Re: [fluid-dev] FluidSynth backend for SDL_mixer

[James Le Cuirot]

Good news, everyone!

Following Matt Giuca's great work on playing files from memory, I've
more or less finished my SDL_mixer patch. The attached patch requires
the latest FluidSynth from Subversion and was created against the latest
SDL_mixer from Mercurial, though it can be applied against 1.2.11. You
can safely ignore the conflict in Makefile.in but remember to run
./autogen.sh in either case. As before, set SDL_SOUNDFONTS to some
path in order to try this out. Note that I've only tried it on Linux
so far.

There's not much to say on the FluidSynth side of things. Matt's
additions made it incredibly simple. :) My only thought was that it
might be nice to allow SoundFonts to be loaded from memory but that can
always be added later.

Having seen how constrained some environments like the PSP can be, I
decided to use the regular old fluid_player_add for files on disk,
otherwise they would have to be copied in memory at least twice after
the initial read. This was more feasible than it was before because now
I'm using the player interface for in-memory files as well. It also
gave me an opportunity to use function pointers for the first time.
That was fun. ;)

It took me some time to decide on the best way to handle SoundFont
selection and when to fall back to Timidity and there are several other
SDL_mixer aspects to discuss but I'll save all that for the SDL mailing
list. I'll post there tomorrow to get some feedback.

Cheers,
James

sdl-mixer-fluidsynth.patch
Description: Text Data