# HG changeset patch
# User Ryan C. Gordon <icculus@icculus.org>
# Date 1010958664 0
# Node ID cbd5e308f12d19feeb0404ec945ec7688bfdaf50
# Parent  9bab949e2318381e010305fc16926431b690da59
Updated.

diff -r 9bab949e2318 -r cbd5e308f12d CHANGELOG
--- a/CHANGELOG	Sun Jan 13 21:34:46 2002 +0000
+++ b/CHANGELOG	Sun Jan 13 21:51:04 2002 +0000
@@ -6,6 +6,8 @@
            playsound to try and wait until SDL has completed playing a given
            sound before closing the audio device. Changed a macro in
            decoders/shn.c to be more uniform with the other decoders.
+           SDL_sound error messages are now maintained on a per-thread basis,
+           and do not interfere with SDL_[GS]etError() anymore.
 01112002 - Committed the rest of Torbjörn's MOD patches, to clean up file
            extension handling.
 01092002 - Torbjörn comes through with a ModPlug-based decoder, which should
diff -r 9bab949e2318 -r cbd5e308f12d TODO
--- a/TODO	Sun Jan 13 21:34:46 2002 +0000
+++ b/TODO	Sun Jan 13 21:51:04 2002 +0000
@@ -14,10 +14,10 @@
 - Rename "test" directory to "playsound".
 - Have default make build playsound, and make install install it.
 - Make some of the SNDDBG() strings more uniform.
-- Shrink buffer in Sound_DecodeAll to fit exact size of sample after decoding.
-- Make Sound_SetError() and Sound_GetError() use their own buffers, so they
-  don't piddle over SDL errors, and vice-versa.
 - Run through ElectricFence, dmalloc, etc...
+- Verify thread safety. SDL_sound should be fully reentrant right now, with
+  the exception of Sound_NewSample*()...starting two sounds at once can cause
+  the internal linked list of currently-decoding samples to scramble.
 
 Decoders to implement:
 - .CDA (Does this need specific decoding? Or is a RWops needed?)