gstreamer/gst/avi/gstavidemux.c

5420 lines
159 KiB
C
Raw Normal View History

/* GStreamer
* Copyright (C) <1999> Erik Walthinsen <omega@temple-baptist.com>
* Copyright (C) <2006> Nokia Corporation (contact <stefan.kost@nokia.com>)
2009-12-04 14:13:12 +00:00
* Copyright (C) <2009-2010> STEricsson <benjamin.gaignard@stericsson.com>
*
* 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.
*/
/* Element-Checklist-Version: 5 */
/**
* SECTION:element-avidemux
*
* Demuxes an .avi file into raw or compressed audio and/or video streams.
*
* This element supports both push and pull-based scheduling, depending on the
* capabilities of the upstream elements.
*
* <refsect2>
* <title>Example launch line</title>
* |[
* gst-launch filesrc location=test.avi ! avidemux name=demux demux.audio_00 ! decodebin ! audioconvert ! audioresample ! autoaudiosink demux.video_00 ! queue ! decodebin ! ffmpegcolorspace ! videoscale ! autovideosink
* ]| Play (parse and decode) an .avi file and try to output it to
* an automatically detected soundcard and videosink. If the AVI file contains
* compressed audio or video data, this will only work if you have the
* right decoder elements/plugins installed.
* </refsect2>
*
* Last reviewed on 2006-12-29 (0.10.6)
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
#include <string.h>
2010-01-13 14:44:58 +00:00
#include <stdio.h>
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
#include "gst/riff/riff-media.h"
#include "gstavidemux.h"
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
#include "avi-ids.h"
#include <gst/gst-i18n-plugin.h>
#include <gst/base/gstadapter.h>
2009-09-22 10:35:30 +00:00
#define DIV_ROUND_UP(s,v) (((s) + ((v)-1)) / (v))
2009-09-22 10:35:30 +00:00
#define GST_AVI_KEYFRAME 1
#define ENTRY_IS_KEYFRAME(e) ((e)->flags == GST_AVI_KEYFRAME)
#define ENTRY_SET_KEYFRAME(e) ((e)->flags = GST_AVI_KEYFRAME)
#define ENTRY_UNSET_KEYFRAME(e) ((e)->flags = 0)
GST_DEBUG_CATEGORY_STATIC (avidemux_debug);
#define GST_CAT_DEFAULT avidemux_debug
static GstStaticPadTemplate sink_templ = GST_STATIC_PAD_TEMPLATE ("sink",
GST_PAD_SINK,
GST_PAD_ALWAYS,
GST_STATIC_CAPS ("video/x-msvideo")
);
static void gst_avi_demux_finalize (GObject * object);
static void gst_avi_demux_reset (GstAviDemux * avi);
#if 0
static const GstEventMask *gst_avi_demux_get_event_mask (GstPad * pad);
#endif
static gboolean gst_avi_demux_handle_src_event (GstPad * pad, GstEvent * event);
static gboolean gst_avi_demux_handle_sink_event (GstPad * pad,
GstEvent * event);
static gboolean gst_avi_demux_push_event (GstAviDemux * avi, GstEvent * event);
#if 0
static const GstFormat *gst_avi_demux_get_src_formats (GstPad * pad);
#endif
static const GstQueryType *gst_avi_demux_get_src_query_types (GstPad * pad);
static gboolean gst_avi_demux_handle_src_query (GstPad * pad, GstQuery * query);
static gboolean gst_avi_demux_src_convert (GstPad * pad, GstFormat src_format,
gint64 src_value, GstFormat * dest_format, gint64 * dest_value);
static gboolean gst_avi_demux_do_seek (GstAviDemux * avi, GstSegment * segment);
static gboolean gst_avi_demux_handle_seek (GstAviDemux * avi, GstPad * pad,
GstEvent * event);
static gboolean gst_avi_demux_handle_seek_push (GstAviDemux * avi, GstPad * pad,
GstEvent * event);
static void gst_avi_demux_loop (GstPad * pad);
static gboolean gst_avi_demux_sink_activate (GstPad * sinkpad);
static gboolean gst_avi_demux_sink_activate_pull (GstPad * sinkpad,
gboolean active);
static gboolean gst_avi_demux_activate_push (GstPad * pad, gboolean active);
static GstFlowReturn gst_avi_demux_chain (GstPad * pad, GstBuffer * buf);
2009-10-14 11:08:47 +00:00
static void gst_avi_demux_set_index (GstElement * element, GstIndex * index);
static GstIndex *gst_avi_demux_get_index (GstElement * element);
static GstStateChangeReturn gst_avi_demux_change_state (GstElement * element,
GstStateChange transition);
2009-12-04 14:13:12 +00:00
static void gst_avi_demux_calculate_durations_from_index (GstAviDemux * avi);
static void gst_avi_demux_get_buffer_info (GstAviDemux * avi,
GstAviStream * stream, guint entry_n, GstClockTime * timestamp,
GstClockTime * ts_end, guint64 * offset, guint64 * offset_end);
static void gst_avi_demux_parse_idit (GstAviDemux * avi, GstBuffer * buf);
/* GObject methods */
2011-04-19 15:20:19 +00:00
#define gst_avi_demux_parent_class parent_class
G_DEFINE_TYPE (GstAviDemux, gst_avi_demux, GST_TYPE_ELEMENT);
static void
2011-04-19 15:20:19 +00:00
gst_avi_demux_class_init (GstAviDemuxClass * klass)
{
2011-04-19 15:20:19 +00:00
GstElementClass *gstelement_class = GST_ELEMENT_CLASS (klass);
GObjectClass *gobject_class = (GObjectClass *) klass;
GstPadTemplate *videosrctempl, *audiosrctempl, *subsrctempl;
GstCaps *audcaps, *vidcaps, *subcaps;
2011-04-19 15:20:19 +00:00
GST_DEBUG_CATEGORY_INIT (avidemux_debug, "avidemux",
0, "Demuxer for AVI streams");
gobject_class->finalize = gst_avi_demux_finalize;
gstelement_class->change_state =
GST_DEBUG_FUNCPTR (gst_avi_demux_change_state);
gstelement_class->set_index = GST_DEBUG_FUNCPTR (gst_avi_demux_set_index);
gstelement_class->get_index = GST_DEBUG_FUNCPTR (gst_avi_demux_get_index);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
audcaps = gst_riff_create_audio_template_caps ();
gst_caps_append (audcaps, gst_caps_new_simple ("audio/x-avi-unknown", NULL));
audiosrctempl = gst_pad_template_new ("audio_%02d",
GST_PAD_SRC, GST_PAD_SOMETIMES, audcaps);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
vidcaps = gst_riff_create_video_template_caps ();
gst_caps_append (vidcaps, gst_riff_create_iavs_template_caps ());
gst_caps_append (vidcaps, gst_caps_new_simple ("video/x-avi-unknown", NULL));
videosrctempl = gst_pad_template_new ("video_%02d",
GST_PAD_SRC, GST_PAD_SOMETIMES, vidcaps);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
subcaps = gst_caps_new_simple ("application/x-subtitle-avi", NULL);
subsrctempl = gst_pad_template_new ("subtitle_%02d",
GST_PAD_SRC, GST_PAD_SOMETIMES, subcaps);
2011-04-19 15:20:19 +00:00
gst_element_class_add_pad_template (gstelement_class, audiosrctempl);
gst_element_class_add_pad_template (gstelement_class, videosrctempl);
gst_element_class_add_pad_template (gstelement_class, subsrctempl);
gst_element_class_add_pad_template (gstelement_class,
gst_static_pad_template_get (&sink_templ));
2011-04-19 15:20:19 +00:00
gst_element_class_set_details_simple (gstelement_class, "Avi demuxer",
"Codec/Demuxer",
"Demultiplex an avi file into audio and video",
"Erik Walthinsen <omega@cse.ogi.edu>, "
"Wim Taymans <wim.taymans@chello.be>, "
"Thijs Vermeir <thijsvermeir@gmail.com>");
}
static void
gst_avi_demux_init (GstAviDemux * avi)
{
Fix memleak with gst_static_pad_template_get(). Original commit message from CVS: * ext/cairo/gsttextoverlay.c: (gst_text_overlay_init): * ext/dv/gstdvdemux.c: (gst_dvdemux_init), (gst_dvdemux_add_pads): * ext/gdk_pixbuf/gstgdkpixbuf.c: (gst_gdk_pixbuf_init): * ext/jpeg/gstjpegdec.c: (gst_jpeg_dec_init), (gst_jpeg_dec_setcaps): * ext/jpeg/gstjpegenc.c: (gst_jpegenc_init): * ext/jpeg/gstsmokedec.c: (gst_smokedec_init): * ext/jpeg/gstsmokeenc.c: (gst_smokeenc_init): * ext/libmng/gstmngdec.c: (gst_mngdec_init), (gst_mngdec_src_getcaps): * ext/libpng/gstpngdec.c: (gst_pngdec_init), (gst_pngdec_caps_create_and_set): * ext/libpng/gstpngenc.c: (gst_pngenc_init): * ext/mikmod/gstmikmod.c: (gst_mikmod_init): * ext/speex/gstspeexdec.c: (gst_speex_dec_init): * gst/alpha/gstalpha.c: (gst_alpha_init): * gst/auparse/gstauparse.c: (gst_au_parse_init): * gst/avi/gstavidemux.c: (gst_avi_demux_init), (gst_avi_demux_handle_src_event), (gst_avi_demux_parse_stream): * gst/cutter/gstcutter.c: (gst_cutter_init): * gst/debug/efence.c: (gst_efence_init), (gst_efence_getrange), (gst_efence_checkgetrange): * gst/debug/negotiation.c: (gst_negotiation_init): * gst/flx/gstflxdec.c: (gst_flxdec_init): * gst/goom/gstgoom.c: (gst_goom_init): * gst/rtp/gstasteriskh263.c: (gst_asteriskh263_init): * gst/rtp/gstrtpL16depay.c: (gst_rtp_L16depay_init): * gst/rtp/gstrtpL16pay.c: (gst_rtpL16pay_init): * gst/rtp/gstrtpamrdepay.c: (gst_rtp_amr_depay_init): * gst/rtp/gstrtpdepay.c: (gst_rtp_depay_init): * gst/rtp/gstrtpmpadepay.c: (gst_rtp_mpa_depay_init): * gst/rtsp/gstrtpdec.c: (gst_rtpdec_init): * gst/smpte/gstsmpte.c: (gst_smpte_init): * gst/wavparse/gstwavparse.c: (gst_wavparse_init), (gst_wavparse_create_sourcepad): Fix memleak with gst_static_pad_template_get(). This uses gst_pad_new_from_static_template() instead. Fixes #333512
2006-03-15 16:17:12 +00:00
avi->sinkpad = gst_pad_new_from_static_template (&sink_templ, "sink");
gst_pad_set_activate_function (avi->sinkpad,
GST_DEBUG_FUNCPTR (gst_avi_demux_sink_activate));
gst_pad_set_activatepull_function (avi->sinkpad,
GST_DEBUG_FUNCPTR (gst_avi_demux_sink_activate_pull));
gst_pad_set_activatepush_function (avi->sinkpad,
GST_DEBUG_FUNCPTR (gst_avi_demux_activate_push));
gst_pad_set_chain_function (avi->sinkpad,
GST_DEBUG_FUNCPTR (gst_avi_demux_chain));
gst_pad_set_event_function (avi->sinkpad,
GST_DEBUG_FUNCPTR (gst_avi_demux_handle_sink_event));
2010-01-15 17:13:24 +00:00
gst_element_add_pad (GST_ELEMENT_CAST (avi), avi->sinkpad);
avi->adapter = gst_adapter_new ();
gst_avi_demux_reset (avi);
}
static void
gst_avi_demux_finalize (GObject * object)
{
GstAviDemux *avi = GST_AVI_DEMUX (object);
GST_DEBUG ("AVI: finalize");
g_object_unref (avi->adapter);
G_OBJECT_CLASS (parent_class)->finalize (object);
}
static void
gst_avi_demux_reset_stream (GstAviDemux * avi, GstAviStream * stream)
{
g_free (stream->strh);
g_free (stream->strf.data);
g_free (stream->name);
g_free (stream->index);
g_free (stream->indexes);
if (stream->initdata)
gst_buffer_unref (stream->initdata);
if (stream->extradata)
gst_buffer_unref (stream->extradata);
if (stream->pad) {
if (stream->exposed) {
gst_pad_set_active (stream->pad, FALSE);
2010-01-15 17:13:24 +00:00
gst_element_remove_pad (GST_ELEMENT_CAST (avi), stream->pad);
} else
gst_object_unref (stream->pad);
}
if (stream->taglist) {
gst_tag_list_free (stream->taglist);
stream->taglist = NULL;
}
memset (stream, 0, sizeof (GstAviStream));
}
static void
gst_avi_demux_reset (GstAviDemux * avi)
{
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
gint i;
GST_DEBUG ("AVI: reset");
for (i = 0; i < avi->num_streams; i++)
gst_avi_demux_reset_stream (avi, &avi->stream[i]);
avi->header_state = GST_AVI_DEMUX_HEADER_TAG_LIST;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
avi->num_streams = 0;
avi->num_v_streams = 0;
avi->num_a_streams = 0;
avi->num_t_streams = 0;
avi->main_stream = -1;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
avi->state = GST_AVI_DEMUX_START;
avi->offset = 0;
avi->building_index = FALSE;
avi->index_offset = 0;
g_free (avi->avih);
avi->avih = NULL;
2009-10-14 11:08:47 +00:00
if (avi->element_index)
gst_object_unref (avi->element_index);
avi->element_index = NULL;
if (avi->seg_event) {
gst_event_unref (avi->seg_event);
avi->seg_event = NULL;
}
if (avi->seek_event) {
gst_event_unref (avi->seek_event);
avi->seek_event = NULL;
}
if (avi->globaltags)
gst_tag_list_free (avi->globaltags);
avi->globaltags = NULL;
2009-06-18 22:36:28 +00:00
avi->got_tags = TRUE; /* we always want to push global tags */
avi->have_eos = FALSE;
2009-10-14 11:08:47 +00:00
avi->seekable = TRUE;
gst_adapter_clear (avi->adapter);
gst_segment_init (&avi->segment, GST_FORMAT_TIME);
}
/* GstElement methods */
#if 0
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static const GstFormat *
gst_avi_demux_get_src_formats (GstPad * pad)
{
GstAviStream *stream = gst_pad_get_element_private (pad);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static const GstFormat src_a_formats[] = {
GST_FORMAT_TIME,
GST_FORMAT_BYTES,
GST_FORMAT_DEFAULT,
0
};
static const GstFormat src_v_formats[] = {
GST_FORMAT_TIME,
GST_FORMAT_DEFAULT,
0
};
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
return (stream->strh->type == GST_RIFF_FCC_auds ?
src_a_formats : src_v_formats);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
#endif
/* assumes stream->strf.auds->av_bps != 0 */
static inline GstClockTime
avi_stream_convert_bytes_to_time_unchecked (GstAviStream * stream,
guint64 bytes)
{
return gst_util_uint64_scale_int (bytes, GST_SECOND,
stream->strf.auds->av_bps);
}
static inline guint64
avi_stream_convert_time_to_bytes_unchecked (GstAviStream * stream,
GstClockTime time)
{
return gst_util_uint64_scale_int (time, stream->strf.auds->av_bps,
GST_SECOND);
}
/* assumes stream->strh->rate != 0 */
static inline GstClockTime
avi_stream_convert_frames_to_time_unchecked (GstAviStream * stream,
guint64 frames)
{
return gst_util_uint64_scale (frames, stream->strh->scale * GST_SECOND,
stream->strh->rate);
}
static inline guint64
avi_stream_convert_time_to_frames_unchecked (GstAviStream * stream,
GstClockTime time)
{
return gst_util_uint64_scale (time, stream->strh->rate,
stream->strh->scale * GST_SECOND);
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static gboolean
gst_avi_demux_src_convert (GstPad * pad,
GstFormat src_format,
gint64 src_value, GstFormat * dest_format, gint64 * dest_value)
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
{
GstAviStream *stream = gst_pad_get_element_private (pad);
gboolean res = TRUE;
GST_LOG_OBJECT (pad,
"Received src_format:%s, src_value:%" G_GUINT64_FORMAT
", dest_format:%s", gst_format_get_name (src_format), src_value,
gst_format_get_name (*dest_format));
if (G_UNLIKELY (src_format == *dest_format)) {
*dest_value = src_value;
goto done;
}
if (G_UNLIKELY (!stream->strh || !stream->strf.data)) {
res = FALSE;
goto done;
}
if (G_UNLIKELY (stream->strh->type == GST_RIFF_FCC_vids &&
(src_format == GST_FORMAT_BYTES
|| *dest_format == GST_FORMAT_BYTES))) {
res = FALSE;
goto done;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
switch (src_format) {
case GST_FORMAT_TIME:
switch (*dest_format) {
case GST_FORMAT_BYTES:
*dest_value = gst_util_uint64_scale_int (src_value,
stream->strf.auds->av_bps, GST_SECOND);
break;
case GST_FORMAT_DEFAULT:
*dest_value =
gst_util_uint64_scale_round (src_value, stream->strh->rate,
stream->strh->scale * GST_SECOND);
break;
default:
res = FALSE;
break;
}
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
case GST_FORMAT_BYTES:
switch (*dest_format) {
case GST_FORMAT_TIME:
if (stream->strf.auds->av_bps != 0) {
*dest_value = avi_stream_convert_bytes_to_time_unchecked (stream,
src_value);
} else
res = FALSE;
break;
default:
res = FALSE;
break;
}
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
case GST_FORMAT_DEFAULT:
switch (*dest_format) {
case GST_FORMAT_TIME:
*dest_value =
avi_stream_convert_frames_to_time_unchecked (stream, src_value);
break;
default:
res = FALSE;
break;
}
break;
default:
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
res = FALSE;
}
done:
GST_LOG_OBJECT (pad,
"Returning res:%d dest_format:%s dest_value:%" G_GUINT64_FORMAT, res,
gst_format_get_name (*dest_format), *dest_value);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
return res;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static const GstQueryType *
gst_avi_demux_get_src_query_types (GstPad * pad)
{
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static const GstQueryType src_types[] = {
GST_QUERY_POSITION,
GST_QUERY_DURATION,
GST_QUERY_SEEKING,
GST_QUERY_CONVERT,
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
0
};
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
return src_types;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static gboolean
gst_avi_demux_handle_src_query (GstPad * pad, GstQuery * query)
{
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
gboolean res = TRUE;
GstAviDemux *avi = GST_AVI_DEMUX (gst_pad_get_parent (pad));
GstAviStream *stream = gst_pad_get_element_private (pad);
if (!stream->strh || !stream->strf.data)
return gst_pad_query_default (pad, query);
switch (GST_QUERY_TYPE (query)) {
case GST_QUERY_POSITION:{
gint64 pos = 0;
GST_DEBUG ("pos query for stream %u: frames %u, bytes %u",
stream->num, stream->current_entry, stream->current_total);
2009-09-23 12:25:08 +00:00
/* FIXME, this looks clumsy */
if (stream->strh->type == GST_RIFF_FCC_auds) {
if (stream->is_vbr) {
/* VBR */
pos = gst_util_uint64_scale ((gint64) stream->current_entry *
stream->strh->scale, GST_SECOND, (guint64) stream->strh->rate);
GST_DEBUG_OBJECT (avi, "VBR convert frame %u, time %"
GST_TIME_FORMAT, stream->current_entry, GST_TIME_ARGS (pos));
} else if (stream->strf.auds->av_bps != 0) {
/* CBR */
pos = gst_util_uint64_scale (stream->current_total, GST_SECOND,
(guint64) stream->strf.auds->av_bps);
GST_DEBUG_OBJECT (avi,
"CBR convert bytes %u, time %" GST_TIME_FORMAT,
stream->current_total, GST_TIME_ARGS (pos));
} else if (stream->idx_n != 0 && stream->total_bytes != 0) {
/* calculate timestamps based on percentage of length */
guint64 xlen = avi->avih->us_frame *
avi->avih->tot_frames * GST_USECOND;
if (stream->is_vbr) {
pos = gst_util_uint64_scale (xlen, stream->current_entry,
stream->idx_n);
GST_DEBUG_OBJECT (avi, "VBR perc convert frame %u, time %"
GST_TIME_FORMAT, stream->current_entry, GST_TIME_ARGS (pos));
} else {
pos = gst_util_uint64_scale (xlen, stream->current_total,
stream->total_bytes);
GST_DEBUG_OBJECT (avi,
"CBR perc convert bytes %u, time %" GST_TIME_FORMAT,
stream->current_total, GST_TIME_ARGS (pos));
}
} else {
/* we don't know */
res = FALSE;
}
} else {
if (stream->strh->rate != 0) {
pos = gst_util_uint64_scale ((guint64) stream->current_entry *
stream->strh->scale, GST_SECOND, (guint64) stream->strh->rate);
} else {
pos = stream->current_entry * avi->avih->us_frame * GST_USECOND;
}
}
if (res) {
GST_DEBUG ("pos query : %" GST_TIME_FORMAT, GST_TIME_ARGS (pos));
gst_query_set_position (query, GST_FORMAT_TIME, pos);
} else
GST_WARNING ("pos query failed");
break;
}
case GST_QUERY_DURATION:
{
GstFormat fmt;
GstClockTime duration;
/* only act on audio or video streams */
if (stream->strh->type != GST_RIFF_FCC_auds &&
stream->strh->type != GST_RIFF_FCC_vids) {
res = FALSE;
break;
}
/* take stream duration, fall back to avih duration */
if ((duration = stream->duration) == -1)
duration = avi->duration;
gst_query_parse_duration (query, &fmt, NULL);
switch (fmt) {
case GST_FORMAT_TIME:
gst_query_set_duration (query, fmt, duration);
break;
case GST_FORMAT_DEFAULT:
{
gint64 dur;
GST_DEBUG_OBJECT (query, "total frames is %" G_GUINT32_FORMAT,
stream->idx_n);
if (stream->idx_n >= 0)
gst_query_set_duration (query, fmt, stream->idx_n);
else if (gst_pad_query_convert (pad, GST_FORMAT_TIME,
duration, fmt, &dur))
gst_query_set_duration (query, fmt, dur);
break;
}
default:
res = FALSE;
break;
}
break;
}
case GST_QUERY_SEEKING:{
GstFormat fmt;
gst_query_parse_seeking (query, &fmt, NULL, NULL, NULL);
if (fmt == GST_FORMAT_TIME) {
gboolean seekable = TRUE;
if (avi->streaming) {
seekable = avi->seekable;
}
gst_query_set_seeking (query, GST_FORMAT_TIME, seekable,
0, stream->duration);
res = TRUE;
}
break;
}
case GST_QUERY_CONVERT:{
GstFormat src_fmt, dest_fmt;
gint64 src_val, dest_val;
gst_query_parse_convert (query, &src_fmt, &src_val, &dest_fmt, &dest_val);
if ((res = gst_avi_demux_src_convert (pad, src_fmt, src_val, &dest_fmt,
&dest_val)))
gst_query_set_convert (query, src_fmt, src_val, dest_fmt, dest_val);
else
res = gst_pad_query_default (pad, query);
break;
}
default:
res = gst_pad_query_default (pad, query);
break;
}
gst_object_unref (avi);
return res;
}
#if 0
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static const GstEventMask *
gst_avi_demux_get_event_mask (GstPad * pad)
{
static const GstEventMask masks[] = {
{GST_EVENT_SEEK, GST_SEEK_METHOD_SET | GST_SEEK_FLAG_KEY_UNIT},
{0,}
};
return masks;
}
#endif
static guint64
gst_avi_demux_seek_streams (GstAviDemux * avi, guint64 offset, gboolean before)
{
GstAviStream *stream;
GstIndexEntry *entry;
gint i;
gint64 val, min = offset;
for (i = 0; i < avi->num_streams; i++) {
stream = &avi->stream[i];
entry = gst_index_get_assoc_entry (avi->element_index, stream->index_id,
before ? GST_INDEX_LOOKUP_BEFORE : GST_INDEX_LOOKUP_AFTER,
GST_ASSOCIATION_FLAG_NONE, GST_FORMAT_BYTES, offset);
if (before) {
if (entry) {
gst_index_entry_assoc_map (entry, GST_FORMAT_BYTES, &val);
GST_DEBUG_OBJECT (avi, "stream %d, previous entry at %"
G_GUINT64_FORMAT, i, val);
if (val < min)
min = val;
}
continue;
}
if (!entry) {
GST_DEBUG_OBJECT (avi, "no position for stream %d, assuming at start", i);
stream->current_entry = 0;
stream->current_total = 0;
continue;
}
gst_index_entry_assoc_map (entry, GST_FORMAT_BYTES, &val);
GST_DEBUG_OBJECT (avi, "stream %d, next entry at %" G_GUINT64_FORMAT,
i, val);
gst_index_entry_assoc_map (entry, GST_FORMAT_TIME, &val);
stream->current_total = val;
gst_index_entry_assoc_map (entry, GST_FORMAT_DEFAULT, &val);
stream->current_entry = val;
}
return min;
}
2009-12-04 14:13:12 +00:00
static guint
gst_avi_demux_index_entry_offset_search (GstAviIndexEntry * entry,
guint64 * offset)
{
if (entry->offset < *offset)
return -1;
else if (entry->offset > *offset)
return 1;
return 0;
}
static guint64
gst_avi_demux_seek_streams_index (GstAviDemux * avi, guint64 offset,
gboolean before)
{
GstAviStream *stream;
GstAviIndexEntry *entry;
gint i;
gint64 val, min = offset;
guint index = 0;
2009-12-04 14:13:12 +00:00
for (i = 0; i < avi->num_streams; i++) {
stream = &avi->stream[i];
/* compensate for chunk header */
offset += 8;
2009-12-04 14:13:12 +00:00
entry =
gst_util_array_binary_search (stream->index, stream->idx_n,
sizeof (GstAviIndexEntry),
(GCompareDataFunc) gst_avi_demux_index_entry_offset_search,
before ? GST_SEARCH_MODE_BEFORE : GST_SEARCH_MODE_AFTER, &offset, NULL);
offset -= 8;
2009-12-04 14:13:12 +00:00
if (entry)
index = entry - stream->index;
if (before) {
if (entry) {
val = stream->index[index].offset;
2009-12-04 14:13:12 +00:00
GST_DEBUG_OBJECT (avi,
"stream %d, previous entry at %" G_GUINT64_FORMAT, i, val);
if (val < min)
min = val;
}
continue;
}
if (!entry) {
GST_DEBUG_OBJECT (avi, "no position for stream %d, assuming at start", i);
stream->current_entry = 0;
stream->current_total = 0;
continue;
}
val = stream->index[index].offset - 8;
2009-12-04 14:13:12 +00:00
GST_DEBUG_OBJECT (avi, "stream %d, next entry at %" G_GUINT64_FORMAT, i,
val);
stream->current_total = stream->index[index].total;
2009-12-04 14:13:12 +00:00
stream->current_entry = index;
}
return min;
}
#define GST_AVI_SEEK_PUSH_DISPLACE (4 * GST_SECOND)
static gboolean
gst_avi_demux_handle_sink_event (GstPad * pad, GstEvent * event)
{
gboolean res = TRUE;
GstAviDemux *avi = GST_AVI_DEMUX (gst_pad_get_parent (pad));
GST_DEBUG_OBJECT (avi,
"have event type %s: %p on sink pad", GST_EVENT_TYPE_NAME (event), event);
switch (GST_EVENT_TYPE (event)) {
2011-06-02 16:47:36 +00:00
case GST_EVENT_SEGMENT:
{
2011-06-02 16:47:36 +00:00
gint64 boffset, offset = 0;
GstSegment segment;
/* some debug output */
2011-06-02 16:47:36 +00:00
gst_event_copy_segment (event, &segment);
GST_DEBUG_OBJECT (avi, "received newsegment %" GST_SEGMENT_FORMAT,
&segment);
/* chain will send initial newsegment after pads have been added */
if (avi->state != GST_AVI_DEMUX_MOVI) {
GST_DEBUG_OBJECT (avi, "still starting, eating event");
goto exit;
}
/* we only expect a BYTE segment, e.g. following a seek */
2011-06-02 16:47:36 +00:00
if (segment.format != GST_FORMAT_BYTES) {
GST_DEBUG_OBJECT (avi, "unsupported segment format, ignoring");
goto exit;
}
2009-12-04 14:13:12 +00:00
if (avi->have_index) {
GstAviIndexEntry *entry;
guint i = 0, index = 0, k = 0;
GstAviStream *stream;
/* compensate chunk header, stored index offset points after header */
2011-06-02 16:47:36 +00:00
boffset = segment.start + 8;
/* find which stream we're on */
do {
stream = &avi->stream[i];
/* find the index for start bytes offset */
entry = gst_util_array_binary_search (stream->index,
stream->idx_n, sizeof (GstAviIndexEntry),
(GCompareDataFunc) gst_avi_demux_index_entry_offset_search,
2011-06-02 16:47:36 +00:00
GST_SEARCH_MODE_AFTER, &boffset, NULL);
if (entry == NULL)
continue;
index = entry - stream->index;
/* we are on the stream with a chunk start offset closest to start */
if (!offset || stream->index[index].offset < offset) {
offset = stream->index[index].offset;
k = i;
}
/* exact match needs no further searching */
2011-06-02 16:47:36 +00:00
if (stream->index[index].offset == segment.start)
break;
} while (++i < avi->num_streams);
2011-06-02 16:47:36 +00:00
boffset -= 8;
offset -= 8;
stream = &avi->stream[k];
/* so we have no idea what is to come, or where we are */
if (!offset) {
GST_WARNING_OBJECT (avi, "insufficient index data, forcing EOS");
goto eos;
}
/* get the ts corresponding to start offset bytes for the stream */
2009-12-04 14:13:12 +00:00
gst_avi_demux_get_buffer_info (avi, stream, index,
2011-06-02 16:47:36 +00:00
(GstClockTime *) & segment.time, NULL, NULL, NULL);
2009-12-04 14:13:12 +00:00
} else if (avi->element_index) {
GstIndexEntry *entry;
/* Let's check if we have an index entry for this position */
entry = gst_index_get_assoc_entry (avi->element_index, avi->index_id,
GST_INDEX_LOOKUP_AFTER, GST_ASSOCIATION_FLAG_NONE,
2011-06-02 16:47:36 +00:00
GST_FORMAT_BYTES, segment.start);
2009-12-04 14:13:12 +00:00
/* we can not go where we have not yet been before ... */
if (!entry) {
GST_WARNING_OBJECT (avi, "insufficient index data, forcing EOS");
goto eos;
}
2011-06-02 16:47:36 +00:00
gst_index_entry_assoc_map (entry, GST_FORMAT_TIME,
(gint64 *) & segment.time);
gst_index_entry_assoc_map (entry, GST_FORMAT_BYTES, &offset);
2009-12-04 14:13:12 +00:00
} else {
GST_WARNING_OBJECT (avi, "no index data, forcing EOS");
goto eos;
}
2011-06-02 16:47:36 +00:00
segment.format = GST_FORMAT_TIME;
segment.start = segment.time;
segment.stop = GST_CLOCK_TIME_NONE;
segment.position = segment.start;
/* rescue duration */
segment.duration = avi->segment.duration;
/* set up segment and send downstream */
2011-06-02 16:47:36 +00:00
gst_segment_copy_into (&segment, &avi->segment);
GST_DEBUG_OBJECT (avi, "Pushing newseg %" GST_SEGMENT_FORMAT, &segment);
gst_avi_demux_push_event (avi, gst_event_new_segment (&segment));
GST_DEBUG_OBJECT (avi, "next chunk expected at %" G_GINT64_FORMAT,
boffset);
/* adjust state for streaming thread accordingly */
2009-12-04 14:13:12 +00:00
if (avi->have_index)
gst_avi_demux_seek_streams_index (avi, offset, FALSE);
else
gst_avi_demux_seek_streams (avi, offset, FALSE);
/* set up streaming thread */
2011-06-02 16:47:36 +00:00
g_assert (offset >= boffset);
avi->offset = boffset;
avi->todrop = offset - boffset;
exit:
gst_event_unref (event);
res = TRUE;
break;
eos:
/* set up for EOS */
avi->have_eos = TRUE;
goto exit;
}
case GST_EVENT_EOS:
{
if (avi->state != GST_AVI_DEMUX_MOVI) {
gst_event_unref (event);
GST_ELEMENT_ERROR (avi, STREAM, DEMUX,
(NULL), ("got eos and didn't receive a complete header object"));
} else if (!gst_avi_demux_push_event (avi, event)) {
GST_ELEMENT_ERROR (avi, STREAM, DEMUX,
(NULL), ("got eos but no streams (yet)"));
}
break;
}
case GST_EVENT_FLUSH_STOP:
{
gint i;
gst_adapter_clear (avi->adapter);
avi->have_eos = FALSE;
for (i = 0; i < avi->num_streams; i++) {
avi->stream[i].last_flow = GST_FLOW_OK;
avi->stream[i].discont = TRUE;
}
/* fall through to default case so that the event gets passed downstream */
}
default:
res = gst_pad_event_default (pad, event);
break;
}
gst_object_unref (avi);
return res;
}
static gboolean
gst_avi_demux_handle_src_event (GstPad * pad, GstEvent * event)
{
gboolean res = TRUE;
GstAviDemux *avi = GST_AVI_DEMUX (gst_pad_get_parent (pad));
GST_DEBUG_OBJECT (avi,
"have event type %s: %p on src pad", GST_EVENT_TYPE_NAME (event), event);
switch (GST_EVENT_TYPE (event)) {
case GST_EVENT_SEEK:
if (!avi->streaming) {
res = gst_avi_demux_handle_seek (avi, pad, event);
} else {
res = gst_avi_demux_handle_seek_push (avi, pad, event);
}
gst_event_unref (event);
break;
case GST_EVENT_QOS:
case GST_EVENT_NAVIGATION:
res = FALSE;
gst_event_unref (event);
break;
default:
res = gst_pad_event_default (pad, event);
break;
}
gst_object_unref (avi);
return res;
}
/* streaming helper (push) */
/*
* gst_avi_demux_peek_chunk_info:
* @avi: Avi object
* @tag: holder for tag
* @size: holder for tag size
*
* Peek next chunk info (tag and size)
*
* Returns: TRUE when one chunk info has been got
*/
static gboolean
gst_avi_demux_peek_chunk_info (GstAviDemux * avi, guint32 * tag, guint32 * size)
{
const guint8 *data = NULL;
if (gst_adapter_available (avi->adapter) < 8)
return FALSE;
2011-04-04 17:17:43 +00:00
data = gst_adapter_map (avi->adapter, 8);
*tag = GST_READ_UINT32_LE (data);
*size = GST_READ_UINT32_LE (data + 4);
2011-04-04 17:17:43 +00:00
gst_adapter_unmap (avi->adapter, 0);
return TRUE;
}
/*
* gst_avi_demux_peek_chunk:
* @avi: Avi object
* @tag: holder for tag
* @size: holder for tag size
*
* Peek enough data for one full chunk
*
* Returns: %TRUE when one chunk has been got
*/
static gboolean
gst_avi_demux_peek_chunk (GstAviDemux * avi, guint32 * tag, guint32 * size)
{
guint32 peek_size = 0;
gint available;
if (!gst_avi_demux_peek_chunk_info (avi, tag, size))
goto peek_failed;
/* size 0 -> empty data buffer would surprise most callers,
* large size -> do not bother trying to squeeze that into adapter,
* so we throw poor man's exception, which can be caught if caller really
* wants to handle 0 size chunk */
if (!(*size) || (*size) >= (1 << 30))
goto strange_size;
peek_size = (*size + 1) & ~1;
available = gst_adapter_available (avi->adapter);
GST_DEBUG_OBJECT (avi,
"Need to peek chunk of %d bytes to read chunk %" GST_FOURCC_FORMAT
", %d bytes available", *size, GST_FOURCC_ARGS (*tag), available);
if (available < (8 + peek_size))
goto need_more;
return TRUE;
/* ERRORS */
peek_failed:
{
GST_INFO_OBJECT (avi, "Failed to peek");
return FALSE;
}
strange_size:
{
GST_INFO_OBJECT (avi,
"Invalid/unexpected chunk size %d for tag %" GST_FOURCC_FORMAT, *size,
GST_FOURCC_ARGS (*tag));
/* chain should give up */
avi->abort_buffering = TRUE;
return FALSE;
}
need_more:
{
GST_INFO_OBJECT (avi, "need more %d < %" G_GUINT32_FORMAT,
available, 8 + peek_size);
return FALSE;
}
}
/* AVI init */
/*
* gst_avi_demux_parse_file_header:
* @element: caller element (used for errors/debug).
* @buf: input data to be used for parsing.
*
* "Open" a RIFF/AVI file. The buffer should be at least 12
* bytes long. Takes ownership of @buf.
*
* Returns: TRUE if the file is a RIFF/AVI file, FALSE otherwise.
* Throws an error, caller should error out (fatal).
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
*/
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
static gboolean
gst_avi_demux_parse_file_header (GstElement * element, GstBuffer * buf)
{
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
guint32 doctype;
GstClockTime stamp;
stamp = gst_util_get_timestamp ();
/* riff_parse posts an error */
if (!gst_riff_parse_file_header (element, buf, &doctype))
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
return FALSE;
if (doctype != GST_RIFF_RIFF_AVI)
goto not_avi;
stamp = gst_util_get_timestamp () - stamp;
GST_DEBUG_OBJECT (element, "header parsing took %" GST_TIME_FORMAT,
GST_TIME_ARGS (stamp));
return TRUE;
/* ERRORS */
not_avi:
{
GST_ELEMENT_ERROR (element, STREAM, WRONG_TYPE, (NULL),
("File is not an AVI file: %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (doctype)));
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
return FALSE;
}
}
/*
* Read AVI file tag when streaming
*/
static GstFlowReturn
gst_avi_demux_stream_init_push (GstAviDemux * avi)
{
if (gst_adapter_available (avi->adapter) >= 12) {
GstBuffer *tmp;
tmp = gst_adapter_take_buffer (avi->adapter, 12);
GST_DEBUG ("Parsing avi header");
2010-01-15 17:13:24 +00:00
if (!gst_avi_demux_parse_file_header (GST_ELEMENT_CAST (avi), tmp)) {
return GST_FLOW_ERROR;
}
GST_DEBUG ("header ok");
avi->offset += 12;
avi->state = GST_AVI_DEMUX_HEADER;
}
return GST_FLOW_OK;
}
/*
* Read AVI file tag
*/
static GstFlowReturn
gst_avi_demux_stream_init_pull (GstAviDemux * avi)
{
GstFlowReturn res;
GstBuffer *buf = NULL;
res = gst_pad_pull_range (avi->sinkpad, avi->offset, 12, &buf);
if (res != GST_FLOW_OK)
return res;
else if (!gst_avi_demux_parse_file_header (GST_ELEMENT_CAST (avi), buf))
goto wrong_header;
avi->offset += 12;
return GST_FLOW_OK;
/* ERRORS */
wrong_header:
{
GST_DEBUG_OBJECT (avi, "error parsing file header");
return GST_FLOW_ERROR;
}
}
/* AVI header handling */
/*
* gst_avi_demux_parse_avih:
2010-01-15 17:15:14 +00:00
* @avi: caller element (used for errors/debug).
* @buf: input data to be used for parsing.
* @avih: pointer to structure (filled in by function) containing
* stream information (such as flags, number of streams, etc.).
*
* Read 'avih' header. Discards buffer after use.
*
* Returns: TRUE on success, FALSE otherwise. Throws an error if
* the header is invalid. The caller should error out
* (fatal).
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
*/
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
static gboolean
2010-01-15 17:15:14 +00:00
gst_avi_demux_parse_avih (GstAviDemux * avi,
GstBuffer * buf, gst_riff_avih ** _avih)
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
{
gst_riff_avih *avih;
2011-04-04 17:17:43 +00:00
gsize size;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
if (buf == NULL)
goto no_buffer;
2011-04-04 17:17:43 +00:00
size = gst_buffer_get_size (buf);
if (size < sizeof (gst_riff_avih))
goto avih_too_small;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
2011-04-04 17:17:43 +00:00
avih = g_malloc (size);
gst_buffer_extract (buf, 0, avih, size);
#if (G_BYTE_ORDER == G_BIG_ENDIAN)
avih->us_frame = GUINT32_FROM_LE (avih->us_frame);
avih->max_bps = GUINT32_FROM_LE (avih->max_bps);
avih->pad_gran = GUINT32_FROM_LE (avih->pad_gran);
avih->flags = GUINT32_FROM_LE (avih->flags);
avih->tot_frames = GUINT32_FROM_LE (avih->tot_frames);
avih->init_frames = GUINT32_FROM_LE (avih->init_frames);
avih->streams = GUINT32_FROM_LE (avih->streams);
avih->bufsize = GUINT32_FROM_LE (avih->bufsize);
avih->width = GUINT32_FROM_LE (avih->width);
avih->height = GUINT32_FROM_LE (avih->height);
avih->scale = GUINT32_FROM_LE (avih->scale);
avih->rate = GUINT32_FROM_LE (avih->rate);
avih->start = GUINT32_FROM_LE (avih->start);
avih->length = GUINT32_FROM_LE (avih->length);
#endif
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* debug stuff */
2010-01-15 17:15:14 +00:00
GST_INFO_OBJECT (avi, "avih tag found:");
GST_INFO_OBJECT (avi, " us_frame %u", avih->us_frame);
GST_INFO_OBJECT (avi, " max_bps %u", avih->max_bps);
GST_INFO_OBJECT (avi, " pad_gran %u", avih->pad_gran);
GST_INFO_OBJECT (avi, " flags 0x%08x", avih->flags);
GST_INFO_OBJECT (avi, " tot_frames %u", avih->tot_frames);
GST_INFO_OBJECT (avi, " init_frames %u", avih->init_frames);
GST_INFO_OBJECT (avi, " streams %u", avih->streams);
GST_INFO_OBJECT (avi, " bufsize %u", avih->bufsize);
GST_INFO_OBJECT (avi, " width %u", avih->width);
GST_INFO_OBJECT (avi, " height %u", avih->height);
GST_INFO_OBJECT (avi, " scale %u", avih->scale);
GST_INFO_OBJECT (avi, " rate %u", avih->rate);
GST_INFO_OBJECT (avi, " start %u", avih->start);
GST_INFO_OBJECT (avi, " length %u", avih->length);
*_avih = avih;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
gst_buffer_unref (buf);
if (avih->us_frame != 0 && avih->tot_frames != 0)
avi->duration =
(guint64) avih->us_frame * (guint64) avih->tot_frames * 1000;
else
avi->duration = GST_CLOCK_TIME_NONE;
2010-01-15 17:15:14 +00:00
GST_INFO_OBJECT (avi, " header duration %" GST_TIME_FORMAT,
GST_TIME_ARGS (avi->duration));
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
return TRUE;
/* ERRORS */
no_buffer:
{
2010-01-15 17:15:14 +00:00
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL), ("No buffer"));
return FALSE;
}
avih_too_small:
{
2010-01-15 17:15:14 +00:00
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Too small avih (%d available, %d needed)",
2011-04-04 17:17:43 +00:00
size, (int) sizeof (gst_riff_avih)));
gst_buffer_unref (buf);
return FALSE;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
/*
* gst_avi_demux_parse_superindex:
* @avi: caller element (used for debugging/errors).
* @buf: input data to use for parsing.
* @locations: locations in the file (byte-offsets) that contain
* the actual indexes (see get_avi_demux_parse_subindex()).
* The array ends with GST_BUFFER_OFFSET_NONE.
*
* Reads superindex (openDML-2 spec stuff) from the provided data.
*
* Returns: TRUE on success, FALSE otherwise. Indexes should be skipped
* on error, but they are not fatal.
*/
static gboolean
gst_avi_demux_parse_superindex (GstAviDemux * avi,
GstBuffer * buf, guint64 ** _indexes)
{
guint8 *data;
guint16 bpe = 16;
guint32 num, i;
guint64 *indexes;
2011-04-04 17:17:43 +00:00
gsize size;
*_indexes = NULL;
if (buf) {
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (buf, &size, NULL, GST_MAP_READ);
} else {
data = NULL;
2011-04-04 17:17:43 +00:00
size = 0;
}
2011-04-04 17:17:43 +00:00
if (size < 24)
goto too_small;
/* check type of index. The opendml2 specs state that
* there should be 4 dwords per array entry. Type can be
* either frame or field (and we don't care). */
if (GST_READ_UINT16_LE (data) != 4 ||
(data[2] & 0xfe) != 0x0 || data[3] != 0x0) {
GST_WARNING_OBJECT (avi,
"Superindex for stream has unexpected "
"size_entry %d (bytes) or flags 0x%02x/0x%02x",
GST_READ_UINT16_LE (data), data[2], data[3]);
bpe = GST_READ_UINT16_LE (data) * 4;
}
num = GST_READ_UINT32_LE (&data[4]);
GST_DEBUG_OBJECT (avi, "got %d indexes", num);
/* this can't work out well ... */
if (num > G_MAXUINT32 >> 1 || bpe < 8) {
goto invalid_params;
}
indexes = g_new (guint64, num + 1);
for (i = 0; i < num; i++) {
if (size < 24 + bpe * (i + 1))
break;
indexes[i] = GST_READ_UINT64_LE (&data[24 + bpe * i]);
GST_DEBUG_OBJECT (avi, "index %d at %" G_GUINT64_FORMAT, i, indexes[i]);
}
indexes[i] = GST_BUFFER_OFFSET_NONE;
*_indexes = indexes;
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
return TRUE;
/* ERRORS */
too_small:
{
GST_ERROR_OBJECT (avi,
"Not enough data to parse superindex (%d available, 24 needed)", size);
2011-04-04 17:17:43 +00:00
if (buf) {
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
2011-04-04 17:17:43 +00:00
}
return FALSE;
}
invalid_params:
{
GST_ERROR_OBJECT (avi, "invalid index parameters (num = %d, bpe = %d)",
num, bpe);
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
return FALSE;
}
}
/* add an entry to the index of a stream. @num should be an estimate of the
* total amount of index entries for all streams and is used to dynamically
* allocate memory for the index entries. */
static inline gboolean
gst_avi_demux_add_index (GstAviDemux * avi, GstAviStream * stream,
guint num, GstAviIndexEntry * entry)
{
/* ensure index memory */
if (G_UNLIKELY (stream->idx_n >= stream->idx_max)) {
guint idx_max = stream->idx_max;
GstAviIndexEntry *new_idx;
/* we need to make some more room */
if (idx_max == 0) {
/* initial size guess, assume each stream has an equal amount of entries,
* overshoot with at least 8K */
idx_max = (num / avi->num_streams) + (8192 / sizeof (GstAviIndexEntry));
} else {
idx_max += 8192 / sizeof (GstAviIndexEntry);
GST_DEBUG_OBJECT (avi, "expanded index from %u to %u",
stream->idx_max, idx_max);
}
new_idx = g_try_renew (GstAviIndexEntry, stream->index, idx_max);
/* out of memory, if this fails stream->index is untouched. */
if (G_UNLIKELY (!new_idx))
return FALSE;
/* use new index */
stream->index = new_idx;
stream->idx_max = idx_max;
}
/* update entry total and stream stats. The entry total can be converted to
* the timestamp of the entry easily. */
if (stream->strh->type == GST_RIFF_FCC_auds) {
gint blockalign;
if (stream->is_vbr) {
entry->total = stream->total_blocks;
} else {
entry->total = stream->total_bytes;
}
blockalign = stream->strf.auds->blockalign;
if (blockalign > 0)
stream->total_blocks += DIV_ROUND_UP (entry->size, blockalign);
else
stream->total_blocks++;
} else {
if (stream->is_vbr) {
entry->total = stream->idx_n;
} else {
entry->total = stream->total_bytes;
}
}
stream->total_bytes += entry->size;
if (ENTRY_IS_KEYFRAME (entry))
stream->n_keyframes++;
/* and add */
GST_LOG_OBJECT (avi,
"Adding stream %u, index entry %d, kf %d, size %u "
", offset %" G_GUINT64_FORMAT ", total %" G_GUINT64_FORMAT, stream->num,
stream->idx_n, ENTRY_IS_KEYFRAME (entry), entry->size, entry->offset,
entry->total);
stream->index[stream->idx_n++] = *entry;
return TRUE;
}
/* given @entry_n in @stream, calculate info such as timestamps and
* offsets for the entry. */
static void
gst_avi_demux_get_buffer_info (GstAviDemux * avi, GstAviStream * stream,
guint entry_n, GstClockTime * timestamp, GstClockTime * ts_end,
guint64 * offset, guint64 * offset_end)
{
GstAviIndexEntry *entry;
entry = &stream->index[entry_n];
if (stream->is_vbr) {
/* VBR stream next timestamp */
if (stream->strh->type == GST_RIFF_FCC_auds) {
if (timestamp)
*timestamp =
avi_stream_convert_frames_to_time_unchecked (stream, entry->total);
if (ts_end)
*ts_end = avi_stream_convert_frames_to_time_unchecked (stream,
entry->total + 1);
} else {
if (timestamp)
*timestamp =
avi_stream_convert_frames_to_time_unchecked (stream, entry_n);
if (ts_end)
*ts_end = avi_stream_convert_frames_to_time_unchecked (stream,
entry_n + 1);
}
} else if (stream->strh->type == GST_RIFF_FCC_auds) {
/* constant rate stream */
if (timestamp)
*timestamp =
avi_stream_convert_bytes_to_time_unchecked (stream, entry->total);
if (ts_end)
*ts_end = avi_stream_convert_bytes_to_time_unchecked (stream,
entry->total + entry->size);
}
if (stream->strh->type == GST_RIFF_FCC_vids) {
/* video offsets are the frame number */
if (offset)
*offset = entry_n;
if (offset_end)
*offset_end = entry_n + 1;
} else {
/* no offsets for audio */
if (offset)
*offset = -1;
if (offset_end)
*offset_end = -1;
}
}
/* collect and debug stats about the indexes for all streams.
* This method is also responsible for filling in the stream duration
* as measured by the amount of index entries.
*
* Returns TRUE if the index is not empty, else FALSE */
static gboolean
gst_avi_demux_do_index_stats (GstAviDemux * avi)
{
guint total_idx = 0;
guint i;
#ifndef GST_DISABLE_GST_DEBUG
guint total_max = 0;
#endif
/* get stream stats now */
for (i = 0; i < avi->num_streams; i++) {
GstAviStream *stream;
if (G_UNLIKELY (!(stream = &avi->stream[i])))
continue;
if (G_UNLIKELY (!stream->strh))
continue;
if (G_UNLIKELY (!stream->index || stream->idx_n == 0))
continue;
/* we interested in the end_ts of the last entry, which is the total
* duration of this stream */
gst_avi_demux_get_buffer_info (avi, stream, stream->idx_n - 1,
NULL, &stream->idx_duration, NULL, NULL);
total_idx += stream->idx_n;
#ifndef GST_DISABLE_GST_DEBUG
total_max += stream->idx_max;
#endif
GST_INFO_OBJECT (avi, "Stream %d, dur %" GST_TIME_FORMAT ", %6u entries, "
"%5u keyframes, entry size = %2u, total size = %10u, allocated %10u",
i, GST_TIME_ARGS (stream->idx_duration), stream->idx_n,
stream->n_keyframes, (guint) sizeof (GstAviIndexEntry),
(guint) (stream->idx_n * sizeof (GstAviIndexEntry)),
(guint) (stream->idx_max * sizeof (GstAviIndexEntry)));
}
total_idx *= sizeof (GstAviIndexEntry);
#ifndef GST_DISABLE_GST_DEBUG
total_max *= sizeof (GstAviIndexEntry);
#endif
GST_INFO_OBJECT (avi, "%u bytes for index vs %u ideally, %u wasted",
total_max, total_idx, total_max - total_idx);
if (total_idx == 0) {
GST_WARNING_OBJECT (avi, "Index is empty !");
return FALSE;
}
return TRUE;
}
/*
* gst_avi_demux_parse_subindex:
* @avi: Avi object
* @buf: input data to use for parsing.
* @stream: stream context.
* @entries_list: a list (returned by the function) containing all the
* indexes parsed in this specific subindex. The first
* entry is also a pointer to allocated memory that needs
* to be free´ed. May be NULL if no supported indexes were
* found.
*
* Reads superindex (openDML-2 spec stuff) from the provided data.
* The buffer should contain a GST_RIFF_TAG_ix?? chunk.
*
* Returns: TRUE on success, FALSE otherwise. Errors are fatal, we
* throw an error, caller should bail out asap.
*/
static gboolean
gst_avi_demux_parse_subindex (GstAviDemux * avi, GstAviStream * stream,
GstBuffer * buf)
{
guint8 *data;
guint16 bpe;
guint32 num, i;
guint64 baseoff;
2011-04-04 17:17:43 +00:00
gsize size;
2011-04-04 17:17:43 +00:00
if (buf == NULL)
return TRUE;
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (buf, &size, NULL, GST_MAP_READ);
/* check size */
if (size < 24)
goto too_small;
/* We don't support index-data yet */
if (data[3] & 0x80)
goto not_implemented;
/* check type of index. The opendml2 specs state that
* there should be 4 dwords per array entry. Type can be
* either frame or field (and we don't care). */
bpe = (data[2] & 0x01) ? 12 : 8;
if (GST_READ_UINT16_LE (data) != bpe / 4 ||
(data[2] & 0xfe) != 0x0 || data[3] != 0x1) {
GST_WARNING_OBJECT (avi,
"Superindex for stream %d has unexpected "
"size_entry %d (bytes) or flags 0x%02x/0x%02x",
stream->num, GST_READ_UINT16_LE (data), data[2], data[3]);
bpe = GST_READ_UINT16_LE (data) * 4;
}
num = GST_READ_UINT32_LE (&data[4]);
baseoff = GST_READ_UINT64_LE (&data[12]);
/* If there's nothing, just return ! */
if (num == 0)
goto empty_index;
GST_INFO_OBJECT (avi, "Parsing subindex, nr_entries = %6d", num);
for (i = 0; i < num; i++) {
GstAviIndexEntry entry;
if (size < 24 + bpe * (i + 1))
break;
/* fill in offset and size. offset contains the keyframe flag in the
* upper bit*/
entry.offset = baseoff + GST_READ_UINT32_LE (&data[24 + bpe * i]);
entry.size = GST_READ_UINT32_LE (&data[24 + bpe * i + 4]);
/* handle flags */
if (stream->strh->type == GST_RIFF_FCC_auds) {
/* all audio frames are keyframes */
ENTRY_SET_KEYFRAME (&entry);
} else {
/* else read flags */
entry.flags = (entry.size & 0x80000000) ? 0 : GST_AVI_KEYFRAME;
}
entry.size &= ~0x80000000;
/* and add */
if (G_UNLIKELY (!gst_avi_demux_add_index (avi, stream, num, &entry)))
goto out_of_mem;
}
2011-04-04 17:17:43 +00:00
done:
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
return TRUE;
/* ERRORS */
too_small:
{
GST_ERROR_OBJECT (avi,
"Not enough data to parse subindex (%d available, 24 needed)", size);
2011-04-04 17:17:43 +00:00
goto done; /* continue */
}
not_implemented:
{
GST_ELEMENT_ERROR (avi, STREAM, NOT_IMPLEMENTED, (NULL),
("Subindex-is-data is not implemented"));
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
return FALSE;
}
empty_index:
{
GST_DEBUG_OBJECT (avi, "the index is empty");
2011-04-04 17:17:43 +00:00
goto done; /* continue */
}
out_of_mem:
{
GST_ELEMENT_ERROR (avi, RESOURCE, NO_SPACE_LEFT, (NULL),
("Cannot allocate memory for %u*%u=%u bytes",
(guint) sizeof (GstAviIndexEntry), num,
(guint) sizeof (GstAviIndexEntry) * num));
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
return FALSE;
}
}
2009-12-04 14:13:12 +00:00
/*
* Create and push a flushing seek event upstream
*/
static gboolean
perform_seek_to_offset (GstAviDemux * demux, guint64 offset)
2009-12-04 14:13:12 +00:00
{
GstEvent *event;
gboolean res = 0;
GST_DEBUG_OBJECT (demux, "Seeking to %" G_GUINT64_FORMAT, offset);
event =
gst_event_new_seek (1.0, GST_FORMAT_BYTES,
GST_SEEK_FLAG_FLUSH | GST_SEEK_FLAG_ACCURATE, GST_SEEK_TYPE_SET, offset,
GST_SEEK_TYPE_NONE, -1);
res = gst_pad_push_event (demux->sinkpad, event);
if (res)
demux->offset = offset;
return res;
}
/*
* Read AVI index when streaming
*/
2009-12-04 14:13:12 +00:00
static gboolean
gst_avi_demux_read_subindexes_push (GstAviDemux * avi)
{
guint32 tag = 0, size;
GstBuffer *buf = NULL;
guint odml_stream;
GST_DEBUG_OBJECT (avi, "read subindexes for %d streams", avi->num_streams);
2009-12-04 14:13:12 +00:00
if (avi->odml_subidxs[avi->odml_subidx] != avi->offset)
return FALSE;
2009-12-04 14:13:12 +00:00
if (!gst_avi_demux_peek_chunk (avi, &tag, &size))
return TRUE;
/* this is the ODML chunk we expect */
odml_stream = avi->odml_stream;
if ((tag != GST_MAKE_FOURCC ('i', 'x', '0' + odml_stream / 10,
'0' + odml_stream % 10)) &&
(tag != GST_MAKE_FOURCC ('0' + odml_stream / 10,
'0' + odml_stream % 10, 'i', 'x'))) {
2009-12-04 14:13:12 +00:00
GST_WARNING_OBJECT (avi, "Not an ix## chunk (%" GST_FOURCC_FORMAT ")",
GST_FOURCC_ARGS (tag));
return FALSE;
}
2009-12-04 14:13:12 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
/* flush chunk header so we get just the 'size' payload data */
gst_adapter_flush (avi->adapter, 8);
buf = gst_adapter_take_buffer (avi->adapter, size);
if (!gst_avi_demux_parse_subindex (avi, &avi->stream[odml_stream], buf))
2009-12-04 14:13:12 +00:00
return FALSE;
/* we parsed the index, go to next subindex */
avi->odml_subidx++;
if (avi->odml_subidxs[avi->odml_subidx] == GST_BUFFER_OFFSET_NONE) {
/* we reached the end of the indexes for this stream, move to the next
* stream to handle the first index */
avi->odml_stream++;
2009-12-04 14:13:12 +00:00
avi->odml_subidx = 0;
if (avi->odml_stream < avi->num_streams) {
/* there are more indexes */
2009-12-04 14:13:12 +00:00
avi->odml_subidxs = avi->stream[avi->odml_stream].indexes;
} else {
/* we're done, get stream stats now */
2009-12-04 14:13:12 +00:00
avi->have_index = gst_avi_demux_do_index_stats (avi);
return TRUE;
}
}
2009-12-04 14:13:12 +00:00
/* seek to next index */
return perform_seek_to_offset (avi, avi->odml_subidxs[avi->odml_subidx]);
}
/*
* Read AVI index
*/
static void
gst_avi_demux_read_subindexes_pull (GstAviDemux * avi)
{
guint32 tag;
GstBuffer *buf;
gint i, n;
GST_DEBUG_OBJECT (avi, "read subindexes for %d streams", avi->num_streams);
for (n = 0; n < avi->num_streams; n++) {
GstAviStream *stream = &avi->stream[n];
if (stream->indexes == NULL)
continue;
for (i = 0; stream->indexes[i] != GST_BUFFER_OFFSET_NONE; i++) {
if (gst_riff_read_chunk (GST_ELEMENT_CAST (avi), avi->sinkpad,
&stream->indexes[i], &tag, &buf) != GST_FLOW_OK)
continue;
else if ((tag != GST_MAKE_FOURCC ('i', 'x', '0' + stream->num / 10,
'0' + stream->num % 10)) &&
(tag != GST_MAKE_FOURCC ('0' + stream->num / 10,
'0' + stream->num % 10, 'i', 'x'))) {
/* Some ODML files (created by god knows what muxer) have a ##ix format
* instead of the 'official' ix##. They are still valid though. */
GST_WARNING_OBJECT (avi, "Not an ix## chunk (%" GST_FOURCC_FORMAT ")",
GST_FOURCC_ARGS (tag));
gst_buffer_unref (buf);
continue;
}
if (!gst_avi_demux_parse_subindex (avi, stream, buf))
continue;
}
g_free (stream->indexes);
stream->indexes = NULL;
}
/* get stream stats now */
avi->have_index = gst_avi_demux_do_index_stats (avi);
}
/*
* gst_avi_demux_riff_parse_vprp:
* @element: caller element (used for debugging/error).
* @buf: input data to be used for parsing, stripped from header.
* @vprp: a pointer (returned by this function) to a filled-in vprp
* structure. Caller should free it.
*
* Parses a video stream´s vprp. This function takes ownership of @buf.
*
* Returns: TRUE if parsing succeeded, otherwise FALSE. The stream
* should be skipped on error, but it is not fatal.
*/
static gboolean
gst_avi_demux_riff_parse_vprp (GstElement * element,
GstBuffer * buf, gst_riff_vprp ** _vprp)
{
gst_riff_vprp *vprp;
gint k;
2011-04-04 17:17:43 +00:00
gsize size;
g_return_val_if_fail (buf != NULL, FALSE);
g_return_val_if_fail (_vprp != NULL, FALSE);
2011-04-04 17:17:43 +00:00
size = gst_buffer_get_size (buf);
if (size < G_STRUCT_OFFSET (gst_riff_vprp, field_info))
goto too_small;
2011-04-04 17:17:43 +00:00
vprp = g_malloc (size);
gst_buffer_extract (buf, 0, vprp, size);
#if (G_BYTE_ORDER == G_BIG_ENDIAN)
vprp->format_token = GUINT32_FROM_LE (vprp->format_token);
vprp->standard = GUINT32_FROM_LE (vprp->standard);
vprp->vert_rate = GUINT32_FROM_LE (vprp->vert_rate);
vprp->hor_t_total = GUINT32_FROM_LE (vprp->hor_t_total);
vprp->vert_lines = GUINT32_FROM_LE (vprp->vert_lines);
vprp->aspect = GUINT32_FROM_LE (vprp->aspect);
vprp->width = GUINT32_FROM_LE (vprp->width);
vprp->height = GUINT32_FROM_LE (vprp->height);
vprp->fields = GUINT32_FROM_LE (vprp->fields);
#endif
/* size checking */
/* calculate fields based on size */
2011-04-04 17:17:43 +00:00
k = (size - G_STRUCT_OFFSET (gst_riff_vprp, field_info)) / vprp->fields;
if (vprp->fields > k) {
GST_WARNING_OBJECT (element,
"vprp header indicated %d fields, only %d available", vprp->fields, k);
vprp->fields = k;
}
if (vprp->fields > GST_RIFF_VPRP_VIDEO_FIELDS) {
GST_WARNING_OBJECT (element,
"vprp header indicated %d fields, at most %d supported", vprp->fields,
GST_RIFF_VPRP_VIDEO_FIELDS);
vprp->fields = GST_RIFF_VPRP_VIDEO_FIELDS;
}
#if (G_BYTE_ORDER == G_BIG_ENDIAN)
for (k = 0; k < vprp->fields; k++) {
gst_riff_vprp_video_field_desc *fd;
fd = &vprp->field_info[k];
fd->compressed_bm_height = GUINT32_FROM_LE (fd->compressed_bm_height);
fd->compressed_bm_width = GUINT32_FROM_LE (fd->compressed_bm_width);
fd->valid_bm_height = GUINT32_FROM_LE (fd->valid_bm_height);
fd->valid_bm_width = GUINT16_FROM_LE (fd->valid_bm_width);
fd->valid_bm_x_offset = GUINT16_FROM_LE (fd->valid_bm_x_offset);
fd->valid_bm_y_offset = GUINT32_FROM_LE (fd->valid_bm_y_offset);
fd->video_x_t_offset = GUINT32_FROM_LE (fd->video_x_t_offset);
fd->video_y_start = GUINT32_FROM_LE (fd->video_y_start);
}
#endif
/* debug */
GST_INFO_OBJECT (element, "vprp tag found in context vids:");
GST_INFO_OBJECT (element, " format_token %d", vprp->format_token);
GST_INFO_OBJECT (element, " standard %d", vprp->standard);
GST_INFO_OBJECT (element, " vert_rate %d", vprp->vert_rate);
GST_INFO_OBJECT (element, " hor_t_total %d", vprp->hor_t_total);
GST_INFO_OBJECT (element, " vert_lines %d", vprp->vert_lines);
GST_INFO_OBJECT (element, " aspect %d:%d", vprp->aspect >> 16,
vprp->aspect & 0xffff);
GST_INFO_OBJECT (element, " width %d", vprp->width);
GST_INFO_OBJECT (element, " height %d", vprp->height);
GST_INFO_OBJECT (element, " fields %d", vprp->fields);
for (k = 0; k < vprp->fields; k++) {
gst_riff_vprp_video_field_desc *fd;
fd = &(vprp->field_info[k]);
GST_INFO_OBJECT (element, " field %u description:", k);
GST_INFO_OBJECT (element, " compressed_bm_height %d",
fd->compressed_bm_height);
GST_INFO_OBJECT (element, " compressed_bm_width %d",
fd->compressed_bm_width);
GST_INFO_OBJECT (element, " valid_bm_height %d",
fd->valid_bm_height);
GST_INFO_OBJECT (element, " valid_bm_width %d", fd->valid_bm_width);
GST_INFO_OBJECT (element, " valid_bm_x_offset %d",
fd->valid_bm_x_offset);
GST_INFO_OBJECT (element, " valid_bm_y_offset %d",
fd->valid_bm_y_offset);
GST_INFO_OBJECT (element, " video_x_t_offset %d",
fd->video_x_t_offset);
GST_INFO_OBJECT (element, " video_y_start %d", fd->video_y_start);
}
gst_buffer_unref (buf);
*_vprp = vprp;
return TRUE;
/* ERRORS */
too_small:
{
GST_ERROR_OBJECT (element,
"Too small vprp (%d available, at least %d needed)",
2011-04-04 17:17:43 +00:00
size, (int) G_STRUCT_OFFSET (gst_riff_vprp, field_info));
gst_buffer_unref (buf);
return FALSE;
}
}
static void
gst_avi_demux_expose_streams (GstAviDemux * avi, gboolean force)
{
guint i;
GST_DEBUG_OBJECT (avi, "force : %d", force);
for (i = 0; i < avi->num_streams; i++) {
GstAviStream *stream = &avi->stream[i];
if (force || stream->idx_n != 0) {
2011-06-02 16:47:36 +00:00
GST_LOG_OBJECT (avi, "Adding pad %s" GST_PTR_FORMAT,
GST_PAD_NAME (stream->pad));
gst_element_add_pad ((GstElement *) avi, stream->pad);
if (avi->element_index)
gst_index_get_writer_id (avi->element_index,
GST_OBJECT_CAST (stream->pad), &stream->index_id);
stream->exposed = TRUE;
if (avi->main_stream == -1)
avi->main_stream = i;
} else {
GST_WARNING_OBJECT (avi, "Stream #%d doesn't have any entry, removing it",
i);
gst_avi_demux_reset_stream (avi, stream);
}
}
}
/* buf contains LIST chunk data, and will be padded to even size,
* since some buggy files do not account for the padding of chunks
* within a LIST in the size of the LIST */
static inline void
gst_avi_demux_roundup_list (GstAviDemux * avi, GstBuffer ** buf)
{
2011-04-04 17:17:43 +00:00
gsize size;
size = gst_buffer_get_size (*buf);
if (G_UNLIKELY (size & 1)) {
GstBuffer *obuf;
2011-04-04 17:17:43 +00:00
guint8 *data;
2011-04-04 17:17:43 +00:00
GST_DEBUG_OBJECT (avi, "rounding up dubious list size %d", size);
obuf = gst_buffer_new_and_alloc (size + 1);
data = gst_buffer_map (obuf, NULL, NULL, GST_MAP_WRITE);
gst_buffer_extract (*buf, 0, data, size);
/* assume 0 padding, at least makes outcome deterministic */
data[size] = 0;
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (obuf, data, size + 1);
gst_buffer_replace (buf, obuf);
}
}
/*
* gst_avi_demux_parse_stream:
* @avi: calling element (used for debugging/errors).
* @buf: input buffer used to parse the stream.
*
* Parses all subchunks in a strl chunk (which defines a single
* stream). Discards the buffer after use. This function will
* increment the stream counter internally.
*
* Returns: whether the stream was identified successfully.
* Errors are not fatal. It does indicate the stream
* was skipped.
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
*/
static gboolean
gst_avi_demux_parse_stream (GstAviDemux * avi, GstBuffer * buf)
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
{
GstAviStream *stream;
GstElementClass *klass;
GstPadTemplate *templ;
GstBuffer *sub = NULL;
guint offset = 4;
guint32 tag = 0;
gchar *codec_name = NULL, *padname = NULL;
const gchar *tag_name;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
GstCaps *caps = NULL;
GstPad *pad;
GstElement *element;
gboolean got_strh = FALSE, got_strf = FALSE, got_vprp = FALSE;
gst_riff_vprp *vprp = NULL;
element = GST_ELEMENT_CAST (avi);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
GST_DEBUG_OBJECT (avi, "Parsing stream");
gst_avi_demux_roundup_list (avi, &buf);
if (avi->num_streams >= GST_AVI_DEMUX_MAX_STREAMS) {
GST_WARNING_OBJECT (avi,
"maximum no of streams (%d) exceeded, ignoring stream",
GST_AVI_DEMUX_MAX_STREAMS);
gst_buffer_unref (buf);
/* not a fatal error, let's say */
return TRUE;
}
stream = &avi->stream[avi->num_streams];
/* initial settings */
stream->idx_duration = GST_CLOCK_TIME_NONE;
stream->hdr_duration = GST_CLOCK_TIME_NONE;
stream->duration = GST_CLOCK_TIME_NONE;
while (gst_riff_parse_chunk (element, buf, &offset, &tag, &sub)) {
/* sub can be NULL if the chunk is empty */
2009-08-04 09:39:39 +00:00
if (sub == NULL) {
GST_DEBUG_OBJECT (avi, "ignoring empty chunk %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (tag));
continue;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
switch (tag) {
case GST_RIFF_TAG_strh:
{
gst_riff_strh *strh;
if (got_strh) {
GST_WARNING_OBJECT (avi, "Ignoring additional strh chunk");
break;
}
if (!gst_riff_parse_strh (element, sub, &stream->strh)) {
2009-08-04 09:39:39 +00:00
/* ownership given away */
sub = NULL;
GST_WARNING_OBJECT (avi, "Failed to parse strh chunk");
goto fail;
}
2009-08-04 09:39:39 +00:00
sub = NULL;
strh = stream->strh;
/* sanity check; stream header frame rate matches global header
* frame duration */
if (stream->strh->type == GST_RIFF_FCC_vids) {
GstClockTime s_dur;
GstClockTime h_dur = avi->avih->us_frame * GST_USECOND;
s_dur = gst_util_uint64_scale (GST_SECOND, strh->scale, strh->rate);
GST_DEBUG_OBJECT (avi, "verifying stream framerate %d/%d, "
"frame duration = %d ms", strh->rate, strh->scale,
2009-09-22 13:03:20 +00:00
(gint) (s_dur / GST_MSECOND));
if (h_dur > (10 * GST_MSECOND) && (s_dur > 10 * h_dur)) {
strh->rate = GST_SECOND / GST_USECOND;
strh->scale = h_dur / GST_USECOND;
GST_DEBUG_OBJECT (avi, "correcting stream framerate to %d/%d",
strh->rate, strh->scale);
}
}
/* determine duration as indicated by header */
stream->hdr_duration = gst_util_uint64_scale ((guint64) strh->length *
strh->scale, GST_SECOND, (guint64) strh->rate);
GST_INFO ("Stream duration according to header: %" GST_TIME_FORMAT,
GST_TIME_ARGS (stream->hdr_duration));
if (stream->hdr_duration == 0)
stream->hdr_duration = GST_CLOCK_TIME_NONE;
got_strh = TRUE;
break;
}
case GST_RIFF_TAG_strf:
{
gboolean res = FALSE;
if (got_strf) {
GST_WARNING_OBJECT (avi, "Ignoring additional strf chunk");
break;
}
if (!got_strh) {
GST_ERROR_OBJECT (avi, "Found strf chunk before strh chunk");
goto fail;
}
switch (stream->strh->type) {
case GST_RIFF_FCC_vids:
stream->is_vbr = TRUE;
res = gst_riff_parse_strf_vids (element, sub,
&stream->strf.vids, &stream->extradata);
2009-08-04 09:39:39 +00:00
sub = NULL;
GST_DEBUG_OBJECT (element, "marking video as VBR, res %d", res);
break;
case GST_RIFF_FCC_auds:
res =
gst_riff_parse_strf_auds (element, sub, &stream->strf.auds,
&stream->extradata);
sub = NULL;
if (!res)
break;
stream->is_vbr = (stream->strh->samplesize == 0)
&& stream->strh->scale > 1
&& stream->strf.auds->blockalign != 1;
GST_DEBUG_OBJECT (element, "marking audio as VBR:%d, res %d",
stream->is_vbr, res);
/* we need these or we have no way to come up with timestamps */
if ((!stream->is_vbr && !stream->strf.auds->av_bps) ||
(stream->is_vbr && (!stream->strh->scale ||
!stream->strh->rate))) {
GST_WARNING_OBJECT (element,
"invalid audio header, ignoring stream");
goto fail;
}
/* some more sanity checks */
if (stream->is_vbr) {
if (stream->strf.auds->blockalign <= 4) {
/* that would mean (too) many frames per chunk,
* so not likely set as expected */
GST_DEBUG_OBJECT (element,
"suspicious blockalign %d for VBR audio; "
"overriding to 1 frame per chunk",
stream->strf.auds->blockalign);
/* this should top any likely value */
stream->strf.auds->blockalign = (1 << 12);
}
}
break;
case GST_RIFF_FCC_iavs:
stream->is_vbr = TRUE;
res = gst_riff_parse_strf_iavs (element, sub,
&stream->strf.iavs, &stream->extradata);
2009-08-04 09:39:39 +00:00
sub = NULL;
GST_DEBUG_OBJECT (element, "marking iavs as VBR, res %d", res);
break;
case GST_RIFF_FCC_txts:
/* nothing to parse here */
stream->is_vbr = (stream->strh->samplesize == 0)
&& (stream->strh->scale > 1);
res = TRUE;
break;
default:
GST_ERROR_OBJECT (avi,
"Don´t know how to handle stream type %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (stream->strh->type));
break;
}
2009-08-04 09:39:39 +00:00
if (sub) {
gst_buffer_unref (sub);
sub = NULL;
}
if (!res)
goto fail;
got_strf = TRUE;
break;
}
case GST_RIFF_TAG_vprp:
{
if (got_vprp) {
GST_WARNING_OBJECT (avi, "Ignoring additional vprp chunk");
break;
}
if (!got_strh) {
GST_ERROR_OBJECT (avi, "Found vprp chunk before strh chunk");
goto fail;
}
if (!got_strf) {
GST_ERROR_OBJECT (avi, "Found vprp chunk before strf chunk");
goto fail;
}
if (!gst_avi_demux_riff_parse_vprp (element, sub, &vprp)) {
GST_WARNING_OBJECT (avi, "Failed to parse vprp chunk");
/* not considered fatal */
g_free (vprp);
vprp = NULL;
} else
got_vprp = TRUE;
2009-08-04 09:39:39 +00:00
sub = NULL;
break;
}
gst/: Add MS RLE support. I added some functions to read out strf chunks into strf chunks and the data behind it. Thi... Original commit message from CVS: reviewed by: <delete if not using a buddy> * gst-libs/gst/riff/riff-media.c: (gst_riff_create_video_caps_with_data), (gst_riff_create_video_caps), (gst_riff_create_audio_caps), (gst_riff_create_video_template_caps), (gst_riff_create_audio_template_caps): * gst-libs/gst/riff/riff-media.h: * gst-libs/gst/riff/riff-read.c: (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_vids): * gst-libs/gst/riff/riff-read.h: * gst/avi/gstavidemux.c: (gst_avi_demux_add_stream): Add MS RLE support. I added some functions to read out strf chunks into strf chunks and the data behind it. This is usually color palettes (as in RLE, but also in 8-bit RGB). Also use those during caps creation. Lastly, add ADPCM (similar to wavparse - which should eventually be rifflib based). * gst/matroska/matroska-demux.c: (gst_matroska_demux_class_init), (gst_matroska_demux_init), (gst_matroska_demux_reset): * gst/matroska/matroska-demux.h: Remove placeholders for some prehistoric tagging system. Didn't add support for any tag system really anyway. * gst/qtdemux/qtdemux.c: Add support for audio/x-m4a (MPEG-4) through spider. * gst/wavparse/gstwavparse.c: (gst_wavparse_parse_fmt), (gst_wavparse_loop): ADPCM support (#135862). Increase max. buffer size because we cannot split buffers for ADPCM (screws references) and I've seen files with 2048 byte chunks. 4096 seems safe for now.
2004-04-16 01:20:44 +00:00
case GST_RIFF_TAG_strd:
if (stream->initdata)
gst_buffer_unref (stream->initdata);
stream->initdata = sub;
2009-08-04 09:39:39 +00:00
sub = NULL;
gst/: Add MS RLE support. I added some functions to read out strf chunks into strf chunks and the data behind it. Thi... Original commit message from CVS: reviewed by: <delete if not using a buddy> * gst-libs/gst/riff/riff-media.c: (gst_riff_create_video_caps_with_data), (gst_riff_create_video_caps), (gst_riff_create_audio_caps), (gst_riff_create_video_template_caps), (gst_riff_create_audio_template_caps): * gst-libs/gst/riff/riff-media.h: * gst-libs/gst/riff/riff-read.c: (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_vids): * gst-libs/gst/riff/riff-read.h: * gst/avi/gstavidemux.c: (gst_avi_demux_add_stream): Add MS RLE support. I added some functions to read out strf chunks into strf chunks and the data behind it. This is usually color palettes (as in RLE, but also in 8-bit RGB). Also use those during caps creation. Lastly, add ADPCM (similar to wavparse - which should eventually be rifflib based). * gst/matroska/matroska-demux.c: (gst_matroska_demux_class_init), (gst_matroska_demux_init), (gst_matroska_demux_reset): * gst/matroska/matroska-demux.h: Remove placeholders for some prehistoric tagging system. Didn't add support for any tag system really anyway. * gst/qtdemux/qtdemux.c: Add support for audio/x-m4a (MPEG-4) through spider. * gst/wavparse/gstwavparse.c: (gst_wavparse_parse_fmt), (gst_wavparse_loop): ADPCM support (#135862). Increase max. buffer size because we cannot split buffers for ADPCM (screws references) and I've seen files with 2048 byte chunks. 4096 seems safe for now.
2004-04-16 01:20:44 +00:00
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
case GST_RIFF_TAG_strn:
g_free (stream->name);
if (sub != NULL) {
2011-04-04 17:17:43 +00:00
gchar *bdata;
gsize bsize;
bdata = gst_buffer_map (sub, &bsize, NULL, GST_MAP_READ);
stream->name = g_strndup (bdata, bsize);
gst_buffer_unmap (sub, bdata, bsize);
gst_buffer_unref (sub);
sub = NULL;
} else {
stream->name = g_strdup ("");
}
GST_DEBUG_OBJECT (avi, "stream name: %s", stream->name);
break;
case GST_RIFF_IDIT:
gst_avi_demux_parse_idit (avi, sub);
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
default:
if (tag == GST_MAKE_FOURCC ('i', 'n', 'd', 'x') ||
tag == GST_MAKE_FOURCC ('i', 'x', '0' + avi->num_streams / 10,
'0' + avi->num_streams % 10)) {
g_free (stream->indexes);
gst_avi_demux_parse_superindex (avi, sub, &stream->indexes);
stream->superindex = TRUE;
2009-08-04 09:39:39 +00:00
sub = NULL;
break;
}
GST_WARNING_OBJECT (avi,
"Unknown stream header tag %" GST_FOURCC_FORMAT ", ignoring",
GST_FOURCC_ARGS (tag));
/* fall-through */
case GST_RIFF_TAG_JUNQ:
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
case GST_RIFF_TAG_JUNK:
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
2009-08-04 09:39:39 +00:00
if (sub != NULL) {
gst_buffer_unref (sub);
sub = NULL;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
if (!got_strh) {
GST_WARNING_OBJECT (avi, "Failed to find strh chunk");
goto fail;
}
if (!got_strf) {
GST_WARNING_OBJECT (avi, "Failed to find strf chunk");
goto fail;
}
/* get class to figure out the template */
klass = GST_ELEMENT_GET_CLASS (avi);
/* we now have all info, let´s set up a pad and a caps and be done */
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* create stream name + pad */
switch (stream->strh->type) {
case GST_RIFF_FCC_vids:{
guint32 fourcc;
fourcc = (stream->strf.vids->compression) ?
stream->strf.vids->compression : stream->strh->fcc_handler;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
padname = g_strdup_printf ("video_%02d", avi->num_v_streams);
templ = gst_element_class_get_pad_template (klass, "video_%02d");
caps = gst_riff_create_video_caps (fourcc, stream->strh,
stream->strf.vids, stream->extradata, stream->initdata, &codec_name);
if (!caps) {
caps = gst_caps_new_simple ("video/x-avi-unknown", "fourcc",
2011-08-22 10:24:15 +00:00
G_TYPE_INT, fourcc, NULL);
} else if (got_vprp && vprp) {
guint32 aspect_n, aspect_d;
gint n, d;
aspect_n = vprp->aspect >> 16;
aspect_d = vprp->aspect & 0xffff;
/* calculate the pixel aspect ratio using w/h and aspect ratio */
n = aspect_n * stream->strf.vids->height;
d = aspect_d * stream->strf.vids->width;
if (n && d)
gst_caps_set_simple (caps, "pixel-aspect-ratio", GST_TYPE_FRACTION,
n, d, NULL);
/* very local, not needed elsewhere */
g_free (vprp);
vprp = NULL;
}
tag_name = GST_TAG_VIDEO_CODEC;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
avi->num_v_streams++;
break;
}
case GST_RIFF_FCC_auds:{
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
padname = g_strdup_printf ("audio_%02d", avi->num_a_streams);
templ = gst_element_class_get_pad_template (klass, "audio_%02d");
caps = gst_riff_create_audio_caps (stream->strf.auds->format,
stream->strh, stream->strf.auds, stream->extradata,
stream->initdata, &codec_name);
if (!caps) {
caps = gst_caps_new_simple ("audio/x-avi-unknown", "codec_id",
G_TYPE_INT, stream->strf.auds->format, NULL);
}
tag_name = GST_TAG_AUDIO_CODEC;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
avi->num_a_streams++;
break;
}
case GST_RIFF_FCC_iavs:{
guint32 fourcc = stream->strh->fcc_handler;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
padname = g_strdup_printf ("video_%02d", avi->num_v_streams);
templ = gst_element_class_get_pad_template (klass, "video_%02d");
caps = gst_riff_create_iavs_caps (fourcc, stream->strh,
stream->strf.iavs, stream->extradata, stream->initdata, &codec_name);
if (!caps) {
caps = gst_caps_new_simple ("video/x-avi-unknown", "fourcc",
2011-08-22 10:24:15 +00:00
G_TYPE_INT, fourcc, NULL);
}
tag_name = GST_TAG_VIDEO_CODEC;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
avi->num_v_streams++;
break;
}
case GST_RIFF_FCC_txts:{
padname = g_strdup_printf ("subtitle_%02d", avi->num_t_streams);
templ = gst_element_class_get_pad_template (klass, "subtitle_%02d");
caps = gst_caps_new_simple ("application/x-subtitle-avi", NULL);
tag_name = NULL;
avi->num_t_streams++;
break;
}
default:
g_return_val_if_reached (FALSE);
}
/* no caps means no stream */
if (!caps) {
GST_ERROR_OBJECT (element, "Did not find caps for stream %s", padname);
goto fail;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
GST_DEBUG_OBJECT (element, "codec-name=%s",
(codec_name ? codec_name : "NULL"));
GST_DEBUG_OBJECT (element, "caps=%" GST_PTR_FORMAT, caps);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* set proper settings and add it */
if (stream->pad)
gst_object_unref (stream->pad);
pad = stream->pad = gst_pad_new_from_template (templ, padname);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
g_free (padname);
gst_pad_use_fixed_caps (pad);
#if 0
gst_pad_set_formats_function (pad,
GST_DEBUG_FUNCPTR (gst_avi_demux_get_src_formats));
gst_pad_set_event_mask_function (pad,
GST_DEBUG_FUNCPTR (gst_avi_demux_get_event_mask));
#endif
gst_pad_set_event_function (pad,
GST_DEBUG_FUNCPTR (gst_avi_demux_handle_src_event));
gst_pad_set_query_type_function (pad,
GST_DEBUG_FUNCPTR (gst_avi_demux_get_src_query_types));
gst_pad_set_query_function (pad,
GST_DEBUG_FUNCPTR (gst_avi_demux_handle_src_query));
#if 0
gst_pad_set_convert_function (pad,
GST_DEBUG_FUNCPTR (gst_avi_demux_src_convert));
#endif
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
stream->num = avi->num_streams;
stream->start_entry = 0;
stream->step_entry = 0;
stream->stop_entry = 0;
stream->current_entry = -1;
stream->current_total = 0;
stream->last_flow = GST_FLOW_OK;
stream->discont = TRUE;
stream->total_bytes = 0;
stream->total_blocks = 0;
stream->n_keyframes = 0;
stream->idx_n = 0;
stream->idx_max = 0;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
gst_pad_set_element_private (pad, stream);
avi->num_streams++;
gst_pad_set_active (pad, TRUE);
2011-06-02 16:47:36 +00:00
gst_pad_push_event (pad, gst_event_new_caps (caps));
gst_caps_unref (caps);
/* make tags */
if (codec_name) {
if (!stream->taglist)
stream->taglist = gst_tag_list_new ();
avi->got_tags = TRUE;
gst_tag_list_add (stream->taglist, GST_TAG_MERGE_APPEND, tag_name,
codec_name, NULL);
g_free (codec_name);
}
2009-08-04 09:39:39 +00:00
gst_buffer_unref (buf);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
return TRUE;
/* ERRORS */
fail:
{
/* unref any mem that may be in use */
if (buf)
gst_buffer_unref (buf);
if (sub)
gst_buffer_unref (sub);
g_free (vprp);
g_free (codec_name);
gst_avi_demux_reset_stream (avi, stream);
avi->num_streams++;
return FALSE;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
/*
* gst_avi_demux_parse_odml:
* @avi: calling element (used for debug/error).
* @buf: input buffer to be used for parsing.
*
* Read an openDML-2.0 extension header. Fills in the frame number
* in the avi demuxer object when reading succeeds.
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
*/
static void
gst_avi_demux_parse_odml (GstAviDemux * avi, GstBuffer * buf)
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
{
guint32 tag = 0;
guint offset = 4;
GstBuffer *sub = NULL;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
while (gst_riff_parse_chunk (GST_ELEMENT_CAST (avi), buf, &offset, &tag,
&sub)) {
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
switch (tag) {
case GST_RIFF_TAG_dmlh:{
gst_riff_dmlh dmlh, *_dmlh;
2011-04-04 17:17:43 +00:00
gsize size;
guint8 *data;
/* sub == NULL is possible and means an empty buffer */
2011-04-04 17:17:43 +00:00
if (sub == NULL)
goto next;
data = gst_buffer_map (sub, &size, NULL, GST_MAP_READ);
/* check size */
if (size < sizeof (gst_riff_dmlh)) {
GST_ERROR_OBJECT (avi,
"DMLH entry is too small (%d bytes, %d needed)",
size, (int) sizeof (gst_riff_dmlh));
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (sub, data, size);
goto next;
}
2011-04-04 17:17:43 +00:00
_dmlh = (gst_riff_dmlh *) data;
dmlh.totalframes = GST_READ_UINT32_LE (&_dmlh->totalframes);
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (sub, data, size);
GST_INFO_OBJECT (avi, "dmlh tag found: totalframes: %u",
dmlh.totalframes);
avi->avih->tot_frames = dmlh.totalframes;
goto next;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
default:
GST_WARNING_OBJECT (avi,
"Unknown tag %" GST_FOURCC_FORMAT " in ODML header",
GST_FOURCC_ARGS (tag));
/* fall-through */
case GST_RIFF_TAG_JUNQ:
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
case GST_RIFF_TAG_JUNK:
next:
/* skip and move to next chunk */
if (sub) {
gst_buffer_unref (sub);
sub = NULL;
}
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
}
if (buf)
gst_buffer_unref (buf);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
/* Index helper */
static guint
gst_avi_demux_index_last (GstAviDemux * avi, GstAviStream * stream)
{
return stream->idx_n;
}
/* find a previous entry in the index with the given flags */
static guint
gst_avi_demux_index_prev (GstAviDemux * avi, GstAviStream * stream,
2009-09-22 10:35:30 +00:00
guint last, gboolean keyframe)
{
GstAviIndexEntry *entry;
guint i;
for (i = last; i > 0; i--) {
entry = &stream->index[i - 1];
2009-09-22 10:35:30 +00:00
if (!keyframe || ENTRY_IS_KEYFRAME (entry)) {
return i - 1;
}
}
return 0;
}
static guint
gst_avi_demux_index_next (GstAviDemux * avi, GstAviStream * stream,
2009-09-22 10:35:30 +00:00
guint last, gboolean keyframe)
{
GstAviIndexEntry *entry;
gint i;
for (i = last + 1; i < stream->idx_n; i++) {
entry = &stream->index[i];
2009-09-22 10:35:30 +00:00
if (!keyframe || ENTRY_IS_KEYFRAME (entry)) {
return i;
}
}
return stream->idx_n - 1;
}
static guint
gst_avi_demux_index_entry_search (GstAviIndexEntry * entry, guint64 * total)
{
if (entry->total < *total)
return -1;
else if (entry->total > *total)
return 1;
return 0;
}
/*
* gst_avi_demux_index_for_time:
* @avi: Avi object
* @stream: the stream
* @time: a time position
*
* Finds the index entry which time is less or equal than the requested time.
* Try to avoid binary search when we can convert the time to an index
* position directly (for example for video frames with a fixed duration).
*
* Returns: the found position in the index.
*/
static guint
gst_avi_demux_index_for_time (GstAviDemux * avi,
GstAviStream * stream, guint64 time)
{
guint index = -1;
guint64 total;
GST_LOG_OBJECT (avi, "search time:%" GST_TIME_FORMAT, GST_TIME_ARGS (time));
/* easy (and common) cases */
if (time == 0 || stream->idx_n == 0)
return 0;
if (time >= stream->idx_duration)
return stream->idx_n - 1;
/* figure out where we need to go. For that we convert the time to an
* index entry or we convert it to a total and then do a binary search. */
if (stream->is_vbr) {
/* VBR stream next timestamp */
if (stream->strh->type == GST_RIFF_FCC_auds) {
total = avi_stream_convert_time_to_frames_unchecked (stream, time);
} else {
index = avi_stream_convert_time_to_frames_unchecked (stream, time);
}
} else {
/* constant rate stream */
total = avi_stream_convert_time_to_bytes_unchecked (stream, time);
}
if (index == -1) {
GstAviIndexEntry *entry;
/* no index, find index with binary search on total */
GST_LOG_OBJECT (avi, "binary search for entry with total %"
G_GUINT64_FORMAT, total);
entry = gst_util_array_binary_search (stream->index,
stream->idx_n, sizeof (GstAviIndexEntry),
(GCompareDataFunc) gst_avi_demux_index_entry_search,
GST_SEARCH_MODE_BEFORE, &total, NULL);
if (entry == NULL) {
GST_LOG_OBJECT (avi, "not found, assume index 0");
index = 0;
} else {
index = entry - stream->index;
GST_LOG_OBJECT (avi, "found at %u", index);
}
} else {
GST_LOG_OBJECT (avi, "converted time to index %u", index);
}
return index;
}
static inline GstAviStream *
gst_avi_demux_stream_for_id (GstAviDemux * avi, guint32 id)
{
guint stream_nr;
GstAviStream *stream;
/* get the stream for this entry */
stream_nr = CHUNKID_TO_STREAMNR (id);
if (G_UNLIKELY (stream_nr >= avi->num_streams)) {
GST_WARNING_OBJECT (avi, "invalid stream nr %d", stream_nr);
return NULL;
}
stream = &avi->stream[stream_nr];
if (G_UNLIKELY (!stream->strh)) {
GST_WARNING_OBJECT (avi, "Unhandled stream %d, skipping", stream_nr);
return NULL;
}
return stream;
}
/*
* gst_avi_demux_parse_index:
* @avi: calling element (used for debugging/errors).
* @buf: buffer containing the full index.
*
* Read index entries from the provided buffer.
* The buffer should contain a GST_RIFF_TAG_idx1 chunk.
*/
static gboolean
gst_avi_demux_parse_index (GstAviDemux * avi, GstBuffer * buf)
{
guint8 *data;
2011-04-04 17:17:43 +00:00
gsize size;
guint i, num, n;
gst_riff_index_entry *index;
GstClockTime stamp;
GstAviStream *stream;
GstAviIndexEntry entry;
guint32 id;
if (!buf)
return FALSE;
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (buf, &size, NULL, GST_MAP_READ);
stamp = gst_util_get_timestamp ();
/* see how many items in the index */
num = size / sizeof (gst_riff_index_entry);
if (num == 0)
goto empty_list;
GST_INFO_OBJECT (avi, "Parsing index, nr_entries = %6d", num);
index = (gst_riff_index_entry *) data;
/* figure out if the index is 0 based or relative to the MOVI start */
entry.offset = GST_READ_UINT32_LE (&index[0].offset);
if (entry.offset < avi->offset) {
avi->index_offset = avi->offset + 8;
GST_DEBUG ("index_offset = %" G_GUINT64_FORMAT, avi->index_offset);
} else {
avi->index_offset = 0;
GST_DEBUG ("index is 0 based");
}
for (i = 0, n = 0; i < num; i++) {
id = GST_READ_UINT32_LE (&index[i].id);
entry.offset = GST_READ_UINT32_LE (&index[i].offset);
/* some sanity checks */
if (G_UNLIKELY (id == GST_RIFF_rec || id == 0 ||
(entry.offset == 0 && n > 0)))
continue;
/* get the stream for this entry */
stream = gst_avi_demux_stream_for_id (avi, id);
if (G_UNLIKELY (!stream))
continue;
/* handle offset and size */
entry.offset += avi->index_offset + 8;
entry.size = GST_READ_UINT32_LE (&index[i].size);
/* handle flags */
if (stream->strh->type == GST_RIFF_FCC_auds) {
/* all audio frames are keyframes */
2009-09-22 10:35:30 +00:00
ENTRY_SET_KEYFRAME (&entry);
} else {
2009-09-22 10:35:30 +00:00
guint32 flags;
/* else read flags */
flags = GST_READ_UINT32_LE (&index[i].flags);
if (flags & GST_RIFF_IF_KEYFRAME) {
ENTRY_SET_KEYFRAME (&entry);
} else {
2009-09-22 10:35:30 +00:00
ENTRY_UNSET_KEYFRAME (&entry);
}
}
/* and add */
if (G_UNLIKELY (!gst_avi_demux_add_index (avi, stream, num, &entry)))
goto out_of_mem;
n++;
}
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
/* get stream stats now */
avi->have_index = gst_avi_demux_do_index_stats (avi);
stamp = gst_util_get_timestamp () - stamp;
GST_DEBUG_OBJECT (avi, "index parsing took %" GST_TIME_FORMAT,
GST_TIME_ARGS (stamp));
return TRUE;
/* ERRORS */
empty_list:
{
GST_DEBUG_OBJECT (avi, "empty index");
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
return FALSE;
}
out_of_mem:
{
GST_ELEMENT_ERROR (avi, RESOURCE, NO_SPACE_LEFT, (NULL),
("Cannot allocate memory for %u*%u=%u bytes",
(guint) sizeof (GstAviIndexEntry), num,
(guint) sizeof (GstAviIndexEntry) * num));
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
gst_buffer_unref (buf);
return FALSE;
}
}
/*
* gst_avi_demux_stream_index:
* @avi: avi demuxer object.
*
* Seeks to index and reads it.
*/
static void
gst_avi_demux_stream_index (GstAviDemux * avi)
{
GstFlowReturn res;
guint64 offset = avi->offset;
GstBuffer *buf;
guint32 tag;
guint32 size;
2011-04-04 17:17:43 +00:00
gsize bsize;
guint8 *bdata;
GST_DEBUG ("demux stream index at offset %" G_GUINT64_FORMAT, offset);
/* get chunk information */
res = gst_pad_pull_range (avi->sinkpad, offset, 8, &buf);
if (res != GST_FLOW_OK)
goto pull_failed;
2011-04-04 17:17:43 +00:00
bdata = gst_buffer_map (buf, &bsize, NULL, GST_MAP_READ);
if (bsize < 8)
goto too_small;
/* check tag first before blindy trying to read 'size' bytes */
2011-04-04 17:17:43 +00:00
tag = GST_READ_UINT32_LE (bdata);
size = GST_READ_UINT32_LE (bdata + 4);
if (tag == GST_RIFF_TAG_LIST) {
/* this is the movi tag */
GST_DEBUG_OBJECT (avi, "skip LIST chunk, size %" G_GUINT32_FORMAT,
2009-09-21 23:28:54 +00:00
(8 + GST_ROUND_UP_2 (size)));
offset += 8 + GST_ROUND_UP_2 (size);
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, bdata, bsize);
gst_buffer_unref (buf);
2011-04-04 17:17:43 +00:00
res = gst_pad_pull_range (avi->sinkpad, offset, 8, &buf);
if (res != GST_FLOW_OK)
goto pull_failed;
2011-04-04 17:17:43 +00:00
bdata = gst_buffer_map (buf, &bsize, NULL, GST_MAP_READ);
if (bsize < 8)
goto too_small;
2011-04-04 17:17:43 +00:00
tag = GST_READ_UINT32_LE (bdata);
size = GST_READ_UINT32_LE (bdata + 4);
}
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, bdata, bsize);
gst_buffer_unref (buf);
if (tag != GST_RIFF_TAG_idx1)
goto no_index;
if (!size)
goto zero_index;
GST_DEBUG ("index found at offset %" G_GUINT64_FORMAT, offset);
/* read chunk, advance offset */
if (gst_riff_read_chunk (GST_ELEMENT_CAST (avi),
avi->sinkpad, &offset, &tag, &buf) != GST_FLOW_OK)
return;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
GST_DEBUG ("will parse index chunk size %u for tag %"
2011-04-04 17:17:43 +00:00
GST_FOURCC_FORMAT, gst_buffer_get_size (buf), GST_FOURCC_ARGS (tag));
gst_avi_demux_parse_index (avi, buf);
#ifndef GST_DISABLE_GST_DEBUG
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* debug our indexes */
{
gint i;
GstAviStream *stream;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
for (i = 0; i < avi->num_streams; i++) {
stream = &avi->stream[i];
GST_DEBUG_OBJECT (avi, "stream %u: %u frames, %" G_GINT64_FORMAT " bytes",
i, stream->idx_n, stream->total_bytes);
}
}
#endif
return;
/* ERRORS */
pull_failed:
{
GST_DEBUG_OBJECT (avi,
"pull range failed: pos=%" G_GUINT64_FORMAT " size=8", offset);
return;
}
too_small:
{
GST_DEBUG_OBJECT (avi, "Buffer is too small");
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, bdata, bsize);
gst_buffer_unref (buf);
return;
}
no_index:
{
GST_WARNING_OBJECT (avi,
"No index data (idx1) after movi chunk, but %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (tag));
return;
}
zero_index:
{
GST_WARNING_OBJECT (avi, "Empty index data (idx1) after movi chunk");
return;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
2009-12-04 14:13:12 +00:00
/*
* gst_avi_demux_stream_index_push:
* @avi: avi demuxer object.
*
* Read index.
*/
static void
gst_avi_demux_stream_index_push (GstAviDemux * avi)
{
guint64 offset = avi->idx1_offset;
GstBuffer *buf;
guint32 tag;
guint32 size;
GST_DEBUG ("demux stream index at offset %" G_GUINT64_FORMAT, offset);
/* get chunk information */
if (!gst_avi_demux_peek_chunk (avi, &tag, &size))
return;
/* check tag first before blindly trying to read 'size' bytes */
if (tag == GST_RIFF_TAG_LIST) {
/* this is the movi tag */
GST_DEBUG_OBJECT (avi, "skip LIST chunk, size %" G_GUINT32_FORMAT,
(8 + GST_ROUND_UP_2 (size)));
avi->idx1_offset = offset + 8 + GST_ROUND_UP_2 (size);
/* issue seek to allow chain function to handle it and return! */
perform_seek_to_offset (avi, avi->idx1_offset);
2009-12-04 14:13:12 +00:00
return;
}
if (tag != GST_RIFF_TAG_idx1)
goto no_index;
GST_DEBUG ("index found at offset %" G_GUINT64_FORMAT, offset);
/* flush chunk header */
gst_adapter_flush (avi->adapter, 8);
/* read chunk payload */
buf = gst_adapter_take_buffer (avi->adapter, size);
if (!buf)
goto pull_failed;
/* advance offset */
offset += 8 + GST_ROUND_UP_2 (size);
GST_DEBUG ("will parse index chunk size %u for tag %"
2011-04-04 17:17:43 +00:00
GST_FOURCC_FORMAT, gst_buffer_get_size (buf), GST_FOURCC_ARGS (tag));
2009-12-04 14:13:12 +00:00
avi->offset = avi->first_movi_offset;
2009-12-04 14:13:12 +00:00
gst_avi_demux_parse_index (avi, buf);
#ifndef GST_DISABLE_GST_DEBUG
/* debug our indexes */
{
gint i;
GstAviStream *stream;
for (i = 0; i < avi->num_streams; i++) {
stream = &avi->stream[i];
GST_DEBUG_OBJECT (avi, "stream %u: %u frames, %" G_GINT64_FORMAT " bytes",
i, stream->idx_n, stream->total_bytes);
}
}
#endif
return;
/* ERRORS */
pull_failed:
{
GST_DEBUG_OBJECT (avi,
"taking data from adapter failed: pos=%" G_GUINT64_FORMAT " size=%u",
offset, size);
return;
}
no_index:
{
GST_WARNING_OBJECT (avi,
"No index data (idx1) after movi chunk, but %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (tag));
return;
}
}
/*
* gst_avi_demux_peek_tag:
*
* Returns the tag and size of the next chunk
*/
static GstFlowReturn
gst_avi_demux_peek_tag (GstAviDemux * avi, guint64 offset, guint32 * tag,
guint * size)
{
GstFlowReturn res = GST_FLOW_OK;
GstBuffer *buf = NULL;
2011-04-04 17:17:43 +00:00
gsize bufsize;
guint8 *bufdata;
res = gst_pad_pull_range (avi->sinkpad, offset, 8, &buf);
if (res != GST_FLOW_OK)
goto pull_failed;
2011-04-04 17:17:43 +00:00
bufdata = gst_buffer_map (buf, &bufsize, NULL, GST_MAP_READ);
if (bufsize != 8)
goto wrong_size;
*tag = GST_READ_UINT32_LE (bufdata);
*size = GST_READ_UINT32_LE (bufdata + 4);
GST_LOG_OBJECT (avi, "Tag[%" GST_FOURCC_FORMAT "] (size:%d) %"
G_GINT64_FORMAT " -- %" G_GINT64_FORMAT, GST_FOURCC_ARGS (*tag),
*size, offset + 8, offset + 8 + (gint64) * size);
done:
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, bufdata, bufsize);
gst_buffer_unref (buf);
return res;
/* ERRORS */
pull_failed:
{
GST_DEBUG_OBJECT (avi, "pull_ranged returned %s", gst_flow_get_name (res));
return res;
}
wrong_size:
{
GST_DEBUG_OBJECT (avi, "got %d bytes which is <> 8 bytes", bufsize);
res = GST_FLOW_ERROR;
goto done;
}
}
/*
* gst_avi_demux_next_data_buffer:
*
* Returns the offset and size of the next buffer
* Position is the position of the buffer (after tag and size)
*/
static GstFlowReturn
gst_avi_demux_next_data_buffer (GstAviDemux * avi, guint64 * offset,
guint32 * tag, guint * size)
{
guint64 off = *offset;
guint _size = 0;
GstFlowReturn res;
do {
res = gst_avi_demux_peek_tag (avi, off, tag, &_size);
if (res != GST_FLOW_OK)
break;
if (*tag == GST_RIFF_TAG_LIST || *tag == GST_RIFF_TAG_RIFF)
off += 8 + 4; /* skip tag + size + subtag */
else {
*offset = off + 8;
*size = _size;
break;
}
} while (TRUE);
return res;
}
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/*
* gst_avi_demux_stream_scan:
* @avi: calling element (used for debugging/errors).
*
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
* Scan the file for all chunks to "create" a new index.
* pull-range based
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
*/
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
static gboolean
gst_avi_demux_stream_scan (GstAviDemux * avi)
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
{
GstFlowReturn res;
GstAviStream *stream;
guint64 pos = 0;
guint64 length;
gint64 tmplength;
guint32 tag = 0;
guint num;
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
/* FIXME:
* - implement non-seekable source support.
*/
GST_DEBUG_OBJECT (avi, "Creating index");
/* get the size of the file */
if (!gst_pad_query_peer_duration (avi->sinkpad, GST_FORMAT_BYTES, &tmplength))
return FALSE;
length = tmplength;
/* guess the total amount of entries we expect */
num = 16000;
while (TRUE) {
GstAviIndexEntry entry;
guint size = 0;
/* start reading data buffers to find the id and offset */
res = gst_avi_demux_next_data_buffer (avi, &pos, &tag, &size);
if (G_UNLIKELY (res != GST_FLOW_OK))
break;
/* get stream */
stream = gst_avi_demux_stream_for_id (avi, tag);
if (G_UNLIKELY (!stream))
goto next;
/* we can't figure out the keyframes, assume they all are */
entry.flags = GST_AVI_KEYFRAME;
entry.offset = pos;
entry.size = size;
/* and add to the index of this stream */
if (G_UNLIKELY (!gst_avi_demux_add_index (avi, stream, num, &entry)))
goto out_of_mem;
next:
/* update position */
pos += GST_ROUND_UP_2 (size);
if (G_UNLIKELY (pos > length)) {
GST_WARNING_OBJECT (avi,
"Stopping index lookup since we are further than EOF");
break;
}
}
/* collect stats */
avi->have_index = gst_avi_demux_do_index_stats (avi);
return TRUE;
/* ERRORS */
out_of_mem:
{
GST_ELEMENT_ERROR (avi, RESOURCE, NO_SPACE_LEFT, (NULL),
("Cannot allocate memory for %u*%u=%u bytes",
(guint) sizeof (GstAviIndexEntry), num,
(guint) sizeof (GstAviIndexEntry) * num));
return FALSE;
}
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
}
static void
gst_avi_demux_calculate_durations_from_index (GstAviDemux * avi)
{
guint i;
GstClockTime total;
GstAviStream *stream;
total = GST_CLOCK_TIME_NONE;
/* all streams start at a timestamp 0 */
for (i = 0; i < avi->num_streams; i++) {
GstClockTime duration, hduration;
gst_riff_strh *strh;
stream = &avi->stream[i];
if (G_UNLIKELY (!stream || !stream->idx_n || !(strh = stream->strh)))
continue;
/* get header duration for the stream */
hduration = stream->hdr_duration;
/* index duration calculated during parsing */
duration = stream->idx_duration;
/* now pick a good duration */
if (GST_CLOCK_TIME_IS_VALID (duration)) {
/* index gave valid duration, use that */
GST_INFO ("Stream %p duration according to index: %" GST_TIME_FORMAT,
stream, GST_TIME_ARGS (duration));
} else {
/* fall back to header info to calculate a duration */
duration = hduration;
}
GST_INFO ("Setting duration of stream #%d to %" GST_TIME_FORMAT,
i, GST_TIME_ARGS (duration));
/* set duration for the stream */
stream->duration = duration;
/* find total duration */
if (total == GST_CLOCK_TIME_NONE ||
(GST_CLOCK_TIME_IS_VALID (duration) && duration > total))
total = duration;
}
if (GST_CLOCK_TIME_IS_VALID (total) && (total > 0)) {
/* now update the duration for those streams where we had none */
for (i = 0; i < avi->num_streams; i++) {
stream = &avi->stream[i];
if (!GST_CLOCK_TIME_IS_VALID (stream->duration)
|| stream->duration == 0) {
stream->duration = total;
GST_INFO ("Stream %p duration according to total: %" GST_TIME_FORMAT,
stream, GST_TIME_ARGS (total));
}
}
}
/* and set the total duration in the segment. */
GST_INFO ("Setting total duration to: %" GST_TIME_FORMAT,
GST_TIME_ARGS (total));
2011-06-02 16:47:36 +00:00
avi->segment.duration = total;
}
/* returns FALSE if there are no pads to deliver event to,
* otherwise TRUE (whatever the outcome of event sending),
* takes ownership of the event. */
static gboolean
gst_avi_demux_push_event (GstAviDemux * avi, GstEvent * event)
{
gboolean result = FALSE;
gint i;
GST_DEBUG_OBJECT (avi, "sending %s event to %d streams",
GST_EVENT_TYPE_NAME (event), avi->num_streams);
for (i = 0; i < avi->num_streams; i++) {
GstAviStream *stream = &avi->stream[i];
if (stream->pad) {
result = TRUE;
gst_pad_push_event (stream->pad, gst_event_ref (event));
}
}
gst_event_unref (event);
return result;
}
2009-10-14 11:08:47 +00:00
static void
gst_avi_demux_check_seekability (GstAviDemux * avi)
{
GstQuery *query;
gboolean seekable = FALSE;
gint64 start = -1, stop = -1;
query = gst_query_new_seeking (GST_FORMAT_BYTES);
if (!gst_pad_peer_query (avi->sinkpad, query)) {
GST_DEBUG_OBJECT (avi, "seeking query failed");
goto done;
}
gst_query_parse_seeking (query, NULL, &seekable, &start, &stop);
/* try harder to query upstream size if we didn't get it the first time */
if (seekable && stop == -1) {
GST_DEBUG_OBJECT (avi, "doing duration query to fix up unset stop");
gst_pad_query_peer_duration (avi->sinkpad, GST_FORMAT_BYTES, &stop);
2009-10-14 11:08:47 +00:00
}
/* if upstream doesn't know the size, it's likely that it's not seekable in
* practice even if it technically may be seekable */
if (seekable && (start != 0 || stop <= start)) {
GST_DEBUG_OBJECT (avi, "seekable but unknown start/stop -> disable");
seekable = FALSE;
}
done:
GST_INFO_OBJECT (avi, "seekable: %d (%" G_GUINT64_FORMAT " - %"
G_GUINT64_FORMAT ")", seekable, start, stop);
avi->seekable = seekable;
gst_query_unref (query);
}
/*
* Read AVI headers when streaming
*/
static GstFlowReturn
gst_avi_demux_stream_header_push (GstAviDemux * avi)
{
GstFlowReturn ret = GST_FLOW_OK;
guint32 tag = 0;
guint32 ltag = 0;
guint32 size = 0;
const guint8 *data;
GstBuffer *buf = NULL, *sub = NULL;
guint offset = 4;
gint i;
GstTagList *tags = NULL;
2011-04-04 17:17:43 +00:00
guint8 fourcc[4];
GST_DEBUG ("Reading and parsing avi headers: %d", avi->header_state);
switch (avi->header_state) {
case GST_AVI_DEMUX_HEADER_TAG_LIST:
if (gst_avi_demux_peek_chunk (avi, &tag, &size)) {
2009-09-21 23:28:54 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
if (tag != GST_RIFF_TAG_LIST)
goto header_no_list;
gst_adapter_flush (avi->adapter, 8);
/* Find the 'hdrl' LIST tag */
GST_DEBUG ("Reading %d bytes", size);
buf = gst_adapter_take_buffer (avi->adapter, size);
2011-04-04 17:17:43 +00:00
gst_buffer_extract (buf, 0, fourcc, 4);
if (GST_READ_UINT32_LE (fourcc) != GST_RIFF_LIST_hdrl)
goto header_no_hdrl;
/* mind padding */
if (size & 1)
gst_adapter_flush (avi->adapter, 1);
GST_DEBUG ("'hdrl' LIST tag found. Parsing next chunk");
gst_avi_demux_roundup_list (avi, &buf);
/* the hdrl starts with a 'avih' header */
2010-01-15 17:13:24 +00:00
if (!gst_riff_parse_chunk (GST_ELEMENT_CAST (avi), buf, &offset, &tag,
&sub))
goto header_no_avih;
if (tag != GST_RIFF_TAG_avih)
goto header_no_avih;
2010-01-15 17:15:14 +00:00
if (!gst_avi_demux_parse_avih (avi, sub, &avi->avih))
goto header_wrong_avih;
GST_DEBUG_OBJECT (avi, "AVI header ok, reading elemnts from header");
/* now, read the elements from the header until the end */
2010-01-15 17:13:24 +00:00
while (gst_riff_parse_chunk (GST_ELEMENT_CAST (avi), buf, &offset, &tag,
&sub)) {
/* sub can be NULL on empty tags */
if (!sub)
continue;
switch (tag) {
case GST_RIFF_TAG_LIST:
2011-04-04 17:17:43 +00:00
if (gst_buffer_get_size (sub) < 4)
goto next;
2011-04-04 17:17:43 +00:00
gst_buffer_extract (sub, 0, fourcc, 4);
switch (GST_READ_UINT32_LE (fourcc)) {
case GST_RIFF_LIST_strl:
if (!(gst_avi_demux_parse_stream (avi, sub))) {
2009-08-04 09:39:39 +00:00
sub = NULL;
GST_ELEMENT_WARNING (avi, STREAM, DEMUX, (NULL),
("failed to parse stream, ignoring"));
goto next;
}
2009-08-04 09:39:39 +00:00
sub = NULL;
goto next;
case GST_RIFF_LIST_odml:
gst_avi_demux_parse_odml (avi, sub);
2009-08-04 09:39:39 +00:00
sub = NULL;
break;
default:
GST_WARNING_OBJECT (avi,
"Unknown list %" GST_FOURCC_FORMAT " in AVI header",
2011-04-04 17:17:43 +00:00
GST_FOURCC_ARGS (GST_READ_UINT32_LE (fourcc)));
/* fall-through */
case GST_RIFF_TAG_JUNQ:
case GST_RIFF_TAG_JUNK:
goto next;
2011-04-04 17:17:43 +00:00
break;
}
case GST_RIFF_IDIT:
gst_avi_demux_parse_idit (avi, sub);
goto next;
default:
GST_WARNING_OBJECT (avi,
"Unknown off %d tag %" GST_FOURCC_FORMAT " in AVI header",
offset, GST_FOURCC_ARGS (tag));
/* fall-through */
case GST_RIFF_TAG_JUNQ:
case GST_RIFF_TAG_JUNK:
next:
/* move to next chunk */
2009-08-04 09:39:39 +00:00
if (sub)
gst_buffer_unref (sub);
sub = NULL;
break;
}
}
gst_buffer_unref (buf);
GST_DEBUG ("elements parsed");
/* check parsed streams */
if (avi->num_streams == 0) {
goto no_streams;
} else if (avi->num_streams != avi->avih->streams) {
GST_WARNING_OBJECT (avi,
"Stream header mentioned %d streams, but %d available",
avi->avih->streams, avi->num_streams);
}
GST_DEBUG ("Get junk and info next");
avi->header_state = GST_AVI_DEMUX_HEADER_INFO;
} else {
/* Need more data */
return ret;
}
/* fall-though */
case GST_AVI_DEMUX_HEADER_INFO:
GST_DEBUG_OBJECT (avi, "skipping junk between header and data ...");
while (TRUE) {
if (gst_adapter_available (avi->adapter) < 12)
return GST_FLOW_OK;
2011-04-04 17:17:43 +00:00
data = gst_adapter_map (avi->adapter, 12);
tag = GST_READ_UINT32_LE (data);
size = GST_READ_UINT32_LE (data + 4);
ltag = GST_READ_UINT32_LE (data + 8);
2011-04-04 17:17:43 +00:00
gst_adapter_unmap (avi->adapter, 0);
if (tag == GST_RIFF_TAG_LIST) {
switch (ltag) {
case GST_RIFF_LIST_movi:
gst_adapter_flush (avi->adapter, 12);
2009-12-04 14:13:12 +00:00
if (!avi->first_movi_offset)
avi->first_movi_offset = avi->offset;
avi->offset += 12;
2009-12-04 14:13:12 +00:00
avi->idx1_offset = avi->offset + size - 4;
goto skipping_done;
case GST_RIFF_LIST_INFO:
GST_DEBUG ("Found INFO chunk");
if (gst_avi_demux_peek_chunk (avi, &tag, &size)) {
GST_DEBUG ("got size %d", size);
avi->offset += 12;
gst_adapter_flush (avi->adapter, 12);
if (size > 4) {
buf = gst_adapter_take_buffer (avi->adapter, size - 4);
/* mind padding */
if (size & 1)
gst_adapter_flush (avi->adapter, 1);
2010-01-15 17:13:24 +00:00
gst_riff_parse_info (GST_ELEMENT_CAST (avi), buf, &tags);
if (tags) {
if (avi->globaltags) {
gst_tag_list_insert (avi->globaltags, tags,
GST_TAG_MERGE_REPLACE);
} else {
avi->globaltags = tags;
}
}
tags = NULL;
gst_buffer_unref (buf);
2009-09-21 23:28:54 +00:00
avi->offset += GST_ROUND_UP_2 (size) - 4;
} else {
GST_DEBUG ("skipping INFO LIST prefix");
}
} else {
/* Need more data */
return GST_FLOW_OK;
}
break;
default:
if (gst_avi_demux_peek_chunk (avi, &tag, &size)) {
2009-09-21 23:28:54 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
gst_adapter_flush (avi->adapter, 8 + GST_ROUND_UP_2 (size));
// ??? goto iterate; ???
} else {
/* Need more data */
return GST_FLOW_OK;
}
break;
}
} else {
if (gst_avi_demux_peek_chunk (avi, &tag, &size)) {
2009-09-21 23:28:54 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
gst_adapter_flush (avi->adapter, 8 + GST_ROUND_UP_2 (size));
//goto iterate;
} else {
/* Need more data */
return GST_FLOW_OK;
}
}
}
break;
default:
GST_WARNING ("unhandled header state: %d", avi->header_state);
break;
}
skipping_done:
GST_DEBUG_OBJECT (avi, "skipping done ... (streams=%u, stream[0].indexes=%p)",
avi->num_streams, avi->stream[0].indexes);
GST_DEBUG ("Found movi chunk. Starting to stream data");
avi->state = GST_AVI_DEMUX_MOVI;
/* no indexes in push mode, but it still sets some variables */
gst_avi_demux_calculate_durations_from_index (avi);
gst_avi_demux_expose_streams (avi, TRUE);
/* prepare all streams for index 0 */
for (i = 0; i < avi->num_streams; i++)
avi->stream[i].current_entry = 0;
/* create initial NEWSEGMENT event */
if (avi->seg_event)
gst_event_unref (avi->seg_event);
2011-06-02 16:47:36 +00:00
avi->seg_event = gst_event_new_segment (&avi->segment);
2009-10-14 11:08:47 +00:00
gst_avi_demux_check_seekability (avi);
/* at this point we know all the streams and we can signal the no more
* pads signal */
GST_DEBUG_OBJECT (avi, "signaling no more pads");
2010-01-15 17:13:24 +00:00
gst_element_no_more_pads (GST_ELEMENT_CAST (avi));
return GST_FLOW_OK;
/* ERRORS */
no_streams:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL), ("No streams found"));
return GST_FLOW_ERROR;
}
header_no_list:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Invalid AVI header (no LIST at start): %"
GST_FOURCC_FORMAT, GST_FOURCC_ARGS (tag)));
return GST_FLOW_ERROR;
}
header_no_hdrl:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Invalid AVI header (no hdrl at start): %"
GST_FOURCC_FORMAT, GST_FOURCC_ARGS (tag)));
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
header_no_avih:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Invalid AVI header (no avih at start): %"
GST_FOURCC_FORMAT, GST_FOURCC_ARGS (tag)));
if (sub)
gst_buffer_unref (sub);
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
header_wrong_avih:
{
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
}
static void
gst_avi_demux_add_date_tag (GstAviDemux * avi, gint y, gint m, gint d,
gint h, gint min, gint s)
{
GDate *date;
GstDateTime *dt;
date = g_date_new_dmy (d, m, y);
if (!g_date_valid (date)) {
/* bogus date */
GST_WARNING_OBJECT (avi, "Refusing to add invalid date %d-%d-%d", y, m, d);
g_date_free (date);
return;
}
dt = gst_date_time_new_local_time (y, m, d, h, min, s);
if (avi->globaltags == NULL)
avi->globaltags = gst_tag_list_new ();
gst_tag_list_add (avi->globaltags, GST_TAG_MERGE_REPLACE, GST_TAG_DATE, date,
NULL);
g_date_free (date);
if (dt) {
gst_tag_list_add (avi->globaltags, GST_TAG_MERGE_REPLACE, GST_TAG_DATE_TIME,
dt, NULL);
gst_date_time_unref (dt);
}
}
static void
gst_avi_demux_parse_idit_nums_only (GstAviDemux * avi, gchar * data)
{
gint y, m, d;
gint hr = 0, min = 0, sec = 0;
gint ret;
GST_DEBUG ("data : '%s'", data);
ret = sscanf (data, "%d:%d:%d %d:%d:%d", &y, &m, &d, &hr, &min, &sec);
if (ret < 3) {
/* Attempt YYYY/MM/DD/ HH:MM variant (found in CASIO cameras) */
ret = sscanf (data, "%04d/%02d/%02d/ %d:%d", &y, &m, &d, &hr, &min);
if (ret < 3) {
GST_WARNING_OBJECT (avi, "Failed to parse IDIT tag");
return;
}
}
gst_avi_demux_add_date_tag (avi, y, m, d, hr, min, sec);
}
static gint
get_month_num (gchar * data, guint size)
{
if (g_ascii_strncasecmp (data, "jan", 3) == 0) {
return 1;
} else if (g_ascii_strncasecmp (data, "feb", 3) == 0) {
return 2;
} else if (g_ascii_strncasecmp (data, "mar", 3) == 0) {
return 3;
} else if (g_ascii_strncasecmp (data, "apr", 3) == 0) {
return 4;
} else if (g_ascii_strncasecmp (data, "may", 3) == 0) {
return 5;
} else if (g_ascii_strncasecmp (data, "jun", 3) == 0) {
return 6;
} else if (g_ascii_strncasecmp (data, "jul", 3) == 0) {
return 7;
} else if (g_ascii_strncasecmp (data, "aug", 3) == 0) {
return 8;
} else if (g_ascii_strncasecmp (data, "sep", 3) == 0) {
return 9;
} else if (g_ascii_strncasecmp (data, "oct", 3) == 0) {
return 10;
} else if (g_ascii_strncasecmp (data, "nov", 3) == 0) {
return 11;
} else if (g_ascii_strncasecmp (data, "dec", 3) == 0) {
return 12;
}
return 0;
}
static void
gst_avi_demux_parse_idit_text (GstAviDemux * avi, gchar * data)
{
gint year, month, day;
gint hour, min, sec;
gint ret;
gchar weekday[4];
gchar monthstr[4];
ret = sscanf (data, "%3s %3s %d %d:%d:%d %d", weekday, monthstr, &day, &hour,
&min, &sec, &year);
if (ret != 7) {
GST_WARNING_OBJECT (avi, "Failed to parse IDIT tag");
return;
}
month = get_month_num (monthstr, strlen (monthstr));
gst_avi_demux_add_date_tag (avi, year, month, day, hour, min, sec);
}
static void
gst_avi_demux_parse_idit (GstAviDemux * avi, GstBuffer * buf)
{
2011-04-04 17:17:43 +00:00
gchar *data, *ptr;
gsize size, left;
gchar *safedata = NULL;
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (buf, &size, NULL, GST_MAP_READ);
/*
* According to:
* http://www.eden-foundation.org/products/code/film_date_stamp/index.html
*
* This tag could be in one of the below formats
* 2005:08:17 11:42:43
* THU OCT 26 16:46:04 2006
* Mon Mar 3 09:44:56 2008
*
* FIXME: Our date tag doesn't include hours
*/
/* skip eventual initial whitespace */
2011-04-04 17:17:43 +00:00
ptr = data;
left = size;
while (left > 0 && g_ascii_isspace (ptr[0])) {
ptr++;
left--;
}
2011-04-04 17:17:43 +00:00
if (left == 0) {
goto non_parsable;
}
/* make a safe copy to add a \0 to the end of the string */
2011-04-04 17:17:43 +00:00
safedata = g_strndup (ptr, left);
/* test if the first char is a alpha or a number */
2011-04-04 17:17:43 +00:00
if (g_ascii_isdigit (ptr[0])) {
gst_avi_demux_parse_idit_nums_only (avi, safedata);
g_free (safedata);
return;
2011-04-04 17:17:43 +00:00
} else if (g_ascii_isalpha (ptr[0])) {
gst_avi_demux_parse_idit_text (avi, safedata);
g_free (safedata);
return;
}
g_free (safedata);
non_parsable:
GST_WARNING_OBJECT (avi, "IDIT tag has no parsable info");
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/*
* Read full AVI headers.
*/
static GstFlowReturn
gst_avi_demux_stream_header_pull (GstAviDemux * avi)
{
GstFlowReturn res;
GstBuffer *buf, *sub = NULL;
guint32 tag;
guint offset = 4;
GstElement *element = GST_ELEMENT_CAST (avi);
GstClockTime stamp;
GstTagList *tags = NULL;
2011-04-04 17:17:43 +00:00
guint8 fourcc[4];
stamp = gst_util_get_timestamp ();
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* the header consists of a 'hdrl' LIST tag */
res = gst_riff_read_chunk (element, avi->sinkpad, &avi->offset, &tag, &buf);
if (res != GST_FLOW_OK)
goto pull_range_failed;
else if (tag != GST_RIFF_TAG_LIST)
goto no_list;
2011-04-04 17:17:43 +00:00
else if (gst_buffer_get_size (buf) < 4)
goto no_header;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
GST_DEBUG_OBJECT (avi, "parsing headers");
/* Find the 'hdrl' LIST tag */
2011-04-04 17:17:43 +00:00
gst_buffer_extract (buf, 0, fourcc, 4);
while (GST_READ_UINT32_LE (fourcc) != GST_RIFF_LIST_hdrl) {
GST_LOG_OBJECT (avi, "buffer contains %" GST_FOURCC_FORMAT,
2011-04-04 17:17:43 +00:00
GST_FOURCC_ARGS (GST_READ_UINT32_LE (fourcc)));
/* Eat up */
gst_buffer_unref (buf);
/* read new chunk */
res = gst_riff_read_chunk (element, avi->sinkpad, &avi->offset, &tag, &buf);
if (res != GST_FLOW_OK)
goto pull_range_failed;
else if (tag != GST_RIFF_TAG_LIST)
goto no_list;
2011-04-04 17:17:43 +00:00
else if (gst_buffer_get_size (buf) < 4)
goto no_header;
2011-04-04 17:17:43 +00:00
gst_buffer_extract (buf, 0, fourcc, 4);
}
GST_DEBUG_OBJECT (avi, "hdrl LIST tag found");
gst_avi_demux_roundup_list (avi, &buf);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* the hdrl starts with a 'avih' header */
if (!gst_riff_parse_chunk (element, buf, &offset, &tag, &sub))
goto no_avih;
else if (tag != GST_RIFF_TAG_avih)
goto no_avih;
2010-01-15 17:15:14 +00:00
else if (!gst_avi_demux_parse_avih (avi, sub, &avi->avih))
goto invalid_avih;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
GST_DEBUG_OBJECT (avi, "AVI header ok, reading elements from header");
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* now, read the elements from the header until the end */
while (gst_riff_parse_chunk (element, buf, &offset, &tag, &sub)) {
2011-04-04 17:17:43 +00:00
gsize size;
guint8 *data;
/* sub can be NULL on empty tags */
if (!sub)
continue;
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (sub, &size, NULL, GST_MAP_READ);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
switch (tag) {
case GST_RIFF_TAG_LIST:
2011-04-04 17:17:43 +00:00
if (size < 4)
goto next;
2011-04-04 17:17:43 +00:00
switch (GST_READ_UINT32_LE (data)) {
case GST_RIFF_LIST_strl:
if (!(gst_avi_demux_parse_stream (avi, sub))) {
GST_ELEMENT_WARNING (avi, STREAM, DEMUX, (NULL),
2009-08-04 09:39:59 +00:00
("failed to parse stream, ignoring"));
sub = NULL;
}
2009-08-04 09:39:39 +00:00
sub = NULL;
goto next;
case GST_RIFF_LIST_odml:
gst_avi_demux_parse_odml (avi, sub);
2009-08-04 09:39:39 +00:00
sub = NULL;
break;
case GST_RIFF_LIST_INFO:
2011-04-04 17:17:43 +00:00
gst_buffer_resize (sub, 4, -1);
gst_riff_parse_info (element, sub, &tags);
if (tags) {
if (avi->globaltags) {
gst_tag_list_insert (avi->globaltags, tags,
GST_TAG_MERGE_REPLACE);
} else {
avi->globaltags = tags;
}
}
tags = NULL;
break;
default:
GST_WARNING_OBJECT (avi,
"Unknown list %" GST_FOURCC_FORMAT " in AVI header",
2011-04-04 17:17:43 +00:00
GST_FOURCC_ARGS (GST_READ_UINT32_LE (data)));
GST_MEMDUMP_OBJECT (avi, "Unknown list", data, size);
/* fall-through */
case GST_RIFF_TAG_JUNQ:
case GST_RIFF_TAG_JUNK:
goto next;
}
break;
case GST_RIFF_IDIT:
gst_avi_demux_parse_idit (avi, sub);
goto next;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
default:
GST_WARNING_OBJECT (avi,
"Unknown tag %" GST_FOURCC_FORMAT " in AVI header at off %d",
GST_FOURCC_ARGS (tag), offset);
2011-04-04 17:17:43 +00:00
GST_MEMDUMP_OBJECT (avi, "Unknown tag", data, size);
/* fall-through */
case GST_RIFF_TAG_JUNQ:
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
case GST_RIFF_TAG_JUNK:
next:
2011-04-04 17:17:43 +00:00
if (sub) {
gst_buffer_unmap (sub, data, size);
gst_buffer_unref (sub);
2011-04-04 17:17:43 +00:00
}
sub = NULL;
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
}
gst_buffer_unref (buf);
GST_DEBUG ("elements parsed");
/* check parsed streams */
if (avi->num_streams == 0)
goto no_streams;
else if (avi->num_streams != avi->avih->streams) {
GST_WARNING_OBJECT (avi,
"Stream header mentioned %d streams, but %d available",
avi->avih->streams, avi->num_streams);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
GST_DEBUG_OBJECT (avi, "skipping junk between header and data, offset=%"
G_GUINT64_FORMAT, avi->offset);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* Now, find the data (i.e. skip all junk between header and data) */
do {
guint size;
2011-04-04 17:17:43 +00:00
gsize bsize;
guint8 *data;
guint32 tag, ltag;
res = gst_pad_pull_range (avi->sinkpad, avi->offset, 12, &buf);
if (res != GST_FLOW_OK) {
GST_DEBUG_OBJECT (avi, "pull_range failure while looking for tags");
goto pull_range_failed;
2011-04-04 17:17:43 +00:00
} else if (gst_buffer_get_size (buf) < 12) {
GST_DEBUG_OBJECT (avi, "got %d bytes which is less than 12 bytes",
2011-04-04 17:17:43 +00:00
gst_buffer_get_size (buf));
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (buf, &bsize, NULL, GST_MAP_READ);
tag = GST_READ_UINT32_LE (data);
size = GST_READ_UINT32_LE (data + 4);
ltag = GST_READ_UINT32_LE (data + 8);
GST_DEBUG ("tag %" GST_FOURCC_FORMAT ", size %u",
GST_FOURCC_ARGS (tag), size);
2011-04-04 17:17:43 +00:00
GST_MEMDUMP ("Tag content", data, bsize);
gst_buffer_unmap (buf, data, bsize);
gst_buffer_unref (buf);
switch (tag) {
case GST_RIFF_TAG_LIST:{
switch (ltag) {
case GST_RIFF_LIST_movi:
GST_DEBUG_OBJECT (avi,
"Reached the 'movi' tag, we're done with skipping");
goto skipping_done;
case GST_RIFF_LIST_INFO:
res =
gst_riff_read_chunk (element, avi->sinkpad, &avi->offset, &tag,
&buf);
if (res != GST_FLOW_OK) {
GST_DEBUG_OBJECT (avi, "couldn't read INFO chunk");
goto pull_range_failed;
}
2011-04-04 17:17:43 +00:00
GST_DEBUG ("got size %u", gst_buffer_get_size (buf));
if (size < 4) {
GST_DEBUG ("skipping INFO LIST prefix");
avi->offset += (4 - GST_ROUND_UP_2 (size));
gst_buffer_unref (buf);
continue;
}
2011-04-04 17:17:43 +00:00
sub = gst_buffer_copy_region (buf, GST_BUFFER_COPY_ALL, 4, -1);
gst_riff_parse_info (element, sub, &tags);
if (tags) {
if (avi->globaltags) {
gst_tag_list_insert (avi->globaltags, tags,
GST_TAG_MERGE_REPLACE);
} else {
avi->globaltags = tags;
}
}
tags = NULL;
if (sub) {
gst_buffer_unref (sub);
sub = NULL;
}
gst_buffer_unref (buf);
/* gst_riff_read_chunk() has already advanced avi->offset */
break;
default:
GST_WARNING_OBJECT (avi,
"Skipping unknown list tag %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (ltag));
2009-09-21 23:28:54 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
break;
}
}
break;
default:
GST_WARNING_OBJECT (avi, "Skipping unknown tag %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (tag));
/* Fall-through */
case GST_MAKE_FOURCC ('J', 'U', 'N', 'Q'):
case GST_MAKE_FOURCC ('J', 'U', 'N', 'K'):
/* Only get buffer for debugging if the memdump is needed */
if (gst_debug_category_get_threshold (GST_CAT_DEFAULT) >= 9) {
res = gst_pad_pull_range (avi->sinkpad, avi->offset, size, &buf);
if (res != GST_FLOW_OK) {
GST_DEBUG_OBJECT (avi, "couldn't read INFO chunk");
goto pull_range_failed;
}
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (buf, &bsize, NULL, GST_MAP_READ);
GST_MEMDUMP ("Junk", data, bsize);
gst_buffer_unmap (buf, data, bsize);
gst_buffer_unref (buf);
}
2009-09-21 23:28:54 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
break;
}
} while (1);
skipping_done:
GST_DEBUG_OBJECT (avi, "skipping done ... (streams=%u, stream[0].indexes=%p)",
avi->num_streams, avi->stream[0].indexes);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
/* create or read stream index (for seeking) */
if (avi->stream[0].indexes != NULL) {
/* we read a super index already (gst_avi_demux_parse_superindex() ) */
gst_avi_demux_read_subindexes_pull (avi);
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
}
if (!avi->have_index) {
if (avi->avih->flags & GST_RIFF_AVIH_HASINDEX)
gst_avi_demux_stream_index (avi);
/* still no index, scan */
if (!avi->have_index) {
gst_avi_demux_stream_scan (avi);
/* still no index.. this is a fatal error for now.
* FIXME, we should switch to plain push mode without seeking
* instead of failing. */
if (!avi->have_index)
goto no_index;
}
}
/* use the indexes now to construct nice durations */
gst_avi_demux_calculate_durations_from_index (avi);
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
gst_avi_demux_expose_streams (avi, FALSE);
/* do initial seek to the default segment values */
gst_avi_demux_do_seek (avi, &avi->segment);
2011-06-02 16:47:36 +00:00
/* create initial NEWSEGMENT event */
if (avi->seg_event)
gst_event_unref (avi->seg_event);
2011-06-02 16:47:36 +00:00
avi->seg_event = gst_event_new_segment (&avi->segment);
stamp = gst_util_get_timestamp () - stamp;
GST_DEBUG_OBJECT (avi, "pulling header took %" GST_TIME_FORMAT,
GST_TIME_ARGS (stamp));
/* at this point we know all the streams and we can signal the no more
* pads signal */
GST_DEBUG_OBJECT (avi, "signaling no more pads");
gst_element_no_more_pads (GST_ELEMENT_CAST (avi));
return GST_FLOW_OK;
/* ERRORS */
no_list:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Invalid AVI header (no LIST at start): %"
GST_FOURCC_FORMAT, GST_FOURCC_ARGS (tag)));
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
no_header:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Invalid AVI header (no hdrl at start): %"
GST_FOURCC_FORMAT, GST_FOURCC_ARGS (tag)));
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
no_avih:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Invalid AVI header (no avih at start): %"
GST_FOURCC_FORMAT, GST_FOURCC_ARGS (tag)));
if (sub)
gst_buffer_unref (sub);
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
invalid_avih:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Invalid AVI header (cannot parse avih at start)"));
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
no_streams:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL), ("No streams found"));
return GST_FLOW_ERROR;
}
no_index:
{
GST_WARNING ("file without or too big index");
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("Could not get/create index"));
return GST_FLOW_ERROR;
}
pull_range_failed:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL),
("pull_range flow reading header: %s", gst_flow_get_name (res)));
return GST_FLOW_ERROR;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
}
/* move a stream to @index */
static void
gst_avi_demux_move_stream (GstAviDemux * avi, GstAviStream * stream,
GstSegment * segment, guint index)
{
GST_DEBUG_OBJECT (avi, "Move stream %d to %u", stream->num, index);
if (segment->rate < 0.0) {
guint next_key;
/* Because we don't know the frame order we need to push from the prev keyframe
* to the next keyframe. If there is a smart decoder downstream he will notice
* that there are too many encoded frames send and return UNEXPECTED when there
* are enough decoded frames to fill the segment. */
2009-09-22 10:35:30 +00:00
next_key = gst_avi_demux_index_next (avi, stream, index, TRUE);
/* FIXME, we go back to 0, we should look at segment.start. We will however
* stop earlier when the see the timestamp < segment.start */
stream->start_entry = 0;
stream->step_entry = index;
stream->current_entry = index;
stream->stop_entry = next_key;
GST_DEBUG_OBJECT (avi, "reverse seek: start %u, step %u, stop %u",
stream->start_entry, stream->step_entry, stream->stop_entry);
} else {
stream->start_entry = index;
stream->step_entry = index;
stream->stop_entry = gst_avi_demux_index_last (avi, stream);
}
if (stream->current_entry != index) {
GST_DEBUG_OBJECT (avi, "Move DISCONT from %u to %u",
stream->current_entry, index);
stream->current_entry = index;
stream->discont = TRUE;
}
/* update the buffer info */
gst_avi_demux_get_buffer_info (avi, stream, index,
&stream->current_timestamp, &stream->current_ts_end,
&stream->current_offset, &stream->current_offset_end);
GST_DEBUG_OBJECT (avi, "Moved to %u, ts %" GST_TIME_FORMAT
", ts_end %" GST_TIME_FORMAT ", off %" G_GUINT64_FORMAT
", off_end %" G_GUINT64_FORMAT, index,
GST_TIME_ARGS (stream->current_timestamp),
GST_TIME_ARGS (stream->current_ts_end), stream->current_offset,
stream->current_offset_end);
2009-12-04 14:13:12 +00:00
GST_DEBUG_OBJECT (avi, "Seeking to offset %" G_GUINT64_FORMAT,
stream->index[index].offset);
}
/*
* Do the actual seeking.
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
*/
static gboolean
gst_avi_demux_do_seek (GstAviDemux * avi, GstSegment * segment)
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
{
GstClockTime seek_time;
gboolean keyframe;
guint i, index;
GstAviStream *stream;
2011-06-02 16:47:36 +00:00
seek_time = segment->position;
keyframe = ! !(segment->flags & GST_SEEK_FLAG_KEY_UNIT);
GST_DEBUG_OBJECT (avi, "seek to: %" GST_TIME_FORMAT
" keyframe seeking:%d", GST_TIME_ARGS (seek_time), keyframe);
/* FIXME, this code assumes the main stream with keyframes is stream 0,
* which is mostly correct... */
stream = &avi->stream[avi->main_stream];
/* get the entry index for the requested position */
index = gst_avi_demux_index_for_time (avi, stream, seek_time);
GST_DEBUG_OBJECT (avi, "Got entry %u", index);
/* check if we are already on a keyframe */
if (!ENTRY_IS_KEYFRAME (&stream->index[index])) {
GST_DEBUG_OBJECT (avi, "not keyframe, searching back");
/* now go to the previous keyframe, this is where we should start
* decoding from. */
2009-09-22 10:35:30 +00:00
index = gst_avi_demux_index_prev (avi, stream, index, TRUE);
GST_DEBUG_OBJECT (avi, "previous keyframe at %u", index);
}
/* move the main stream to this position */
gst_avi_demux_move_stream (avi, stream, segment, index);
if (keyframe) {
/* when seeking to a keyframe, we update the result seek time
* to the time of the keyframe. */
seek_time = stream->current_timestamp;
GST_DEBUG_OBJECT (avi, "keyframe adjusted to %" GST_TIME_FORMAT,
GST_TIME_ARGS (seek_time));
}
2011-06-02 16:47:36 +00:00
/* the seek time is also the position and stream time when going
2009-09-22 16:18:20 +00:00
* forwards */
2011-06-02 16:47:36 +00:00
segment->position = seek_time;
2009-09-22 16:18:20 +00:00
if (segment->rate > 0.0)
segment->time = seek_time;
/* now set DISCONT and align the other streams */
for (i = 0; i < avi->num_streams; i++) {
GstAviStream *ostream;
ostream = &avi->stream[i];
2009-10-18 14:02:01 +00:00
if ((ostream == stream) || (ostream->index == NULL))
continue;
/* get the entry index for the requested position */
index = gst_avi_demux_index_for_time (avi, ostream, seek_time);
/* move to previous keyframe */
if (!ENTRY_IS_KEYFRAME (&ostream->index[index]))
2009-09-22 10:35:30 +00:00
index = gst_avi_demux_index_prev (avi, ostream, index, TRUE);
gst_avi_demux_move_stream (avi, ostream, segment, index);
}
GST_DEBUG_OBJECT (avi, "done seek to: %" GST_TIME_FORMAT,
GST_TIME_ARGS (seek_time));
return TRUE;
}
/*
* Handle seek event in pull mode.
*/
static gboolean
gst_avi_demux_handle_seek (GstAviDemux * avi, GstPad * pad, GstEvent * event)
{
gdouble rate;
GstFormat format;
GstSeekFlags flags;
GstSeekType cur_type = GST_SEEK_TYPE_NONE, stop_type;
gint64 cur, stop;
gboolean flush;
gboolean update;
GstSegment seeksegment = { 0, };
2009-09-22 16:18:20 +00:00
gint i;
if (event) {
GST_DEBUG_OBJECT (avi, "doing seek with event");
gst_event_parse_seek (event, &rate, &format, &flags,
&cur_type, &cur, &stop_type, &stop);
/* we have to have a format as the segment format. Try to convert
* if not. */
if (format != GST_FORMAT_TIME) {
gboolean res = TRUE;
if (cur_type != GST_SEEK_TYPE_NONE)
res = gst_pad_query_convert (pad, format, cur, GST_FORMAT_TIME, &cur);
if (res && stop_type != GST_SEEK_TYPE_NONE)
res = gst_pad_query_convert (pad, format, stop, GST_FORMAT_TIME, &stop);
if (!res)
goto no_format;
format = GST_FORMAT_TIME;
}
GST_DEBUG_OBJECT (avi,
"seek requested: rate %g cur %" GST_TIME_FORMAT " stop %"
GST_TIME_FORMAT, rate, GST_TIME_ARGS (cur), GST_TIME_ARGS (stop));
/* FIXME: can we do anything with rate!=1.0 */
} else {
GST_DEBUG_OBJECT (avi, "doing seek without event");
flags = 0;
rate = 1.0;
}
/* save flush flag */
flush = flags & GST_SEEK_FLAG_FLUSH;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
if (flush) {
2009-09-22 16:18:20 +00:00
GstEvent *fevent = gst_event_new_flush_start ();
/* for a flushing seek, we send a flush_start on all pads. This will
* eventually stop streaming with a WRONG_STATE. We can thus eventually
* take the STREAM_LOCK. */
GST_DEBUG_OBJECT (avi, "sending flush start");
2009-09-22 16:18:20 +00:00
gst_avi_demux_push_event (avi, gst_event_ref (fevent));
gst_pad_push_event (avi->sinkpad, fevent);
} else {
/* a non-flushing seek, we PAUSE the task so that we can take the
* STREAM_LOCK */
GST_DEBUG_OBJECT (avi, "non flushing seek, pausing task");
gst_pad_pause_task (avi->sinkpad);
}
/* wait for streaming to stop */
GST_DEBUG_OBJECT (avi, "wait for streaming to stop");
GST_PAD_STREAM_LOCK (avi->sinkpad);
/* copy segment, we need this because we still need the old
* segment when we close the current segment. */
memcpy (&seeksegment, &avi->segment, sizeof (GstSegment));
if (event) {
GST_DEBUG_OBJECT (avi, "configuring seek");
2011-06-02 16:47:36 +00:00
gst_segment_do_seek (&seeksegment, rate, format, flags,
cur_type, cur, stop_type, stop, &update);
}
2011-06-02 16:47:36 +00:00
/* do the seek, seeksegment.position contains the new position, this
* actually never fails. */
gst_avi_demux_do_seek (avi, &seeksegment);
if (flush) {
GstEvent *fevent = gst_event_new_flush_stop (TRUE);
GST_DEBUG_OBJECT (avi, "sending flush stop");
2009-09-22 16:18:20 +00:00
gst_avi_demux_push_event (avi, gst_event_ref (fevent));
gst_pad_push_event (avi->sinkpad, fevent);
}
/* now update the real segment info */
memcpy (&avi->segment, &seeksegment, sizeof (GstSegment));
/* post the SEGMENT_START message when we do segmented playback */
if (avi->segment.flags & GST_SEEK_FLAG_SEGMENT) {
2010-01-15 17:13:24 +00:00
gst_element_post_message (GST_ELEMENT_CAST (avi),
gst_message_new_segment_start (GST_OBJECT_CAST (avi),
2011-06-02 16:47:36 +00:00
avi->segment.format, avi->segment.position));
}
/* queue the segment event for the streaming thread. */
if (avi->seg_event)
gst_event_unref (avi->seg_event);
2011-06-02 16:47:36 +00:00
avi->seg_event = gst_event_new_segment (&avi->segment);
if (!avi->streaming) {
gst_pad_start_task (avi->sinkpad, (GstTaskFunction) gst_avi_demux_loop,
avi->sinkpad);
}
/* reset the last flow and mark discont, seek is always DISCONT */
for (i = 0; i < avi->num_streams; i++) {
GST_DEBUG_OBJECT (avi, "marking DISCONT");
avi->stream[i].last_flow = GST_FLOW_OK;
avi->stream[i].discont = TRUE;
}
GST_PAD_STREAM_UNLOCK (avi->sinkpad);
return TRUE;
/* ERRORS */
no_format:
{
GST_DEBUG_OBJECT (avi, "unsupported format given, seek aborted.");
return FALSE;
}
}
2009-12-04 14:13:12 +00:00
/*
* Handle seek event in push mode.
2009-12-04 14:13:12 +00:00
*/
static gboolean
2009-12-04 14:13:12 +00:00
avi_demux_handle_seek_push (GstAviDemux * avi, GstPad * pad, GstEvent * event)
{
gdouble rate;
GstFormat format;
GstSeekFlags flags;
2009-12-04 14:13:12 +00:00
GstSeekType cur_type = GST_SEEK_TYPE_NONE, stop_type;
gint64 cur, stop;
2009-12-04 14:13:12 +00:00
gboolean keyframe;
GstAviStream *stream;
guint index;
guint n, str_num;
guint64 min_offset;
GstSegment seeksegment;
gboolean update;
2009-12-04 14:13:12 +00:00
/* check we have the index */
if (!avi->have_index) {
GST_DEBUG_OBJECT (avi, "no seek index built, seek aborted.");
return FALSE;
} else {
GST_DEBUG_OBJECT (avi, "doing push-based seek with event");
}
gst_event_parse_seek (event, &rate, &format, &flags,
&cur_type, &cur, &stop_type, &stop);
2009-12-04 14:13:12 +00:00
if (format != GST_FORMAT_TIME) {
gboolean res = TRUE;
2009-12-04 14:13:12 +00:00
if (cur_type != GST_SEEK_TYPE_NONE)
res = gst_pad_query_convert (pad, format, cur, GST_FORMAT_TIME, &cur);
2009-12-04 14:13:12 +00:00
if (res && stop_type != GST_SEEK_TYPE_NONE)
res = gst_pad_query_convert (pad, format, stop, GST_FORMAT_TIME, &stop);
2009-12-04 14:13:12 +00:00
if (!res) {
GST_DEBUG_OBJECT (avi, "unsupported format given, seek aborted.");
return FALSE;
}
format = GST_FORMAT_TIME;
2009-12-04 14:13:12 +00:00
}
/* let gst_segment handle any tricky stuff */
GST_DEBUG_OBJECT (avi, "configuring seek");
memcpy (&seeksegment, &avi->segment, sizeof (GstSegment));
2011-06-02 16:47:36 +00:00
gst_segment_do_seek (&seeksegment, rate, format, flags,
cur_type, cur, stop_type, stop, &update);
keyframe = ! !(flags & GST_SEEK_FLAG_KEY_UNIT);
2011-06-02 16:47:36 +00:00
cur = seeksegment.position;
GST_DEBUG_OBJECT (avi,
"Seek requested: ts %" GST_TIME_FORMAT " stop %" GST_TIME_FORMAT
", kf %u, rate %lf", GST_TIME_ARGS (cur), GST_TIME_ARGS (stop), keyframe,
rate);
if (rate < 0) {
GST_DEBUG_OBJECT (avi, "negative rate seek not supported in push mode");
return FALSE;
}
2009-12-04 14:13:12 +00:00
/* FIXME, this code assumes the main stream with keyframes is stream 0,
* which is mostly correct... */
str_num = avi->main_stream;
stream = &avi->stream[str_num];
2009-12-04 14:13:12 +00:00
/* get the entry index for the requested position */
index = gst_avi_demux_index_for_time (avi, stream, cur);
GST_DEBUG_OBJECT (avi, "str %u: Found entry %u for %" GST_TIME_FORMAT,
str_num, index, GST_TIME_ARGS (cur));
2009-12-04 14:13:12 +00:00
/* check if we are already on a keyframe */
if (!ENTRY_IS_KEYFRAME (&stream->index[index])) {
GST_DEBUG_OBJECT (avi, "Entry is not a keyframe - searching back");
2009-12-04 14:13:12 +00:00
/* now go to the previous keyframe, this is where we should start
* decoding from. */
index = gst_avi_demux_index_prev (avi, stream, index, TRUE);
GST_DEBUG_OBJECT (avi, "Found previous keyframe at %u", index);
2009-12-04 14:13:12 +00:00
}
gst_avi_demux_get_buffer_info (avi, stream, index,
&stream->current_timestamp, &stream->current_ts_end,
&stream->current_offset, &stream->current_offset_end);
/* re-use cur to be the timestamp of the seek as it _will_ be */
cur = stream->current_timestamp;
min_offset = stream->index[index].offset;
avi->seek_kf_offset = min_offset - 8;
GST_DEBUG_OBJECT (avi,
"Seek to: ts %" GST_TIME_FORMAT " (on str %u, idx %u, offset %"
G_GUINT64_FORMAT ")", GST_TIME_ARGS (stream->current_timestamp), str_num,
index, min_offset);
2009-12-04 14:13:12 +00:00
for (n = 0; n < avi->num_streams; n++) {
GstAviStream *str = &avi->stream[n];
guint idx;
2009-12-04 14:13:12 +00:00
if (n == avi->main_stream)
continue;
2009-12-04 14:13:12 +00:00
/* get the entry index for the requested position */
idx = gst_avi_demux_index_for_time (avi, str, cur);
GST_DEBUG_OBJECT (avi, "str %u: Found entry %u for %" GST_TIME_FORMAT, n,
idx, GST_TIME_ARGS (cur));
2009-12-04 14:13:12 +00:00
/* check if we are already on a keyframe */
if (!ENTRY_IS_KEYFRAME (&str->index[idx])) {
GST_DEBUG_OBJECT (avi, "Entry is not a keyframe - searching back");
2009-12-04 14:13:12 +00:00
/* now go to the previous keyframe, this is where we should start
* decoding from. */
idx = gst_avi_demux_index_prev (avi, str, idx, TRUE);
GST_DEBUG_OBJECT (avi, "Found previous keyframe at %u", idx);
2009-12-04 14:13:12 +00:00
}
gst_avi_demux_get_buffer_info (avi, str, idx,
&str->current_timestamp, &str->current_ts_end,
&str->current_offset, &str->current_offset_end);
if (str->index[idx].offset < min_offset) {
min_offset = str->index[idx].offset;
2009-12-04 14:13:12 +00:00
GST_DEBUG_OBJECT (avi,
"Found an earlier offset at %" G_GUINT64_FORMAT ", str %u",
min_offset, n);
2009-12-04 14:13:12 +00:00
str_num = n;
stream = str;
index = idx;
}
}
GST_DEBUG_OBJECT (avi,
"Seek performed: str %u, offset %" G_GUINT64_FORMAT ", idx %u, ts %"
GST_TIME_FORMAT ", ts_end %" GST_TIME_FORMAT ", off %" G_GUINT64_FORMAT
", off_end %" G_GUINT64_FORMAT, str_num, min_offset, index,
2009-12-04 14:13:12 +00:00
GST_TIME_ARGS (stream->current_timestamp),
GST_TIME_ARGS (stream->current_ts_end), stream->current_offset,
stream->current_offset_end);
/* index data refers to data, not chunk header (for pull mode convenience) */
min_offset -= 8;
GST_DEBUG_OBJECT (avi, "seeking to chunk at offset %" G_GUINT64_FORMAT,
min_offset);
if (!perform_seek_to_offset (avi, min_offset)) {
2009-12-04 14:13:12 +00:00
GST_DEBUG_OBJECT (avi, "seek event failed!");
return FALSE;
}
2009-12-04 14:13:12 +00:00
return TRUE;
}
/*
* Handle whether we can perform the seek event or if we have to let the chain
* function handle seeks to build the seek indexes first.
*/
static gboolean
gst_avi_demux_handle_seek_push (GstAviDemux * avi, GstPad * pad,
GstEvent * event)
{
/* check for having parsed index already */
if (!avi->have_index) {
2010-03-11 16:02:44 +00:00
guint64 offset = 0;
gboolean building_index;
2009-12-04 14:13:12 +00:00
GST_OBJECT_LOCK (avi);
2009-12-04 14:13:12 +00:00
/* handle the seek event in the chain function */
avi->state = GST_AVI_DEMUX_SEEK;
/* copy the event */
if (avi->seek_event)
gst_event_unref (avi->seek_event);
avi->seek_event = gst_event_ref (event);
2009-12-04 14:13:12 +00:00
/* set the building_index flag so that only one thread can setup the
* structures for index seeking. */
building_index = avi->building_index;
if (!building_index) {
avi->building_index = TRUE;
2009-12-04 14:13:12 +00:00
if (avi->stream[0].indexes) {
avi->odml_stream = 0;
avi->odml_subidxs = avi->stream[avi->odml_stream].indexes;
offset = avi->odml_subidxs[0];
} else {
offset = avi->idx1_offset;
}
}
GST_OBJECT_UNLOCK (avi);
2009-12-04 14:13:12 +00:00
if (!building_index) {
2009-12-04 14:13:12 +00:00
/* seek to the first subindex or legacy index */
GST_INFO_OBJECT (avi,
"Seeking to legacy index/first subindex at %" G_GUINT64_FORMAT,
offset);
return perform_seek_to_offset (avi, offset);
2009-12-04 14:13:12 +00:00
}
/* FIXME: we have to always return true so that we don't block the seek
* thread.
* Note: maybe it is OK to return true if we're still building the index */
return TRUE;
}
2009-12-04 14:13:12 +00:00
return avi_demux_handle_seek_push (avi, pad, event);
}
/*
* Helper for gst_avi_demux_invert()
*/
static inline void
swap_line (guint8 * d1, guint8 * d2, guint8 * tmp, gint bytes)
{
memcpy (tmp, d1, bytes);
memcpy (d1, d2, bytes);
memcpy (d2, tmp, bytes);
}
#define gst_avi_demux_is_uncompressed(fourcc) \
(fourcc && \
(fourcc == GST_RIFF_DIB || \
fourcc == GST_RIFF_rgb || \
fourcc == GST_RIFF_RGB || fourcc == GST_RIFF_RAW))
/*
* Invert DIB buffers... Takes existing buffer and
* returns either the buffer or a new one (with old
* one dereferenced).
* FIXME: can't we preallocate tmp? and remember stride, bpp?
*/
static GstBuffer *
gst_avi_demux_invert (GstAviStream * stream, GstBuffer * buf)
{
GstStructure *s;
gint y, w, h;
gint bpp, stride;
2011-04-04 17:17:43 +00:00
guint8 *tmp = NULL, *data;
gsize size;
2011-06-02 16:47:36 +00:00
GstCaps *caps;
if (stream->strh->type != GST_RIFF_FCC_vids)
return buf;
if (!gst_avi_demux_is_uncompressed (stream->strh->fcc_handler)) {
return buf; /* Ignore non DIB buffers */
}
2011-06-02 16:47:36 +00:00
caps = gst_pad_get_current_caps (stream->pad);
s = gst_caps_get_structure (caps, 0);
gst_caps_unref (caps);
if (!gst_structure_get_int (s, "bpp", &bpp)) {
GST_WARNING ("Failed to retrieve depth from caps");
return buf;
}
if (stream->strf.vids == NULL) {
GST_WARNING ("Failed to retrieve vids for stream");
return buf;
}
h = stream->strf.vids->height;
w = stream->strf.vids->width;
stride = w * (bpp / 8);
buf = gst_buffer_make_writable (buf);
2011-04-04 17:17:43 +00:00
data = gst_buffer_map (buf, &size, NULL, GST_MAP_READWRITE);
if (size < (stride * h)) {
GST_WARNING ("Buffer is smaller than reported Width x Height x Depth");
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
return buf;
}
tmp = g_malloc (stride);
for (y = 0; y < h / 2; y++) {
2011-04-04 17:17:43 +00:00
swap_line (data + stride * y, data + stride * (h - 1 - y), tmp, stride);
}
g_free (tmp);
2011-04-04 17:17:43 +00:00
gst_buffer_unmap (buf, data, size);
return buf;
}
2009-10-14 11:08:47 +00:00
static void
gst_avi_demux_add_assoc (GstAviDemux * avi, GstAviStream * stream,
GstClockTime timestamp, guint64 offset, gboolean keyframe)
{
/* do not add indefinitely for open-ended streaming */
if (G_UNLIKELY (avi->element_index && avi->seekable)) {
GST_LOG_OBJECT (avi, "adding association %" GST_TIME_FORMAT "-> %"
G_GUINT64_FORMAT, GST_TIME_ARGS (timestamp), offset);
gst_index_add_association (avi->element_index, avi->index_id,
keyframe ? GST_ASSOCIATION_FLAG_KEY_UNIT :
GST_ASSOCIATION_FLAG_DELTA_UNIT, GST_FORMAT_TIME, timestamp,
GST_FORMAT_BYTES, offset, NULL);
/* current_entry is DEFAULT (frame #) */
2009-10-14 11:08:47 +00:00
gst_index_add_association (avi->element_index, stream->index_id,
keyframe ? GST_ASSOCIATION_FLAG_KEY_UNIT :
GST_ASSOCIATION_FLAG_DELTA_UNIT, GST_FORMAT_TIME, timestamp,
GST_FORMAT_BYTES, offset, GST_FORMAT_DEFAULT, stream->current_entry,
NULL);
2009-10-14 11:08:47 +00:00
}
}
/*
* Returns the aggregated GstFlowReturn.
*/
static GstFlowReturn
gst_avi_demux_combine_flows (GstAviDemux * avi, GstAviStream * stream,
GstFlowReturn ret)
{
guint i;
gboolean unexpected = FALSE, not_linked = TRUE;
/* store the value */
stream->last_flow = ret;
/* any other error that is not-linked or eos can be returned right away */
if (G_LIKELY (ret != GST_FLOW_UNEXPECTED && ret != GST_FLOW_NOT_LINKED))
goto done;
/* only return NOT_LINKED if all other pads returned NOT_LINKED */
for (i = 0; i < avi->num_streams; i++) {
GstAviStream *ostream = &avi->stream[i];
ret = ostream->last_flow;
/* no unexpected or unlinked, return */
if (G_LIKELY (ret != GST_FLOW_UNEXPECTED && ret != GST_FLOW_NOT_LINKED))
goto done;
/* we check to see if we have at least 1 unexpected or all unlinked */
unexpected |= (ret == GST_FLOW_UNEXPECTED);
not_linked &= (ret == GST_FLOW_NOT_LINKED);
}
/* when we get here, we all have unlinked or unexpected */
if (not_linked)
ret = GST_FLOW_NOT_LINKED;
else if (unexpected)
ret = GST_FLOW_UNEXPECTED;
done:
GST_LOG_OBJECT (avi, "combined %s to return %s",
gst_flow_get_name (stream->last_flow), gst_flow_get_name (ret));
return ret;
}
/* move @stream to the next position in its index */
static GstFlowReturn
gst_avi_demux_advance (GstAviDemux * avi, GstAviStream * stream,
GstFlowReturn ret)
{
guint old_entry, new_entry;
old_entry = stream->current_entry;
/* move forwards */
new_entry = old_entry + 1;
/* see if we reached the end */
if (new_entry >= stream->stop_entry) {
if (avi->segment.rate < 0.0) {
if (stream->step_entry == stream->start_entry) {
/* we stepped all the way to the start, eos */
GST_DEBUG_OBJECT (avi, "reverse reached start %u", stream->start_entry);
goto eos;
}
/* backwards, stop becomes step, find a new step */
stream->stop_entry = stream->step_entry;
stream->step_entry = gst_avi_demux_index_prev (avi, stream,
2009-09-22 10:35:30 +00:00
stream->stop_entry, TRUE);
GST_DEBUG_OBJECT (avi,
"reverse playback jump: start %u, step %u, stop %u",
stream->start_entry, stream->step_entry, stream->stop_entry);
/* and start from the previous keyframe now */
new_entry = stream->step_entry;
} else {
/* EOS */
GST_DEBUG_OBJECT (avi, "forward reached stop %u", stream->stop_entry);
goto eos;
}
}
if (new_entry != old_entry) {
stream->current_entry = new_entry;
stream->current_total = stream->index[new_entry].total;
if (new_entry == old_entry + 1) {
GST_DEBUG_OBJECT (avi, "moved forwards from %u to %u",
old_entry, new_entry);
/* we simply moved one step forwards, reuse current info */
stream->current_timestamp = stream->current_ts_end;
stream->current_offset = stream->current_offset_end;
gst_avi_demux_get_buffer_info (avi, stream, new_entry,
NULL, &stream->current_ts_end, NULL, &stream->current_offset_end);
} else {
/* we moved DISCONT, full update */
gst_avi_demux_get_buffer_info (avi, stream, new_entry,
&stream->current_timestamp, &stream->current_ts_end,
&stream->current_offset, &stream->current_offset_end);
/* and MARK discont for this stream */
stream->last_flow = GST_FLOW_OK;
stream->discont = TRUE;
2009-09-22 16:18:20 +00:00
GST_DEBUG_OBJECT (avi, "Moved from %u to %u, ts %" GST_TIME_FORMAT
", ts_end %" GST_TIME_FORMAT ", off %" G_GUINT64_FORMAT
", off_end %" G_GUINT64_FORMAT, old_entry, new_entry,
GST_TIME_ARGS (stream->current_timestamp),
GST_TIME_ARGS (stream->current_ts_end), stream->current_offset,
stream->current_offset_end);
}
}
return ret;
/* ERROR */
eos:
{
GST_DEBUG_OBJECT (avi, "we are EOS");
/* setting current_timestamp to -1 marks EOS */
stream->current_timestamp = -1;
return GST_FLOW_UNEXPECTED;
}
}
2009-09-22 16:18:20 +00:00
/* find the stream with the lowest current position when going forwards or with
* the highest position when going backwards, this is the stream
* we should push from next */
static gint
gst_avi_demux_find_next (GstAviDemux * avi, gfloat rate)
{
guint64 min_time, max_time;
guint stream_num, i;
max_time = 0;
min_time = G_MAXUINT64;
stream_num = -1;
for (i = 0; i < avi->num_streams; i++) {
guint64 position;
GstAviStream *stream;
stream = &avi->stream[i];
/* ignore streams that finished */
if (stream->last_flow == GST_FLOW_UNEXPECTED)
continue;
2009-09-22 16:18:20 +00:00
position = stream->current_timestamp;
/* position of -1 is EOS */
if (position != -1) {
if (rate > 0.0 && position < min_time) {
min_time = position;
stream_num = i;
} else if (rate < 0.0 && position >= max_time) {
max_time = position;
stream_num = i;
}
}
}
return stream_num;
}
static GstFlowReturn
gst_avi_demux_loop_data (GstAviDemux * avi)
{
GstFlowReturn ret = GST_FLOW_OK;
2009-09-22 16:18:20 +00:00
guint stream_num;
GstAviStream *stream;
gboolean processed = FALSE;
GstBuffer *buf;
guint64 offset, size;
GstClockTime timestamp, duration;
guint64 out_offset, out_offset_end;
gboolean keyframe;
GstAviIndexEntry *entry;
do {
2009-09-22 16:18:20 +00:00
stream_num = gst_avi_demux_find_next (avi, avi->segment.rate);
/* all are EOS */
if (G_UNLIKELY (stream_num == -1)) {
GST_DEBUG_OBJECT (avi, "all streams are EOS");
goto eos;
}
/* we have the stream now */
stream = &avi->stream[stream_num];
/* skip streams without pads */
if (!stream->pad) {
GST_DEBUG_OBJECT (avi, "skipping entry from stream %d without pad",
stream_num);
goto next;
}
/* get the timing info for the entry */
timestamp = stream->current_timestamp;
duration = stream->current_ts_end - timestamp;
out_offset = stream->current_offset;
out_offset_end = stream->current_offset_end;
/* get the entry data info */
entry = &stream->index[stream->current_entry];
offset = entry->offset;
size = entry->size;
keyframe = ENTRY_IS_KEYFRAME (entry);
/* skip empty entries */
if (size == 0) {
GST_DEBUG_OBJECT (avi, "Skipping entry %u (%" G_GUINT64_FORMAT ", %p)",
stream->current_entry, size, stream->pad);
goto next;
}
if (avi->segment.rate > 0.0) {
/* only check this for fowards playback for now */
if (keyframe && GST_CLOCK_TIME_IS_VALID (avi->segment.stop)
&& (timestamp > avi->segment.stop)) {
goto eos_stop;
}
}
GST_LOG ("reading buffer (size=%" G_GUINT64_FORMAT "), stream %d, pos %"
G_GUINT64_FORMAT " (0x%" G_GINT64_MODIFIER "x), kf %d", size,
stream_num, offset, offset, keyframe);
/* FIXME, check large chunks and cut them up */
/* pull in the data */
ret = gst_pad_pull_range (avi->sinkpad, offset, size, &buf);
if (ret != GST_FLOW_OK)
goto pull_failed;
/* check for short buffers, this is EOS as well */
2011-04-04 17:17:43 +00:00
if (gst_buffer_get_size (buf) < size)
goto short_buffer;
/* invert the picture if needed */
buf = gst_avi_demux_invert (stream, buf);
/* mark non-keyframes */
if (keyframe)
GST_BUFFER_FLAG_UNSET (buf, GST_BUFFER_FLAG_DELTA_UNIT);
else
GST_BUFFER_FLAG_SET (buf, GST_BUFFER_FLAG_DELTA_UNIT);
GST_BUFFER_TIMESTAMP (buf) = timestamp;
GST_BUFFER_DURATION (buf) = duration;
GST_BUFFER_OFFSET (buf) = out_offset;
GST_BUFFER_OFFSET_END (buf) = out_offset_end;
/* mark discont when pending */
if (stream->discont) {
GST_DEBUG_OBJECT (avi, "setting DISCONT flag");
GST_BUFFER_FLAG_SET (buf, GST_BUFFER_FLAG_DISCONT);
stream->discont = FALSE;
}
2009-10-14 11:08:47 +00:00
gst_avi_demux_add_assoc (avi, stream, timestamp, offset, keyframe);
/* update current position in the segment */
2011-06-02 16:47:36 +00:00
avi->segment.position = timestamp;
GST_DEBUG_OBJECT (avi, "Pushing buffer of size %u, ts %"
GST_TIME_FORMAT ", dur %" GST_TIME_FORMAT ", off %" G_GUINT64_FORMAT
", off_end %" G_GUINT64_FORMAT,
2011-04-04 17:17:43 +00:00
gst_buffer_get_size (buf), GST_TIME_ARGS (timestamp),
GST_TIME_ARGS (duration), out_offset, out_offset_end);
ret = gst_pad_push (stream->pad, buf);
/* mark as processed, we increment the frame and byte counters then
* leave the while loop and return the GstFlowReturn */
processed = TRUE;
if (avi->segment.rate < 0) {
if (timestamp > avi->segment.stop && ret == GST_FLOW_UNEXPECTED) {
/* In reverse playback we can get a GST_FLOW_UNEXPECTED when
* we are at the end of the segment, so we just need to jump
* back to the previous section. */
GST_DEBUG_OBJECT (avi, "downstream has reached end of segment");
ret = GST_FLOW_OK;
}
}
next:
/* move to next item */
ret = gst_avi_demux_advance (avi, stream, ret);
/* combine flows */
ret = gst_avi_demux_combine_flows (avi, stream, ret);
} while (!processed);
beach:
return ret;
/* special cases */
eos:
{
GST_DEBUG_OBJECT (avi, "No samples left for any streams - EOS");
ret = GST_FLOW_UNEXPECTED;
goto beach;
}
eos_stop:
{
GST_LOG_OBJECT (avi, "Found keyframe after segment,"
" setting EOS (%" GST_TIME_FORMAT " > %" GST_TIME_FORMAT ")",
GST_TIME_ARGS (timestamp), GST_TIME_ARGS (avi->segment.stop));
ret = GST_FLOW_UNEXPECTED;
/* move to next stream */
goto next;
}
pull_failed:
{
GST_DEBUG_OBJECT (avi, "pull range failed: pos=%" G_GUINT64_FORMAT
" size=%" G_GUINT64_FORMAT, offset, size);
goto beach;
}
short_buffer:
{
GST_WARNING_OBJECT (avi, "Short read at offset %" G_GUINT64_FORMAT
", only got %d/%" G_GUINT64_FORMAT " bytes (truncated file?)", offset,
2011-04-04 17:17:43 +00:00
gst_buffer_get_size (buf), size);
gst_buffer_unref (buf);
ret = GST_FLOW_UNEXPECTED;
goto beach;
}
}
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
/*
* Read data. If we have an index it delegates to
* gst_avi_demux_process_next_entry().
*/
static GstFlowReturn
gst_avi_demux_stream_data (GstAviDemux * avi)
{
guint32 tag = 0;
guint32 size = 0;
gint stream_nr = 0;
GstFlowReturn res = GST_FLOW_OK;
if (G_UNLIKELY (avi->have_eos)) {
/* Clean adapter, we're done */
gst_adapter_clear (avi->adapter);
return GST_FLOW_UNEXPECTED;
}
if (G_UNLIKELY (avi->todrop)) {
guint drop;
if ((drop = gst_adapter_available (avi->adapter))) {
if (drop > avi->todrop)
drop = avi->todrop;
GST_DEBUG_OBJECT (avi, "Dropping %d bytes", drop);
gst_adapter_flush (avi->adapter, drop);
avi->todrop -= drop;
avi->offset += drop;
}
}
/* Iterate until need more data, so adapter won't grow too much */
while (1) {
if (G_UNLIKELY (!gst_avi_demux_peek_chunk_info (avi, &tag, &size))) {
return GST_FLOW_OK;
}
GST_DEBUG ("Trying chunk (%" GST_FOURCC_FORMAT "), size %d",
GST_FOURCC_ARGS (tag), size);
if (G_LIKELY ((tag & 0xff) >= '0' && (tag & 0xff) <= '9' &&
((tag >> 8) & 0xff) >= '0' && ((tag >> 8) & 0xff) <= '9')) {
GST_LOG ("Chunk ok");
} else if ((tag & 0xffff) == (('x' << 8) | 'i')) {
GST_DEBUG ("Found sub-index tag");
if (gst_avi_demux_peek_chunk (avi, &tag, &size) || size == 0) {
/* accept 0 size buffer here */
avi->abort_buffering = FALSE;
GST_DEBUG (" skipping %d bytes for now", size);
gst_adapter_flush (avi->adapter, 8 + GST_ROUND_UP_2 (size));
}
return GST_FLOW_OK;
} else if (tag == GST_RIFF_TAG_RIFF) {
/* RIFF tags can appear in ODML files, just jump over them */
if (gst_adapter_available (avi->adapter) >= 12) {
GST_DEBUG ("Found RIFF tag, skipping RIFF header");
gst_adapter_flush (avi->adapter, 12);
continue;
}
return GST_FLOW_OK;
} else if (tag == GST_RIFF_TAG_idx1) {
GST_DEBUG ("Found index tag");
if (gst_avi_demux_peek_chunk (avi, &tag, &size) || size == 0) {
/* accept 0 size buffer here */
avi->abort_buffering = FALSE;
GST_DEBUG (" skipping %d bytes for now", size);
gst_adapter_flush (avi->adapter, 8 + GST_ROUND_UP_2 (size));
}
return GST_FLOW_OK;
} else if (tag == GST_RIFF_TAG_LIST) {
/* movi chunks might be grouped in rec list */
if (gst_adapter_available (avi->adapter) >= 12) {
GST_DEBUG ("Found LIST tag, skipping LIST header");
gst_adapter_flush (avi->adapter, 12);
continue;
}
return GST_FLOW_OK;
} else if (tag == GST_RIFF_TAG_JUNK || tag == GST_RIFF_TAG_JUNQ) {
/* rec list might contain JUNK chunks */
GST_DEBUG ("Found JUNK tag");
if (gst_avi_demux_peek_chunk (avi, &tag, &size) || size == 0) {
/* accept 0 size buffer here */
avi->abort_buffering = FALSE;
GST_DEBUG (" skipping %d bytes for now", size);
gst_adapter_flush (avi->adapter, 8 + GST_ROUND_UP_2 (size));
}
return GST_FLOW_OK;
} else {
GST_DEBUG ("No more stream chunks, send EOS");
avi->have_eos = TRUE;
return GST_FLOW_UNEXPECTED;
}
if (G_UNLIKELY (!gst_avi_demux_peek_chunk (avi, &tag, &size))) {
/* supposedly one hopes to catch a nicer chunk later on ... */
/* FIXME ?? give up here rather than possibly ending up going
* through the whole file */
if (avi->abort_buffering) {
avi->abort_buffering = FALSE;
if (size) {
gst_adapter_flush (avi->adapter, 8);
return GST_FLOW_OK;
}
} else {
return GST_FLOW_OK;
}
}
GST_DEBUG ("chunk ID %" GST_FOURCC_FORMAT ", size %u",
GST_FOURCC_ARGS (tag), size);
stream_nr = CHUNKID_TO_STREAMNR (tag);
if (G_UNLIKELY (stream_nr < 0 || stream_nr >= avi->num_streams)) {
/* recoverable */
GST_WARNING ("Invalid stream ID %d (%" GST_FOURCC_FORMAT ")",
stream_nr, GST_FOURCC_ARGS (tag));
2009-09-21 23:28:54 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
gst_adapter_flush (avi->adapter, 8 + GST_ROUND_UP_2 (size));
} else {
GstAviStream *stream;
GstClockTime next_ts = 0;
GstBuffer *buf = NULL;
2009-10-14 11:08:47 +00:00
guint64 offset;
gboolean saw_desired_kf = stream_nr != avi->main_stream
|| avi->offset >= avi->seek_kf_offset;
if (stream_nr == avi->main_stream && avi->offset == avi->seek_kf_offset) {
GST_DEBUG_OBJECT (avi, "Desired keyframe reached");
avi->seek_kf_offset = 0;
}
if (saw_desired_kf) {
gst_adapter_flush (avi->adapter, 8);
/* get buffer */
if (size) {
buf = gst_adapter_take_buffer (avi->adapter, GST_ROUND_UP_2 (size));
/* patch the size */
2011-04-04 17:17:43 +00:00
gst_buffer_resize (buf, 0, size);
} else {
buf = NULL;
}
} else {
GST_DEBUG_OBJECT (avi,
"Desired keyframe not yet reached, flushing chunk");
gst_adapter_flush (avi->adapter, 8 + GST_ROUND_UP_2 (size));
}
2009-10-14 11:08:47 +00:00
offset = avi->offset;
2009-09-21 23:28:54 +00:00
avi->offset += 8 + GST_ROUND_UP_2 (size);
stream = &avi->stream[stream_nr];
/* set delay (if any)
if (stream->strh->init_frames == stream->current_frame &&
stream->delay == 0)
stream->delay = next_ts;
*/
/* parsing of corresponding header may have failed */
if (G_UNLIKELY (!stream->pad)) {
GST_WARNING_OBJECT (avi, "no pad for stream ID %" GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (tag));
if (buf)
gst_buffer_unref (buf);
} else {
/* get time of this buffer */
gst_pad_query_position (stream->pad, GST_FORMAT_TIME,
(gint64 *) & next_ts);
2009-10-14 11:08:47 +00:00
gst_avi_demux_add_assoc (avi, stream, next_ts, offset, FALSE);
/* increment our positions */
stream->current_entry++;
stream->current_total += size;
2009-09-21 16:09:12 +00:00
/* update current position in the segment */
2011-06-02 16:47:36 +00:00
avi->segment.position = next_ts;
if (saw_desired_kf && buf) {
GstClockTime dur_ts = 0;
/* invert the picture if needed */
buf = gst_avi_demux_invert (stream, buf);
gst_pad_query_position (stream->pad, GST_FORMAT_TIME,
(gint64 *) & dur_ts);
GST_BUFFER_TIMESTAMP (buf) = next_ts;
GST_BUFFER_DURATION (buf) = dur_ts - next_ts;
if (stream->strh->type == GST_RIFF_FCC_vids) {
GST_BUFFER_OFFSET (buf) = stream->current_entry - 1;
GST_BUFFER_OFFSET_END (buf) = stream->current_entry;
} else {
GST_BUFFER_OFFSET (buf) = GST_BUFFER_OFFSET_NONE;
GST_BUFFER_OFFSET_END (buf) = GST_BUFFER_OFFSET_NONE;
}
GST_DEBUG_OBJECT (avi,
"Pushing buffer with time=%" GST_TIME_FORMAT ", duration %"
GST_TIME_FORMAT ", offset %" G_GUINT64_FORMAT
" and size %d over pad %s", GST_TIME_ARGS (next_ts),
GST_TIME_ARGS (GST_BUFFER_DURATION (buf)),
GST_BUFFER_OFFSET (buf), size, GST_PAD_NAME (stream->pad));
/* mark discont when pending */
if (G_UNLIKELY (stream->discont)) {
GST_DEBUG_OBJECT (avi, "Setting DISCONT");
GST_BUFFER_FLAG_SET (buf, GST_BUFFER_FLAG_DISCONT);
stream->discont = FALSE;
}
res = gst_pad_push (stream->pad, buf);
buf = NULL;
/* combine flows */
res = gst_avi_demux_combine_flows (avi, stream, res);
if (G_UNLIKELY (res != GST_FLOW_OK)) {
GST_DEBUG ("Push failed; %s", gst_flow_get_name (res));
return res;
}
}
}
}
}
return res;
}
/*
* Send pending tags.
*/
static void
push_tag_lists (GstAviDemux * avi)
{
guint i;
2009-09-23 12:25:08 +00:00
GstTagList *tags;
if (!avi->got_tags)
return;
GST_DEBUG_OBJECT (avi, "Pushing pending tag lists");
2009-09-23 12:25:08 +00:00
for (i = 0; i < avi->num_streams; i++) {
GstAviStream *stream = &avi->stream[i];
GstPad *pad = stream->pad;
tags = stream->taglist;
if (pad && tags) {
GST_DEBUG_OBJECT (pad, "Tags: %" GST_PTR_FORMAT, tags);
gst_element_found_tags_for_pad (GST_ELEMENT_CAST (avi), pad, tags);
stream->taglist = NULL;
}
2009-09-23 12:25:08 +00:00
}
2009-06-18 22:36:28 +00:00
2009-09-23 12:25:08 +00:00
if (!(tags = avi->globaltags))
tags = gst_tag_list_new ();
2009-06-18 22:36:28 +00:00
2009-09-23 12:25:08 +00:00
gst_tag_list_add (tags, GST_TAG_MERGE_REPLACE,
2009-06-18 22:36:28 +00:00
GST_TAG_CONTAINER_FORMAT, "AVI", NULL);
2009-09-23 12:25:08 +00:00
GST_DEBUG_OBJECT (avi, "Global tags: %" GST_PTR_FORMAT, tags);
gst_element_found_tags (GST_ELEMENT_CAST (avi), tags);
2009-06-18 22:36:28 +00:00
avi->globaltags = NULL;
avi->got_tags = FALSE;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
static void
gst_avi_demux_loop (GstPad * pad)
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
{
GstFlowReturn res;
GstAviDemux *avi = GST_AVI_DEMUX (GST_PAD_PARENT (pad));
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
switch (avi->state) {
case GST_AVI_DEMUX_START:
2009-09-23 12:25:08 +00:00
res = gst_avi_demux_stream_init_pull (avi);
if (G_UNLIKELY (res != GST_FLOW_OK)) {
GST_WARNING ("stream_init flow: %s", gst_flow_get_name (res));
goto pause;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
avi->state = GST_AVI_DEMUX_HEADER;
/* fall-through */
case GST_AVI_DEMUX_HEADER:
2009-09-23 12:25:08 +00:00
res = gst_avi_demux_stream_header_pull (avi);
if (G_UNLIKELY (res != GST_FLOW_OK)) {
GST_WARNING ("stream_header flow: %s", gst_flow_get_name (res));
goto pause;
}
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
avi->state = GST_AVI_DEMUX_MOVI;
ext/flac/gstflacdec.c: Only return true if we actually filled something in. Prevents player applications from showing... Original commit message from CVS: * ext/flac/gstflacdec.c: (gst_flacdec_src_query): Only return true if we actually filled something in. Prevents player applications from showing a random length for flac files. * gst-libs/gst/riff/riff-read.c: (gst_riff_read_class_init), (gst_riff_read_use_event), (gst_riff_read_handle_event), (gst_riff_read_seek), (gst_riff_read_skip), (gst_riff_read_strh), (gst_riff_read_strf_vids_with_data), (gst_riff_read_strf_auds_with_data), (gst_riff_read_strf_iavs): OK, ok, so I implemented event handling. Apparently it's normal that we receive random events at random points without asking for it. * gst/avi/gstavidemux.c: (gst_avi_demux_reset), (gst_avi_demux_src_convert), (gst_avi_demux_handle_src_query), (gst_avi_demux_handle_src_event), (gst_avi_demux_stream_index), (gst_avi_demux_sync), (gst_avi_demux_stream_scan), (gst_avi_demux_massage_index), (gst_avi_demux_stream_header), (gst_avi_demux_handle_seek), (gst_avi_demux_process_next_entry), (gst_avi_demux_stream_data), (gst_avi_demux_loop): * gst/avi/gstavidemux.h: Implement non-lineair chunk handling and subchunk processing. The first solves playback of AVI files where the audio and video data of individual buffers that we read are not synchronized. This should not happen according to the wonderful AVI specs, but of course it does happen in reality. It is also a prerequisite for the second. Subchunk processing allows us to cut chunks in small pieces and process each of these pieces separately. This is required because I've seen several AVI files with incredibly large audio chunks, even some files with only one audio chunk for the whole file. This allows for proper playback including seeking. This patch is supposed to fix all AVI A/V sync issues. * gst/flx/gstflxdec.c: (gst_flxdec_class_init), (flx_decode_chunks), (flx_decode_color), (gst_flxdec_loop): Work. * gst/modplug/gstmodplug.cc: Proper return value setting for the query() function. * gst/playback/gstplaybasebin.c: (setup_source): Being in non-playing state (after, e.g., EOS) is not necessarily a bad thing. Allow for that. This fixes playback of short files. They don't actually playback fully now, because the clock already runs. This means that small files (<500kB) with a small length (<2sec) will still not or barely play. Other files, such as mod or flx, will work correctly, however.
2004-09-29 09:45:40 +00:00
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
case GST_AVI_DEMUX_MOVI:
if (G_UNLIKELY (avi->seg_event)) {
gst_avi_demux_push_event (avi, avi->seg_event);
avi->seg_event = NULL;
}
if (G_UNLIKELY (avi->got_tags)) {
push_tag_lists (avi);
}
/* process each index entry in turn */
res = gst_avi_demux_loop_data (avi);
/* pause when error */
if (G_UNLIKELY (res != GST_FLOW_OK)) {
GST_INFO ("stream_movi flow: %s", gst_flow_get_name (res));
goto pause;
}
break;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
default:
GST_ERROR_OBJECT (avi, "unknown state %d", avi->state);
res = GST_FLOW_ERROR;
goto pause;
}
return;
/* ERRORS */
pause:{
gboolean push_eos = FALSE;
GST_LOG_OBJECT (avi, "pausing task, reason %s", gst_flow_get_name (res));
gst_pad_pause_task (avi->sinkpad);
if (res == GST_FLOW_UNEXPECTED) {
/* handle end-of-stream/segment */
/* so align our position with the end of it, if there is one
* this ensures a subsequent will arrive at correct base/acc time */
if (avi->segment.rate > 0.0 &&
GST_CLOCK_TIME_IS_VALID (avi->segment.stop))
avi->segment.position = avi->segment.stop;
else if (avi->segment.rate < 0.0)
avi->segment.position = avi->segment.start;
if (avi->segment.flags & GST_SEEK_FLAG_SEGMENT) {
gint64 stop;
if ((stop = avi->segment.stop) == -1)
stop = avi->segment.duration;
GST_INFO_OBJECT (avi, "sending segment_done");
gst_element_post_message
2010-01-15 17:13:24 +00:00
(GST_ELEMENT_CAST (avi),
gst_message_new_segment_done (GST_OBJECT_CAST (avi),
GST_FORMAT_TIME, stop));
} else {
push_eos = TRUE;
}
} else if (res == GST_FLOW_NOT_LINKED || res < GST_FLOW_UNEXPECTED) {
/* for fatal errors we post an error message, wrong-state is
* not fatal because it happens due to flushes and only means
* that we should stop now. */
GST_ELEMENT_ERROR (avi, STREAM, FAILED,
(_("Internal data stream error.")),
("streaming stopped, reason %s", gst_flow_get_name (res)));
push_eos = TRUE;
}
if (push_eos) {
GST_INFO_OBJECT (avi, "sending eos");
if (!gst_avi_demux_push_event (avi, gst_event_new_eos ()) &&
(res == GST_FLOW_UNEXPECTED)) {
GST_ELEMENT_ERROR (avi, STREAM, DEMUX,
(NULL), ("got eos but no streams (yet)"));
}
}
}
}
static GstFlowReturn
gst_avi_demux_chain (GstPad * pad, GstBuffer * buf)
{
GstFlowReturn res;
GstAviDemux *avi = GST_AVI_DEMUX (GST_PAD_PARENT (pad));
gint i;
if (GST_BUFFER_IS_DISCONT (buf)) {
GST_DEBUG_OBJECT (avi, "got DISCONT");
gst_adapter_clear (avi->adapter);
/* mark all streams DISCONT */
for (i = 0; i < avi->num_streams; i++)
avi->stream[i].discont = TRUE;
}
2011-04-04 17:17:43 +00:00
GST_DEBUG ("Store %d bytes in adapter", gst_buffer_get_size (buf));
gst_adapter_push (avi->adapter, buf);
switch (avi->state) {
case GST_AVI_DEMUX_START:
if ((res = gst_avi_demux_stream_init_push (avi)) != GST_FLOW_OK) {
GST_WARNING ("stream_init flow: %s", gst_flow_get_name (res));
break;
}
break;
case GST_AVI_DEMUX_HEADER:
if ((res = gst_avi_demux_stream_header_push (avi)) != GST_FLOW_OK) {
GST_WARNING ("stream_header flow: %s", gst_flow_get_name (res));
break;
}
break;
case GST_AVI_DEMUX_MOVI:
if (G_UNLIKELY (avi->seg_event)) {
gst_avi_demux_push_event (avi, avi->seg_event);
avi->seg_event = NULL;
}
if (G_UNLIKELY (avi->got_tags)) {
push_tag_lists (avi);
}
res = gst_avi_demux_stream_data (avi);
break;
2009-12-04 14:13:12 +00:00
case GST_AVI_DEMUX_SEEK:
{
GstEvent *event;
2009-12-04 14:13:12 +00:00
res = GST_FLOW_OK;
/* obtain and parse indexes */
if (avi->stream[0].indexes && !gst_avi_demux_read_subindexes_push (avi))
2009-12-04 14:13:12 +00:00
/* seek in subindex read function failed */
goto index_failed;
2009-12-04 14:13:12 +00:00
if (!avi->stream[0].indexes && !avi->have_index
&& avi->avih->flags & GST_RIFF_AVIH_HASINDEX)
gst_avi_demux_stream_index_push (avi);
if (avi->have_index) {
/* use the indexes now to construct nice durations */
gst_avi_demux_calculate_durations_from_index (avi);
} else {
/* still parsing indexes */
break;
}
GST_OBJECT_LOCK (avi);
event = avi->seek_event;
avi->seek_event = NULL;
GST_OBJECT_UNLOCK (avi);
2009-12-04 14:13:12 +00:00
/* calculate and perform seek */
if (!avi_demux_handle_seek_push (avi, avi->sinkpad, event))
goto seek_failed;
gst_event_unref (event);
2009-12-04 14:13:12 +00:00
avi->state = GST_AVI_DEMUX_MOVI;
break;
}
default:
GST_ELEMENT_ERROR (avi, STREAM, FAILED, (NULL),
("Illegal internal state"));
res = GST_FLOW_ERROR;
break;
}
GST_DEBUG_OBJECT (avi, "state: %d res:%s", avi->state,
gst_flow_get_name (res));
if (G_UNLIKELY (avi->abort_buffering))
goto abort_buffering;
return res;
/* ERRORS */
index_failed:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL), ("failed to read indexes"));
return GST_FLOW_ERROR;
}
seek_failed:
{
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL), ("push mode seek failed"));
return GST_FLOW_ERROR;
}
abort_buffering:
{
avi->abort_buffering = FALSE;
GST_ELEMENT_ERROR (avi, STREAM, DEMUX, (NULL), ("unhandled buffer size"));
return GST_FLOW_ERROR;
}
}
static gboolean
gst_avi_demux_sink_activate (GstPad * sinkpad)
{
2011-06-02 16:47:36 +00:00
GstQuery *query;
gboolean pull_mode;
query = gst_query_new_scheduling ();
if (!gst_pad_peer_query (sinkpad, query)) {
gst_query_unref (query);
goto activate_push;
}
gst_query_parse_scheduling (query, &pull_mode, NULL, NULL, NULL, NULL, NULL);
2011-07-04 09:06:54 +00:00
gst_query_unref (query);
2011-06-02 16:47:36 +00:00
if (!pull_mode)
goto activate_push;
GST_DEBUG_OBJECT (sinkpad, "activating pull");
return gst_pad_activate_pull (sinkpad, TRUE);
activate_push:
{
GST_DEBUG_OBJECT (sinkpad, "activating push");
return gst_pad_activate_push (sinkpad, TRUE);
}
}
static gboolean
gst_avi_demux_sink_activate_pull (GstPad * sinkpad, gboolean active)
{
GstAviDemux *avi = GST_AVI_DEMUX (GST_OBJECT_PARENT (sinkpad));
if (active) {
avi->streaming = FALSE;
return gst_pad_start_task (sinkpad, (GstTaskFunction) gst_avi_demux_loop,
sinkpad);
} else {
return gst_pad_stop_task (sinkpad);
}
}
static gboolean
gst_avi_demux_activate_push (GstPad * pad, gboolean active)
{
GstAviDemux *avi = GST_AVI_DEMUX (GST_OBJECT_PARENT (pad));
if (active) {
GST_DEBUG ("avi: activating push/chain function");
avi->streaming = TRUE;
#if 0
/* create index for some push based seeking if not provided */
GST_OBJECT_LOCK (avi);
if (!avi->element_index) {
GST_DEBUG_OBJECT (avi, "creating index");
avi->element_index = gst_index_factory_make ("memindex");
}
GST_OBJECT_UNLOCK (avi);
/* object lock might be taken again */
2010-01-15 17:13:24 +00:00
gst_index_get_writer_id (avi->element_index, GST_OBJECT_CAST (avi),
&avi->index_id);
#endif
} else {
GST_DEBUG ("avi: deactivating push/chain function");
}
return TRUE;
}
2009-10-14 11:08:47 +00:00
static void
gst_avi_demux_set_index (GstElement * element, GstIndex * index)
{
GstAviDemux *avi = GST_AVI_DEMUX (element);
GST_OBJECT_LOCK (avi);
if (avi->element_index)
gst_object_unref (avi->element_index);
if (index) {
avi->element_index = gst_object_ref (index);
} else {
avi->element_index = NULL;
}
GST_OBJECT_UNLOCK (avi);
/* object lock might be taken again */
if (index)
2010-01-15 17:13:24 +00:00
gst_index_get_writer_id (index, GST_OBJECT_CAST (element), &avi->index_id);
2009-10-14 11:08:47 +00:00
GST_DEBUG_OBJECT (avi, "Set index %" GST_PTR_FORMAT, avi->element_index);
}
static GstIndex *
gst_avi_demux_get_index (GstElement * element)
{
GstIndex *result = NULL;
GstAviDemux *avi = GST_AVI_DEMUX (element);
GST_OBJECT_LOCK (avi);
if (avi->element_index)
result = gst_object_ref (avi->element_index);
GST_OBJECT_UNLOCK (avi);
GST_DEBUG_OBJECT (avi, "Returning index %" GST_PTR_FORMAT, result);
return result;
}
static GstStateChangeReturn
gst_avi_demux_change_state (GstElement * element, GstStateChange transition)
{
GstStateChangeReturn ret;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
GstAviDemux *avi = GST_AVI_DEMUX (element);
switch (transition) {
case GST_STATE_CHANGE_READY_TO_PAUSED:
avi->streaming = FALSE;
gst_segment_init (&avi->segment, GST_FORMAT_TIME);
break;
default:
break;
}
ret = GST_ELEMENT_CLASS (parent_class)->change_state (element, transition);
if (ret == GST_STATE_CHANGE_FAILURE)
goto done;
switch (transition) {
case GST_STATE_CHANGE_PAUSED_TO_READY:
avi->have_index = FALSE;
Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first b... Original commit message from CVS: Riff, EBML, fourcc etc. work. Not fully finished, but better than what we used to have and definately worth a first broad testing. I've revived rifflib. Rifflib used to be a bytestream-for-riff, which just dup'ed bytestream. I've rewritten rifflib to be a modern riff- chunk parser that uses bytestream fully, plus adds some extra functions so that riff file parsing becomes extremely easy. It also contains some small usability functions for strh/strf and metadata parsing. Note that it doesn't use the new tagging yet, that's a TODO. Avidemux has been rewritten to use this. I think we all agreed that avidemux was pretty much a big mess, which is because it used all sort of bytestream magic all around the place. It was just ugly. This is a lot nicer, very complete and safe. I think this is far more robust than what the old avidemux could ever have been. Of course, it might contain bugs, please let me know. EBML writing has also been implemented. This is useful for matroska. I'm intending to modify avidemux (with a riffwriter) similarly. Maybe I'll change wavparse/-enc too to use rifflib. Lastly, several plugins have been modified to use rifflib's fourcc parsing instead of their own. this puts fourcc parsing in one central place, which should make it a lot simpler to add new fourccs. We might want to move this to its own lib instead of rifflib. Enjoy!
2003-12-07 20:00:41 +00:00
gst_avi_demux_reset (avi);
break;
default:
break;
}
done:
return ret;
}