[Apologies if you are seeing this twice.  My news/mail gateway appears
to be misbehaving. -wsr]

I've just picked up a sound card (Telesound by Televideo).  The box
essentially claims that the card emulates every sound producing
standard back to Jerico Trumpets.  It claims compatability with "Sound
Blaster Pro", "Ad Lib", "Windows Sound system", "MPU-401", "MPC2", and
"Windows 3.1".  It also mentions Plug-and-Pray.  With all this crap
you'd think the would just work with at least one of the drivers in
netbsd-current (either sb or wss).  Well, er, not quite.

What I realize now is that PNP requires a small amount of glue to be
added to the sound drivers that want to use it.  The glue is there for
the sb driver but not for the wss driver.  

Seeing how the sb driver was pnp capable, I tried that one first.
It didn't recognize the chip.  Ok, use the pnp printf output to 
find the magic token that the card spits out - "OPT9250".

    isapnp0: <OPTi Audio 16, OPT9250, OPTi Audio 16> port 
	0x220/16,0x530/8,0x380/12,0xf8d/16 irq 5 drq 0,0 not configured

I added this to the recognized chips at the top of sb_isapnp.c,
recompiled and rebooted.  Better.  Now it doesn't like the DMA choice
that PNP stuck it with.

    sb0 at isapnp0 port 0x220/16,0x530/8,0x380/12,0xf8d/16 irq 5 drq 0,0
    sb0: configured dma chan 0 invalid
    sb0: sbmatch failed

Sigh.  I don't know if I should hotwire this objection out, or if I
need to hack pnp to assign a different DMA.  Can anyone point me to a
good PNP doc?  Does the card present a list of acceptable choices to
the pnp driver and the driver is supposed to pick from that list?  Is
pnp just choosing the first free resource when it should really
restrict the pool of answers somehow?

Is anyone else working on this sort of thing?  Perhaps we can trade
notes.  I found some nice docs on the OPTI web site.  I did manage to
find a 30-page hardware doc for the main chip (an OPTI 82C925).
Despite the lower number from the "standard" 82C928, 82C929 that
various drivers expect, this appears to be a much newer chip.  The
prelim.  specs all bear a Jan '96 date.  The chip appears to be an
all-singing-all-dancing one-chip solution that does anything you want;
if you can figure out how to program it that is.

BTW. The card is dirt cheap -- approx. $45 at Fry's.  If it works, it
sure beats the $200 offerings from the soundblaster folks.

-wolfgang
-- 
Wolfgang Rupprecht    <wolfgang@wsrcc.com>     http://www.wsrcc.com/
Sick of spam?  Check out all the crap we've gotten. http://www.wsrcc.com/spam/