]> git.sesse.net Git - casparcg/blobdiff - modules/ffmpeg/producer/audio/audio_decoder.cpp
2.0.2: - graph: Fixed potential buffer overflow.
[casparcg] / modules / ffmpeg / producer / audio / audio_decoder.cpp
index 13ede2f734238635e608495fa256e541d1d74346..1ac7fdeaa796fe61c3db3c1602f62b11e747ca5a 100644 (file)
@@ -1,34 +1,45 @@
 /*\r
-* copyright (c) 2010 Sveriges Television AB <info@casparcg.com>\r
+* Copyright (c) 2011 Sveriges Television AB <info@casparcg.com>\r
 *\r
-*  This file is part of CasparCG.\r
+* This file is part of CasparCG (www.casparcg.com).\r
 *\r
-*    CasparCG is free software: you can redistribute it and/or modify\r
-*    it under the terms of the GNU General Public License as published by\r
-*    the Free Software Foundation, either version 3 of the License, or\r
-*    (at your option) any later version.\r
+* CasparCG is free software: you can redistribute it and/or modify\r
+* it under the terms of the GNU General Public License as published by\r
+* the Free Software Foundation, either version 3 of the License, or\r
+* (at your option) any later version.\r
 *\r
-*    CasparCG is distributed in the hope that it will be useful,\r
-*    but WITHOUT ANY WARRANTY; without even the implied warranty of\r
-*    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the\r
-*    GNU General Public License for more details.\r
-\r
-*    You should have received a copy of the GNU General Public License\r
-*    along with CasparCG.  If not, see <http://www.gnu.org/licenses/>.\r
+* CasparCG is distributed in the hope that it will be useful,\r
+* but WITHOUT ANY WARRANTY; without even the implied warranty of\r
+* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the\r
+* GNU General Public License for more details.\r
+*\r
+* You should have received a copy of the GNU General Public License\r
+* along with CasparCG. If not, see <http://www.gnu.org/licenses/>.\r
 *\r
+* Author: Robert Nagy, ronag89@gmail.com\r
 */\r
+\r
 #include "../../stdafx.h"\r
 \r
 #include "audio_decoder.h"\r
 \r
+#include "audio_resampler.h"\r
+\r
+#include "../util/util.h"\r
+#include "../../ffmpeg_error.h"\r
+\r
+#include <core/video_format.h>\r
+\r
+#include <tbb/cache_aligned_allocator.h>\r
+\r
+#include <queue>\r
+\r
 #if defined(_MSC_VER)\r
 #pragma warning (push)\r
 #pragma warning (disable : 4244)\r
 #endif\r
 extern "C" \r
 {\r
-       #define __STDC_CONSTANT_MACROS\r
-       #define __STDC_LIMIT_MACROS\r
        #include <libavformat/avformat.h>\r
        #include <libavcodec/avcodec.h>\r
 }\r
@@ -36,96 +47,107 @@ extern "C"
 #pragma warning (pop)\r
 #endif\r
 \r
-namespace caspar {\r
-\r
+namespace caspar { namespace ffmpeg {\r
+       \r
 struct audio_decoder::implementation : boost::noncopyable\r
 {      \r
-       input&                                                  input_;\r
-       AVCodecContext&                                 codec_context_;         \r
-       const core::video_format_desc   format_desc_;\r
+       int                                                                                                                     index_;\r
+       const safe_ptr<AVCodecContext>                                                          codec_context_;         \r
+       const core::video_format_desc                                                           format_desc_;\r
 \r
-       std::vector<int16_t, tbb::cache_aligned_allocator<int16_t>>                     current_chunk_; \r
+       audio_resampler                                                                                         resampler_;\r
 \r
-       size_t                                                  frame_number_;\r
-       bool                                                    wait_for_eof_;\r
+       std::vector<int8_t,  tbb::cache_aligned_allocator<int8_t>>      buffer1_;\r
 \r
-       std::vector<int16_t, tbb::cache_aligned_allocator<int16_t>> buffer_;\r
-public:\r
-       explicit implementation(input& input, const core::video_format_desc& format_desc) \r
-               : input_(input)\r
-               , codec_context_(*input_.get_audio_codec_context())\r
-               , format_desc_(format_desc)     \r
-               , frame_number_(0)\r
-               , wait_for_eof_(false)\r
-               , buffer_(4*format_desc_.audio_sample_rate*2+FF_INPUT_BUFFER_PADDING_SIZE/2, 0)\r
-       {\r
-               if(codec_context_.sample_rate != static_cast<int>(format_desc_.audio_sample_rate) || \r
-                  codec_context_.channels != static_cast<int>(format_desc_.audio_channels))\r
-               {       \r
-                       BOOST_THROW_EXCEPTION(\r
-                               file_read_error()  <<\r
-                               msg_info("Invalid sample-rate or number of channels.") <<\r
-                               arg_value_info(boost::lexical_cast<std::string>(codec_context_.sample_rate)) << \r
-                               arg_name_info("codec_context"));\r
-               }\r
-       }\r
-               \r
-       std::deque<std::pair<int, std::vector<int16_t>>> receive()\r
-       {\r
-               std::deque<std::pair<int, std::vector<int16_t>>> result;\r
-               \r
-               std::shared_ptr<AVPacket> pkt;\r
-               for(int n = 0; n < 32 && result.empty() && input_.try_pop_audio_packet(pkt); ++n)       \r
-                       result = decode(pkt);\r
+       std::queue<safe_ptr<AVPacket>>                                                          packets_;\r
 \r
-               return result;\r
+       const int64_t                                                                                           nb_frames_;\r
+       tbb::atomic<size_t>                                                                                     file_frame_number_;\r
+public:\r
+       explicit implementation(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) \r
+               : format_desc_(format_desc)     \r
+               , codec_context_(open_codec(*context, AVMEDIA_TYPE_AUDIO, index_))\r
+               , resampler_(format_desc.audio_channels,        codec_context_->channels,\r
+                                        format_desc.audio_sample_rate, codec_context_->sample_rate,\r
+                                        AV_SAMPLE_FMT_S32,                             codec_context_->sample_fmt)\r
+               , buffer1_(AVCODEC_MAX_AUDIO_FRAME_SIZE*2)\r
+               , nb_frames_(0)//context->streams[index_]->nb_frames)\r
+       {               \r
+               file_frame_number_ = 0;\r
+               CASPAR_LOG(debug) << "[audio_decoder] " << context->streams[index_]->codec->codec->long_name;      \r
        }\r
 \r
-       std::deque<std::pair<int, std::vector<int16_t>>> decode(const std::shared_ptr<AVPacket>& audio_packet)\r
+       void push(const std::shared_ptr<AVPacket>& packet)\r
        {                       \r
-               std::deque<std::pair<int, std::vector<int16_t>>> result;\r
-\r
-               if(!audio_packet) // eof\r
-               {       \r
-                       avcodec_flush_buffers(&codec_context_);\r
-                       current_chunk_.clear();\r
-                       frame_number_ = 0;\r
-                       wait_for_eof_ = false;\r
-                       return result;\r
+               if(!packet)\r
+                       return;\r
+\r
+               if(packet->stream_index == index_ || packet->data == nullptr)\r
+                       packets_.push(make_safe_ptr(packet));\r
+       }       \r
+       \r
+       std::shared_ptr<core::audio_buffer> poll()\r
+       {\r
+               if(packets_.empty())\r
+                       return nullptr;\r
+                               \r
+               auto packet = packets_.front();\r
+\r
+               if(packet->data == nullptr)\r
+               {\r
+                       packets_.pop();\r
+                       file_frame_number_ = static_cast<size_t>(packet->pos);\r
+                       avcodec_flush_buffers(codec_context_.get());\r
+                       return flush_audio();\r
                }\r
 \r
-               if(wait_for_eof_)\r
-                       return result;\r
-                                               \r
-               int written_bytes = buffer_.size()-FF_INPUT_BUFFER_PADDING_SIZE/2;\r
-               const int errn = avcodec_decode_audio3(&codec_context_, buffer_.data(), &written_bytes, audio_packet.get());\r
-               if(errn < 0)\r
-               {       \r
-                       BOOST_THROW_EXCEPTION(\r
-                               invalid_operation() <<\r
-                               boost::errinfo_api_function("avcodec_decode_audio2") <<\r
-                               boost::errinfo_errno(AVUNERROR(errn)));\r
-               }\r
+               auto audio = decode(*packet);\r
+\r
+               if(packet->size == 0)                                   \r
+                       packets_.pop();\r
+\r
+               return audio;\r
+       }\r
+\r
+       std::shared_ptr<core::audio_buffer> decode(AVPacket& pkt)\r
+       {               \r
+               buffer1_.resize(AVCODEC_MAX_AUDIO_FRAME_SIZE*2);\r
+               int written_bytes = buffer1_.size() - FF_INPUT_BUFFER_PADDING_SIZE;\r
+               \r
+               int ret = THROW_ON_ERROR2(avcodec_decode_audio3(codec_context_.get(), reinterpret_cast<int16_t*>(buffer1_.data()), &written_bytes, &pkt), "[audio_decoder]");\r
 \r
-               current_chunk_.insert(current_chunk_.end(), buffer_.begin(), buffer_.begin() + written_bytes/2);\r
+               // There might be several frames in one packet.\r
+               pkt.size -= ret;\r
+               pkt.data += ret;\r
+                       \r
+               buffer1_.resize(written_bytes);\r
 \r
-               const auto last = current_chunk_.end() - current_chunk_.size() % format_desc_.audio_samples_per_frame;\r
+               buffer1_ = resampler_.resample(std::move(buffer1_));\r
                \r
-               for(auto it = current_chunk_.begin(); it != last; it += format_desc_.audio_samples_per_frame)           \r
-                       result.push_back(std::make_pair(frame_number_++, std::vector<int16_t>(it, it + format_desc_.audio_samples_per_frame)));         \r
+               const auto n_samples = buffer1_.size() / av_get_bytes_per_sample(AV_SAMPLE_FMT_S32);\r
+               const auto samples = reinterpret_cast<int32_t*>(buffer1_.data());\r
 \r
-               current_chunk_.erase(current_chunk_.begin(), last);\r
+               ++file_frame_number_;\r
 \r
-               return result;\r
+               return std::make_shared<core::audio_buffer>(samples, samples + n_samples);\r
        }\r
 \r
-       void restart()\r
+       bool ready() const\r
        {\r
-               wait_for_eof_ = true;\r
+               return packets_.size() > 10;\r
+       }\r
+\r
+       uint32_t nb_frames() const\r
+       {\r
+               return 0;//std::max<int64_t>(nb_frames_, file_frame_number_);\r
        }\r
 };\r
 \r
-audio_decoder::audio_decoder(input& input, const core::video_format_desc& format_desc) : impl_(new implementation(input, format_desc)){}\r
-std::deque<std::pair<int, std::vector<int16_t>>> audio_decoder::receive(){return impl_->receive();}\r
-void audio_decoder::restart(){impl_->restart();}\r
-}
\ No newline at end of file
+audio_decoder::audio_decoder(const safe_ptr<AVFormatContext>& context, const core::video_format_desc& format_desc) : impl_(new implementation(context, format_desc)){}\r
+void audio_decoder::push(const std::shared_ptr<AVPacket>& packet){impl_->push(packet);}\r
+bool audio_decoder::ready() const{return impl_->ready();}\r
+std::shared_ptr<core::audio_buffer> audio_decoder::poll(){return impl_->poll();}\r
+uint32_t audio_decoder::nb_frames() const{return impl_->nb_frames();}\r
+uint32_t audio_decoder::file_frame_number() const{return impl_->file_frame_number_;}\r
+\r
+}}
\ No newline at end of file