In Linux, Makefiles can be used to cross-compile with the MingW headers, but they aren't well prepared to generate native code at the moment. It should be fixable with some effort.
**With MSVC**: open *./vgmstream_full.sln* and compile in Visual Studio, or use MSBuild in the command line. See the foobar2000 section for dependencies and CMD examples.
Open *./vgmstream_full.sln* as a base, which expects the above dependencies. Then, depending on your VS version (like VS2015) you may need to manually do the following (in *Debug* and *Release* configurations):
VS2013/VS2015/VS2017 may not be compatible with the SDK in release mode when some options are enabled due to compiler bugs.
FDK-AAC/QAAC can be safely disabled by removing *VGM_USE_MP4V2* and *VGM_USE_FDKAAC* in the compiler/linker options and the project dependencies, MAIATRAC3 too, as FFmpeg is used instead to support their codecs.
You can also use the command line to compile with MSBuild, if you don't want to touch the .vcxproj files, register VS2015 after trial, or only have VC++/MSBuild tools.
The plugin itself works with Audacious 3.5 or higher. New Audacious releases can break plugin compatibility so it may not work with the latest version unless adapted first.
FFmpeg and other external libraries aren't enabled, thus some formats are not supported. libvorbis and libmpg123 can be disabled with -DVGM_DISABLE_VORBIS and -DVGM_DISABLE_MPEG.
Windows builds aren't supported at the moment (should be possible but there are complex dependency chains).
Terminal example, assuming a Ubuntu-based Linux distribution:
vgmstream uses C (C89 when possible), and C++ for the foobar2000 and Audacious plugins.
C is restricted to features VS2010 can understand. This mainly means means declaring variables at the start of a { .. } block (declare+initialize is fine, as long as it doesn't reference variables declared in that block) and avoiding C99 like variable-length arrays (but others like // comments are fine).
vgmstream works by parsing a music stream header (*meta/*), preparing/demuxing data (*layout/*) and decoding the compressed data into listenable PCM samples (*coding/*).
- init tries all parsers (metas) until one works *[init_vgmstream]*
- parser reads header (channels, sample rate, loop points) and set ups a VGMSTREAM struct + layout/coding, if the format is correct *[init_vgmstream_(format-name)]*
- player gets total_samples to play, based on the number of loops and other settings *[get_vgmstream_play_samples]*
- player asks to fill a small sample buffer *[render_vgmstream]*
- *src/meta/(format-name).c*: create new init_vgmstream_(format-name) parser that tests the extension and header id, and reads all needed info from the stream header and inits the VGMSTREAM
The new meta is usually named after the format's header id or main extension, possibly with prepended platform. Each file should parse one format (regardless of accepted extensions or decoders used) for consistency, but variations can be found as code evolved. Differents formats can use the same extension, this is not a problem as long as the header id or some other validation tells them apart. If the format is headerless and the extension isn't unique enough it may need a generic GENH/TXTH header instead of direct support.
A STREAMFILE is passed to init_vgmstream_(format-name) function, and I/O must be done using its functions and not STDIO/FILEs, as this lets plugins do their own I/O. This includes reading data from the header or opening other STREAMFILEs (if the header has companion files that need to be parsed).
When a parser is successful (allocates VGMSTREAM and sets values) it also needs to open and assign to the VGMSTREAM one or several STREAMFILEs (usually reopens the one passed, but could be any other file) to do I/O during decode. The STREAMFILE passed to the meta will be discarded and must not be reused.
If it supports subsongs it should read and handle the stream index (subsong number) in the passed STREAMFILE, and report the number of subsongs in the VGMSTREAM, to signal the plugins this feature. The index is 1-based (first subsong is 1, not 0).
