Date: Mon, 11 Oct 2004 15:17:27 +0300 (EEST)

From: Hannu Savolainen
Subject: Re: SDL uses obsolete OSS features

I did some work on getting OSS to work better with SDL. There have been
some problems with select which should be fixed now.

I'm having some problems in understanding what is the purpose of the
DSP_WaitAudio() routine. I added a return to the very beginning of this
routine and commendted out the define for USE_BLOCKING_WRITES. At least
lbreakout2 seems to work as well as earlier. The latencies are the same.

An ordinary blocking write does exactly the same thing than DSP_WaitAudio
does. So I would recommend using the USE_BLOCKING_WRITES approach and
removing everything from the DSP_WaitAudio routine. Also enabling
USE_BLOCKING_WRITES makes it possible to simplify DSP_PlayAudio() because
you don't need to handle the partial writes (the do-while loop).

Attached is a patch against SDL-1.2.7. After these changes SDL will use
OSS as it's designed to be used (make it as simple as possible). This code
should work with all OSS implementations because it uses only the very
fundamental features that have been there since the jurassic times.

--HG--
extra : convert_revision : svn%3Ac70aab31-4412-0410-b14c-859654838e24/trunk%40969

src/audio/dsp/SDL_dspaudio.h

@@ -43,10 +43,6 @@ struct SDL_PrivateAudioData {
 	/* Raw mixing buffer */
 	Uint8 *mixbuf;
 	int    mixlen;
-
-	/* Support for audio timing using a timer, in addition to select() */
-	float frame_ticks;
-	float next_frame;
 };
 #define FUDGE_TICKS	10	/* The scheduler overhead ticks per frame */