SDL_mixer is an audio library that supports:
SDL_sound(formally SDL_voice) was developed for the purpose of dividing 1. decode/decompress audio data, 2. mixing them. SDL_sound supports the former one.
The decoded/decompressed data by SDL_sound SHOULD be passed to SDL_mixer, but there is no support of SDL_mixer that handling decoded/decompressed data from SDL_sound (yet).
I have an experimental code(patch) that enables SDL_sound as a part of SDL_mixer's driver(decoder).
Nothing special. You can use normal SDL_mixer functions like Mix_LoadMus(). The only one this patch do is bypassing decoding process to SDL_sound.
Nothing. Simply sending decoded PCM data to internal mixing loop of SDL_mixer.