DWORD help

[Andrew Fenn]

I have written the following but when I have dwUserIndex as 0,1,2,3 or
any number, doesn't really matter, it always returns
ERROR_BAD_ARGUMENTS.

DWORD XInputGetState(DWORD dwUserIndex, XINPUT_STATE* pState)
{
    FIXME("Stub - Controller: %d \n", dwUserIndex);

    if (dwUserIndex >= 0 && dwUserIndex < 4) {
        return ERROR_DEVICE_NOT_CONNECTED;
        /* If controller is connected return ERROR_SUCCESS */
    }
    return ERROR_BAD_ARGUMENTS;
}

[Juan Lang]

Hi Andrew,
> I have written the following but when I have dwUserIndex as 0,1,2,3 or
> any number, doesn't really matter, it always returns
> ERROR_BAD_ARGUMENTS.

Not sure why it's not returning ERROR_DEVICE_NOT_CONNECTED for you.
I'll just point out:
> DWORD XInputGetState(DWORD dwUserIndex, XINPUT_STATE* pState)
> {
>    FIXME("Stub - Controller: %d \n", dwUserIndex);
>
> if (dwUserIndex >= 0 && dwUserIndex < 4) {
>        return ERROR_DEVICE_NOT_CONNECTED;

The first part of this is always true, since DWORD is unsigned.  You
can write this more simply as:
  if (dwUserIndex < 4) { ... }

--Juan

[Andrew Fenn]

Thanks Juan. It's fixed now.

On Sat, Oct 25, 2008 at 5:13 PM, Juan Lang <juan.lang(a)gmail.com> wrote:
> Hi Andrew,
> > I have written the following but when I have dwUserIndex as 0,1,2,3 or
> > any number, doesn't really matter, it always returns
> > ERROR_BAD_ARGUMENTS.
>
> Not sure why it's not returning ERROR_DEVICE_NOT_CONNECTED for you.
> I'll just point out:
> > DWORD XInputGetState(DWORD dwUserIndex, XINPUT_STATE* pState)
> > {
> >    FIXME("Stub - Controller: %d \n", dwUserIndex);
> >
> > if (dwUserIndex >= 0 && dwUserIndex < 4) {
> >        return ERROR_DEVICE_NOT_CONNECTED;
>
> The first part of this is always true, since DWORD is unsigned.  You
> can write this more simply as:
>  if (dwUserIndex < 4) { ... }
>
> --Juan

[Stefan Dösinger]

> -----Original Message-----
> From: wine-devel-bounces(a)winehq.org [mailto:wine-devel-
> bounces(a)winehq.org] On Behalf Of Andrew Fenn
> Sent: Saturday, October 25, 2008 6:43 PM
> To: Juan Lang
> Cc: wine-devel(a)winehq.org
> Subject: Re: DWORD help
>
> Thanks Juan. It's fixed now.
> Are you sure? If the calling convention is wrong(as suggested in my other
> mail), the result could be random depending on what happens to be on the
> stack or the registers.

[Andrew Fenn]

Yes, it was a combination of a simple mistake and not compiling wine
again before issuing the make test.

I very much doubt it's a failing of wine, anyway I have posted the
patches in my other thread about xinput if you still think there's a
problem with wine.

Regards,
Andrew

On Sat, Oct 25, 2008 at 4:54 PM, Stefan Dösinger <stefan(a)codeweavers.com> wrote:
> > -----Original Message-----
> > From: wine-devel-bounces(a)winehq.org [mailto:wine-devel-
> > bounces(a)winehq.org] On Behalf Of Andrew Fenn
> > Sent: Saturday, October 25, 2008 6:43 PM
> > To: Juan Lang
> > Cc: wine-devel(a)winehq.org
> > Subject: Re: DWORD help
> >
> > Thanks Juan. It's fixed now.
> > Are you sure? If the calling convention is wrong(as suggested in my other
> > mail), the result could be random depending on what happens to be on the
> > stack or the registers.

[Stefan Dösinger]

> > I have written the following but when I have dwUserIndex as 0,1,2,3
> > or
> > any number, doesn't really matter, it always returns
> > ERROR_BAD_ARGUMENTS.
>
> Not sure why it's not returning ERROR_DEVICE_NOT_CONNECTED for you.
> I'll just point out:
> Probably the calling convention is wrong. Usually Windows API functions have
> to be declared like this:

DWORD WINAPI XInputGetState(DWORD dwUserIndex, XINPUT_STATE* pState);

The difference is (among others) that the first parameter is passed in the
ECX register instead of the stack

[Henri Verbeet]

2008/10/25 Stefan Dösinger <stefan(a)codeweavers.com>:
> The difference is (among others) that the first parameter is passed in the
> ECX register instead of the stack

No, that would be thiscall. The main difference between stdcall and
cdecl is who's responsible for popping arguments from the stack
(callee vs. caller). Getting the calling convention wrong would screw
up the stack after the call, but the call itself would still get the
right parameters. Besides, you would notice the wrong value in the
FIXME.