[Openal] Music replay.

[steve]

I'm playing around with a MIDIfile player under Linux using
OpenAL because I ultimately want music that I can attach
to a 3D object - and I'm taking cues from the music to drive
animation.  I have a MIDIfile loader and sequencer and all
that stuff and I have a loader for the sound-fonts that
all of the standard Linux MIDIfile players (timidity,
wildmidi, etc) use.  I plan to release the code as
an OpenSource library once I have it working.

So - the plan was to load each of my sound-font waveforms
into its own alBuffer - and use alSources to play them.

The tricky part turns out to be playing notes at the
correct frequency.

Each instrument sound in the sound font has a bunch of
waveforms - each one appropriate to notes played at a
specified range of frequencies - so the first step is
to pick the sample that most closely matches the
frequency of the note I'm trying to play - then I
kick off an alSource with that alBuffer - adjusting
the AL_PITCH to get the precise frequency I need:

alSourcef(handle,AL_PITCH,note_freq/sample_freq);

...but here's the snag.  OpenAL is telling me that
any AL_PITCH greater than 2.0 is an error!  That
means that I can't boost the frequency by more than
one octave...that *SUCKS*!

This makes no sense - but I've peered into the Linux
sources and yes, indeed it clamps AL_PITCH to 2.0 and
flags an error.

A comment in the code says this is contrary to the
OpenAL specification...but that's not true - the
specification says that implementations are free to
limit the range of AL_PITCH.

So while the Linux driver is horribly limited, it
doesn't seem to be illegal.

My questions:

1) Is there a way to query OpenAL's AL_PITCH limits
    so my code can at least avoid getting an error?
    What's to say that if I get it working with pitch
    limits of 2.0, I won't one day find a player with
    a limit of 1.5?  If the API allows there to be a
    limit that's allowed to be different between
    implementations, there MUST be an alGet for it.

2) Is there a better way to write my music player?

3) If I do push my software to use AL_PITCH up to
    2.0 - but not beyond - will I run into problems
    when doppler pushes the overall frequency higher
    than that?

4) Is there a was I could fix the driver to increase
    that limit?   Does anyone understand WHY there is
    a limit - does anyone know where I should be looking
    in the sources to fix it.  If I simply remove the
    clamp (and another one in al_filter.c), the sound
    plays OK up to an AL_PITCH of maybe 3.0 - but higher
    than that I get some nasty replay artifacts that
    suggest there is an aliasing problem.

5) In desperation, I tried lying to OpenAL about the
    sampling rate of the alBuffer (eg doubling the
    true number) so that I could reduce the AL_PITCH
    settings proportionately.  This works to some
    extent - but there seems to be some kind of limit
    on the maximum value for the alBuffer's sampling
    rate - and beyond that some very strange symptoms
    start cropping up.  Is there some kind of hidden
    upper limit to the alBuffer sample frequency?

This is very frustrating!