Bug 206216 – PointerEvent has 0 as pressure when the input device is not pressure-sensitive

• Component: UI Events
• Status: NEW

[Akihiko Odaki]

W3C's pointer events level 2 specification states:
> For hardware and platforms that do not support pressure, the value MUST be 0.5 when in the active buttons state and 0 otherwise.
https://www.w3.org/TR/2019/REC-pointerevents2-20190404/

However, WebKit sets 0 as pressure in such case. The behavior is confirmed with:
Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_2) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/13.0.4 Safari/605.1.15

Apparently, WebKit refers to pressure property of NSEvent. The documentation states:
> For input devices that aren’t pressure-sensitive, the value is either 0.0 or 1.0.
https://developer.apple.com/documentation/appkit/nsevent/1534543-pressure

[Radar WebKit Bug Importer]

<rdar://problem/58596321>

[Akihiko Odaki]

Created attachment 388210 [details]
Patch

[Darin Adler]

Comment on attachment 388210 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=388210&action=review

> Source/WebCore/dom/PointerEvent.cpp:115
> +    , m_pressure(mouseEvent.buttons() ? 0.5 : 0)

I am not 100% sure this is right. The specification calls for all pointerup events having pressure of 0. Are we guaranteed that buttons() will be 0 for *all* pointerup events? The test case covers only a simple case of a mouse with a single button.

> LayoutTests/ChangeLog:11
> +        * pointerevents/mouse/pointer-button-and-buttons-expected.txt: Removed.
> +        * pointerevents/mouse/pointer-button-and-buttons.html: Removed.
> +        * pointerevents/mouse/pointer-button-buttons-and-pressure-expected.txt: Added.
> +        * pointerevents/mouse/pointer-button-buttons-and-pressure.html: Added.

Why are we renaming this test case?

[Darin Adler]

(In reply to Akihiko Odaki from comment #0)
> Apparently, WebKit refers to pressure property of NSEvent. The documentation
> states:
> > For input devices that aren’t pressure-sensitive, the value is either 0.0 or 1.0.
> https://developer.apple.com/documentation/appkit/nsevent/1534543-pressure

Aside from the bug fixed here, this implies that on *Cocoa* platforms, for non-pressure-sensitive devices, WebKit is setting 1.0 as the pressure, where the standard instead requests/requires that we set it to 0.5. May be challenging to fix if NSEvent does not tell us whether the device is pressure-sensitive. It would be good if someone investigated this and created tests.

[Antoine Quint]

(In reply to Darin Adler from comment #3)
> Comment on attachment 388210 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=388210&action=review
> 
> > Source/WebCore/dom/PointerEvent.cpp:115
> > +    , m_pressure(mouseEvent.buttons() ? 0.5 : 0)
> 
> I am not 100% sure this is right. The specification calls for all pointerup
> events having pressure of 0. Are we guaranteed that buttons() will be 0 for
> *all* pointerup events? The test case covers only a simple case of a mouse
> with a single button.

Maybe it would be better to use the provided event type to set the pressure value.

[Akihiko Odaki]

(In reply to Darin Adler from comment #3)
> Comment on attachment 388210 [details]
> Patch
> 
> View in context:
> https://bugs.webkit.org/attachment.cgi?id=388210&action=review
> 
> > Source/WebCore/dom/PointerEvent.cpp:115
> > +    , m_pressure(mouseEvent.buttons() ? 0.5 : 0)
> 
> I am not 100% sure this is right. The specification calls for all pointerup
> events having pressure of 0. Are we guaranteed that buttons() will be 0 for
> *all* pointerup events? The test case covers only a simple case of a mouse
> with a single button.

Yes, the specification states:
> A user agent MUST fire a pointer event named pointerup when a pointer leaves the active buttons state.

> 
> > LayoutTests/ChangeLog:11
> > +        * pointerevents/mouse/pointer-button-and-buttons-expected.txt: Removed.
> > +        * pointerevents/mouse/pointer-button-and-buttons.html: Removed.
> > +        * pointerevents/mouse/pointer-button-buttons-and-pressure-expected.txt: Added.
> > +        * pointerevents/mouse/pointer-button-buttons-and-pressure.html: Added.
> 
> Why are we renaming this test case?

Because it is now testing pressure property too. The name simply represents tested properties.

(In reply to Darin Adler from comment #4)

> (In reply to Akihiko Odaki from comment #0)
> > Apparently, WebKit refers to pressure property of NSEvent. The documentation
> > states:
> > > For input devices that aren’t pressure-sensitive, the value is either 0.0 or 1.0.
> > https://developer.apple.com/documentation/appkit/nsevent/1534543-pressure
> 
> Aside from the bug fixed here, this implies that on *Cocoa* platforms, for
> non-pressure-sensitive devices, WebKit is setting 1.0 as the pressure, where
> the standard instead requests/requires that we set it to 0.5. May be
> challenging to fix if NSEvent does not tell us whether the device is
> pressure-sensitive. It would be good if someone investigated this and
> created tests.

I wrote that statement before carefully looking at the code and it is misleading. In reality, pressure property of NSEvent is completely ignored because MouseEvent provided to PointerEvent to create a new instance does not have pressure information.

(In reply to Antoine Quint from comment #5)
> (In reply to Darin Adler from comment #3)
> > Comment on attachment 388210 [details]
> > Patch
> > 
> > View in context:
> > https://bugs.webkit.org/attachment.cgi?id=388210&action=review
> > 
> > > Source/WebCore/dom/PointerEvent.cpp:115
> > > +    , m_pressure(mouseEvent.buttons() ? 0.5 : 0)
> > 
> > I am not 100% sure this is right. The specification calls for all pointerup
> > events having pressure of 0. Are we guaranteed that buttons() will be 0 for
> > *all* pointerup events? The test case covers only a simple case of a mouse
> > with a single button.
> 
> Maybe it would be better to use the provided event type to set the pressure
> value.

The specification states it is guaranteed, so I do not think it makes difference. We can still add an assertion, but such an assertion should cover all (including one generated by pressure-sensitive device) PointerEvents and I think it is out of scope of this bug.