Add AviUtl2 input plugin (#126)

* Add AviUtl2 input plugin

* Use 16-bit PCM output for AviUtl2 audio

* Add AviUtl2 multi-track stream selection

* Preserve AviUtl2 track options on rebuild failure

* Release unused AviUtl2 reader handlers on close

* Match libavsmash decoder streams by track id

* Avoid lwlibav fallback after libavsmash stream selection

* Probe AviUtl2 MP4 tracks in l-smash order

* Reject invalid libavsmash track numbers

Implemented #119

Author: msg7086

AviUtl2/include/au2/AudioOutputAdapter.hpp

@@ -0,0 +1,17 @@
+#pragma once
+
+#include "au2/ReaderOptions.hpp"
+
+#include "../../common/audio_output.h"
+
+#include <windows.h>
+
+extern "C" {
+#include <libavcodec/avcodec.h>
+}
+
+namespace au2 {
+
+int setup_audio_rendering(lw_audio_output_handler_t* aohp, AVCodecContext* ctx, AudioOptions* opt, WAVEFORMATEX* format);
+
+} // namespace au2

AviUtl2/include/au2/InputPlugin.hpp

@@ -0,0 +1,18 @@
+#pragma once
+
+#ifndef NOMINMAX
+#define NOMINMAX
+#endif
+
+#include <windows.h>
+#include <mmsystem.h>
+
+#include "input2.h"
+
+namespace au2 {
+
+bool initialize_plugin(DWORD version) noexcept;
+void uninitialize_plugin() noexcept;
+INPUT_PLUGIN_TABLE* input_plugin_table() noexcept;
+
+} // namespace au2

AviUtl2/include/au2/InputSession.hpp

@@ -0,0 +1,49 @@
+#pragma once
+
+#include "au2/ReaderBackend.hpp"
+#include "input2.h"
+
+#include <memory>
+#include <string>
+#include <vector>
+
+namespace au2 {
+
+struct InputTrackList {
+    struct Track {
+        int stream_index = -1;
+        int media_track_index = -1;
+    };
+    std::vector<Track> video;
+    std::vector<Track> audio;
+};
+
+class InputSession {
+public:
+    static std::unique_ptr<InputSession> open(LPCWSTR path);
+
+    ~InputSession();
+
+    InputSession(const InputSession&) = delete;
+    InputSession& operator=(const InputSession&) = delete;
+
+    bool info(INPUT_INFO& out) const noexcept;
+    int read_video(int frame, void* dst) noexcept;
+    int read_audio(int start, int length, void* dst) noexcept;
+    bool is_keyframe(int frame) noexcept;
+    int set_track(int type, int index) noexcept;
+
+private:
+    InputSession(std::string path, ReaderOptions options, InputTrackList tracks) noexcept;
+
+    bool rebuild_core() noexcept;
+    int track_count(int type) const noexcept;
+
+    std::string path_;
+    ReaderOptions options_ {};
+    InputTrackList tracks_;
+    SessionCore core_ {};
+    int audio_delay_ = 0;
+};
+
+} // namespace au2

AviUtl2/include/au2/Path.hpp

@@ -0,0 +1,11 @@
+#pragma once
+
+#include <string>
+#include <string_view>
+
+namespace au2 {
+
+std::string wide_to_utf8(std::wstring_view text);
+bool has_script_extension(std::wstring_view path);
+
+} // namespace au2

AviUtl2/include/au2/ReaderBackend.hpp

@@ -0,0 +1,57 @@
+#pragma once
+
+#include "au2/ReaderOptions.hpp"
+
+#include <windows.h>
+#include <mmreg.h>
+
+namespace au2 {
+
+struct SessionCore;
+
+struct ReaderCallbacks {
+    ReaderType type = ReaderType::None;
+    void* (*open_file)(char* file_name, ReaderOptions* opt) = nullptr;
+    int (*get_video_track)(SessionCore* session, VideoOptions* opt) = nullptr;
+    int (*get_audio_track)(SessionCore* session, AudioOptions* opt) = nullptr;
+    void (*destroy_disposable)(void* private_stuff) = nullptr;
+    int (*read_video)(SessionCore* session, int sample_number, void* buf) = nullptr;
+    int (*read_audio)(SessionCore* session, int start, int wanted_length, void* buf) = nullptr;
+    int (*is_keyframe)(SessionCore* session, int sample_number) = nullptr;
+    int (*delay_audio)(SessionCore* session, int* start, int wanted_length, int audio_delay) = nullptr;
+    void (*video_cleanup)(SessionCore* session) = nullptr;
+    void (*audio_cleanup)(SessionCore* session) = nullptr;
+    void (*close_file)(void* private_stuff) = nullptr;
+};
+
+struct SessionCore {
+    void* global_private = nullptr;
+    void (*close_file)(void* private_stuff) = nullptr;
+
+    ReaderType video_reader = ReaderType::None;
+    void* video_private = nullptr;
+    BITMAPINFOHEADER video_format {};
+    int framerate_num = 0;
+    int framerate_den = 1;
+    uint32_t video_sample_count = 0;
+    int (*read_video)(SessionCore* session, int sample_number, void* buf) = nullptr;
+    int (*is_keyframe)(SessionCore* session, int sample_number) = nullptr;
+    void (*video_cleanup)(SessionCore* session) = nullptr;
+    void (*close_video_file)(void* private_stuff) = nullptr;
+
+    ReaderType audio_reader = ReaderType::None;
+    void* audio_private = nullptr;
+    WAVEFORMATEXTENSIBLE audio_format {};
+    uint32_t audio_pcm_sample_count = 0;
+    int (*read_audio)(SessionCore* session, int start, int wanted_length, void* buf) = nullptr;
+    int (*delay_audio)(SessionCore* session, int* start, int wanted_length, int audio_delay) = nullptr;
+    void (*audio_cleanup)(SessionCore* session) = nullptr;
+    void (*close_audio_file)(void* private_stuff) = nullptr;
+};
+
+const ReaderCallbacks& libavsmash_reader() noexcept;
+const ReaderCallbacks& lwlibav_reader() noexcept;
+
+void au2_log_noop(lw_log_handler_t* lhp, lw_log_level level, const char* message);
+
+} // namespace au2

AviUtl2/include/au2/ReaderOptions.hpp

@@ -0,0 +1,71 @@
+#pragma once
+
+#include <cstdint>
+#include <windows.h>
+#include <mmreg.h>
+
+#include "../../common/utils.h"
+
+#define AU2_MAKE_PITCH(bits) ((((bits) + 31) & ~31) >> 3)
+#define AU2_PREFERRED_DECODER_NAMES_BUFSIZE 512
+
+namespace au2 {
+
+enum class ReaderType {
+    None = 0,
+    Libavsmash = 1,
+    Lwlibav = 2,
+};
+
+enum class OutputColorspace {
+    Auto = 0,
+    Yuy2 = 0,
+    Rgb24 = 1,
+    Rgba = 2,
+    Yc48 = 3,
+    Lw48 = 4,
+};
+
+struct VideoOptions {
+    int seek_mode = 0;
+    int forward_seek_threshold = 10;
+    int scaler = 0;
+    int apply_repeat_flag = 1;
+    int field_dominance = 0;
+    struct {
+        int active = 0;
+        int framerate_num = 60000;
+        int framerate_den = 1001;
+    } vfr2cfr;
+    OutputColorspace colorspace = OutputColorspace::Auto;
+};
+
+enum {
+    MIX_LEVEL_INDEX_CENTER = 0,
+    MIX_LEVEL_INDEX_SURROUND,
+    MIX_LEVEL_INDEX_LFE,
+};
+
+struct AudioOptions {
+    uint64_t channel_layout = 0;
+    int sample_rate = 0;
+    int mix_level[3] = { 71, 71, 0 };
+};
+
+struct ReaderOptions {
+    int threads = 0;
+    int av_sync = 1;
+    char preferred_decoder_names_buf[AU2_PREFERRED_DECODER_NAMES_BUFSIZE] {};
+    const char** preferred_decoder_names = nullptr;
+    int no_create_index = 0;
+    int force_video = 0;
+    int force_video_index = -1;
+    int force_audio = 0;
+    int force_audio_index = -1;
+    int libavsmash_video_media_index = 0;
+    int libavsmash_audio_media_index = 0;
+    VideoOptions video;
+    AudioOptions audio;
+};
+
+} // namespace au2

AviUtl2/include/au2/VideoOutputAdapter.hpp

@@ -0,0 +1,33 @@
+#pragma once
+
+#include "au2/ReaderOptions.hpp"
+
+#include "../../common/video_output.h"
+
+#include <windows.h>
+
+extern "C" {
+#include <libavutil/frame.h>
+}
+
+namespace au2 {
+
+typedef void FuncConvertColorspace(lw_video_output_handler_t* vohp, AVFrame* picture, void* buf);
+typedef int FuncConvertColorspaceResult(lw_video_output_handler_t* vohp, AVFrame* picture, uint8_t* buf);
+
+struct VideoOutputHandler {
+    int output_linesize = 0;
+    uint32_t output_frame_size = 0;
+    uint8_t* back_ground = nullptr;
+    uint8_t* another_chroma = nullptr;
+    uint32_t another_chroma_size = 0;
+    AVFrame* yuv444p16 = nullptr;
+    int (*convert_colorspace)(lw_video_output_handler_t* vohp, AVFrame* picture, uint8_t* buf) = nullptr;
+};
+
+int setup_video_rendering(lw_video_output_handler_t* vohp, VideoOptions* opt, BITMAPINFOHEADER* format, int output_width,
+    int output_height, enum AVPixelFormat input_pixel_format);
+
+int convert_colorspace(lw_video_output_handler_t* vohp, AVFrame* picture, uint8_t* buf);
+
+} // namespace au2

AviUtl2/include/au2/input2.h

@@ -0,0 +1,42 @@
+#pragma once
+
+#include <windows.h>
+#include <mmsystem.h>
+
+struct INPUT_INFO {
+    int flag;
+    static constexpr int FLAG_VIDEO = 1;
+    static constexpr int FLAG_AUDIO = 2;
+    static constexpr int FLAG_TIME_TO_FRAME = 16;
+    int rate;
+    int scale;
+    int n;
+    BITMAPINFOHEADER* format;
+    int format_size;
+    int audio_n;
+    WAVEFORMATEX* audio_format;
+    int audio_format_size;
+};
+
+typedef void* INPUT_HANDLE;
+
+struct INPUT_PLUGIN_TABLE {
+    int flag;
+    static constexpr int FLAG_VIDEO = 1;
+    static constexpr int FLAG_AUDIO = 2;
+    static constexpr int FLAG_CONCURRENT = 16;
+    static constexpr int FLAG_MULTI_TRACK = 32;
+    LPCWSTR name;
+    LPCWSTR filefilter;
+    LPCWSTR information;
+    INPUT_HANDLE (*func_open)(LPCWSTR file);
+    bool (*func_close)(INPUT_HANDLE ih);
+    bool (*func_info_get)(INPUT_HANDLE ih, INPUT_INFO* iip);
+    int (*func_read_video)(INPUT_HANDLE ih, int frame, void* buf);
+    int (*func_read_audio)(INPUT_HANDLE ih, int start, int length, void* buf);
+    bool (*func_config)(HWND hwnd, HINSTANCE dll_hinst);
+    int (*func_set_track)(INPUT_HANDLE ih, int type, int index);
+    static constexpr int TRACK_TYPE_VIDEO = 0;
+    static constexpr int TRACK_TYPE_AUDIO = 1;
+    int (*func_time_to_frame)(INPUT_HANDLE ih, double time);
+};