'll cause segfaults... */
if (len == -1 && in_plugin->id == CODEC_ID_AAC) {
GST_ELEMENT_ERROR (ffmpegdec, STREAM, DECODE, (NULL),
("Decoding of AAC stream by FFMPEG failed."));
*ret = GST_FLOW_ERROR;
}
beach:
GST_DEBUG_OBJECT (ffmpegdec, "return flow %d, out %p, len %d",
*ret, *outbuf, len);
return len;
/* ERRORS */
clipped:
{
GST_DEBUG_OBJECT (ffmpegdec, "buffer clipped");
gst_buffer_unref (*outbuf);
*outbuf = NULL;
goto beach;
}
}
/* gst_ffmpegdec_frame:
* ffmpegdec:
* data: pointer to the data to decode
* size: size of data in bytes
* got_data: 0 if no data was decoded, != 0 otherwise.
* in_time: timestamp of data
* in_duration: duration of data
* ret: GstFlowReturn to return in the chain function
*
* Decode the given frame and pushes it downstream.
*
* Returns: Number of bytes used in decoding, -1 on error/failure.
*/
static gint
gst_ffmpegdec_frame (GstFFMpegDec * ffmpegdec,
guint8 * data, guint size, gint * got_data,
GstClockTime in_timestamp, GstClockTime in_duration, gint64 in_offset,
GstFlowReturn * ret)
{
GstFFMpegDecClass *oclass;
GstBuffer *outbuf = NULL;
gint have_data = 0, len = 0;
if (G_UNLIKELY (ffmpegdec->context->codec == NULL))
goto no_codec;
GST_LOG_OBJECT (ffmpegdec,
"data:%p, size:%d, offset:%" G_GINT64_FORMAT ", ts:%" GST_TIME_FORMAT
", dur:%" GST_TIME_FORMAT, data, size, in_offset,
GST_TIME_ARGS (in_timestamp), GST_TIME_ARGS (in_duration));
*ret = GST_FLOW_OK;
ffmpegdec->context->frame_number++;
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
switch (oclass->in_plugin->type) {
case CODEC_TYPE_VIDEO:
len =
gst_ffmpegdec_video_frame (ffmpegdec, data, size, in_timestamp,
in_duration, in_offset, &outbuf, ret);
break;
case CODEC_TYPE_AUDIO:
len =
gst_ffmpegdec_audio_frame (ffmpegdec, oclass->in_plugin, data, size,
in_timestamp, in_duration, in_offset, &outbuf, ret);
/* if we did not get an output buffer and we have a pending discont, don't
* clear the input timestamps, we will put them on the next buffer because
* else we might create the first buffer with a very big timestamp gap. */
if (outbuf == NULL && ffmpegdec->discont) {
GST_DEBUG_OBJECT (ffmpegdec, "no buffer but keeping timestamp");
ffmpegdec->clear_ts = FALSE;
}
break;
default:
GST_ERROR_OBJECT (ffmpegdec, "Asked to decode non-audio/video frame !");
g_assert_not_reached ();
break;
}
if (outbuf)
have_data = 1;
if (len < 0 || have_data < 0) {
GST_WARNING_OBJECT (ffmpegdec,
"ffdec_%s: decoding error (len: %d, have_data: %d)",
oclass->in_plugin->name, len, have_data);
*got_data = 0;
goto beach;
} else if (len == 0 && have_data == 0) {
*got_data = 0;
goto beach;
} else {
/* this is where I lost my last clue on ffmpeg... */
*got_data = 1;
}
if (outbuf) {
GST_LOG_OBJECT (ffmpegdec,
"Decoded data, now pushing buffer %p with offset %" G_GINT64_FORMAT
", timestamp %" GST_TIME_FORMAT " and duration %" GST_TIME_FORMAT,
outbuf, GST_BUFFER_OFFSET (outbuf),
GST_TIME_ARGS (GST_BUFFER_TIMESTAMP (outbuf)),
GST_TIME_ARGS (GST_BUFFER_DURATION (outbuf)));
/* mark pending discont */
if (ffmpegdec->discont) {
GST_BUFFER_FLAG_SET (outbuf, GST_BUFFER_FLAG_DISCONT);
ffmpegdec->discont = FALSE;
}
/* set caps */
gst_buffer_set_caps (outbuf, GST_PAD_CAPS (ffmpegdec->srcpad));
if (ffmpegdec->segment.rate > 0.0) {
/* and off we go */
*ret = gst_pad_push (ffmpegdec->srcpad, outbuf);
} else {
/* reverse playback, queue frame till later when we get a discont. */
GST_DEBUG_OBJECT (ffmpegdec, "queued frame");
ffmpegdec->queued = g_list_prepend (ffmpegdec->queued, outbuf);
*ret = GST_FLOW_OK;
}
} else {
GST_DEBUG_OBJECT (ffmpegdec, "We didn't get a decoded buffer");
}
beach:
return len;
/* ERRORS */
no_codec:
{
GST_ERROR_OBJECT (ffmpegdec, "no codec context");
return -1;
}
}
static void
gst_ffmpegdec_drain (GstFFMpegDec * ffmpegdec)
{
GstFFMpegDecClass *oclass;
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
if (oclass->in_plugin->capabilities & CODEC_CAP_DELAY) {
gint have_data, len, try = 0;
GST_LOG_OBJECT (ffmpegdec,
"codec has delay capabilities, calling until ffmpeg has drained everything");
do {
GstFlowReturn ret;
len = gst_ffmpegdec_frame (ffmpegdec, NULL, 0, &have_data,
GST_CLOCK_TIME_NONE, GST_CLOCK_TIME_NONE, -1, &ret);
if (len < 0 || have_data == 0)
break;
} while (try++ < 10);
}
if (ffmpegdec->segment.rate < 0.0) {
/* if we have some queued frames for reverse playback, flush them now */
flush_queued (ffmpegdec);
}
}
static void
gst_ffmpegdec_flush_pcache (GstFFMpegDec * ffmpegdec)
{
if (ffmpegdec->pcache) {
gst_buffer_unref (ffmpegdec->pcache);
ffmpegdec->pcache = NULL;
}
if (ffmpegdec->pctx) {
GstFFMpegDecClass *oclass;
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
av_parser_close (ffmpegdec->pctx);
ffmpegdec->pctx = av_parser_init (oclass->in_plugin->id);
}
}
static gboolean
gst_ffmpegdec_sink_event (GstPad * pad, GstEvent * event)
{
GstFFMpegDec *ffmpegdec;
GstFFMpegDecClass *oclass;
gboolean ret = FALSE;
ffmpegdec = (GstFFMpegDec *) gst_pad_get_parent (pad);
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
GST_DEBUG_OBJECT (ffmpegdec, "Handling %s event",
GST_EVENT_TYPE_NAME (event));
switch (GST_EVENT_TYPE (event)) {
case GST_EVENT_EOS:
{
gst_ffmpegdec_drain (ffmpegdec);
break;
}
case GST_EVENT_FLUSH_STOP:
{
if (ffmpegdec->opened) {
avcodec_flush_buffers (ffmpegdec->context);
}
ffmpegdec->last_out = GST_CLOCK_TIME_NONE;
ffmpegdec->next_ts = GST_CLOCK_TIME_NONE;
gst_ffmpegdec_reset_qos (ffmpegdec);
gst_ffmpegdec_flush_pcache (ffmpegdec);
ffmpegdec->waiting_for_key = TRUE;
gst_segment_init (&ffmpegdec->segment, GST_FORMAT_TIME);
clear_queued (ffmpegdec);
break;
}
case GST_EVENT_NEWSEGMENT:
{
gboolean update;
GstFormat fmt;
gint64 start, stop, time;
gdouble rate, arate;
gst_event_parse_new_segment_full (event, &update, &rate, &arate, &fmt,
&start, &stop, &time);
switch (fmt) {
case GST_FORMAT_TIME:
/* fine, our native segment format */
break;
case GST_FORMAT_BYTES:
{
gint bit_rate;
bit_rate = ffmpegdec->context->bit_rate;
/* convert to time or fail */
if (!bit_rate)
goto no_bitrate;
GST_DEBUG_OBJECT (ffmpegdec, "bitrate: %d", bit_rate);
/* convert values to TIME */
if (start != -1)
start = gst_util_uint64_scale_int (start, GST_SECOND, bit_rate);
if (stop != -1)
stop = gst_util_uint64_scale_int (stop, GST_SECOND, bit_rate);
if (time != -1)
time = gst_util_uint64_scale_int (time, GST_SECOND, bit_rate);
/* unref old event */
gst_event_unref (event);
/* create new converted time segment */
fmt = GST_FORMAT_TIME;
/* FIXME, bitrate is not good enough too find a good stop, let's
* hope start and time were 0... meh. */
stop = -1;
event = gst_event_new_new_segment (update, rate, fmt,
start, stop, time);
break;
}
default:
/* invalid format */
goto invalid_format;
}
/* drain pending frames before trying to use the new segment, queued
* buffers belonged to the previous segment. */
if (ffmpegdec->context->codec)
gst_ffmpegdec_drain (ffmpegdec);
GST_DEBUG_OBJECT (ffmpegdec,
"NEWSEGMENT in time start %" GST_TIME_FORMAT " -- stop %"
GST_TIME_FORMAT, GST_TIME_ARGS (start), GST_TIME_ARGS (stop));
/* and store the values */
gst_segment_set_newsegment_full (&ffmpegdec->segment, update,
rate, arate, fmt, start, stop, time);
break;
}
default:
break;
}
/* and push segment downstream */
ret = gst_pad_push_event (ffmpegdec->srcpad, event);
done:
gst_object_unref (ffmpegdec);
return ret;
/* ERRORS */
no_bitrate:
{
GST_WARNING_OBJECT (ffmpegdec, "no bitrate to convert BYTES to TIME");
gst_event_unref (event);
goto done;
}
invalid_format:
{
GST_WARNING_OBJECT (ffmpegdec, "unknown format received in NEWSEGMENT");
gst_event_unref (event);
goto done;
}
}
static GstFlowReturn
gst_ffmpegdec_chain (GstPad * pad, GstBuffer * inbuf)
{
GstFFMpegDec *ffmpegdec;
GstFFMpegDecClass *oclass;
guint8 *data, *bdata, *pdata;
gint size, bsize, len, have_data;
GstFlowReturn ret = GST_FLOW_OK;
GstClockTime in_timestamp, in_duration;
gboolean discont;
gint64 in_offset;
ffmpegdec = (GstFFMpegDec *) (GST_PAD_PARENT (pad));
if (G_UNLIKELY (!ffmpegdec->opened))
goto not_negotiated;
discont = GST_BUFFER_IS_DISCONT (inbuf);
/* The discont flags marks a buffer that is not continuous with the previous
* buffer. This means we need to clear whatever data we currently have. We
* currently also wait for a new keyframe, which might be suboptimal in the
* case of a network error, better show the errors than to drop all data.. */
if (G_UNLIKELY (discont)) {
GST_DEBUG_OBJECT (ffmpegdec, "received DISCONT");
/* drain what we have queued */
gst_ffmpegdec_drain (ffmpegdec);
gst_ffmpegdec_flush_pcache (ffmpegdec);
avcodec_flush_buffers (ffmpegdec->context);
ffmpegdec->waiting_for_key = TRUE;
ffmpegdec->discont = TRUE;
ffmpegdec->last_out = GST_CLOCK_TIME_NONE;
ffmpegdec->next_ts = GST_CLOCK_TIME_NONE;
}
/* by default we clear the input timestamp after decoding each frame so that
* interpollation can work. */
ffmpegdec->clear_ts = TRUE;
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
/* do early keyframe check pretty bad to rely on the keyframe flag in the
* source for this as it might not even be parsed (UDP/file/..). */
if (G_UNLIKELY (ffmpegdec->waiting_for_key)) {
if (GST_BUFFER_FLAG_IS_SET (inbuf, GST_BUFFER_FLAG_DELTA_UNIT) &&
oclass->in_plugin->type != CODEC_TYPE_AUDIO)
goto skip_keyframe;
GST_DEBUG_OBJECT (ffmpegdec, "got keyframe");
ffmpegdec->waiting_for_key = FALSE;
}
in_timestamp = GST_BUFFER_TIMESTAMP (inbuf);
in_duration = GST_BUFFER_DURATION (inbuf);
in_offset = GST_BUFFER_OFFSET (inbuf);
GST_LOG_OBJECT (ffmpegdec,
"Received new data of size %d, offset:%" G_GINT64_FORMAT ", ts:%"
GST_TIME_FORMAT ", dur:%" GST_TIME_FORMAT, GST_BUFFER_OFFSET (inbuf),
GST_BUFFER_SIZE (inbuf), GST_TIME_ARGS (in_timestamp),
GST_TIME_ARGS (in_duration));
/* parse cache joining. If there is cached data, its timestamp will be what we
* send to the parse. */
if (ffmpegdec->pcache) {
/* use timestamp and duration of what is in the cache */
in_timestamp = GST_BUFFER_TIMESTAMP (ffmpegdec->pcache);
in_duration = GST_BUFFER_DURATION (ffmpegdec->pcache);
in_offset = GST_BUFFER_OFFSET (ffmpegdec->pcache);
/* join with previous data */
inbuf = gst_buffer_join (ffmpegdec->pcache, inbuf);
GST_LOG_OBJECT (ffmpegdec,
"joined parse cache, inbuf now has offset %" G_GINT64_FORMAT ", ts:%"
GST_TIME_FORMAT, in_offset, GST_TIME_ARGS (in_timestamp));
/* no more cached data, we assume we can consume the complete cache */
ffmpegdec->pcache = NULL;
}
/* workarounds, functions write to buffers:
* libavcodec/svq1.c:svq1_decode_frame writes to the given buffer.
* libavcodec/svq3.c:svq3_decode_slice_header too.
* ffmpeg devs know about it and will fix it (they said). */
if (oclass->in_plugin->id == CODEC_ID_SVQ1 ||
oclass->in_plugin->id == CODEC_ID_SVQ3) {
inbuf = gst_buffer_make_writable (inbuf);
}
bdata = GST_BUFFER_DATA (inbuf);
bsize = GST_BUFFER_SIZE (inbuf);
do {
/* parse, if at all possible */
if (ffmpegdec->pctx) {
gint res;
GST_LOG_OBJECT (ffmpegdec,
"Calling av_parser_parse with offset %" G_GINT64_FORMAT ", ts:%"
GST_TIME_FORMAT, in_offset, GST_TIME_ARGS (in_timestamp));
/* feed the parser. We store the raw gstreamer timestamp because
* converting it to ffmpeg timestamps can corrupt it if the framerate is
* wrong. */
res = av_parser_parse (ffmpegdec->pctx, ffmpegdec->context,
&data, &size, bdata, bsize, in_timestamp, in_timestamp);
GST_LOG_OBJECT (ffmpegdec,
"parser returned res %d and size %d", res, size);
/* store pts for get_buffer */
ffmpegdec->in_ts = ffmpegdec->pctx->pts;
GST_LOG_OBJECT (ffmpegdec, "consuming %d bytes. ts:%"
GST_TIME_FORMAT, size, GST_TIME_ARGS (ffmpegdec->pctx->pts));
if (res) {
/* there is output, set pointers for next round. */
bsize -= res;
bdata += res;
} else {
/* Parser did not consume any data, make sure we don't clear the
* timestamp for the next round */
ffmpegdec->clear_ts = FALSE;
}
/* if there is no output, we must break and wait for more data. also the
* timestamp in the context is not updated. */
if (size == 0) {
if (bsize > 0)
continue;
else
break;
}
} else {
data = bdata;
size = bsize;
ffmpegdec->in_ts = in_timestamp;
ffmpegdec->in_offset = in_offset;
}
if (ffmpegdec->do_padding) {
/* add padding */
if (ffmpegdec->padded_size <= size + FF_INPUT_BUFFER_PADDING_SIZE) {
ffmpegdec->padded_size = size + FF_INPUT_BUFFER_PADDING_SIZE;
ffmpegdec->padded =
g_realloc (ffmpegdec->padded, ffmpegdec->padded_size);
GST_LOG_OBJECT (ffmpegdec, "resized padding buffer to %d",
ffmpegdec->padded_size);
}
memcpy (ffmpegdec->padded, data, size);
memset (ffmpegdec->padded + size, 0, FF_INPUT_BUFFER_PADDING_SIZE);
pdata = ffmpegdec->padded;
} else {
pdata = data;
}
/* decode a frame of audio/video now */
len =
gst_ffmpegdec_frame (ffmpegdec, pdata, size, &have_data, in_timestamp,
in_duration, in_offset, &ret);
if (ret != GST_FLOW_OK) {
GST_LOG_OBJECT (ffmpegdec, "breaking because of flow ret %s",
gst_flow_get_name (ret));
/* bad flow retun, make sure we discard all data and exit */
bsize = 0;
break;
}
if (!ffmpegdec->pctx) {
if (len == 0 && !have_data) {
/* nothing was decoded, this could be because no data was available or
* because we were skipping frames.
* If we have no context we must exit and wait for more data, we keep the
* data we tried. */
GST_LOG_OBJECT (ffmpegdec, "Decoding didn't return any data, breaking");
break;
} else if (len < 0) {
/* a decoding error happened, we must break and try again with next data. */
GST_LOG_OBJECT (ffmpegdec, "Decoding error, breaking");
bsize = 0;
break;
}
/* prepare for the next round, for codecs with a context we did this
* already when using the parser. */
bsize -= len;
bdata += len;
} else {
if (len == 0) {
/* nothing was decoded, this could be because no data was available or
* because we were skipping frames. Since we have a parser we can
* continue with the next frame */
GST_LOG_OBJECT (ffmpegdec,
"Decoding didn't return any data, trying next");
} else if (len < 0) {
/* we have a context that will bring us to the next frame */
GST_LOG_OBJECT (ffmpegdec, "Decoding error, trying next");
}
}
/* make sure we don't use the same old timestamp for the next frame and let
* the interpollation take care of it. */
if (ffmpegdec->clear_ts) {
in_timestamp = GST_CLOCK_TIME_NONE;
in_duration = GST_CLOCK_TIME_NONE;
in_offset = GST_BUFFER_OFFSET_NONE;
} else {
ffmpegdec->clear_ts = TRUE;
}
GST_LOG_OBJECT (ffmpegdec, "Before (while bsize>0). bsize:%d , bdata:%p",
bsize, bdata);
} while (bsize > 0);
/* keep left-over */
if (ffmpegdec->pctx && bsize > 0) {
in_timestamp = GST_BUFFER_TIMESTAMP (inbuf);
in_offset = GST_BUFFER_OFFSET (inbuf);
GST_LOG_OBJECT (ffmpegdec,
"Keeping %d bytes of data with offset %" G_GINT64_FORMAT ", timestamp %"
GST_TIME_FORMAT, bsize, in_offset, GST_TIME_ARGS (in_timestamp));
ffmpegdec->pcache = gst_buffer_create_sub (inbuf,
GST_BUFFER_SIZE (inbuf) - bsize, bsize);
/* we keep timestamp, even though all we really know is that the correct
* timestamp is not below the one from inbuf */
GST_BUFFER_TIMESTAMP (ffmpegdec->pcache) = in_timestamp;
GST_BUFFER_OFFSET (ffmpegdec->pcache) = in_offset;
} else if (bsize > 0) {
GST_DEBUG_OBJECT (ffmpegdec, "Dropping %d bytes of data", bsize);
}
gst_buffer_unref (inbuf);
return ret;
/* ERRORS */
not_negotiated:
{
oclass = (GstFFMpegDecClass *) (G_OBJECT_GET_CLASS (ffmpegdec));
GST_ELEMENT_ERROR (ffmpegdec, CORE, NEGOTIATION, (NULL),
("ffdec_%s: input format was not set before data start",
oclass->in_plugin->name));
gst_buffer_unref (inbuf);
return GST_FLOW_NOT_NEGOTIATED;
}
skip_keyframe:
{
GST_DEBUG_OBJECT (ffmpegdec, "skipping non keyframe");
gst_buffer_unref (inbuf);
return GST_FLOW_OK;
}
}
static GstStateChangeReturn
gst_ffmpegdec_change_state (GstElement * element, GstStateChange transition)
{
GstFFMpegDec *ffmpegdec = (GstFFMpegDec *) element;
GstStateChangeReturn ret;
ret = GST_ELEMENT_CLASS (parent_class)->change_state (element, transition);
switch (transition) {
case GST_STATE_CHANGE_PAUSED_TO_READY:
GST_OBJECT_LOCK (ffmpegdec);
gst_ffmpegdec_close (ffmpegdec);
GST_OBJECT_UNLOCK (ffmpegdec);
clear_queued (ffmpegdec);
g_free (ffmpegdec->padded);
ffmpegdec->padded = NULL;
ffmpegdec->padded_size = 0;
break;
default:
break;
}
return ret;
}
static void
gst_ffmpegdec_set_property (GObject * object,
guint prop_id, const GValue * value, GParamSpec * pspec)
{
GstFFMpegDec *ffmpegdec = (GstFFMpegDec *) object;
switch (prop_id) {
case PROP_LOWRES:
ffmpegdec->lowres = ffmpegdec->context->lowres = g_value_get_enum (value);
break;
case PROP_SKIPFRAME:
ffmpegdec->hurry_up = ffmpegdec->context->hurry_up =
g_value_get_enum (value);
break;
case PROP_DIRECT_RENDERING:
ffmpegdec->direct_rendering = g_value_get_boolean (value);
break;
case PROP_DO_PADDING:
ffmpegdec->do_padding = g_value_get_boolean (value);
break;
case PROP_DEBUG_MV:
ffmpegdec->debug_mv = ffmpegdec->context->debug_mv =
g_value_get_boolean (value);
break;
case PROP_CROP:
ffmpegdec->crop = g_value_get_boolean (value);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
static void
gst_ffmpegdec_get_property (GObject * object,
guint prop_id, GValue * value, GParamSpec * pspec)
{
GstFFMpegDec *ffmpegdec = (GstFFMpegDec *) object;
switch (prop_id) {
case PROP_LOWRES:
g_value_set_enum (value, ffmpegdec->context->lowres);
break;
case PROP_SKIPFRAME:
g_value_set_enum (value, ffmpegdec->context->hurry_up);
break;
case PROP_DIRECT_RENDERING:
g_value_set_boolean (value, ffmpegdec->direct_rendering);
break;
case PROP_DO_PADDING:
g_value_set_boolean (value, ffmpegdec->do_padding);
break;
case PROP_DEBUG_MV:
g_value_set_boolean (value, ffmpegdec->context->debug_mv);
break;
case PROP_CROP:
g_value_set_boolean (value, ffmpegdec->crop);
break;
default:
G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
break;
}
}
gboolean
gst_ffmpegdec_register (GstPlugin * plugin)
{
GTypeInfo typeinfo = {
sizeof (GstFFMpegDecClass),
(GBaseInitFunc) gst_ffmpegdec_base_init,
NULL,
(GClassInitFunc) gst_ffmpegdec_class_init,
NULL,
NULL,
sizeof (GstFFMpegDec),
0,
(GInstanceInitFunc) gst_ffmpegdec_init,
};
GType type;
AVCodec *in_plugin;
gint rank;
in_plugin = av_codec_next (NULL);
GST_LOG ("Registering decoders");
while (in_plugin) {
GstFFMpegDecClassParams *params;
GstCaps *srccaps = NULL, *sinkcaps = NULL;
gchar *type_name;
gchar *plugin_name;
/* only decoders */
if (!in_plugin->decode) {
goto next;
}
/* no quasi-codecs, please */
if (in_plugin->id == CODEC_ID_RAWVIDEO ||
(in_plugin->id >= CODEC_ID_PCM_S16LE &&
in_plugin->id <= CODEC_ID_PCM_F64LE)) {
goto next;
}
/* No decoders depending on external libraries (we don't build them, but
* people who build against an external ffmpeg might have them.
* We have native gstreamer plugins for all of those libraries anyway. */
if (!strncmp (in_plugin->name, "lib", 3)) {
GST_DEBUG
("Not using external library decoder %s. Use the gstreamer-native ones instead.",
in_plugin->name);
goto next;
}
/* No vdpau plugins until we can figure out how to properly use them
* outside of ffmpeg. */
if (!strcmp (in_plugin->name, "h264_vdpau") ||
!strcmp (in_plugin->name, "mpeg_vdpau") ||
!strcmp (in_plugin->name, "mpeg1_vdpau") ||
!strcmp (in_plugin->name, "vc1_vdpau") ||
!strcmp (in_plugin->name, "wmv3_vdpau")) {
GST_DEBUG
("Ignoring VDPAU decoder %s. We can't handle this outside of ffmpeg",
in_plugin->name);
goto next;
}
GST_DEBUG ("Trying plugin %s [%s]", in_plugin->name, in_plugin->long_name);
/* no codecs for which we're GUARANTEED to have better alternatives */
/* MPEG1VIDEO : the mpeg2video decoder is preferred */
/* MP1 : Use MP3 for decoding */
/* MP2 : Use MP3 for decoding */
if (!strcmp (in_plugin->name, "gif") ||
!strcmp (in_plugin->name, "vorbis") ||
!strcmp (in_plugin->name, "mpeg1video") ||
!strcmp (in_plugin->name, "wavpack") ||
!strcmp (in_plugin->name, "mp1") ||
!strcmp (in_plugin->name, "mp2") ||
!strcmp (in_plugin->name, "libfaad") ||
!strcmp (in_plugin->name, "mpeg4aac")) {
GST_LOG ("Ignoring decoder %s", in_plugin->name);
goto next;
}
/* first make sure we've got a supported type */
sinkcaps = gst_ffmpeg_codecid_to_caps (in_plugin->id, NULL, FALSE);
if (!sinkcaps) {
GST_WARNING ("Couldn't get sink caps for decoder '%s'", in_plugin->name);
goto next;
}
if (in_plugin->type == CODEC_TYPE_VIDEO) {
srccaps = gst_caps_from_string ("video/x-raw-rgb; video/x-raw-yuv");
} else {
srccaps = gst_ffmpeg_codectype_to_audio_caps (NULL,
in_plugin->id, FALSE, in_plugin);
}
if (!srccaps) {
GST_WARNING ("Couldn't get source caps for decoder %s", in_plugin->name);
goto next;
}
/* construct the type */
plugin_name = g_strdup ((gchar *) in_plugin->name);
g_strdelimit (plugin_name, NULL, '_');
type_name = g_strdup_printf ("ffdec_%s", plugin_name);
g_free (plugin_name);
/* if it's already registered, drop it */
if (g_type_from_name (type_name)) {
g_free (type_name);
goto next;
}
params = g_new0 (GstFFMpegDecClassParams, 1);
params->in_plugin = in_plugin;
params->srccaps = gst_caps_ref (srccaps);
params->sinkcaps = gst_caps_ref (sinkcaps);
/* create the gtype now */
type = g_type_register_static (GST_TYPE_ELEMENT, type_name, &typeinfo, 0);
g_type_set_qdata (type, GST_FFDEC_PARAMS_QDATA, (gpointer) params);
/* (Ronald) MPEG-4 gets a higher priority because it has been well-
* tested and by far outperforms divxdec/xviddec - so we prefer it.
* msmpeg4v3 same, as it outperforms divxdec for divx3 playback.
* VC1/WMV3 are not working and thus unpreferred for now. */
switch (in_plugin->id) {
case CODEC_ID_MPEG4:
case CODEC_ID_MSMPEG4V3:
case CODEC_ID_H264:
case CODEC_ID_RA_144:
case CODEC_ID_RA_288:
case CODEC_ID_RV10:
case CODEC_ID_RV20:
case CODEC_ID_RV30:
case CODEC_ID_RV40:
case CODEC_ID_COOK:
rank = GST_RANK_PRIMARY;
break;
case CODEC_ID_DVVIDEO:
/* we have a good dv decoder, fast on both ppc as well as x86. they say
libdv's quality is better though. leave as secondary.
note: if you change this, see the code in gstdv.c in good/ext/dv. */
rank = GST_RANK_SECONDARY;
break;
case CODEC_ID_AAC:
/* The ffmpeg AAC decoder isn't complete, and there's no way to figure out
* before decoding whether it will support the given stream or not.
* We therefore set it to NONE until it can handle the full specs. */
rank = GST_RANK_NONE;
break;
default:
rank = GST_RANK_MARGINAL;
break;
}
if (!gst_element_register (plugin, type_name, rank, type)) {
g_warning ("Failed to register %s", type_name);
g_free (type_name);
return FALSE;
}
g_free (type_name);
next:
if (sinkcaps)
gst_caps_unref (sinkcaps);
if (srccaps)
gst_caps_unref (srccaps);
in_plugin = av_codec_next (in_plugin);
}
GST_LOG ("Finished Registering decoders");
return TRUE;
}
Reference in a new issue View git blame Copy permalink