Date: Thu, 17 Jun 2004 11:38:51 -0700 (PDT) From: Eric Wing <ewing2121@yahoo.com> Subject: New OS X patch (was Re: [SDL] Bug with inverted mouse coordinates in I have a new patch for OS X I would like to submit. First, it appears no further action has been taken on my fix from Apple on the OpenGL windowed mode mouse inversion problem. The fix would reunify the code, and no longer require case checking for which version of the OS you are running. This is probably a good fix because the behavior with the old code could change again with future versions of the OS, so those fixes are included in this new patch. But in addition, when I was at Apple, I asked them about the ability to distinguish between the modifier keys on the left and right sides of the keyboard (e.g. Left Shift, Right Shift, Left/Right Alt, L/R Cmd, L/R Ctrl). They told me that starting with Panther, the OS began supporting this feature. This has always been a source of annoyance for me when bringing a program that comes from Windows or Linux to OS X when the keybindings happened to need distinguishable left-side and right-side keys. So the rest of the patch I am submitting contains new code to support this feature on Panther (and presumably later versions of the OS). So after removing the OS version checks for the mouse inversion problem, I reused the OS version checks to activate the Left/Right detection of modifier keys. If you are running Panther (or above), the new code will attempt to distinguish between sides. For the older OS's, the code path reverts to the original code. I've tested with Panther on a G4 Cube, G5 dual processor, and Powerbook Rev C. The Cube and G5 keyboards demonstrated the ability to distinguish between sides. The Powerbook seems to only have left-side keys, but the patch was still able to handle it by producing the same results as before the patch. I also wanted to test a non-Apple keyboard. Unfortunately, I don't have any PC USB keyboards. However, I was able to borrow a Sun Microsystems USB keyboard, so I tried that out on the G5, and I got the correct behavior for left and right sides. I'm expecting that if it worked with a Sun keyboard, most other keyboards should work with no problems.

/* Test the thread and mutex locking functions 
   Also exercises the system's signal/thread interaction
*/

#include <signal.h>
#include <stdio.h>
#include <stdlib.h>

#include "SDL.h"
#include "SDL_mutex.h"
#include "SDL_thread.h"

static SDL_mutex *mutex = NULL;
static Uint32 mainthread;
static SDL_Thread *threads[6];

void printid(void)
{
	printf("Process %u:  exiting\n", SDL_ThreadID());
}
	
void terminate(int sig)
{
	printf("Process %u:  raising SIGTERM\n", SDL_ThreadID());
	raise(SIGTERM);
}
void closemutex(int sig)
{
	Uint32 id = SDL_ThreadID();
	int i;
	printf("Process %u:  Cleaning up...\n", id == mainthread ? 0 : id);
	for ( i=0; i<6; ++i )
		SDL_KillThread(threads[i]);
	SDL_DestroyMutex(mutex);
	exit(sig);
}
int Run(void *data)
{
	if ( SDL_ThreadID() == mainthread )
		signal(SIGTERM, closemutex);
	while ( 1 ) {
		printf("Process %u ready to work\n", SDL_ThreadID());
		if ( SDL_mutexP(mutex) < 0 ) {
			fprintf(stderr, "Couldn't lock mutex: %s", SDL_GetError());
			exit(1);
		}
		printf("Process %u, working!\n", SDL_ThreadID());
		SDL_Delay(1*1000);
		printf("Process %u, done!\n", SDL_ThreadID());
		if ( SDL_mutexV(mutex) < 0 ) {
			fprintf(stderr, "Couldn't unlock mutex: %s", SDL_GetError());
			exit(1);
		}
		/* If this sleep isn't done, then threads may starve */
		SDL_Delay(10);
	}
	return(0);
}

int main(int argc, char *argv[])
{
	int i;
	int maxproc = 6;

	/* Load the SDL library */
	if ( SDL_Init(0) < 0 ) {
		fprintf(stderr, "%s\n", SDL_GetError());
		exit(1);
	}
	atexit(SDL_Quit);

	if ( (mutex=SDL_CreateMutex()) == NULL ) {
		fprintf(stderr, "Couldn't create mutex: %s\n", SDL_GetError());
		exit(1);
	}

	mainthread = SDL_ThreadID();
	printf("Main thread: %u\n", mainthread);
	atexit(printid);
	for ( i=0; i<maxproc; ++i ) {
		if ( (threads[i]=SDL_CreateThread(Run, NULL)) == NULL )
			fprintf(stderr, "Couldn't create thread!\n");
	}
	signal(SIGINT, terminate);
	Run(NULL);

	return(0);	/* Never reached */
}