Re: [dinput][joystick_linuxinput] make axes work, that are "not in a row"

[Vitaliy Margolen]

Christoph Frick wrote:
> with Vitaliy's patches i tried my games yesterday and everything seemed
> ok (RBR works again). but my clutch on the wheel did not work and also

That's a good news. I was hopping not to brake too many things while
reorganizing the dinput.
> @@ -128,11 +122,20 @@ struct JoystickImpl
>    /* The 'parent' DInput */
>    IDirectInputImpl               *dinput;
> - /* autodetecting ranges per axe by following movement */
> - LONG				havemax[8];
> - LONG				havemin[8];

I could never understood why we need these? In other joystick_linux the
range assumed to be 0 - 0xffff.
> /* joystick private */
>    /* what range and deadzone the game wants */
> - LONG				wantmin[ABS_MAX];
> - LONG				wantmax[ABS_MAX];
> - LONG				deadz[ABS_MAX];
> - LONG				wantmin[8];
> - LONG				wantmax[8];
> - LONG				deadz[8];

Can we move these into separate struct like it's done in joystick_linux?
It's much cleaner end eventually will allow a better integration of two.

And can you use spaces instead of tabs please?
> @@ -395,19 +373,29 @@ static JoystickImpl *alloc_device(REFGUID rguid, const void *jvt, IDirectInputIm
>      /* Supported Axis & POVs should map 1-to-1 */
>      for (i = 0; i < 8; i++)
>      {
> -
>    if (!test_bit(newDevice->joydev->absbits, i)) continue;
> -
>    if (!test_bit(newDevice->joydev->absbits, i)) {
> -
>      newDevice->axes[i] = -1;
> -
>      continue;
> -
>    }

Can you please don't brake the good readable style? Please put that
curly bracket into separate line, and use 4 space indentation.
> - fake_current_js_state(newDevice);
> -

Can you explain what this suppose to do? Why do we need to "fake current
state" everywhere?
> -
>        int axe = This->axes[ie.code];

It's an "axis" not an "axe" :)

Vitaliy.

[dinput][joystick_linuxinput] make axes work, that are "not in a row"

[Christoph Frick]

On Thu, Jan 11, 2007 at 10:25:51AM -0700, Vitaliy Margolen wrote:
> > @@ -128,11 +122,20 @@ struct JoystickImpl
> >    /* The 'parent' DInput */
> >    IDirectInputImpl               *dinput;
> > - /* autodetecting ranges per axe by following movement */
> > - LONG				havemax[8];
> > - LONG				havemin[8];
>
> I could never understood why we need these? In other joystick_linux the
> range assumed to be 0 - 0xffff.

linux/input.h reports us the proper ranges from the device. these
numbers will also be used as teh want(min|max). so what this numbers are
good for. apps that want the input as "pure" as it gets dont
SetProperty some something and just get what the hardware tells them
(e.g. RBR does so).
> > /* joystick private */
> >    /* what range and deadzone the game wants */
> > - LONG				wantmin[ABS_MAX];
> > - LONG				wantmax[ABS_MAX];
> > - LONG				deadz[ABS_MAX];
> > - LONG				wantmin[8];
> > - LONG				wantmax[8];
> > - LONG				deadz[8];
>
> Can we move these into separate struct like it's done in
> joystick_linux?  It's much cleaner end eventually will allow a better
> integration of two.

i will do this.
> And can you use spaces instead of tabs please?

of course
> > @@ -395,19 +373,29 @@ static JoystickImpl *alloc_device(REFGUID rguid, const void *jvt, IDirectInputIm
> >      /* Supported Axis & POVs should map 1-to-1 */
> >      for (i = 0; i < 8; i++)
> >      {
> > -
> >    if (!test_bit(newDevice->joydev->absbits, i)) continue;
> > -
> >    if (!test_bit(newDevice->joydev->absbits, i)) {
> > -
> >      newDevice->axes[i] = -1;
> > -
> >      continue;
> > -
> >    }
>
> Can you please don't brake the good readable style? Please put that
> curly bracket into separate line, and use 4 space indentation.

sorry - to a K&R guy that uses ts=8 as default this all is unreadable
gibberish; i will try my best here.
> > - fake_current_js_state(newDevice);
>
> Can you explain what this suppose to do? Why do we need to "fake
> current state" everywhere?

after e.g. SetProperties and initializing the device we have to put the
values of the DIJOYSTATE2 in a sane state. e.g. GPL set the ranges to
1000-25k; where my pedals are in 0..512; so the app gets in the next
step after the setting a default value of 512 (clutch not pressed) and
gets nuts.
> > -
> >        int axe = This->axes[ie.code];
>
> It's an "axis" not an "axe" :)

it was 0300 and i guess i wished for an axe there ;)

[dinput][joystick_linuxinput] make axes work, that are "not in a row"

[Vitaliy Margolen]

Christoph Frick wrote:
> On Thu, Jan 11, 2007 at 10:25:51AM -0700, Vitaliy Margolen wrote:
> > > @@ -128,11 +122,20 @@ struct JoystickImpl
> > >    /* The 'parent' DInput */
> > >    IDirectInputImpl               *dinput;
> > > - /* autodetecting ranges per axe by following movement */
> > > - LONG				havemax[8];
> > > - LONG				havemin[8];
> >
> > I could never understood why we need these? In other joystick_linux the
> > range assumed to be 0 - 0xffff.
>
> linux/input.h reports us the proper ranges from the device. these
> numbers will also be used as teh want(min|max). so what this numbers are
> good for. apps that want the input as "pure" as it gets dont
> SetProperty some something and just get what the hardware tells them
> (e.g. RBR does so).

What you described is SetProperty(DIPROP_RANGE). What I'm talking about
is the range we get from device itself. In the map_axis we have:

if (val > hmax) This->joydev->havemax[axis] = hmax = val;
    if (val < hmin) This->joydev->havemin[axis] = hmin = val;

We either need to query the device for what it's ranges are, or use
default range (as it's done in joystick_linux). Auto-detecting range
will work only if you move joystick on all axes from min to max. And
even then it still doesn't work here.
I'm using joystick sample from DXSDK and it always shows 0-128-256 range
for accelerator - brake pedals (as a combined Z axis). joystick_linux
shows it correctly - 0..1000 (it sets this range for all axes).
> > > - fake_current_js_state(newDevice);
> >
> > Can you explain what this suppose to do? Why do we need to "fake
> > current state" everywhere?
>
> after e.g. SetProperties and initializing the device we have to put the
> values of the DIJOYSTATE2 in a sane state. e.g. GPL set the ranges to
> 1000-25k; where my pedals are in 0..512; so the app gets in the next
> step after the setting a default value of 512 (clutch not pressed) and
> gets nuts.

Right. But can we query device for the current axes position? They are
analog devices, and should provide absolute position.

Vitaliy.

[dinput][joystick_linuxinput] make axes work, that are "not in a row"

[Christoph Frick]

On Thu, Jan 11, 2007 at 01:10:21PM -0700, Vitaliy Margolen wrote:
> What you described is SetProperty(DIPROP_RANGE). What I'm talking about
> is the range we get from device itself. In the map_axis we have:
> if (val > hmax) This->joydev->havemax[axis] = hmax = val;
> if (val < hmin) This->joydev->havemin[axis] = hmin = val;

to my knowledge this code is useless. it got copied over from the
/dev/js branch.
> We either need to query the device for what it's ranges are, or use
> default range (as it's done in joystick_linux).

we allready query it. see the code in alloc_device where the
have(min|max) is set.
> Right. But can we query device for the current axes position? They are
> analog devices, and should provide absolute position.

well instead of using ...[AXE_ABS] (another axe) we could use the
current value from the device. what it is worth for is this:

i bet nearly all apps just call alloc_device and SetProperty in the
beginning or when some config is involved. so AXE_ABS is for sure
between MIN/MAX (kernel should assert that) and in nearly all cases
represent the "null"-state of the device. so if someone fiddles with
this input device between find_joydevs and alloc/setprop the value in
DJOYSTATE2 would be wrong.

IMHO its not worth the hastle there.

[dinput][joystick_linuxinput] make axes work, that are "not in a row"

[Christoph Frick]

On Thu, Jan 11, 2007 at 10:25:51AM -0700, Vitaliy Margolen wrote:

attached a patch with all the suggestions for Vitaliy.
> Christoph Frick wrote:
> > with Vitaliy's patches i tried my games yesterday and everything seemed
> > ok (RBR works again). but my clutch on the wheel did not work and also
>
> That's a good news. I was hopping not to brake too many things while
> reorganizing the dinput.
> > @@ -128,11 +122,20 @@ struct JoystickImpl
> >    /* The 'parent' DInput */
> >    IDirectInputImpl               *dinput;
> > - /* autodetecting ranges per axe by following movement */
> > - LONG				havemax[8];
> > - LONG				havemin[8];
>
> I could never understood why we need these? In other joystick_linux the
> range assumed to be 0 - 0xffff.
> > /* joystick private */
> >    /* what range and deadzone the game wants */
> > - LONG				wantmin[ABS_MAX];
> > - LONG				wantmax[ABS_MAX];
> > - LONG				deadz[ABS_MAX];
> > - LONG				wantmin[8];
> > - LONG				wantmax[8];
> > - LONG				deadz[8];
>
> Can we move these into separate struct like it's done in joystick_linux?
> It's much cleaner end eventually will allow a better integration of two.
>
> And can you use spaces instead of tabs please?
> > @@ -395,19 +373,29 @@ static JoystickImpl *alloc_device(REFGUID rguid, const void *jvt, IDirectInputIm
> >      /* Supported Axis & POVs should map 1-to-1 */
> >      for (i = 0; i < 8; i++)
> >      {
> > -
> >    if (!test_bit(newDevice->joydev->absbits, i)) continue;
> > -
> >    if (!test_bit(newDevice->joydev->absbits, i)) {
> > -
> >      newDevice->axes[i] = -1;
> > -
> >      continue;
> > -
> >    }
>
> Can you please don't brake the good readable style? Please put that
> curly bracket into separate line, and use 4 space indentation.
> > - fake_current_js_state(newDevice);
> > -
>
> Can you explain what this suppose to do? Why do we need to "fake current
> state" everywhere?
> > -
> >        int axe = This->axes[ie.code];
>
> It's an "axis" not an "axe" :)