[Bug 40658] New: Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

Bug ID: 40658
           Summary: Joystick POW hat switch rotated for JOYDEVDRIVER=(js)
                    while JOYDEVDRIVER=(event) is Ok
           Product: Wine
           Version: 1.9.9
          Hardware: x86-64
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: directx-dinput
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ---

I have a Saitek AV8R-01 joystick with a POV hat switch.

When using the JOYDEVDRIVER=(event) device (dinput/joystick_linuxinput.c), the
orientation is Ok both in the Wine Control Panel (control.exe) Joystick test
POV area and in the games, i.e when moving the hat switch horizontally to the
right the indicator moves towards the right, when moving the hat switch
vertically up, the indicator moves up.

Hoever the JOYDEVDRIVER=(js) device (dinput/joystick_linux.c) has its
orientation mixed up:

* when moving the hat switch horizontally to the right the indicator moves
down.

* when moving the hat switch vertically up, the indicator moves left.

This happens on wine-1.9.9  on Ubuntu 14.04 (LTS) installed from
http://ppa.launchpad.net/wine/wine-builds/ubuntu trusty main, but is also
present in earlier versions, e.g. 1.6.1

Unfortunately I need to use the (js) device as I cannot get the game to
recognise the rudder and thrust axes with the (event) device.
(BTW, does the (event) device recognize the
[HKEY_CURRENT_USER\Software\Wine\DirectInput] registry key in the same way as
teh (js) device?)

BR,

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

[e-mail] changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #1 from [e-mail] ---
Created attachment 54550
  --> https://bugs.winehq.org/attachment.cgi?id=54550
bug 40658 fix

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #2 from [e-mail] ---
I have compiled wine-1.9.10-206-g698d411 and exchanged the POV x and y axis in
dlls/dinput/joystick_linux.c around line 734 - see attached patch. This
corrects the problem.

BR,

O

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

Bruno Jesus  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |hardware

--- Comment #3 from Bruno Jesus  ---
Is the problem also present in wine control panel? Because you say it works
with event driver but you don't say you can reproduce in the control panel with
the js driver. I ask because I could help testing with a Xbox 360 controller or
a PS2 controller so using the control panel would be trivial to reproduce.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #4 from [e-mail] ---
(In reply to Bruno Jesus from comment #3)
> Is the problem also present in wine control panel? Because you say it works
> with event driver but you don't say you can reproduce in the control panel
> with the js driver. I ask because I could help testing with a Xbox 360
> controller or a PS2 controller so using the control panel would be trivial
> to reproduce.

I see the same behaviour in the wine control panel/joystick test as in the
game.

There is a discrepancy in the POV rotation between the (js) and the (event)
driver, with the (event) driver giving the correct result.

The (js) driver can be fixed by the patch provided previously, exchanging the x
and y orientation.

Br,

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #5 from Bruno Jesus  ---
If I apply your patch from comment 1 it breaks my Xbox 360 gamepad, so there is
something deeper to do here. There should be something else to check to
determine that in your controller the POV is rotated.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #6 from Bruno Jesus  ---
Please run the following commnand in an unpatched wine with both (js) and
(event) drivers enabled for your joystick.

WINEDEBUG=+dinput wine control > /tmp/log.txt

Open the joystick config and press the POV buttons to generate some log events.

And then attach the file /tmp/log.txt. It could have some clues about what is
wrong;

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #7 from [e-mail] ---
Created attachment 62083
  --> https://bugs.winehq.org/attachment.cgi?id=62083
logs of patched and unpatched 3.12 when operating Saitek AV8R POV switch

Content:
Archive:  logs_40658_wine3.13_patched.zip
  Length      Date    Time    Name
---------  ---------- -----   ----
        0  2018-08-16 19:04   tmp/
  3589906  2018-08-16 19:06   tmp/js-3.13_log.txt
  1870358  2018-08-16 18:36   tmp/js-patched_log.txt
   977689  2018-08-16 19:03   tmp/event-3.13_log.txt
   892834  2018-08-16 18:38   tmp/event-patched_log.txt
---------                     -------
  7330787                     5 files

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #8 from [e-mail] ---
(In reply to Bruno Jesus from comment #6)
> Please run the following commnand in an unpatched wine with both (js) and
> (event) drivers enabled for your joystick.
>
> WINEDEBUG=+dinput wine control > /tmp/log.txt
>
> Open the joystick config and press the POV buttons to generate some log
> events.
>
> And then attach the file /tmp/log.txt. It could have some clues about what
> is wrong;

After a long time, I finally found a couple of hours during my vacation to
build a Ubunti Biarch version 3.13 and generate the requested logs. 
(problem and patch are identical to 1.19)

Registry settings are the following:
[HKEY_CURRENT_USER\Software\Wine\DirectInput] 
"Saitek AV8R Joystick (js)"="X,Y,Z,Rz,Rx,POV1" 
"Saitek AV8R Joystick (event)"="X,Y,Z,Rz,Rx,POV1"

[HKEY_CURRENT_USER\Software\Wine\DirectInput\Joysticks] 
"Saitek AV8R Joystick (event)"="disabled"

For the logs,  I moved the AV8R POV switch in the following pattern:

left - forward - right -backward

Except for the unpatched (js) variant, the status indicator in the control.exe
moved as expected.

For the unpatched (js) variant, the following status was observed in
control.exe:
forward - left - right - bottom

BR,

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

https://bugs.winehq.org/show_bug.cgi?id=40658

[e-mail] changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
            Version|1.9.9                       |3.13

--- Comment #9 from [e-mail] ---
Additional Question: what is the GUID of the AV8R joystick device - in the
BoBII game it seems the joystick axes can be assigned based on the joystick
GUID?

In the log it seems that mouse, keyboard, etc. get the same GUID
9e573ed9-7734-0000-8d4a-23903fb6bdf7, while each joystick axis gets a separate
GUID:

0009:trace:dinput:find_joystick_devices Found a joystick on /dev/input/js0:
Saitek AV8R Joystick (js) 
 with 7 axes and 14 buttons 
0009:trace:dinput:fill_joystick_dideviceinstanceW 1100 0x33e7b0 
0009:trace:dinput:joydev_enum_deviceW Enumerating the linux Joystick device:
/dev/input/js0 (Saitek AV8R Joystick (js)) 
0009:trace:dinput:IDirectInputWImpl_EnumDevices   - checking device 3 ('Wine
Linux joystick driver') 
0009:trace:dinput:IDirectInputWImpl_EnumDevices (this=0x16c5d8,0x0004
'DI8DEVCLASS_GAMECTRL',0x7d83a310,0x7d85bb40,0x0001) 
0009:trace:dinput:_dump_EnumDevices_dwFlags  flags: DIEDFL_ATTACHEDONLY  
0009:trace:dinput:IDirectInputWImpl_EnumDevices   - checking device 0 ('Wine
mouse driver') 
0009:trace:dinput:IDirectInputWImpl_EnumDevices   - checking device 0 ('Wine
mouse driver') 
0009:trace:dinput:IDirectInputWImpl_EnumDevices   - checking device 1 ('Wine
keyboard driver') 
0009:trace:dinput:IDirectInputWImpl_EnumDevices   - checking device 1 ('Wine
keyboard driver') 
0009:trace:dinput:IDirectInputWImpl_EnumDevices   - checking device 2 ('Wine
Linux-input joystick driver') 
0009:trace:dinput:IDirectInputWImpl_EnumDevices   - checking device 3 ('Wine
Linux joystick driver') 
0009:trace:dinput:fill_joystick_dideviceinstanceW 1100 0x33e7b0 
0009:trace:dinput:joydev_enum_deviceW Enumerating the linux Joystick device:
/dev/input/js0 (Saitek AV8R Joystick (js)) 
0009:trace:dinput:IDirectInput7WImpl_CreateDeviceEx
(0x16c5d8)->({9e573ed9-7734-0000-8d4a-23903fb6bdf7}, (null), 0x16c9c8, (nil)) 
0009:trace:dinput:mousedev_create_device 0x16c5d8
{9e573ed9-7734-0000-8d4a-23903fb6bdf7} (null) 0x16c9c8 1 
0009:trace:dinput:keyboarddev_create_device 0x16c5d8
{9e573ed9-7734-0000-8d4a-23903fb6bdf7} (null) 0x16c9c8 1 
0009:trace:dinput:joydev_create_device 0x16c5d8
{9e573ed9-7734-0000-8d4a-23903fb6bdf7} (null) 0x16c9c8 1 
0009:trace:dinput:joydev_create_device 0x16c5d8
{9e573ed9-7734-0000-8d4a-23903fb6bdf7} (null) 0x16c9c8 1 
0009:trace:dinput:alloc_device {9e573ed9-7734-0000-8d4a-23903fb6bdf7} 0x16c5d8
0x33e58c 0 
0009:trace:dinput:setup_dinput_options "Saitek AV8R Joystick (js)" =
"X,Y,Z,Rz,Rx,POV1" 
0009:trace:dinput:IDirectInputAImpl_AddRef (0x16c5d8) incrementing from 1 
0009:trace:dinput:fill_joystick_dideviceinstanceW 1100 0x33e0e0 
0009:trace:dinput:_dump_DIDATAFORMAT Dumping DIDATAFORMAT structure: 
0009:trace:dinput:_dump_DIDATAFORMAT   - dwSize: 24 
0009:trace:dinput:_dump_DIDATAFORMAT   - dwObjsize: 16 
0009:trace:dinput:_dump_DIDATAFORMAT   - dwFlags: 0x00000001 (DIDF_ABSAXIS) 
0009:trace:dinput:_dump_DIDATAFORMAT   - dwDataSize: 272 
0009:trace:dinput:_dump_DIDATAFORMAT   - dwNumObjs: 20 
0009:trace:dinput:_dump_DIDATAFORMAT   - Object 0: 
0009:trace:dinput:_dump_DIDATAFORMAT       * GUID:
{a36d02e0-c9f3-11cf-bfc7-444553540000} ('GUID_XAxis') 
0009:trace:dinput:_dump_DIDATAFORMAT       * dwOfs: 0 
0009:trace:dinput:_dump_DIDATAFORMAT       * dwType: 0x00000002 
0009:trace:dinput:_dump_DIDATAFORMAT         Type: DIDFT_ABSAXIS / Instance:  
0 
0009:trace:dinput:_dump_DIDATAFORMAT       * dwFlags: 0x00000000 
0009:trace:dinput:_dump_DIDATAFORMAT          
0009:trace:dinput:_dump_DIDATAFORMAT   - Object 1: 
0009:trace:dinput:_dump_DIDATAFORMAT       * GUID:
{a36d02e1-c9f3-11cf-bfc7-444553540000} ('GUID_YAxis') 
0009:trace:dinput:_dump_DIDATAFORMAT       * dwOfs: 4 
0009:trace:dinput:_dump_DIDATAFORMAT       * dwType: 0x00000102 
0009:trace:dinput:_dump_DIDATAFORMAT         Type: DIDFT_ABSAXIS / Instance:  
1 
0009:trace:dinput:_dump_DIDATAFORMAT       * dwFlags: 0x00000000 
0009:trace:dinput:_dump_DIDATAFORMAT

In dlls/dinput/joystick_linux.c it looks as if the GUID is fixed for all
joystick types:

static const GUID DInput_Wine_Joystick_GUID = { /*
9e573ed9-7734-11d2-8d4a-23903fb6bdf7 */
  0x9e573ed9,
  0x7734,
  0x11d2,
  {0x8d, 0x4a, 0x23, 0x90, 0x3f, 0xb6, 0xbd, 0xf7}
};

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

Paul  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #10 from Paul  ---
Hi.  Ran into this apparently old issue after dusting off a T.Flight Hotas X
from Thrustmaster.  Same exact behavior with the hat switch through the js0
interface as described above.

Present in vanilla wine 5.0.3 and 6.3 in Gentoo, in 32 and 64 bit clean
prefixes.

In game (tested in Everspace) and in wine control panel, the joystick's event
interface (/dev/input/event8) registers proper cardinal directions for the hat
switch.  Switching to the js interface (/dev/input/js0), the hat switch's X and
Y axes are swapped.  As a fix, I can swap the mapping in jstest-gtk, and js0
will behave properly in wine.  Not ideal, since I'd be swapping it back and
forth.

My preference for the js0 interface is because I can adjust deadzones through
jstest-gtk and jscal.  The counterpart event program "evdev-joystick" doesn't
seem to actually affect the deadzone.

I'd pursue the event program's issues, but since js0 works fine outside of
wine, I'd consider this issue an easier bridge to cross.

The joystick has seven axes (identified in jscal as 0, 1, 2, 5, 6, 16, 17) ,
including the two representing the single hat switch on it.  The hat axes are
16 (x axis) and 17 (y axis).

I can successfully remap axes on the joystick through instructions at:
https://wiki.winehq.org/Useful_Registry_Keys
However, it doesn't affect the hat switch's orientation.

I patched wine 6.3, using the patch here as a base, and it fixed the hat switch
cardinals.  According to comment #5 it looks like it breaks other controllers,
so that may not be the solution.  I am attaching logs from WINEDEBUG=+dinput
wine control for the unpatched and patched cases.  I am also attaching the
udevadm info for the joystick.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #11 from Paul  ---
Created attachment 69565
  --> https://bugs.winehq.org/attachment.cgi?id=69565
"wine control" unpatched wine 6.3 with t.flight hotas x

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #12 from Paul  ---
Created attachment 69566
  --> https://bugs.winehq.org/attachment.cgi?id=69566
"wine control" patched wine 6.3 with t.flight hotas x

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #13 from Paul  ---
Created attachment 69567
  --> https://bugs.winehq.org/attachment.cgi?id=69567
udevadm info for hotas x

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

Rémi Bernon  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #14 from Rémi Bernon  ---
The legacy dinput joystick backends are now gone, could you please check if
this is still wrong with current Wine HEAD?

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #15 from Paul  ---
(In reply to Rémi Bernon from comment #14)
> The legacy dinput joystick backends are now gone, could you please check if
> this is still wrong with current Wine HEAD?

I just built Wine from latest today (10/13/2021) on Gentoo.  With the latest, I
see only one joystick option through Wine 'control' now.

I tested one-by-one with the T.master HOTAS X, and the newer T.master HOTAS 4.

Seeing as the option to select the (js0) device is no longer present in Head, I
can't answer to the question of it being fixed.  I can say that with the HOTAS
X, all analog axes and buttons are reported correctly, however the HAT switch's
Y-axis is inverted for some reason.  Its X-axis is correct.

With the T.master HOTAS 4, one axis is not being reported, and the HAT switch
has the same inverted Y-axis.  The HAT's X-axis is correct.  The unreported
analog axis is:

type 3 (EV_ABS), code 7 (ABS_RUDDER)

as reported by 'evtest.'

Because I don't see any suffix for the driver device node in Wine control, I
don't know what software to use in linux to calibrate and adjust deadzones.  I
also don't know what losing the "legacy dinput joystick backends" means for
legacy windows software on Wine.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #16 from Rémi Bernon  ---
Thanks, Wine is now using SDL library by default to access controllers and
joysticks, so it's not possible to tell exactly which Linux subsystem it is
using, but most likely evdev, if it is available.

If SDL support is disabled (through the registry) or not compiled in, Wine is
using evdev directly instead.

Regarding the Y axis inversion, it would be awesome if you could provide a
WINEDEBUG=+plugplay,+hid,+hidp,+hid_report,+dinput log, combined with the
"evtest" output recorded at the same time, and pressing a few HAT switch
direction.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #17 from Paul  ---
Created attachment 70807
  --> https://bugs.winehq.org/attachment.cgi?id=70807
WINEDEBUG=+plugplay,+hid,+hidp,+hid_report,+dinput wine control

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #18 from Paul  ---
Created attachment 70808
  --> https://bugs.winehq.org/attachment.cgi?id=70808
evtest t.master hostas x

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #19 from Paul  ---
(In reply to Rémi Bernon from comment #16)
> Thanks, Wine is now using SDL library by default to access controllers and
> joysticks, so it's not possible to tell exactly which Linux subsystem it is
> using, but most likely evdev, if it is available.
>
> If SDL support is disabled (through the registry) or not compiled in, Wine
> is using evdev directly instead.
>
> Regarding the Y axis inversion, it would be awesome if you could provide a
> WINEDEBUG=+plugplay,+hid,+hidp,+hid_report,+dinput log, combined with the
> "evtest" output recorded at the same time, and pressing a few HAT switch
> direction.

Attachments uploaded per request.  It also looks like the SDL devs are getting
off their butts and fixing the calibration issues inherent in their code:

https://github.com/libsdl-org/SDL/issues/3781

Perhaps the js driver can finally be put to rest.  Before now, it was the only
one that would accept and reflect calibrations to correct the deadzones
arbitrarily put in place.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #20 from Rémi Bernon  ---
Thanks, I'll have a look.

FWIW dinput has some builtin deadzone and saturation calibration features, it
should be implemented but it's not persistent yet or easily configurable
either, only programmatically accessible by any any application which may set
it.

We could think of having it persistent and exposed in joy.cpl to ease the setup
of Wine joysticks.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #21 from Paul  ---
(In reply to Rémi Bernon from comment #20)
> Thanks, I'll have a look.
>
> FWIW dinput has some builtin deadzone and saturation calibration features,
> it should be implemented but it's not persistent yet or easily configurable
> either, only programmatically accessible by any any application which may
> set it.
>
> We could think of having it persistent and exposed in joy.cpl to ease the
> setup of Wine joysticks.

I see.  Since the consensus seems to be that the application should be
controlling the deadzone, I'd be all for leaving it up to that, conditionally
to whether the app does it well!  A good point was made at the link I offered,
which is if the application employs any math to calculate a variable deadzone,
then a presence of a lower-level deadzone would skew the results.  My only goal
in using the js device node was to use 'jscal' to actually eliminate the
built-in deadzones at the time.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #22 from Rémi Bernon  ---
I've sent https://source.winehq.org/patches/data/217348 and
https://source.winehq.org/patches/data/217347 which should hopefully solve the
two issues mentioned above.

The Y axis fix at least is quite obvious, I'm less sure about the missing
RUDDER axis though, maybe it's something that isn't mapped in SDL.

You can also try by disabling SDL, so that Wine directly uses evdev. This can
be done by setting the "Enable SDL" option to 0, as described in
https://wiki.winehq.org/Useful_Registry_Keys, and if that still doesn't work, a
log with the same WINEDEBUG options as above would be really helpful!

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #23 from Paul  ---
(In reply to Rémi Bernon from comment #22)
> I've sent https://source.winehq.org/patches/data/217348 and
> https://source.winehq.org/patches/data/217347 which should hopefully solve
> the two issues mentioned above.
>
> The Y axis fix at least is quite obvious, I'm less sure about the missing
> RUDDER axis though, maybe it's something that isn't mapped in SDL.
>
> You can also try by disabling SDL, so that Wine directly uses evdev. This
> can be done by setting the "Enable SDL" option to 0, as described in
> https://wiki.winehq.org/Useful_Registry_Keys, and if that still doesn't
> work, a log with the same WINEDEBUG options as above would be really helpful!

OK, I used the very same wine head for this round of tests.  I noticed a few
problems with this new SDL/event dependency choice of wine.  In the past (wine
6.12, etc.) I could use the regedit instructions from:

https://wiki.winehq.org/Useful_Registry_Keys

"axes mapping" could reorganize the axes of the HOTAS 4 so I could once again
use the rudder (the kernel driver for it isn't mature due to few people knowing
how to get it out of the sony PS4 compatibility and into PC output, and the
driver identifies more axes than it actually has, throwing other software off)

Using the instructions to enable/disable SDL, this particular registry key
doesn't produce any change in axes.

Specifically, in the past, this would work, and now does not (yes in today's
tests I changed the name of the device to reflect what wine identifies.  This
is from an older prefix):
[Software\Wine\DirectInput] 1615525022
#time=1d716fc2425fc64
"Thrustmaster T.Flight Hotas 4 (event)"="X,Y,Z,-,-,Rx,-,Ry,POV1"

Further, once SDL is disabled in the registry, the input device name is reduced
to an ambiguous "Thrustmaster" which would be a problem if there were more than
one Thrustmaster device connected.

In each test with and without SDL, I sampled each axis, with the rudder axis
the last.

One last thing, with SDL disabled, the HAT has correct directions.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #24 from Paul  ---
Created attachment 70821
  --> https://bugs.winehq.org/attachment.cgi?id=70821
SDL enabled $WINEDEBUG=+plugplay,+hid,+hidp,+hid_report,+dinput wine control

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #25 from Paul  ---
Created attachment 70822
  --> https://bugs.winehq.org/attachment.cgi?id=70822
evtest t.master hotas 4 in tandem with sdlenabled report

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #26 from Paul  ---
Created attachment 70823
  --> https://bugs.winehq.org/attachment.cgi?id=70823
SDL disabled $WINEDEBUG=+plugplay,+hid,+hidp,+hid_report,+dinput wine control

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #27 from Paul  ---
Created attachment 70824
  --> https://bugs.winehq.org/attachment.cgi?id=70824
evtest t.master hotas 4 in tandem with sdldisabled report

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

Julian Rüger  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #28 from Rémi Bernon  ---
Looks like the SDL backend is hitting the absolute axis limit, but the evdev
backend maps the ABS_RUDDER axis correctly and dinput is picking it.

Also note that joy.cpl can currently only display up to 6 axes, so the although
the rudder axis is apparently supported (with SDL disabled), it's not displayed
there. We'd have a improve the control panel interface for it to be.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #29 from Rémi Bernon  ---
Regarding the axis mapping, it's indeed not supported anymore. Maybe we could
restore something like it in the new dinput HID joystick, but I'm not sure it's
a good idea to do it there, it would be better to have all the axis properly
mapped in the first place.

Then SDL already has some sort of mapping for gamepad controllers, and we do
some evdev to HID mapping in winebus.sys, so maybe it's better to do it there.
It's not going to be possible to have a per-application mapping in that case
though.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #30 from Paul  ---
(In reply to Rémi Bernon from comment #29)
> Regarding the axis mapping, it's indeed not supported anymore. Maybe we
> could restore something like it in the new dinput HID joystick, but I'm not
> sure it's a good idea to do it there, it would be better to have all the
> axis properly mapped in the first place.
>
> Then SDL already has some sort of mapping for gamepad controllers, and we do
> some evdev to HID mapping in winebus.sys, so maybe it's better to do it
> there. It's not going to be possible to have a per-application mapping in
> that case though.

Right, it didn't take long in my research to figure out there was an axis limit
in most abstraction layers, hence the registry edit removing them from wine's
sight.  I think if I could connect with someone in charge of the driver, that
could be settled.  Seems like a lot of effort for a sim input device, though. 
I'm convinced the proper way should be by masking the useless axes at the
kernel level, but the event nodes make it much more difficult than the old jsX
nodes.

Seems like it's getting more and more difficult to modify things on the
software side to work with flawed hardware, which exists in the world, and must
be tolerated at times.  I remember conversations like this when the first
modern touchscreens came out.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #31 from Rémi Bernon  ---
I'm definitely not opposed to make the axis mapping configurable somehow, to
help with bogus devices and users, just that doing it in DInput like it was
done before doesn't seem right anymore.

Wine now exposes any host joystick as an HID device internally, and DInput is
only one of the many ways to access them. Applications may use HID APIs
directly, and I know several middleware are doing it instead of DInput which is
kind of considered as legacy.

So having a customizable mapping in winebus.sys instead, for when where we
convert evdev axes to HID usages makes more sense, it's just not implemented
yet.

Of course, that won't cover cases where Wine simply acts as a pass-through,
like with hidraw devices, or where the mapping is already done elsewhere, like
with SDL.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #32 from Rémi Bernon  ---
Of course that should read "help the users", not help with bogus users.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #33 from Paul  ---
(In reply to Rémi Bernon from comment #31)
> I'm definitely not opposed to make the axis mapping configurable somehow, to
> help with bogus devices and users, just that doing it in DInput like it was
> done before doesn't seem right anymore.
>
> Wine now exposes any host joystick as an HID device internally, and DInput
> is only one of the many ways to access them. Applications may use HID APIs
> directly, and I know several middleware are doing it instead of DInput which
> is kind of considered as legacy.
>
> So having a customizable mapping in winebus.sys instead, for when where we
> convert evdev axes to HID usages makes more sense, it's just not implemented
> yet.
>
> Of course, that won't cover cases where Wine simply acts as a pass-through,
> like with hidraw devices, or where the mapping is already done elsewhere,
> like with SDL.

I admit that would take care of the current situation.  Per-application mapping
should be a non-issue with separate prefixes, as is recommended, and should be
further encouraged by distro wikis.

I've run across apps that used raw input (Everspace I think it was?).  The only
way to correct it was through the device node, which loops us back to the
beginning of the discourse.

Of course, hearing wine implement SDL in this way brings me back to all the
terrible times, as an end user, I had trying to make sense of their wiki.  This
lists some up-to-date tools, though I'm not sure how relevant this will be to
wine.
https://github.com/gabomdq/SDL_GameControllerDB
> Of course that should read "help the users", not help with bogus users.

Yeah, we shouldn't help them, stupid non-sentient wannabe humans.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #34 from Rémi Bernon  ---
I sent https://source.winehq.org/patches/data/217914,
https://source.winehq.org/patches/data/217915,
https://source.winehq.org/patches/data/217916 and
https://source.winehq.org/patches/data/217917 to help with SDL axis limit, and
hopefully addressing the remaining issues here, then I believe the bug could be
marked as resolved.

I also opened https://bugs.winehq.org/show_bug.cgi?id=51908 to track the
addition of a custom axis mapping to winebus.sys evdev backend.

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

Rémi Bernon  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           See Also|                            |https://bugs.winehq.org/sho
                   |                            |w_bug.cgi?id=51908

[Bug 40658] Joystick POW hat switch rotated for JOYDEVDRIVER=(js) while JOYDEVDRIVER=(event) is Ok

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=40658

--- Comment #35 from Paul  ---
(In reply to Rémi Bernon from comment #34)
> I sent https://source.winehq.org/patches/data/217914,
> https://source.winehq.org/patches/data/217915,
> https://source.winehq.org/patches/data/217916 and
> https://source.winehq.org/patches/data/217917 to help with SDL axis limit,
> and hopefully addressing the remaining issues here, then I believe the bug
> could be marked as resolved.
>
> I also opened https://bugs.winehq.org/show_bug.cgi?id=51908 to track the
> addition of a custom axis mapping to winebus.sys evdev backend.

Let me know when/how to test the head so we can mark this solved.