Port ffmpeg plugin to FFMpeg 5.0 API

CMakeLists.txt

@@ -126,7 +126,7 @@ if(K3B_ENABLE_MUSICBRAINZ)
 endif()
 
 if(K3B_BUILD_FFMPEG_DECODER_PLUGIN)
-    find_package(FFmpeg)
+    find_package(FFmpeg 4.4.0)
     set_package_properties(FFmpeg PROPERTIES
         PURPOSE "Needed for the K3b FFmpeg decoder plugin which can decode virtually all audio types."
         URL "https://ffmpeg.org/"

config-k3b.h.cmake

@@ -34,8 +34,6 @@
 #cmakedefine HAVE_FFMPEG_AVFORMAT_FIND_STREAM_INFO
 #cmakedefine HAVE_FFMPEG_AVFORMAT_CLOSE_INPUT
 #cmakedefine HAVE_FFMPEG_AVCODEC_OPEN2
-#cmakedefine HAVE_FFMPEG_AVCODEC_DECODE_AUDIO2
-#cmakedefine HAVE_FFMPEG_AVCODEC_DECODE_AUDIO3
 #cmakedefine HAVE_FFMPEG_AVCODEC_DECODE_AUDIO4
 #cmakedefine HAVE_FFMPEG_AVMEDIA_TYPE
 #cmakedefine HAVE_FFMPEG_CODEC_MP3

plugins/decoder/ffmpeg/CMakeLists.txt

 kcoreaddons_add_plugin(k3bffmpegdecoder
     SOURCES k3bffmpegdecoder.cpp k3bffmpegwrapper.cpp
-    INSTALL_NAMESPACE "k3b")
+    INSTALL_NAMESPACE "k3b_plugins")
 
 if(FFMPEG_INCLUDE_DIR_OLD_STYLE)
     message(STATUS "didn't find new ffmpegcodecpath")

plugins/decoder/ffmpeg/k3bffmpegwrapper.cpp

@@ -24,56 +24,34 @@ extern "C" {
 #endif
 }
 
-#include <string.h>
-#include <math.h>
+#include <cstring>
+#include <cmath>
 
 
-#define FFMPEG_CODEC(s) (s->codec)
-
-#ifndef HAVE_FFMPEG_AVFORMAT_OPEN_INPUT
-//      this works because the parameters/options are not used
-#  define avformat_open_input(c,s,f,o) av_open_input_file(c,s,f,0,o)
-#endif
-#ifndef HAVE_FFMPEG_AV_DUMP_FORMAT
-#  define av_dump_format(c,x,f,y) dump_format(c,x,f,y)
-#endif
-#ifndef HAVE_FFMPEG_AVFORMAT_FIND_STREAM_INFO
-#  define avformat_find_stream_info(c,o) av_find_stream_info(c)
-#endif
-#ifndef HAVE_FFMPEG_AVFORMAT_CLOSE_INPUT
-#  define avformat_close_input(c) av_close_input_file(*c)
-#endif
-#ifndef HAVE_FFMPEG_AVCODEC_OPEN2
-#  define avcodec_open2(a,c,o) avcodec_open(a,c)
-#endif
-#ifndef HAVE_FFMPEG_AVMEDIA_TYPE
-#  define AVMEDIA_TYPE_AUDIO CODEC_TYPE_AUDIO
-#endif
 #ifndef HAVE_FFMPEG_CODEC_MP3
 #  define CODEC_ID_MP3 CODEC_ID_MP3LAME
 #endif
 
-K3bFFMpegWrapper* K3bFFMpegWrapper::s_instance = 0;
+K3bFFMpegWrapper* K3bFFMpegWrapper::s_instance = nullptr;
 
 
 class K3bFFMpegFile::Private
 {
 public:
     ::AVFormatContext* formatContext;
-    ::AVCodec* codec;
+    const ::AVCodec* codec;
+    ::AVCodecContext* codecContext;
     ::AVStream *audio_stream;
 
     K3b::Msf length;
 
     // for decoding. ffmpeg requires 16-byte alignment.
-    ::AVFrame* frame;
-    char* outputBufferPos;
-    int outputBufferSize;
-    ::AVPacket packet;
-    quint8* packetData;
-    int packetSize;
+    ::AVFrame* frame = nullptr;
+    ::AVPacket* packet = nullptr;
+    char* outputBufferPos = nullptr;
+    int outputBufferSize = 0;
+    ::AVSampleFormat sampleFormat;
     bool isSpacious;
-    int sampleFormat;
 };
 
 
@@ -81,8 +59,8 @@ K3bFFMpegFile::K3bFFMpegFile( const QString& filename )
     : m_filename(filename)
 {
     d = new Private;
-    d->formatContext = 0;
-    d->codec = 0;
+    d->formatContext = nullptr;
+    d->codec = nullptr;
     d->audio_stream = nullptr;
     d->frame = av_frame_alloc();
 }
@@ -127,38 +105,43 @@ bool K3bFFMpegFile::open()
         }
     }
 
-    // urgh... ugly
-    ::AVCodecContext* codecContext =  FFMPEG_CODEC(d->audio_stream);
-    if( codecContext->codec_type != AVMEDIA_TYPE_AUDIO)
-    {
-        qDebug() << "(K3bFFMpegFile) not a simple audio stream: " << m_filename;
-        return false;
-    }
-
     // get the codec
-    d->codec = ::avcodec_find_decoder(codecContext->codec_id);
+    d->codec = ::avcodec_find_decoder(
+        d->audio_stream->codecpar->codec_id
+    );
     if( !d->codec ) {
         qDebug() << "(K3bFFMpegFile) no codec found for " << m_filename;
         return false;
     }
 
+    // get the codec context
+    d->codecContext = ::avcodec_alloc_context3(d->codec);
+    ::avcodec_parameters_to_context(d->codecContext, d->audio_stream->codecpar);
+    if(d->codecContext->codec_type != AVMEDIA_TYPE_AUDIO) {
+        qDebug() << "(K3bFFMpegFile) not a simple audio stream: " << m_filename;
+        return false;
+    }
+
     // open the codec on our context
     qDebug() << "(K3bFFMpegFile) found codec for " << m_filename;
-    if( ::avcodec_open2( codecContext, d->codec, 0 ) < 0 ) {
+    if( ::avcodec_open2( d->codecContext, d->codec, 0 ) < 0 ) {
         qDebug() << "(K3bFFMpegDecoderFactory) could not open codec.";
         return false;
     }
 
     // determine the length of the stream
-    d->length = K3b::Msf::fromSeconds( (double)d->formatContext->duration / (double)AV_TIME_BASE );
+    d->length = K3b::Msf::fromSeconds(
+        static_cast<double>(d->formatContext->duration) / static_cast<double>(AV_TIME_BASE));
 
     if( d->length == 0 ) {
         qDebug() << "(K3bFFMpegDecoderFactory) invalid length.";
         return false;
     }
 
-    d->sampleFormat = d->audio_stream->codecpar->format;
-    d->isSpacious = ::av_sample_fmt_is_planar((AVSampleFormat)d->sampleFormat) && d->audio_stream->codecpar->channels > 1;
+    d->sampleFormat = d->codecContext->sample_fmt;
+    d->isSpacious = ::av_sample_fmt_is_planar(d->sampleFormat) &&
+        d->codecContext->channels > 1;
+    d->packet = ::av_packet_alloc();
 
     // dump some debugging info
     ::av_dump_format( d->formatContext, 0, m_filename.toLocal8Bit(), 0 );
@@ -170,17 +153,18 @@ bool K3bFFMpegFile::open()
 void K3bFFMpegFile::close()
 {
     d->outputBufferSize = 0;
-    d->packetSize = 0;
-    d->packetData = 0;
+    ::av_packet_free(&d->packet);
 
     if( d->codec ) {
-        ::avcodec_close( FFMPEG_CODEC(d->audio_stream) );
-        d->codec = 0;
+        ::avcodec_close(d->codecContext);
+        d->codec = nullptr;
+        ::avcodec_free_context(&d->codecContext);
+        d->codecContext = nullptr;
     }
 
     if( d->formatContext ) {
         ::avformat_close_input( &d->formatContext );
-        d->formatContext = 0;
+        d->formatContext = nullptr;
     }
 
     d->audio_stream = nullptr;
@@ -195,29 +179,25 @@ K3b::Msf K3bFFMpegFile::length() const
 
 int K3bFFMpegFile::sampleRate() const
 {
-    return d->audio_stream->codecpar->sample_rate;
+    return d->codecContext->sample_rate;
 }
 
 
 int K3bFFMpegFile::channels() const
 {
-    return d->audio_stream->codecpar->channels;
+    return d->codecContext->channels;
 }
 
 
 int K3bFFMpegFile::type() const
 {
-    return d->audio_stream->codecpar->codec_id;
+    return d->codecContext->codec_id;
 }
 
 
 QString K3bFFMpegFile::typeComment() const
 {
     switch( type() ) {
-    case AV_CODEC_ID_WMAV1:
-        return i18n("Windows Media v1");
-    case AV_CODEC_ID_WMAV2:
-        return i18n("Windows Media v2");
     case AV_CODEC_ID_WAVPACK:
         return i18n("WavPack");
     case AV_CODEC_ID_APE:
@@ -233,7 +213,7 @@ QString K3bFFMpegFile::typeComment() const
 QString K3bFFMpegFile::title() const
 {
     // FIXME: is this UTF8 or something??
-    AVDictionaryEntry *ade = av_dict_get( d->formatContext->metadata, "TITLE", NULL, 0 );
+    AVDictionaryEntry *ade = av_dict_get( d->formatContext->metadata, "TITLE", nullptr, 0 );
     return ade && ade->value && ade->value[0] != '\0' ? QString::fromLocal8Bit( ade->value ) : QString();
 }
 
@@ -241,7 +221,7 @@ QString K3bFFMpegFile::title() const
 QString K3bFFMpegFile::author() const
 {
     // FIXME: is this UTF8 or something??
-    AVDictionaryEntry *ade = av_dict_get( d->formatContext->metadata, "ARTIST", NULL, 0 );
+    AVDictionaryEntry *ade = av_dict_get( d->formatContext->metadata, "ARTIST", nullptr, 0 );
     return ade && ade->value && ade->value[0] != '\0' ? QString::fromLocal8Bit( ade->value ) : QString();
 }
 
@@ -249,115 +229,117 @@ QString K3bFFMpegFile::author() const
 QString K3bFFMpegFile::comment() const
 {
     // FIXME: is this UTF8 or something??
-    AVDictionaryEntry *ade = av_dict_get( d->formatContext->metadata, "COMMENT", NULL, 0 );
+    AVDictionaryEntry *ade = av_dict_get( d->formatContext->metadata, "COMMENT", nullptr, 0 );
     return ade && ade->value && ade->value[0] != '\0' ? QString::fromLocal8Bit( ade->value ) : QString();
 }
 
 
 int K3bFFMpegFile::read(char* buf, int bufLen)
 {
-    if (!buf || !d->outputBufferPos)
+    if (!buf ) {
         return -1;
+    }
+
+    if(d->outputBufferSize <= 0)
+        d->outputBufferPos = new char[bufLen];
 
     int ret = fillOutputBuffer();
     if (ret <= 0) {
         return ret;
     }
 
-    int len = qMin(bufLen, d->outputBufferSize);
+    int len = qMin(bufLen, ret);
     ::memcpy(buf, d->outputBufferPos, len);
 
-    if(d->isSpacious && bufLen > d->outputBufferSize)
-        delete[] d->outputBufferPos; // clean up allocated space
-
     // TODO: only swap if needed
-    for(int i=0; i<len-1; i+=2)
-        qSwap(buf[i], buf[i+1]); // BE -> LE
+    for(int i = 0; i < len-1; i += 2)
+        std::swap(buf[i], buf[i+1]); // BE -> LE
 
     d->outputBufferSize -= len;
     if(d->outputBufferSize > 0)
         d->outputBufferPos += len;
+    else
+        delete[] d->outputBufferPos;
+
     return len;
 }
 
 
-// fill d->packetData with data to decode
 int K3bFFMpegFile::readPacket()
 {
-    if( d->packetSize <= 0 ) {
-        ::av_init_packet( &d->packet );
-
-        if( ::av_read_frame( d->formatContext, &d->packet ) < 0 ) {
-            return 0;
-        }
-        d->packetSize = d->packet.size;
-        d->packetData = d->packet.data;
+    if( ::av_read_frame( d->formatContext, d->packet ) < 0 ) {
+        return 0;
     }
 
-    return d->packetSize;
+    return d->packet->size;
 }
 
 
-// decode data in d->packetData and fill d->outputBuffer
 int K3bFFMpegFile::fillOutputBuffer()
 {
+    static int ret = -1;
+
     // decode if the output buffer is empty
-    while(d->outputBufferSize <= 0) {
+    while( d->outputBufferSize <= 0 ) {
+        d->outputBufferSize = 0;
 
-        // make sure we have data to decode
-        if( readPacket() == 0 ) {
-            return 0;
-        }
+        if ( ret >= 0 ) {
+            ret = ::avcodec_receive_frame( d->codecContext, d->frame );
 
-        int gotFrame = 0;
-        int len = ::avcodec_decode_audio4(
-            FFMPEG_CODEC(d->audio_stream),
-            d->frame,
-            &gotFrame,
-            &d->packet );
-
-        if( d->packetSize <= 0 || len < 0 )
-            ::av_packet_unref( &d->packet );
-        if( len < 0 ) {
-            qDebug() << "(K3bFFMpegFile) decoding failed for " << m_filename;
-            return -1;
-        }
+            if ( ret == AVERROR(EAGAIN) || ret == AVERROR_EOF ) {
+                ret = -1;
+                continue;
+            }
+            else if ( ret < 0 ) {
+                qDebug() << "(K3bFFMpegFile) decoding failed for " << m_filename;
+                return -1;
+            }
 
-        if (gotFrame) {
             int nb_s = d->frame->nb_samples;
-            int nb_ch = 2; // copy only two channels even if there're more
-            d->outputBufferSize = nb_s * nb_ch * 2; // 2 means 2 bytes (16bit)
-            d->outputBufferPos = reinterpret_cast<char*>(
-                d->frame->extended_data[0]);
-            if(d->isSpacious) {
-                d->outputBufferPos = new char[d->outputBufferSize];
-                if(d->sampleFormat == AV_SAMPLE_FMT_FLTP) {
-                    int width = sizeof(float); // sample width of float audio
-                    for(int sample=0; sample<nb_s; sample++) {
-                        for(int ch=0; ch<nb_ch; ch++) {
+            const int nb_ch = 2;  // copy only 2 channels even if there're more
+            d->outputBufferSize = nb_s * nb_ch * 2;  // 2 means 2 bytes (16bit)
+
+            if( d->isSpacious ) {
+                if( d->sampleFormat == AV_SAMPLE_FMT_FLTP ) {
+                    constexpr int width = sizeof(float);  // sample width of float audio
+                    for( int sample = 0; sample < nb_s; sample++ ) {
+                        for( int ch = 0; ch < nb_ch; ch++ ) {
                             float val = *(reinterpret_cast<float*>(
                                 d->frame->extended_data[ch] + sample * width));
                             val = ::abs(val) > 1 ? ::copysign(1.0, val) : val;
                             int16_t result = static_cast<int16_t>(
                                 val * 32767.0 + 32768.5) - 32768;
-                            ::memcpy(d->outputBufferPos + (sample*nb_ch+ch) * 2,
-                                     &result,
-                                     2); // 2 is sample width of 16 bit audio
+                            ::memcpy(
+                                d->outputBufferPos + (sample * nb_ch + ch) * 2,
+                                &result,
+                                2  // 2 is sample width of 16 bit audio
+                            );
                         }
                     }
                 } else {
-                    for(int sample=0; sample<nb_s; sample++) {
-                        for(int ch=0; ch<nb_ch; ch++) {
-                            ::memcpy(d->outputBufferPos + (sample*nb_ch+ch) * 2,
-                                     d->frame->extended_data[ch] + sample * 2,
-                                     2); // 16 bit here as well
+                    for(int sample = 0; sample < nb_s; sample++) {
+                        for(int ch = 0; ch < nb_ch; ch++) {
+                            ::memcpy(
+                                d->outputBufferPos + (sample * nb_ch + ch) * 2,
+                                d->frame->extended_data[ch] + sample * 2,
+                                2 // 16 bit here as well
+                            );
                         }
                     }
                 }
             }
+        } else {
+            // make sure we have data to decode
+            if( readPacket() == 0 ) {
+                return 0;
+            }
+            ret = ::avcodec_send_packet( d->codecContext, d->packet );
+            if( ret < 0 ) {
+                qDebug() << "(K3bFFMpegFile) error submitting packet to the decoder";
+                return -1;
+            }
+            continue;
         }
-        d->packetSize -= len;
-        d->packetData += len;
     }
 
     return d->outputBufferSize;
@@ -367,7 +349,6 @@ int K3bFFMpegFile::fillOutputBuffer()
 bool K3bFFMpegFile::seek( const K3b::Msf& msf )
 {
     d->outputBufferSize = 0;
-    d->packetSize = 0;
 
     double seconds = (double)msf.totalFrames()/75.0;
     quint64 timestamp = (quint64)(seconds * (double)AV_TIME_BASE);
@@ -383,13 +364,13 @@ bool K3bFFMpegFile::seek( const K3b::Msf& msf )
 
 K3bFFMpegWrapper::K3bFFMpegWrapper()
 {
-    ::av_register_all();
+    //::av_register_all();
 }
 
 
 K3bFFMpegWrapper::~K3bFFMpegWrapper()
 {
-    s_instance = 0;
+    s_instance = nullptr;
 }
 
 
@@ -413,9 +394,7 @@ K3bFFMpegFile* K3bFFMpegWrapper::open( const QString& filename ) const
         // mp3 being one of them sadly. Most importantly: allow the libsndfile decoder to do
         // its thing.
         //
-        if( file->type() == AV_CODEC_ID_WMAV1 ||
-            file->type() == AV_CODEC_ID_WMAV2 ||
-            file->type() == AV_CODEC_ID_AAC ||
+        if( file->type() == AV_CODEC_ID_AAC ||
             file->type() == AV_CODEC_ID_APE ||
             file->type() == AV_CODEC_ID_WAVPACK )
 #endif
@@ -423,5 +402,5 @@ K3bFFMpegFile* K3bFFMpegWrapper::open( const QString& filename ) const
     }
 
     delete file;
-    return 0;
+    return nullptr;
 }