dinput, wheel/joystick and richard burns rally

[Tuomo Kohvakka]

Hello,

first, appdb entry:

http://appdb.winehq.org/appview.php?iVersionId=5934&iTestingId=6382

As shown there, joystick (that ofcourse really means the wheel :) isn't 
detected by the game with current wine. I decided to take a look at 
that, figuring that it might be fixable.

Well, I managed to "fix" it. Code follows, dlls/dinput/device.c,
create_DataFormat() around line 387 says that

&&
(/* Then check if it accepts any instance id, and if not, if it matches 
Wine's
   * instance id.
   */
((asked_format->rgodf[j].dwType & DIDFT_INSTANCEMASK) == 
DIDFT_ANYINSTANCE) ||

(DIDFT_GETINSTANCE(asked_format->rgodf[j].dwType) == 0x00FF) || /* 
This is mentionned in no DX docs, but it works fine - tested on WinXP */

(DIDFT_GETINSTANCE(asked_format->rgodf[j].dwType) == 
DIDFT_GETINSTANCE(format->wine_df->rgodf[i].dwType)))
         &&

Now, adding

|| ((asked_format->rgodf[j].dwType&0xff) == (format->wine_df->rgodf[i]. 
   =>dwType&0xff))

into middle of that fixes it. It works fine and doesn't seem to break 
anything (didn't try other applications thought).

It took me a couple of hours to figure above ou, a few more comments 
more might have helped! :) (code itself is rather easy to follow)

Now, my problem is that I don't really know anything about directx or 
much about windows anyhow. That means that I don't know if above "fix" 
is really proper way to fix anything. I figure that purpose of this 
function make a mapping from devices axes and buttons (wine 
datastructure) into applications preferred structure (which is used 
later on in fill_DataFormat). Mapping goes wrong without above, as 
application requests "3" in dwType and wine has something like 
0x80ffff03. It's defined in c_dfDIJoystick2, data_formats.c as 
DIDFT_OPTIONAL|DIDFT_AXIS|DIDFT_ANYINSTANCE.

If mapping goes wrong offset_in becomes -1 and that means that default 
values will be used (button not pressed etc). (If I would submit a patch 
containing only comments, some probably wrong guesses, would it be 
accepted?)

Someone here on list probably has more insight on this. What is that 
dwType anyways? :) Is above fix correct, or should dwType be something 
else from the beginning with, or am I just completely lost?

Bottom line ofcourse is that I would like to see a proper fix in wine 
.29 that would allow rbr and wheel co-operate :) Should I just try to 
submit this as a patch?

[Christoph Frick]

On Sun, Jan 07, 2007 at 10:22:34PM +0200, Tuomo Kohvakka wrote:
> http://appdb.winehq.org/appview.php?iVersionId=5934&iTestingId=6382
>
> As shown there, joystick (that ofcourse really means the wheel :)
> isn't detected by the game with current wine. I decided to take a look
> at that, figuring that it might be fixable.

the problem here is somehow deeper within wine. we store a dataformat as
default with the device, that does not represent the informations sent
to the app. so when an app want to build its own data-format using the
infos there.

so the problem was actually fixed some time ago but the fix got lost
with a huge clean up session. so there is actually currently work
undergoing, that fixes the problem (see the mails about RBR before the
holidays).
> &&
> (/* Then check if it accepts any instance id, and if not, if it matches 
> Wine's
> - instance id.
>
> */
> ((asked_format->rgodf[j].dwType & DIDFT_INSTANCEMASK) == 
> DIDFT_ANYINSTANCE) ||
>
> (DIDFT_GETINSTANCE(asked_format->rgodf[j].dwType) == 0x00FF) || /* 
> This is mentionned in no DX docs, but it works fine - tested on WinXP */

imho this already does not make sense..
> (DIDFT_GETINSTANCE(asked_format->rgodf[j].dwType) == 
> DIDFT_GETINSTANCE(format->wine_df->rgodf[i].dwType)))
>         &&
>
> Now, adding
>
> || ((asked_format->rgodf[j].dwType&0xff) == (format->wine_df->rgodf[i]. 
>   =>dwType&0xff))

... but then adding the next "magic" check is wrong. but this is a
gut-feeling and maybe there are already new problems introduced with the
recent patches. for my part i can at least say, that everythings is
working in WINE's head trunk - except RBR with the same problem as
before xmas.

for what i have seen Vitaliy has build the ground to finally make the
same patches for the joysticks as for keyboard and mouse?

[Vitaliy Margolen]

Christoph Frick wrote:
> On Sun, Jan 07, 2007 at 10:22:34PM +0200, Tuomo Kohvakka wrote:
> > http://appdb.winehq.org/appview.php?iVersionId=5934&iTestingId=6382
> >
> > As shown there, joystick (that ofcourse really means the wheel :)
> > isn't detected by the game with current wine. I decided to take a look
> > at that, figuring that it might be fixable.
>
> the problem here is somehow deeper within wine. we store a dataformat as
> default with the device, that does not represent the informations sent
> to the app. so when an app want to build its own data-format using the
> infos there.
>
> so the problem was actually fixed some time ago but the fix got lost
> with a huge clean up session. so there is actually currently work
> undergoing, that fixes the problem (see the mails about RBR before the
> holidays).

That is correct. I already fixed mouse and keyboard the proper way. Now
left to fix joysticks. It seems joysticks are in the worse shape that I
thought. Number of places using wrong object indexes.
> > Now, adding
> >
> > || ((asked_format->rgodf[j].dwType&0xff) == (format->wine_df->rgodf[i]. 
> >   =>dwType&0xff))

Yeah this is wrong. You left only object type and got rid of everything
else, including instance.

Vitaliy.