author Sam Lantinga <>
Mon, 02 Jun 2003 14:50:22 +0000
changeset 632 85e104fe14c2
parent 150 df1d68818edb
child 1361 19418e4422cb
permissions -rw-r--r--
Date: Sun, 1 Jun 2003 15:38:45 -0700 (PDT)
From: Jeff Brown <>
Subject: [patch] SDL-1.2.5 + FreeBSD joystick axes, hat fixes

Hello again! When I sent in some SDL fixes last December, I found out
they'd already been fixed in the CVS version. This time, I checked the
repository before bugging you. =)

I'm using SDL-1.2.5 on a FreeBSD 4.6.2-RELEASE system, and in the course
of getting my multi-analog-axis USB controller (with a hat switch!)
working with d2x-sdl -- the SDL port of the Descent 2 engine -- I came
across a few problems:

1) The second analog stick is reported as a slider in one direction, and
"Rz" in the other. SDL was ignoring the Rz axis, so I added Rx/Ry/Rz to
the set of things SDL considers to be axes.

2) After the above change, the set of JOYAXE_* axes for my gamepad was
{0,1,3,7}; however, d2x-sdl expects the axes to be contiguously numbered
from 0, which seems like a pretty reasonable expectation, rather than
having to scan the entire space of axes that SDL may or may not have.
So, I added a table lookup which maps the JOYAXE_* axis numbers to 0,1,...
in the order they're detected by SDL_SYS_JoystickOpen(), when reporting
them to the application. I also added a function "usage_to_joyaxe()"
which maps the USB HUG_* usage values to JOYAXE_values, since the repeated
case statements testing for HUG_* were getting out of hand.

3) The BSD joystick driver had no hat support, so I added it. It looks
like our USB library can only support one hat switch per device, which
makes life easy.

The patch against SDL-1.2.5 which implements these changes is at:

After applying, SDL's "testjoystick" reports all activity from my gamepad
correctly, and d2x works too (though it needed some other fixes).

Moving on...

There is also a problem with slightly different USBHID library interfaces
on different versions of FreeBSD. I wasn't going to mention this since the
FreeBSD port for SDL-1.2.5 (and not SDL itself) was doing the FreeBSD
version-specific patching, so I e-mailed the port maintainer with this
change. However, I see that you've incorporated the FreeBSD
version-checking stuff into the CVS version of SDL, so now it's relevant
for you too.

The problem is, the FreeBSD #if tests don't work right for FreeBSD
4.6.2-RELEASE. There may be other versions with this problem, but I've
only tested 4.6.2-R. The following patch against your latest CVS version
fixes this:

--- SDL_sysjoystick.c-1.16 Tue Apr 15 09:02:08 2003
+++ SDL_sysjoystick.c Sun Jun 1 15:10:28 2003
@@ -420,6 +420,8 @@
# else
len = hid_report_size(rd, repinfo[repind].kind, r->rid);
# endif
+# elif (__FreeBSD_version == 460002)
+ len = hid_report_size(rd, r->rid, repinfo[repind].kind);
# else
len = hid_report_size(rd, repinfo[repind].kind, &r->rid);

I hope this is all useful to you. I've been getting myself dizzy playing
Descent 2 with it, all morning!

-Jeff Brown

P.S. My USB controller is a Thrustmaster Firestorm Dual Analog 2. That's
probably irrelevant, but I threw it in for completeness.
     2 Wish list for the 1.3 development branch:
     4  * Use /etc/fb.modes, if available, like GGI does
     5  * Add mousewheel events (new unified event architecture?)
     6  * DirectInput joystick support needs to be implemented
     7  * Be able to enumerate and select available audio and video drivers
     8  * Fullscreen video mode support for MacOS X
     9  * Explicit vertical retrace wait (maybe separate from SDL_Flip?)
    10  * Shaped windows, windows without borders
    11  * Multiple windows, multiple display support
    12  * SDL_INIT_EVENTTHREAD on Windows and MacOS?
    13  * Add a timestamp to events
    14  * Use RDTSC for timer resolution on x86 hardware
    15  * Add audio input API
    16  * Add hardware accelerated scaled blit
    17  * Add hardware accelerated alpha blits
    18  * Redesign blitting architecture to allow blit plugins
    20 In the jump from 1.2 to 1.3, we should change the SDL_Rect members to
    21 int and evaluate all the rest of the datatypes.  This is the only place
    22 we should do it though, since the 1.2 series should not break binary
    23 compatibility in this way.
    25 Requests:
    26  * PCM and CDROM volume control (deprecated, but possible)