oma: return meaningful errors
[ffmpeg.git] / libavformat / sierravmd.c
index fdf5b5e..645b99b 100644 (file)
  * Sierra VMD Format Demuxer
  * Copyright (c) 2004 The ffmpeg Project
  *
- * This library is free software; you can redistribute it and/or
+ * This file is part of Libav.
+ *
+ * Libav is free software; you can redistribute it and/or
  * modify it under the terms of the GNU Lesser General Public
  * License as published by the Free Software Foundation; either
- * version 2 of the License, or (at your option) any later version.
+ * version 2.1 of the License, or (at your option) any later version.
  *
- * This library is distributed in the hope that it will be useful,
+ * Libav 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
  * Lesser General Public License for more details.
  *
  * You should have received a copy of the GNU Lesser General Public
- * License along with this library; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
 /**
- * @file sierravmd.c
+ * @file
  * Sierra VMD file demuxer
  * by Vladimir "VAG" Gneushev (vagsoft at mail.ru)
  * for more information on the Sierra VMD file format, visit:
  *   http://www.pcisys.net/~melanson/codecs/
  */
 
+#include "libavutil/channel_layout.h"
+#include "libavutil/intreadwrite.h"
 #include "avformat.h"
+#include "internal.h"
 
 #define VMD_HEADER_SIZE 0x0330
 #define BYTES_PER_FRAME_RECORD 16
 
 typedef struct {
   int stream_index;
-  offset_t frame_offset;
+  int64_t frame_offset;
   unsigned int frame_size;
   int64_t pts;
   int keyframe;
   unsigned char frame_record[BYTES_PER_FRAME_RECORD];
-} vmd_frame_t;
+} vmd_frame;
 
 typedef struct VmdDemuxContext {
     int video_stream_index;
     int audio_stream_index;
 
-    unsigned int audio_type;
-    unsigned int audio_samplerate;
-    unsigned int audio_bits;
-    unsigned int audio_channels;
-
     unsigned int frame_count;
-    vmd_frame_t *frame_table;
+    unsigned int frames_per_block;
+    vmd_frame *frame_table;
     unsigned int current_frame;
+    int is_indeo3;
 
     int sample_rate;
     int64_t audio_sample_counter;
-    int audio_frame_divisor;
-    int audio_block_align;
+    int skiphdr;
 
     unsigned char vmd_header[VMD_HEADER_SIZE];
 } VmdDemuxContext;
 
 static int vmd_probe(AVProbeData *p)
 {
-    if (p->buf_size < 2)
+    int w, h;
+    if (p->buf_size < 16)
         return 0;
-
     /* check if the first 2 bytes of the file contain the appropriate size
      * of a VMD header chunk */
-    if (LE_16(&p->buf[0]) != VMD_HEADER_SIZE - 2)
+    if (AV_RL16(&p->buf[0]) != VMD_HEADER_SIZE - 2)
         return 0;
-
-    /* only return half certainty since this check is a bit sketchy */
-    return AVPROBE_SCORE_MAX / 2;
-}
-
-/* This is a support function to determine the duration, in sample
- * frames, of a particular audio chunk, taking into account silent
- * encodings. */
-static int vmd_calculate_audio_duration(unsigned char *audio_chunk,
-    int audio_chunk_size, int block_align)
-{
-    unsigned char *p = audio_chunk + 16;
-    unsigned char *p_end = audio_chunk + audio_chunk_size;
-    int total_samples = 0;
-    unsigned int sound_flags;
-
-    if (audio_chunk_size < 16)
+    w = AV_RL16(&p->buf[12]);
+    h = AV_RL16(&p->buf[14]);
+    if (!w || w > 2048 || !h || h > 2048)
         return 0;
 
-    sound_flags = LE_32(p);
-    p += 4;
-    while (p < p_end) {
-        total_samples += block_align;
-        if ((sound_flags & 0x01) == 0)
-            p += block_align;
-        sound_flags >>= 1;
-    }
-
-    return total_samples;
+    /* only return half certainty since this check is a bit sketchy */
+    return AVPROBE_SCORE_EXTENSION;
 }
 
-static int vmd_read_header(AVFormatContext *s,
-                           AVFormatParameters *ap)
+static int vmd_read_header(AVFormatContext *s)
 {
-    VmdDemuxContext *vmd = (VmdDemuxContext *)s->priv_data;
-    ByteIOContext *pb = &s->pb;
-    AVStream *st;
+    VmdDemuxContext *vmd = s->priv_data;
+    AVIOContext *pb = s->pb;
+    AVStream *st = NULL, *vst;
     unsigned int toc_offset;
     unsigned char *raw_frame_table;
     int raw_frame_table_size;
-    unsigned char *current_frame_record;
-    offset_t current_offset;
-    int i;
+    int64_t current_offset;
+    int i, j;
     unsigned int total_frames;
-    int64_t video_pts_inc;
-    int64_t current_video_pts = 0;
+    int64_t current_audio_pts = 0;
+    unsigned char chunk[BYTES_PER_FRAME_RECORD];
+    int num, den;
+    int sound_buffers;
 
     /* fetch the main header, including the 2 header length bytes */
-    url_fseek(pb, 0, SEEK_SET);
-    if (get_buffer(pb, vmd->vmd_header, VMD_HEADER_SIZE) != VMD_HEADER_SIZE)
-        return AVERROR_IO;
-
-    vmd->audio_sample_counter = 0;
-    vmd->audio_frame_divisor = 1;
-    vmd->audio_block_align = 1;
-
+    avio_seek(pb, 0, SEEK_SET);
+    if (avio_read(pb, vmd->vmd_header, VMD_HEADER_SIZE) != VMD_HEADER_SIZE)
+        return AVERROR(EIO);
+
+    if(vmd->vmd_header[24] == 'i' && vmd->vmd_header[25] == 'v' && vmd->vmd_header[26] == '3')
+        vmd->is_indeo3 = 1;
+    else
+        vmd->is_indeo3 = 0;
     /* start up the decoders */
-    st = av_new_stream(s, 0);
-    if (!st)
-        return AVERROR_NOMEM;
-    av_set_pts_info(st, 33, 1, 90000);
-    vmd->video_stream_index = st->index;
-    st->codec->codec_type = CODEC_TYPE_VIDEO;
-    st->codec->codec_id = CODEC_ID_VMDVIDEO;
-    st->codec->codec_tag = 0;  /* no fourcc */
-    st->codec->width = LE_16(&vmd->vmd_header[12]);
-    st->codec->height = LE_16(&vmd->vmd_header[14]);
-    st->codec->extradata_size = VMD_HEADER_SIZE;
-    st->codec->extradata = av_mallocz(VMD_HEADER_SIZE + FF_INPUT_BUFFER_PADDING_SIZE);
-    memcpy(st->codec->extradata, vmd->vmd_header, VMD_HEADER_SIZE);
+    vst = avformat_new_stream(s, NULL);
+    if (!vst)
+        return AVERROR(ENOMEM);
+    avpriv_set_pts_info(vst, 33, 1, 10);
+    vmd->video_stream_index = vst->index;
+    vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
+    vst->codec->codec_id = vmd->is_indeo3 ? AV_CODEC_ID_INDEO3 : AV_CODEC_ID_VMDVIDEO;
+    vst->codec->codec_tag = 0;  /* no fourcc */
+    vst->codec->width = AV_RL16(&vmd->vmd_header[12]);
+    vst->codec->height = AV_RL16(&vmd->vmd_header[14]);
+    if(vmd->is_indeo3 && vst->codec->width > 320){
+        vst->codec->width >>= 1;
+        vst->codec->height >>= 1;
+    }
+    vst->codec->extradata_size = VMD_HEADER_SIZE;
+    vst->codec->extradata = av_mallocz(VMD_HEADER_SIZE + FF_INPUT_BUFFER_PADDING_SIZE);
+    memcpy(vst->codec->extradata, vmd->vmd_header, VMD_HEADER_SIZE);
 
     /* if sample rate is 0, assume no audio */
-    vmd->sample_rate = LE_16(&vmd->vmd_header[804]);
+    vmd->sample_rate = AV_RL16(&vmd->vmd_header[804]);
     if (vmd->sample_rate) {
-        st = av_new_stream(s, 0);
+        st = avformat_new_stream(s, NULL);
         if (!st)
-            return AVERROR_NOMEM;
-        av_set_pts_info(st, 33, 1, 90000);
+            return AVERROR(ENOMEM);
         vmd->audio_stream_index = st->index;
-        st->codec->codec_type = CODEC_TYPE_AUDIO;
-        st->codec->codec_id = CODEC_ID_VMDAUDIO;
-        st->codec->codec_tag = 0;  /* no codec tag */
-        st->codec->channels = (vmd->vmd_header[811] & 0x80) ? 2 : 1;
+        st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
+        st->codec->codec_id = AV_CODEC_ID_VMDAUDIO;
+        st->codec->codec_tag = 0;  /* no fourcc */
+        if (vmd->vmd_header[811] & 0x80) {
+            st->codec->channels       = 2;
+            st->codec->channel_layout = AV_CH_LAYOUT_STEREO;
+        } else {
+            st->codec->channels       = 1;
+            st->codec->channel_layout = AV_CH_LAYOUT_MONO;
+        }
         st->codec->sample_rate = vmd->sample_rate;
-        st->codec->block_align = vmd->audio_block_align =
-            LE_16(&vmd->vmd_header[806]);
+        st->codec->block_align = AV_RL16(&vmd->vmd_header[806]);
         if (st->codec->block_align & 0x8000) {
-            st->codec->bits_per_sample = 16;
+            st->codec->bits_per_coded_sample = 16;
             st->codec->block_align = -(st->codec->block_align - 0x10000);
-        } else
-            st->codec->bits_per_sample = 16;
-//            st->codec->bits_per_sample = 8;
+        } else {
+            st->codec->bits_per_coded_sample = 8;
+        }
         st->codec->bit_rate = st->codec->sample_rate *
-            st->codec->bits_per_sample * st->codec->channels;
-
-        /* for calculating pts */
-        vmd->audio_frame_divisor = st->codec->bits_per_sample / 8 /
-            st->codec->channels;
-
-        video_pts_inc = 90000;
-        video_pts_inc *= st->codec->block_align;
-        video_pts_inc /= st->codec->sample_rate;
-    } else {
-        /* if no audio, assume 10 frames/second */
-        video_pts_inc = 90000 / 10;
+            st->codec->bits_per_coded_sample * st->codec->channels;
+
+        /* calculate pts */
+        num = st->codec->block_align;
+        den = st->codec->sample_rate * st->codec->channels;
+        av_reduce(&den, &num, den, num, (1UL<<31)-1);
+        avpriv_set_pts_info(vst, 33, num, den);
+        avpriv_set_pts_info(st, 33, num, den);
     }
 
-    /* skip over the offset table and load the table of contents; don't
-     * care about the offset table since demuxer will calculate those
-     * independently */
-    toc_offset = LE_32(&vmd->vmd_header[812]);
-    vmd->frame_count = LE_16(&vmd->vmd_header[6]);
-    url_fseek(pb, toc_offset + vmd->frame_count * 6, SEEK_SET);
-
-    /* each on-disk VMD frame has an audio part and a video part; demuxer
-     * accounts them separately */
-    if(vmd->sample_rate)
-        vmd->frame_count *= 2;
+    toc_offset = AV_RL32(&vmd->vmd_header[812]);
+    vmd->frame_count = AV_RL16(&vmd->vmd_header[6]);
+    vmd->frames_per_block = AV_RL16(&vmd->vmd_header[18]);
+    avio_seek(pb, toc_offset, SEEK_SET);
+
     raw_frame_table = NULL;
     vmd->frame_table = NULL;
-    raw_frame_table_size = vmd->frame_count * BYTES_PER_FRAME_RECORD;
+    sound_buffers = AV_RL16(&vmd->vmd_header[808]);
+    raw_frame_table_size = vmd->frame_count * 6;
+    if(vmd->frame_count * vmd->frames_per_block >= UINT_MAX / sizeof(vmd_frame) - sound_buffers){
+        av_log(s, AV_LOG_ERROR, "vmd->frame_count * vmd->frames_per_block too large\n");
+        return -1;
+    }
     raw_frame_table = av_malloc(raw_frame_table_size);
-    vmd->frame_table = av_malloc(vmd->frame_count * sizeof(vmd_frame_t));
+    vmd->frame_table = av_malloc((vmd->frame_count * vmd->frames_per_block + sound_buffers) * sizeof(vmd_frame));
     if (!raw_frame_table || !vmd->frame_table) {
         av_free(raw_frame_table);
         av_free(vmd->frame_table);
-        return AVERROR_NOMEM;
+        return AVERROR(ENOMEM);
     }
-    if (get_buffer(pb, raw_frame_table, raw_frame_table_size) !=
+    if (avio_read(pb, raw_frame_table, raw_frame_table_size) !=
         raw_frame_table_size) {
         av_free(raw_frame_table);
         av_free(vmd->frame_table);
-        return AVERROR_IO;
+        return AVERROR(EIO);
     }
 
-    current_offset = LE_32(&vmd->vmd_header[20]);
-    current_frame_record = raw_frame_table;
-    total_frames = vmd->frame_count;
-    i = 0;
-    while (total_frames--) {
-
-        /* if the frame size is 0, do not count the frame and bring the
-         * total frame count down */
-        // note, we limit the size to 1Gb to ensure that we dont end up overflowing the size integer used to allocate the memory
-        vmd->frame_table[i].frame_size = LE_32(&current_frame_record[2]) & 0x3FFFFFFF;
-
-        /* this logic is present so that 0-length audio chunks are not
-         * accounted */
-        if (!vmd->frame_table[i].frame_size) {
-            vmd->frame_count--;  /* one less frame to count */
-            current_frame_record += BYTES_PER_FRAME_RECORD;
-            continue;
-        }
-
-        if (current_frame_record[0] == 0x02)
-            vmd->frame_table[i].stream_index = vmd->video_stream_index;
-        else
-            vmd->frame_table[i].stream_index = vmd->audio_stream_index;
-        vmd->frame_table[i].frame_offset = current_offset;
-        current_offset += vmd->frame_table[i].frame_size;
-        memcpy(vmd->frame_table[i].frame_record, current_frame_record,
-            BYTES_PER_FRAME_RECORD);
-
-        /* figure out the pts for this frame */
-        if (current_frame_record[0] == 0x02) {
-            vmd->frame_table[i].pts = current_video_pts;
-            current_video_pts += video_pts_inc;
-        } else if (current_frame_record[0] == 0x01) {
-            /* figure out the pts during the dispatch phase */
-            vmd->frame_table[i].pts = 0;
+    total_frames = 0;
+    for (i = 0; i < vmd->frame_count; i++) {
+
+        current_offset = AV_RL32(&raw_frame_table[6 * i + 2]);
+
+        /* handle each entry in index block */
+        for (j = 0; j < vmd->frames_per_block; j++) {
+            int type;
+            uint32_t size;
+
+            avio_read(pb, chunk, BYTES_PER_FRAME_RECORD);
+            type = chunk[0];
+            size = AV_RL32(&chunk[2]);
+            if(!size && type != 1)
+                continue;
+            switch(type) {
+            case 1: /* Audio Chunk */
+                if (!st) break;
+                /* first audio chunk contains several audio buffers */
+                vmd->frame_table[total_frames].frame_offset = current_offset;
+                vmd->frame_table[total_frames].stream_index = vmd->audio_stream_index;
+                vmd->frame_table[total_frames].frame_size = size;
+                memcpy(vmd->frame_table[total_frames].frame_record, chunk, BYTES_PER_FRAME_RECORD);
+                vmd->frame_table[total_frames].pts = current_audio_pts;
+                total_frames++;
+                if(!current_audio_pts)
+                    current_audio_pts += sound_buffers - 1;
+                else
+                    current_audio_pts++;
+                break;
+            case 2: /* Video Chunk */
+                vmd->frame_table[total_frames].frame_offset = current_offset;
+                vmd->frame_table[total_frames].stream_index = vmd->video_stream_index;
+                vmd->frame_table[total_frames].frame_size = size;
+                memcpy(vmd->frame_table[total_frames].frame_record, chunk, BYTES_PER_FRAME_RECORD);
+                vmd->frame_table[total_frames].pts = i;
+                total_frames++;
+                break;
+            }
+            current_offset += size;
         }
-
-        current_frame_record += BYTES_PER_FRAME_RECORD;
-        i++;
     }
 
     av_free(raw_frame_table);
 
     vmd->current_frame = 0;
+    vmd->frame_count = total_frames;
 
     return 0;
 }
@@ -255,45 +240,38 @@ static int vmd_read_header(AVFormatContext *s,
 static int vmd_read_packet(AVFormatContext *s,
                            AVPacket *pkt)
 {
-    VmdDemuxContext *vmd = (VmdDemuxContext *)s->priv_data;
-    ByteIOContext *pb = &s->pb;
+    VmdDemuxContext *vmd = s->priv_data;
+    AVIOContext *pb = s->pb;
     int ret = 0;
-    vmd_frame_t *frame;
+    vmd_frame *frame;
 
     if (vmd->current_frame >= vmd->frame_count)
-        return AVERROR_IO;
+        return AVERROR(EIO);
 
     frame = &vmd->frame_table[vmd->current_frame];
     /* position the stream (will probably be there already) */
-    url_fseek(pb, frame->frame_offset, SEEK_SET);
+    avio_seek(pb, frame->frame_offset, SEEK_SET);
 
     if (av_new_packet(pkt, frame->frame_size + BYTES_PER_FRAME_RECORD))
-        return AVERROR_NOMEM;
-    pkt->pos= url_ftell(pb);
+        return AVERROR(ENOMEM);
+    pkt->pos= avio_tell(pb);
     memcpy(pkt->data, frame->frame_record, BYTES_PER_FRAME_RECORD);
-    ret = get_buffer(pb, pkt->data + BYTES_PER_FRAME_RECORD,
-        frame->frame_size);
+    if(vmd->is_indeo3 && frame->frame_record[0] == 0x02)
+        ret = avio_read(pb, pkt->data, frame->frame_size);
+    else
+        ret = avio_read(pb, pkt->data + BYTES_PER_FRAME_RECORD,
+            frame->frame_size);
 
     if (ret != frame->frame_size) {
         av_free_packet(pkt);
-        ret = AVERROR_IO;
+        ret = AVERROR(EIO);
     }
     pkt->stream_index = frame->stream_index;
-    if (frame->frame_record[0] == 0x02)
-        pkt->pts = frame->pts;
-    else {
-        pkt->pts = vmd->audio_sample_counter;
-        pkt->pts *= 90000;
-        pkt->pts /= vmd->sample_rate;
-//        pkt->pts /= vmd->audio_frame_divisor;
-        vmd->audio_sample_counter += vmd_calculate_audio_duration(
-            pkt->data, pkt->size, vmd->audio_block_align);
-
-    }
-av_log(NULL, AV_LOG_INFO, " dispatching %s frame with %d bytes and pts %"PRId64" (%0.1f sec)\n",
-  (frame->frame_record[0] == 0x02) ? "video" : "audio",
-  frame->frame_size + BYTES_PER_FRAME_RECORD,
-  pkt->pts, (float)(pkt->pts / 90000.0));
+    pkt->pts = frame->pts;
+    av_log(s, AV_LOG_DEBUG, " dispatching %s frame with %d bytes and pts %"PRId64"\n",
+            (frame->frame_record[0] == 0x02) ? "video" : "audio",
+            frame->frame_size + BYTES_PER_FRAME_RECORD,
+            pkt->pts);
 
     vmd->current_frame++;
 
@@ -302,25 +280,19 @@ av_log(NULL, AV_LOG_INFO, " dispatching %s frame with %d bytes and pts %"PRId64"
 
 static int vmd_read_close(AVFormatContext *s)
 {
-    VmdDemuxContext *vmd = (VmdDemuxContext *)s->priv_data;
+    VmdDemuxContext *vmd = s->priv_data;
 
     av_free(vmd->frame_table);
 
     return 0;
 }
 
-static AVInputFormat vmd_iformat = {
-    "vmd",
-    "Sierra VMD format",
-    sizeof(VmdDemuxContext),
-    vmd_probe,
-    vmd_read_header,
-    vmd_read_packet,
-    vmd_read_close,
+AVInputFormat ff_vmd_demuxer = {
+    .name           = "vmd",
+    .long_name      = NULL_IF_CONFIG_SMALL("Sierra VMD"),
+    .priv_data_size = sizeof(VmdDemuxContext),
+    .read_probe     = vmd_probe,
+    .read_header    = vmd_read_header,
+    .read_packet    = vmd_read_packet,
+    .read_close     = vmd_read_close,
 };
-
-int vmd_init(void)
-{
-    av_register_input_format(&vmd_iformat);
-    return 0;
-}