gstreamer/gst-libs/gst/riff/riff-read.c
Wim Taymans 1dae961cbf Plugin port to 0.9, ogg/theora playback should work in the seek example now.
Original commit message from CVS:
Plugin port to 0.9, ogg/theora playback should work in the seek
example now.
Removed old examples.
Removed old oggvorbisenc, renamed rawvorbisenc to vorbisenc as
explained in 0.9 TODO doc.
2005-03-31 09:43:49 +00:00

634 lines
20 KiB
C
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/* GStreamer RIFF I/O
* Copyright (C) 2003 Ronald Bultje <rbultje@ronald.bitfreak.net>
*
* riff-read.c: RIFF input file parsing
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the
* Free Software Foundation, Inc., 59 Temple Place - Suite 330,
* Boston, MA 02111-1307, USA.
*/
#ifdef HAVE_CONFIG_H
#include "config.h"
#endif
#include <string.h>
#include <gst/gstutils.h>
#include "riff-read.h"
GST_DEBUG_CATEGORY_EXTERN (riff_debug);
#define GST_CAT_DEFAULT riff_debug
/**
* gst_riff_read_chunk:
* @element: caller element (used for debugging).
* @pad: pad to pull data from.
* @offset: offset to pull from, incremented by this function.
* @tag: fourcc of the chunk (returned by this function).
* @chunk_data: buffer (returned by this function).
*
* Reads a single chunk of data.
*
* Returns: flow status.
*/
GstFlowReturn
gst_riff_read_chunk (GstElement * element,
GstPad * pad, guint64 * _offset, guint32 * tag, GstBuffer ** _chunk_data)
{
GstBuffer *buf;
GstFlowReturn res;
guint size;
guint64 offset = *_offset;
if ((res = gst_pad_pull_range (pad, offset, 8, &buf)) != GST_FLOW_OK)
return res;
else if (!buf || GST_BUFFER_SIZE (buf) < 8) {
if (buf)
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
*tag = GST_READ_UINT32_LE (GST_BUFFER_DATA (buf));
size = GST_READ_UINT32_LE (GST_BUFFER_DATA (buf) + 4);
gst_buffer_unref (buf);
if ((res = gst_pad_pull_range (pad, offset + 8, size, &buf)) != GST_FLOW_OK)
return res;
else if (!buf || GST_BUFFER_SIZE (buf) < size) {
if (buf)
gst_buffer_unref (buf);
return GST_FLOW_ERROR;
}
*_chunk_data = buf;
*_offset += 8 + ((size + 1) & ~1);
return GST_FLOW_OK;
}
/**
* gst_riff_parse_chunk:
* @element: caller element (used for debugging).
* @buf: input buffer.
* @offset: offset in the buffer in the caller. Is incremented
* by the read size by this function.
* @fourcc: fourcc (returned by this function0 of the chunk.
* @chunk_data: buffer (returned by the function) containing the
* chunk data.
*
* Reads a single chunk.
*
* Returns: the fourcc tag of this chunk, or 0 on error.
*/
gboolean
gst_riff_parse_chunk (GstElement * element, GstBuffer * buf,
guint * _offset, guint32 * _fourcc, GstBuffer ** chunk_data)
{
guint size;
guint32 fourcc;
guint8 *data;
guint offset = *_offset;
*chunk_data = NULL;
*_fourcc = 0;
if (!buf || GST_BUFFER_SIZE (buf) < offset + 8) {
GST_DEBUG_OBJECT (element,
"Failed to parse chunk header (offset %d, %d available, %d needed)",
offset, buf ? GST_BUFFER_DATA (buf) : 0, 8);
return FALSE;
}
/* read header */
data = GST_BUFFER_DATA (buf) + offset;
fourcc = GST_READ_UINT32_LE (data);
size = GST_READ_UINT32_LE (data + 4);
if (GST_BUFFER_SIZE (buf) < size + 8 + offset) {
GST_DEBUG_OBJECT (element,
"Needed chunk data (%d) is more than available (%d), shortcutting",
size, GST_BUFFER_SIZE (buf) - 8 - offset);
size = GST_BUFFER_SIZE (buf) - 8 - offset;
}
if (size)
*chunk_data = gst_buffer_create_sub (buf, offset + 8, size);
else
*chunk_data = NULL;
*_fourcc = fourcc;
*_offset += 8 + ((size + 1) & ~1);
return TRUE;
}
/**
* gst_riff_parse_file_header:
* @element: caller element (used for debugging/error).
* @buf: input buffer from which the file header will be parsed,
* should be at least 12 bytes long.
* @doctype: a fourcc (returned by this function) to indicate the
* type of document (according to the header).
*
* Reads the first few bytes from the provided buffer, checks
* if this stream is a RIFF stream, and determines document type.
* The input data is discarded after use.
*
* Returns: FALSE if this is not a RIFF stream (in which case the
* caller should error out; we already throw an error), or TRUE
* if it is.
*/
gboolean
gst_riff_parse_file_header (GstElement * element,
GstBuffer * buf, guint32 * doctype)
{
guint8 *data = GST_BUFFER_DATA (buf);
guint32 tag;
if (!buf || GST_BUFFER_SIZE (buf) < 12) {
GST_ELEMENT_ERROR (element, STREAM, WRONG_TYPE, (NULL),
("Not enough data to parse RIFF header (%d available, %d needed)",
buf ? GST_BUFFER_SIZE (buf) : 0, 12));
if (buf)
gst_buffer_unref (buf);
return FALSE;
}
tag = GST_READ_UINT32_LE (data);
if (tag != GST_RIFF_TAG_RIFF) {
GST_ELEMENT_ERROR (element, STREAM, WRONG_TYPE, (NULL),
("Stream is no RIFF stream: " GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (tag)));
gst_buffer_unref (buf);
return FALSE;
}
*doctype = GST_READ_UINT32_LE (data + 8);
gst_buffer_unref (buf);
return TRUE;
}
/**
* gst_riff_parse_strh:
* @element: caller element (used for debugging/error).
* @buf: input data to be used for parsing, stripped from header.
* @strh: a pointer (returned by this function) to a filled-in
* strh structure. Caller should free it.
*
* Parses a strh structure from input data. The input data is
* discarded after use.
*
* Returns: TRUE if parsing succeeded, otherwise FALSE. The stream
* should be skipped on error, but it is not fatal.
*/
gboolean
gst_riff_parse_strh (GstElement * element,
GstBuffer * buf, gst_riff_strh ** _strh)
{
gst_riff_strh *strh;
if (!buf || GST_BUFFER_SIZE (buf) < sizeof (gst_riff_strh)) {
GST_ERROR_OBJECT (element,
"Too small strh (%d available, %d needed)",
buf ? GST_BUFFER_SIZE (buf) : 0, (int) sizeof (gst_riff_strh));
if (buf)
gst_buffer_unref (buf);
return FALSE;
}
strh = g_memdup (GST_BUFFER_DATA (buf), GST_BUFFER_SIZE (buf));
gst_buffer_unref (buf);
#if (G_BYTE_ORDER == G_BIG_ENDIAN)
strh->type = GUINT32_FROM_LE (strh->type);
strh->fcc_handler = GUINT32_FROM_LE (strh->fcc_handler);
strh->flags = GUINT32_FROM_LE (strh->flags);
strh->priority = GUINT32_FROM_LE (strh->priority);
strh->init_frames = GUINT32_FROM_LE (strh->init_frames);
strh->scale = GUINT32_FROM_LE (strh->scale);
strh->rate = GUINT32_FROM_LE (strh->rate);
strh->start = GUINT32_FROM_LE (strh->start);
strh->length = GUINT32_FROM_LE (strh->length);
strh->bufsize = GUINT32_FROM_LE (strh->bufsize);
strh->quality = GUINT32_FROM_LE (strh->quality);
strh->samplesize = GUINT32_FROM_LE (strh->samplesize);
#endif
/* avoid divisions by zero */
if (!strh->scale)
strh->scale = 1;
if (!strh->rate)
strh->rate = 1;
/* debug */
GST_INFO_OBJECT (element, "strh tag found:");
GST_INFO_OBJECT (element, " type " GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (strh->type));
GST_INFO_OBJECT (element, " fcc_handler " GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (strh->fcc_handler));
GST_INFO_OBJECT (element, " flags 0x%08x", strh->flags);
GST_INFO_OBJECT (element, " priority %d", strh->priority);
GST_INFO_OBJECT (element, " init_frames %d", strh->init_frames);
GST_INFO_OBJECT (element, " scale %d", strh->scale);
GST_INFO_OBJECT (element, " rate %d", strh->rate);
GST_INFO_OBJECT (element, " start %d", strh->start);
GST_INFO_OBJECT (element, " length %d", strh->length);
GST_INFO_OBJECT (element, " bufsize %d", strh->bufsize);
GST_INFO_OBJECT (element, " quality %d", strh->quality);
GST_INFO_OBJECT (element, " samplesize %d", strh->samplesize);
*_strh = strh;
return TRUE;
}
/**
* gst_riff_parse_strf_vids:
* @element: caller element (used for debugging/error).
* @buf: input data to be used for parsing, stripped from header.
* @strf: a pointer (returned by this function) to a filled-in
* strf/vids structure. Caller should free it.
* @data: a pointer (returned by this function) to a buffer
* containing extradata for this particular stream (e.g.
* palette, codec initialization data).
*
* Parses a video stream´s strf structure plus optionally some
* extradata from input data. The input data is discarded after
* use.
*
* Returns: TRUE if parsing succeeded, otherwise FALSE. The stream
* should be skipped on error, but it is not fatal.
*/
gboolean
gst_riff_parse_strf_vids (GstElement * element,
GstBuffer * buf, gst_riff_strf_vids ** _strf, GstBuffer ** data)
{
gst_riff_strf_vids *strf;
if (!buf || GST_BUFFER_SIZE (buf) < sizeof (gst_riff_strf_vids)) {
GST_ERROR_OBJECT (element,
"Too small strf_vids (%d available, %d needed)",
buf ? GST_BUFFER_SIZE (buf) : 0, (int) sizeof (gst_riff_strf_vids));
if (buf)
gst_buffer_unref (buf);
return FALSE;
}
strf = g_memdup (GST_BUFFER_DATA (buf), GST_BUFFER_SIZE (buf));
#if (G_BYTE_ORDER == G_BIG_ENDIAN)
strf->size = GUINT32_FROM_LE (strf->size);
strf->width = GUINT32_FROM_LE (strf->width);
strf->height = GUINT32_FROM_LE (strf->height);
strf->planes = GUINT16_FROM_LE (strf->planes);
strf->bit_cnt = GUINT16_FROM_LE (strf->bit_cnt);
strf->compression = GUINT32_FROM_LE (strf->compression);
strf->image_size = GUINT32_FROM_LE (strf->image_size);
strf->xpels_meter = GUINT32_FROM_LE (strf->xpels_meter);
strf->ypels_meter = GUINT32_FROM_LE (strf->ypels_meter);
strf->num_colors = GUINT32_FROM_LE (strf->num_colors);
strf->imp_colors = GUINT32_FROM_LE (strf->imp_colors);
#endif
/* size checking */
*data = NULL;
if (strf->size > GST_BUFFER_SIZE (buf)) {
GST_WARNING_OBJECT (element,
"strf_vids header gave %d bytes data, only %d available",
strf->size, GST_BUFFER_SIZE (buf));
strf->size = GST_BUFFER_SIZE (buf);
} else if (strf->size < GST_BUFFER_SIZE (buf)) {
*data = gst_buffer_create_sub (buf, strf->size,
GST_BUFFER_SIZE (buf) - strf->size);
}
/* debug */
GST_INFO_OBJECT (element, "strf tag found in context vids:");
GST_INFO_OBJECT (element, " size %d", strf->size);
GST_INFO_OBJECT (element, " width %d", strf->width);
GST_INFO_OBJECT (element, " height %d", strf->height);
GST_INFO_OBJECT (element, " planes %d", strf->planes);
GST_INFO_OBJECT (element, " bit_cnt %d", strf->bit_cnt);
GST_INFO_OBJECT (element, " compression " GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (strf->compression));
GST_INFO_OBJECT (element, " image_size %d", strf->image_size);
GST_INFO_OBJECT (element, " xpels_meter %d", strf->xpels_meter);
GST_INFO_OBJECT (element, " ypels_meter %d", strf->ypels_meter);
GST_INFO_OBJECT (element, " num_colors %d", strf->num_colors);
GST_INFO_OBJECT (element, " imp_colors %d", strf->imp_colors);
if (*data)
GST_INFO_OBJECT (element, " %d bytes extradata", GST_BUFFER_SIZE (*data));
gst_buffer_unref (buf);
*_strf = strf;
return TRUE;
}
/**
* gst_riff_parse_strf_auds:
* @element: caller element (used for debugging/error).
* @buf: input data to be used for parsing, stripped from header.
* @strf: a pointer (returned by this function) to a filled-in
* strf/auds structure. Caller should free it.
* @data: a pointer (returned by this function) to a buffer
* containing extradata for this particular stream (e.g.
* codec initialization data).
*
* Parses an audio stream´s strf structure plus optionally some
* extradata from input data. The input data is discarded after
* use.
*
* Returns: TRUE if parsing succeeded, otherwise FALSE. The stream
* should be skipped on error, but it is not fatal.
*/
gboolean
gst_riff_parse_strf_auds (GstElement * element,
GstBuffer * buf, gst_riff_strf_auds ** _strf, GstBuffer ** data)
{
gst_riff_strf_auds *strf;
if (!buf || GST_BUFFER_SIZE (buf) < sizeof (gst_riff_strf_auds)) {
GST_ERROR_OBJECT (element,
"Too small strf_auds (%d available, %d needed)",
buf ? GST_BUFFER_SIZE (buf) : 0, (int) sizeof (gst_riff_strf_auds));
if (buf)
gst_buffer_unref (buf);
return FALSE;
}
strf = g_memdup (GST_BUFFER_DATA (buf), GST_BUFFER_SIZE (buf));
#if (G_BYTE_ORDER == G_BIG_ENDIAN)
strf->format = GUINT16_FROM_LE (strf->format);
strf->channels = GUINT16_FROM_LE (strf->channels);
strf->rate = GUINT32_FROM_LE (strf->rate);
strf->av_bps = GUINT32_FROM_LE (strf->av_bps);
strf->blockalign = GUINT16_FROM_LE (strf->blockalign);
strf->size = GUINT16_FROM_LE (strf->size);
#endif
/* size checking */
*data = NULL;
if (GST_BUFFER_SIZE (buf) > sizeof (gst_riff_strf_auds) + 2) {
gint len;
len = GST_READ_UINT16_LE (&GST_BUFFER_DATA (buf)[16]);
if (len + 2 + sizeof (gst_riff_strf_auds) > GST_BUFFER_SIZE (buf)) {
GST_WARNING_OBJECT (element,
"Extradata indicated %d bytes, but only %d available",
len, GST_BUFFER_SIZE (buf) - 2 - sizeof (gst_riff_strf_auds));
len = GST_BUFFER_SIZE (buf) - 2 - sizeof (gst_riff_strf_auds);
}
*data = gst_buffer_create_sub (buf, sizeof (gst_riff_strf_auds) + 2, len);
}
/* debug */
GST_INFO_OBJECT (element, "strf tag found in context auds:");
GST_INFO_OBJECT (element, " format %d", strf->format);
GST_INFO_OBJECT (element, " channels %d", strf->channels);
GST_INFO_OBJECT (element, " rate %d", strf->rate);
GST_INFO_OBJECT (element, " av_bps %d", strf->av_bps);
GST_INFO_OBJECT (element, " blockalign %d", strf->blockalign);
GST_INFO_OBJECT (element, " size %d", strf->size);
if (*data)
GST_INFO_OBJECT (element, " %d bytes extradata", GST_BUFFER_SIZE (*data));
gst_buffer_unref (buf);
*_strf = strf;
return TRUE;
}
/**
* gst_riff_parse_strf_iavs:
* @element: caller element (used for debugging/error).
* @buf: input data to be used for parsing, stripped from header.
* @strf: a pointer (returned by this function) to a filled-in
* strf/iavs structure. Caller should free it.
* @data: a pointer (returned by this function) to a buffer
* containing extradata for this particular stream (e.g.
* codec initialization data).
*
* Parses a interleaved (also known as "complex") stream´s strf
* structure plus optionally some extradata from input data. The
* input data is discarded after use.
*
* Returns: TRUE if parsing succeeded, otherwise FALSE.
*/
gboolean
gst_riff_parse_strf_iavs (GstElement * element,
GstBuffer * buf, gst_riff_strf_iavs ** _strf, GstBuffer ** data)
{
gst_riff_strf_iavs *strf;
if (!buf || GST_BUFFER_SIZE (buf) < sizeof (gst_riff_strf_iavs)) {
GST_ERROR_OBJECT (element,
"Too small strf_iavs (%d available, %d needed)",
buf ? GST_BUFFER_SIZE (buf) : 0, (int) sizeof (gst_riff_strf_iavs));
gst_buffer_unref (buf);
return FALSE;
}
strf = g_memdup (GST_BUFFER_DATA (buf), GST_BUFFER_SIZE (buf));
gst_buffer_unref (buf);
#if (G_BYTE_ORDER == G_BIG_ENDIAN)
strf->DVAAuxSrc = GUINT32_FROM_LE (strf->DVAAuxSrc);
strf->DVAAuxCtl = GUINT32_FROM_LE (strf->DVAAuxCtl);
strf->DVAAuxSrc1 = GUINT32_FROM_LE (strf->DVAAuxSrc1);
strf->DVAAuxCtl1 = GUINT32_FROM_LE (strf->DVAAuxCtl1);
strf->DVVAuxSrc = GUINT32_FROM_LE (strf->DVVAuxSrc);
strf->DVVAuxCtl = GUINT32_FROM_LE (strf->DVVAuxCtl);
strf->DVReserved1 = GUINT32_FROM_LE (strf->DVReserved1);
strf->DVReserved2 = GUINT32_FROM_LE (strf->DVReserved2);
#endif
/* debug */
GST_INFO_OBJECT (element, "strf tag found in context iavs:");
GST_INFO_OBJECT (element, " DVAAuxSrc %08x", strf->DVAAuxSrc);
GST_INFO_OBJECT (element, " DVAAuxCtl %08x", strf->DVAAuxCtl);
GST_INFO_OBJECT (element, " DVAAuxSrc1 %08x", strf->DVAAuxSrc1);
GST_INFO_OBJECT (element, " DVAAuxCtl1 %08x", strf->DVAAuxCtl1);
GST_INFO_OBJECT (element, " DVVAuxSrc %08x", strf->DVVAuxSrc);
GST_INFO_OBJECT (element, " DVVAuxCtl %08x", strf->DVVAuxCtl);
GST_INFO_OBJECT (element, " DVReserved1 %08x", strf->DVReserved1);
GST_INFO_OBJECT (element, " DVReserved2 %08x", strf->DVReserved2);
*_strf = strf;
*data = NULL;
return TRUE;
}
/**
* gst_riff_parse_info:
* @element: caller element (used for debugging/error).
* @buf: input data to be used for parsing, stripped from header.
* @taglist: a pointer to a taglist (returned by this function)
* containing information about this stream. May be
* NULL if no supported tags were found.
*
* Parses stream metadata from input data. The input data is
* discarded after use.
*/
void
gst_riff_parse_info (GstElement * element,
GstBuffer * buf, GstTagList ** _taglist)
{
guint8 *data;
guint size, tsize;
guint32 tag;
const gchar *type;
gchar *name;
GstTagList *taglist;
gboolean have_tags = FALSE;
if (!buf) {
*_taglist = NULL;
return;
}
data = GST_BUFFER_DATA (buf);
size = GST_BUFFER_SIZE (buf);
taglist = gst_tag_list_new ();
while (size > 8) {
tag = GST_READ_UINT32_LE (data);
tsize = GST_READ_UINT32_LE (data + 4);
size -= 8;
data += 8;
if (tsize > size) {
GST_WARNING_OBJECT (element,
"Tagsize %d is larger than available data %d", tsize, size);
tsize = size;
}
/* find out the type of metadata */
switch (tag) {
case GST_RIFF_INFO_IARL:
type = GST_TAG_LOCATION;
break;
case GST_RIFF_INFO_IART:
type = GST_TAG_ARTIST;
break;
case GST_RIFF_INFO_ICMS:
type = NULL; /*"Commissioner"; */
break;
case GST_RIFF_INFO_ICMT:
type = GST_TAG_COMMENT;
break;
case GST_RIFF_INFO_ICOP:
type = GST_TAG_COPYRIGHT;
break;
case GST_RIFF_INFO_ICRD:
type = GST_TAG_DATE;
break;
case GST_RIFF_INFO_ICRP:
type = NULL; /*"Cropped"; */
break;
case GST_RIFF_INFO_IDIM:
type = NULL; /*"Dimensions"; */
break;
case GST_RIFF_INFO_IDPI:
type = NULL; /*"Dots per Inch"; */
break;
case GST_RIFF_INFO_IENG:
type = NULL; /*"Engineer"; */
break;
case GST_RIFF_INFO_IGNR:
type = GST_TAG_GENRE;
break;
case GST_RIFF_INFO_IKEY:
type = NULL; /*"Keywords"; */ ;
break;
case GST_RIFF_INFO_ILGT:
type = NULL; /*"Lightness"; */
break;
case GST_RIFF_INFO_IMED:
type = NULL; /*"Medium"; */
break;
case GST_RIFF_INFO_INAM:
type = GST_TAG_TITLE;
break;
case GST_RIFF_INFO_IPLT:
type = NULL; /*"Palette"; */
break;
case GST_RIFF_INFO_IPRD:
type = NULL; /*"Product"; */
break;
case GST_RIFF_INFO_ISBJ:
type = NULL; /*"Subject"; */
break;
case GST_RIFF_INFO_ISFT:
type = GST_TAG_ENCODER;
break;
case GST_RIFF_INFO_ISHP:
type = NULL; /*"Sharpness"; */
break;
case GST_RIFF_INFO_ISRC:
type = GST_TAG_ISRC;
break;
case GST_RIFF_INFO_ISRF:
type = NULL; /*"Source Form"; */
break;
case GST_RIFF_INFO_ITCH:
type = NULL; /*"Technician"; */
break;
default:
type = NULL;
GST_WARNING_OBJECT (element,
"Unknown INFO (metadata) tag entry " GST_FOURCC_FORMAT,
GST_FOURCC_ARGS (tag));
break;
}
if (type) {
if (data[0] != '\0') {
/* read, NULL-terminate */
name = g_new (gchar, tsize + 1);
name[tsize] = '\0';
memcpy (name, data, tsize);
/* add to list */
have_tags = TRUE;
gst_tag_list_add (taglist, GST_TAG_MERGE_APPEND, type, name, NULL);
g_free (name);
}
}
data += tsize;
size -= tsize;
}
if (have_tags) {
*_taglist = taglist;
} else {
*_taglist = NULL;
gst_tag_list_free (taglist);
}
return;
}