Wish I had an analog joystick.......

Anyone bother to test this?

I tried this to see if it would fix the calibration issue with Strike Commander. It didn't, still gives failure with gameport joystick(Thrustmaster F-22/TQS)

Rewiring a DB-15 bracket for my motherboard's analog joystick/MIDI pin header is still on my to-do list...

Hi guys,
I tried the dll on my PC for DosBox 0.65 under Win ME. It doesn't improve the gameport issue there. As soon as I define a 2axis Joystick in the .conf and connect the Joystick, CPU load is rising to 100% as soon as I start DosBox - I do not even need to load a game...
CPU is a AMD Sempron 2200+.

I just reviewed that thing and found out it is completely wrong... update soon.

Ok, this time it has some effect 😉 (file is in first post)

It can be fixed from Dosbox side.
The patch reduces SDL_PollEvent to being called 50 times a second which should be more than sufficient - humans are awfully slow devices 😁.
This even frees up a little bit of CPU time.

Cool, thanks.
Any idea why the polling is that high by default?

Ok, i was assuming that you modified something that has
an effect on the polling inside SDL_PollEvent, so my
question is pointless.
Anyways the SDL_ACTIVEEVENT should be handled as soon as
possible in certain cases, maybe the input events can be
separated from the rest in some way.

Hmm which cases would that be? Mouse? It will only be delayed up to 20 milliseconds.

the SDL_ACTIVEEVENT deals (for example) with mode switches caused by the OS (alt-tab)
it must be handled fast else the video graphics will render in an invalid buffer and it will crash.

That is it!
Gameport joystick (Wingman Extreme) works perfectly now on all games, that I tested so far.
I didn't see a crash, but I usually don't Alt-Tab my game away, while I'm smoking a Mig down the sky...

Thanks very much!

> but I usually don't Alt-Tab my game away

Need not be user-forced, can happen especially when some program
forces the focus while dosbox is in fullscreen mode (popup messages,
messenger programs etc.)

If you use the sdl.dll from up there you are safe anyway as it only limits the joystick.

Maybe you can put up the patch for sdl as well, imho it's the better solution to the problem.

That would be in SDL_events.c:

1/* Run the system dependent event loops */
2static int SDL_joy_skip2=0;
3
4void SDL_PumpEvents(void)
5{
6	int now;	
7	if ( !SDL_EventThread ) {
8		SDL_VideoDevice *video = current_video;
9		SDL_VideoDevice *this  = current_video;
10
11		/* Get events from the video subsystem */
12		if ( video ) {
13			video->PumpEvents(this);
14		}
15
16		/* Queue pending key-repeat events */
17		SDL_CheckKeyRepeat();
18
19#ifndef DISABLE_JOYSTICK
20		/* Check for joystick state change */
21		now=SDL_GetTicks();
22		if ( SDL_numjoysticks && (SDL_eventstate & SDL_JOYEVENTMASK) &&
23				((now-SDL_joy_skip2)>50) ) {
24			SDL_JoystickUpdate();
25			SDL_joy_skip2=now;
26		}
27		
28		/*if ( SDL_numjoysticks && (SDL_eventstate & SDL_JOYEVENTMASK) ) {
29			SDL_JoystickUpdate();
30		}*/
31#endif
32	}
33}

(here it is set to 20 Hz)

It's just that you can increase the cycles noticeable (2000 cycles or so with dyncore on my xp2400) with the Dosbox patch.
Edit: even without joystick.

Thanks for the diff!

> It's just that you can increase the cycles noticeable (2000 cycles or so
> with dyncore on my xp2400) with the Dosbox patch.

Yes i know, i don't like it being called that often, too, but it was hard to
avoid the crashing when alt-tabbing away from a fullscreen dosbox
(and it still is not really bulletproof), so it should be handled with care
as far as possible.

Maybe the joystick-polling interval can be reduced (from the
dosbox side) without affecting the rest, something like this:

1void GFX_Events() {
2	int time=SDL_GetTicks();
3	if (time-poll_delay>20) {
4		poll_delay=time;
5		SDL_EventState(SDL_JOYEVENTMASK,SDL_ENABLE);
6	} else SDL_EventState(SDL_JOYEVENTMASK,SDL_IGNORE);

I have no idea why but this code makes the keyboard freak out and doesn't help smoothing CPU usage.