Add missing #includes for *INT64_MAX and *INT64_C
[ffmpeg.git] / libavformat / oggenc.c
index f947269..e96cde7 100644 (file)
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include <stdint.h>
+
 #include "libavutil/crc.h"
 #include "libavutil/mathematics.h"
+#include "libavutil/opt.h"
 #include "libavutil/random_seed.h"
 #include "libavcodec/xiph.h"
 #include "libavcodec/bytestream.h"
@@ -33,6 +36,7 @@
 #define MAX_PAGE_SIZE 65025
 
 typedef struct {
+    int64_t start_granule;
     int64_t granule;
     int stream_index;
     uint8_t flags;
@@ -63,9 +67,31 @@ typedef struct OGGPageList {
 } OGGPageList;
 
 typedef struct {
+    const AVClass *class;
     OGGPageList *page_list;
+    int pref_size; ///< preferred page size (0 => fill all segments)
+    int64_t pref_duration;      ///< preferred page duration (0 => fill all segments)
 } OGGContext;
 
+#define OFFSET(x) offsetof(OGGContext, x)
+#define PARAM AV_OPT_FLAG_ENCODING_PARAM
+
+static const AVOption options[] = {
+    { "pagesize", "preferred page size in bytes (deprecated)",
+        OFFSET(pref_size), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, MAX_PAGE_SIZE, PARAM },
+    { "page_duration", "preferred page duration, in microseconds",
+        OFFSET(pref_duration), AV_OPT_TYPE_INT, { .i64 = 1000000 }, 0, INT64_MAX, PARAM },
+    { NULL },
+};
+
+static const AVClass ogg_muxer_class = {
+    .class_name = "Ogg muxer",
+    .item_name  = av_default_item_name,
+    .option     = options,
+    .version    = LIBAVUTIL_VERSION_INT,
+};
+
+
 static void ogg_update_checksum(AVFormatContext *s, AVIOContext *pb, int64_t crc_offset)
 {
     int64_t pos = avio_tell(pb);
@@ -157,6 +183,7 @@ static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
         return AVERROR(ENOMEM);
     l->page = oggstream->page;
 
+    oggstream->page.start_granule = oggstream->page.granule;
     oggstream->page_count++;
     ogg_reset_cur_page(oggstream);
 
@@ -172,15 +199,17 @@ static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
 }
 
 static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
-                           uint8_t *data, unsigned size, int64_t granule)
+                           uint8_t *data, unsigned size, int64_t granule,
+                           int header)
 {
     OGGStreamContext *oggstream = st->priv_data;
+    OGGContext *ogg = s->priv_data;
     int total_segments = size / 255 + 1;
     uint8_t *p = data;
     int i, segments, len, flush = 0;
 
     // Handles VFR by flushing page because this frame needs to have a timestamp
-    if (st->codec->codec_id == CODEC_ID_THEORA &&
+    if (st->codec->codec_id == AV_CODEC_ID_THEORA && !header &&
         ogg_granule_to_timestamp(oggstream, granule) >
         ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1) {
         if (oggstream->page.granule != -1)
@@ -188,6 +217,12 @@ static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
         flush = 1;
     }
 
+    // avoid a continued page
+    if (!header && oggstream->page.size > 0 &&
+        MAX_PAGE_SIZE - oggstream->page.size < size) {
+        ogg_buffer_page(s, oggstream);
+    }
+
     for (i = 0; i < total_segments; ) {
         OGGPage *page = &oggstream->page;
 
@@ -210,8 +245,19 @@ static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
         if (i == total_segments)
             page->granule = granule;
 
-        if (page->segments_count == 255) {
-            ogg_buffer_page(s, oggstream);
+        if (!header) {
+            AVStream *st = s->streams[page->stream_index];
+
+            int64_t start = av_rescale_q(page->start_granule, st->time_base,
+                                         AV_TIME_BASE_Q);
+            int64_t next  = av_rescale_q(page->granule, st->time_base,
+                                         AV_TIME_BASE_Q);
+
+            if (page->segments_count == 255 ||
+                (ogg->pref_size     > 0 && page->size   >= ogg->pref_size) ||
+                (ogg->pref_duration > 0 && next - start >= ogg->pref_duration)) {
+                ogg_buffer_page(s, oggstream);
+            }
         }
     }
 
@@ -313,23 +359,61 @@ static int ogg_build_speex_headers(AVCodecContext *avctx,
     return 0;
 }
 
+#define OPUS_HEADER_SIZE 19
+
+static int ogg_build_opus_headers(AVCodecContext *avctx,
+                                  OGGStreamContext *oggstream, int bitexact,
+                                  AVDictionary **m)
+{
+    uint8_t *p;
+
+    if (avctx->extradata_size < OPUS_HEADER_SIZE)
+        return -1;
+
+    /* first packet: Opus header */
+    p = av_mallocz(avctx->extradata_size);
+    if (!p)
+        return AVERROR(ENOMEM);
+    oggstream->header[0] = p;
+    oggstream->header_len[0] = avctx->extradata_size;
+    bytestream_put_buffer(&p, avctx->extradata, avctx->extradata_size);
+
+    /* second packet: VorbisComment */
+    p = ogg_write_vorbiscomment(8, bitexact, &oggstream->header_len[1], m, 0);
+    if (!p)
+        return AVERROR(ENOMEM);
+    oggstream->header[1] = p;
+    bytestream_put_buffer(&p, "OpusTags", 8);
+
+    return 0;
+}
+
 static int ogg_write_header(AVFormatContext *s)
 {
+    OGGContext *ogg = s->priv_data;
     OGGStreamContext *oggstream;
     int i, j;
 
+    if (ogg->pref_size)
+        av_log(s, AV_LOG_WARNING, "The pagesize option is deprecated\n");
+
     for (i = 0; i < s->nb_streams; i++) {
         AVStream *st = s->streams[i];
         unsigned serial_num = i;
 
         if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
-            avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
+            if (st->codec->codec_id == AV_CODEC_ID_OPUS)
+                /* Opus requires a fixed 48kHz clock */
+                avpriv_set_pts_info(st, 64, 1, 48000);
+            else
+                avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
         else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
             avpriv_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
-        if (st->codec->codec_id != CODEC_ID_VORBIS &&
-            st->codec->codec_id != CODEC_ID_THEORA &&
-            st->codec->codec_id != CODEC_ID_SPEEX  &&
-            st->codec->codec_id != CODEC_ID_FLAC) {
+        if (st->codec->codec_id != AV_CODEC_ID_VORBIS &&
+            st->codec->codec_id != AV_CODEC_ID_THEORA &&
+            st->codec->codec_id != AV_CODEC_ID_SPEEX  &&
+            st->codec->codec_id != AV_CODEC_ID_FLAC   &&
+            st->codec->codec_id != AV_CODEC_ID_OPUS) {
             av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
             return -1;
         }
@@ -353,7 +437,7 @@ static int ogg_write_header(AVFormatContext *s)
         oggstream->serial_num = serial_num;
 
         st->priv_data = oggstream;
-        if (st->codec->codec_id == CODEC_ID_FLAC) {
+        if (st->codec->codec_id == AV_CODEC_ID_FLAC) {
             int err = ogg_build_flac_headers(st->codec, oggstream,
                                              st->codec->flags & CODEC_FLAG_BITEXACT,
                                              &s->metadata);
@@ -362,7 +446,7 @@ static int ogg_write_header(AVFormatContext *s)
                 av_freep(&st->priv_data);
                 return err;
             }
-        } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
+        } else if (st->codec->codec_id == AV_CODEC_ID_SPEEX) {
             int err = ogg_build_speex_headers(st->codec, oggstream,
                                               st->codec->flags & CODEC_FLAG_BITEXACT,
                                               &s->metadata);
@@ -371,14 +455,23 @@ static int ogg_write_header(AVFormatContext *s)
                 av_freep(&st->priv_data);
                 return err;
             }
+        } else if (st->codec->codec_id == AV_CODEC_ID_OPUS) {
+            int err = ogg_build_opus_headers(st->codec, oggstream,
+                                             st->codec->flags & CODEC_FLAG_BITEXACT,
+                                             &s->metadata);
+            if (err) {
+                av_log(s, AV_LOG_ERROR, "Error writing Opus headers\n");
+                av_freep(&st->priv_data);
+                return err;
+            }
         } else {
             uint8_t *p;
-            const char *cstr = st->codec->codec_id == CODEC_ID_VORBIS ? "vorbis" : "theora";
-            int header_type = st->codec->codec_id == CODEC_ID_VORBIS ? 3 : 0x81;
-            int framing_bit = st->codec->codec_id == CODEC_ID_VORBIS ? 1 : 0;
+            const char *cstr = st->codec->codec_id == AV_CODEC_ID_VORBIS ? "vorbis" : "theora";
+            int header_type = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 3 : 0x81;
+            int framing_bit = st->codec->codec_id == AV_CODEC_ID_VORBIS ? 1 : 0;
 
             if (avpriv_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
-                                      st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
+                                      st->codec->codec_id == AV_CODEC_ID_VORBIS ? 30 : 42,
                                       oggstream->header, oggstream->header_len) < 0) {
                 av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
                 av_freep(&st->priv_data);
@@ -395,7 +488,7 @@ static int ogg_write_header(AVFormatContext *s)
             bytestream_put_byte(&p, header_type);
             bytestream_put_buffer(&p, cstr, 6);
 
-            if (st->codec->codec_id == CODEC_ID_THEORA) {
+            if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
                 /** KFGSHIFT is the width of the less significant section of the granule position
                     The less significant section is the frame count since the last keyframe */
                 oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
@@ -409,7 +502,7 @@ static int ogg_write_header(AVFormatContext *s)
     for (j = 0; j < s->nb_streams; j++) {
         OGGStreamContext *oggstream = s->streams[j]->priv_data;
         ogg_buffer_data(s, s->streams[j], oggstream->header[0],
-                        oggstream->header_len[0], 0);
+                        oggstream->header_len[0], 0, 1);
         oggstream->page.flags |= 2; // bos
         ogg_buffer_page(s, oggstream);
     }
@@ -419,10 +512,13 @@ static int ogg_write_header(AVFormatContext *s)
         for (i = 1; i < 3; i++) {
             if (oggstream && oggstream->header_len[i])
                 ogg_buffer_data(s, st, oggstream->header[i],
-                                oggstream->header_len[i], 0);
+                                oggstream->header_len[i], 0, 1);
         }
         ogg_buffer_page(s, oggstream);
     }
+
+    oggstream->page.start_granule = AV_NOPTS_VALUE;
+
     return 0;
 }
 
@@ -455,7 +551,7 @@ static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
     int ret;
     int64_t granule;
 
-    if (st->codec->codec_id == CODEC_ID_THEORA) {
+    if (st->codec->codec_id == AV_CODEC_ID_THEORA) {
         int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
         int pframe_count;
         if (pkt->flags & AV_PKT_FLAG_KEY)
@@ -467,10 +563,15 @@ static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
             pframe_count = 0;
         }
         granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
-    } else
+    } else if (st->codec->codec_id == AV_CODEC_ID_OPUS)
+        granule = pkt->pts + pkt->duration + av_rescale_q(st->codec->delay, (AVRational){ 1, st->codec->sample_rate }, st->time_base);
+    else
         granule = pkt->pts + pkt->duration;
 
-    ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule);
+    if (oggstream->page.start_granule == AV_NOPTS_VALUE)
+        oggstream->page.start_granule = pkt->pts;
+
+    ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule, 0);
     if (ret < 0)
         return ret;
 
@@ -485,17 +586,22 @@ static int ogg_write_trailer(AVFormatContext *s)
 {
     int i;
 
-    /* flush current page */
-    for (i = 0; i < s->nb_streams; i++)
-        ogg_buffer_page(s, s->streams[i]->priv_data);
+    /* flush current page if needed */
+    for (i = 0; i < s->nb_streams; i++) {
+        OGGStreamContext *oggstream = s->streams[i]->priv_data;
+
+        if (oggstream->page.size > 0)
+            ogg_buffer_page(s, oggstream);
+    }
 
     ogg_write_pages(s, 1);
 
     for (i = 0; i < s->nb_streams; i++) {
         AVStream *st = s->streams[i];
         OGGStreamContext *oggstream = st->priv_data;
-        if (st->codec->codec_id == CODEC_ID_FLAC ||
-            st->codec->codec_id == CODEC_ID_SPEEX) {
+        if (st->codec->codec_id == AV_CODEC_ID_FLAC ||
+            st->codec->codec_id == AV_CODEC_ID_SPEEX ||
+            st->codec->codec_id == AV_CODEC_ID_OPUS) {
             av_free(oggstream->header[0]);
         }
         av_freep(&oggstream->header[1]);
@@ -508,11 +614,13 @@ AVOutputFormat ff_ogg_muxer = {
     .name              = "ogg",
     .long_name         = NULL_IF_CONFIG_SMALL("Ogg"),
     .mime_type         = "application/ogg",
-    .extensions        = "ogg,ogv,spx",
+    .extensions        = "ogg,ogv,spx,opus",
     .priv_data_size    = sizeof(OGGContext),
-    .audio_codec       = CODEC_ID_FLAC,
-    .video_codec       = CODEC_ID_THEORA,
+    .audio_codec       = AV_CODEC_ID_FLAC,
+    .video_codec       = AV_CODEC_ID_THEORA,
     .write_header      = ogg_write_header,
     .write_packet      = ogg_write_packet,
     .write_trailer     = ogg_write_trailer,
+    .flags             = AVFMT_TS_NEGATIVE,
+    .priv_class        = &ogg_muxer_class,
 };