[sdlsound] Re: Valgrind testing.

[Ryan C. Gordon]

> This is the patch I was talking about in my last email. Does it fix the
> remaining TiMidity leaks?

I have applied this patch (and then took my obnoxious printf() out, too).
Thanks!

I get this a lot from Timidity...it's not a leak, and it's probably
harmless, but it does suggest a bug (and if, on some platform, that
invalid read goes over a memory page boundary, it _would_ segfault).

The MIDI in question:
   http://icculus.org/SDL_sound/downloads/example_sounds/piratespenzmodmajgenral.mid

==6295== Thread 2:
==6295== Invalid read of size 2
==6295==    at 0x40255547: rs_loop (resample.c:129)
==6295==    by 0x40256366: resample_voice (resample.c:542)
==6295==    by 0x402500B4: mix_voice (mix.c:535)
==6295==    by 0x40252FB4: do_compute_data (playmidi.c:611)
==6295==    Address 0x437C3064 is 0 bytes after a block of size 59768 alloc'd
==6295==    at 0x4004314B: malloc (vg_clientfuncs.c:100)
==6295==    by 0x4024C95E: safe_malloc (common.c:97)
==6295==    by 0x4024D82F: load_instrument (instrum.c:396)
==6295==    by 0x4024DFF6: fill_bank (instrum.c:555)


If you're bored, that's the only problem I'm seeing with Valgrind and midi
playback now.

--ryan.