[Bug 48252] New: Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

Bug ID: 48252
           Summary: Input is dropped while drawing perfectly-vertical
                    lines if tablet is set to 'Relative' mode
           Product: Wine
           Version: 4.21
          Hardware: x86
                OS: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: wintab32
          Assignee: [e-mail]
          Reporter: [e-mail]
      Distribution: ArchLinux

Hello there, I'm having issues in Clip Studio Paint where if the tablet is set
to relative mode input is completely lost if the lines are perfectly vertical.

I've tried this in both Wine 4.21, and Wine 4.21 staging, while in staging the
terminal outputs "002c:fixme:wintab32:motion_event Negative orAltitude
detected" for each motion event.

This issue has existed for a number of Wine releases, likely dating back even
further than wine 3.x, terminal output however has been somewhat new.

Video of the problem, recorded using Clip Studio Paint Pro v1.9.4 running on
the latest Wine 4.21 staging from Manjaro repository:
https://streamable.com/jxi6u

A download for Clip Studio Paint (v1.9.5) can be found at: 
https://www.clipstudio.net/en/dl

(Although recorded in v1.9.4, the issue has been present for a number of
releases, and is still present in the latest release v1.9.5.)

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

amiire  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|wintab32                    |-unknown
                 CC|                            |[e-mail]
                   |                            |, [e-mail]
            Product|Wine                        |Wine-staging

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

amiire  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|-unknown                    |wintab32
            Product|Wine-staging                |Wine

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

--- Comment #1 from Alistair Leslie-Hughes  ---
Can you please attach a +wintab32 log?

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

Alistair Leslie-Hughes  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |https://www.clipstudio.net/
                   |                            |en/dl
           Keywords|                            |download

--- Comment #2 from Alistair Leslie-Hughes  ---
I've just updated the patchset in staging to just print a FIXME and not fail
when a Negative orAltitude is detected.  I'm not expect this to fix the issue
but at least it a starting point.

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

--- Comment #3 from amiire  ---
Created attachment 65954
  --> https://bugs.winehq.org/attachment.cgi?id=65954
output log

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

--- Comment #4 from amiire  ---
(In reply to Alistair Leslie-Hughes from comment #1)
> Can you please attach a +wintab32 log?

Sure thing.

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

--- Comment #5 from amiire  ---
Created attachment 65963
  --> https://bugs.winehq.org/attachment.cgi?id=65963
+wintab32 log

Here is actual the +wintab32 log, my apologies.

Also perhaps its of some significance to mention my tablet is has a rotation
configuration of 'half' (so it is upside down).

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

--- Comment #6 from amiire  ---
The mouse no longer lags in the latest release, however lines aren't drawn
during this period. I'll link a recording of how it looks soon.

[Bug 48252] Input is dropped while drawing perfectly-vertical lines if tablet is set to 'Relative' mode

[WineHQ Bugzilla]

https://bugs.winehq.org/show_bug.cgi?id=48252

[e-mail] changed:

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

--- Comment #7 from [e-mail] ---
Created attachment 70457
  --> https://bugs.winehq.org/attachment.cgi?id=70457
fix for negative orAltitude

The XEvents' axis_data array only updates values that have changed, and sets
its size according to the maximum index of the changed axis +1.

For instance; the pointer's X value corresponds to the indice [0], the Y [1],
the NormalPressure [2], and so on. 
If only the X value of the pointer changes then the reported size of the
axis_data will be 1, axis_data[0] will hold the new X value of the pointer, and
any indice after [0] will have stale data from the previous frame.

However if only the NormalPressure changes, then the reported size of the
axis_data will be 3 and axis_data[2] will report the new NormalPressure,
however the indices for the X and Y in the axis_data will be left unaltered as
that axis has reported no change. Previously, the struct was cleared before use
at each motion_event, button_event, and proximity_event, due to this whenever
there was a change in a higher axis, but the the lower axis, such as X or Y
were unchanged the pointer would erratically teleport to the top-left corner
(0,0).

To fix this bug I added moved the memsets to the
X11DRV_AttachEventQueueToTablet so that the struct would only be cleared once
after initalization and not after each event callback, allowing for the
previous positions to be retained when only the pressure, or Y-axis changes.

Furthermore I added simple checks to only update axis that were changed, so
that axis with an index greater than the reported axis_data size will not
modified. This fixes the the negative OrAltitude bug.