revise has("touch"), add has("touch-events"), has("pointer-events"), has("MSPointer")

[Sebastien Pereira]

Note: this ticket may impact dojo, dijit and dojox modules.

Pointer Events are being developed and integrated on Chrome [1] and Firefox [2]. Until supported release of Chrome/FF implement those features, feature detection in dojo should not be a problem. However we should prepare for changes:

1. navigator.pointerEnabled is not anymore part of the Pointer Events specification [3].
2. navigator.maxTouchPoints will be available on non-IE

Dojo has("touch") condition to return true is "ontouchstart" in document || window.navigator.msMaxTouchPoints > 0.

The meaning of has("touch") should be clarified as "user agent is running on a device with a touchscreen" and condition window.navigator.maxTouchPoints > 0 should be added to has("touch") definition.

Also, we should introduce new has flags to avoid confusion and replace current tests based on navigator.pointerEnabled and navigator.MSPointerEnabled

has("touch") ===> "ontouchstart" in document || window.navigator.maxTouchPoints > 0 || window.navigator.msMaxTouchPoints > 0
has("touch-events") ===> "ontouchstart" in document
has("pointer-events") ===> navigator.maxTouchPoints || navigator.msMaxTouchPoints
has("MSPointer") ===> navigator.msMaxTouchPoints

With these flags we should be able to cover all use cases where has("touch")/navigator.pointerEnabled/navigator.msPointerEnabled are used in dojo, dijit and dojox modules.

=== [1] Chromium ​Issue 196799: Support for Pointer Events ​Issue 258459: touch-action CSS property

[2] Firefox ​Bug 822898 - Implement pointer events ​Bug 795567 - Implement touch-action CSS property for Pointer Events

[3] ​Pointer Events W3C Editor's Draft

[Sebastien Pereira]

Pull requests Dojo: ​https://github.com/dojo/dojo/pull/84 Dijit: ​https://github.com/dojo/dijit/pull/47 Dojox: ​https://github.com/dojo/dojox/pull/103 Calendar: ​https://github.com/damiengarbarino/dojo-calendar/pull/95

[Bill Keese <bill@…>]

In b25b080d6fb0e1b9e64ba48e303d207614f80ada/dojo:

Error: Processor CommitTicketReference failed

Unsupported version control system "git": Can't find an appropriate component, maybe the corresponding plugin was not enabled? 

[Bill Keese <bill@…>]

In 2dcb4e325792caca5bd4ee053a8d7e390c619fff/dijit:

Error: Processor CommitTicketReference failed

Unsupported version control system "git": Can't find an appropriate component, maybe the corresponding plugin was not enabled? 

[Patrick Ruzand <pruzand@…>]

In 184f71630051e64b864024b703df17f6d7f23507/dojox:

Error: Processor CommitTicketReference failed

Unsupported version control system "git": Can't find an appropriate component, maybe the corresponding plugin was not enabled? 

[bill]

seb - I'm seeing test_touch.html failures on IE10, see ​https://www.dropbox.com/s/myijmnynmmcr5qt/Screenshot%202014-05-10%2006.59.46.gif, hopefully just a problem w/the test needing to be updated.

[Bill Keese <bill@…>]

In f73dd454bdfce56069fd1072fb6d8d6438ccf2d8/dojo:

Error: Processor CommitTicketReference failed

Unsupported version control system "git": Can't find an appropriate component, maybe the corresponding plugin was not enabled? 

[Bill Keese <bill@…>]

In 9bde2f924856aaccbd62500b62ad310c016c6692/dijit:

Error: Processor CommitTicketReference failed

Unsupported version control system "git": Can't find an appropriate component, maybe the corresponding plugin was not enabled? 

[Sebastien Brunot]

The current fix, as discussed with spereira, has a side effect: because Chrome 35 (at least on mac OS X) do implement maxTouchPoints in the navigator object but do not implements pointer events, dojo/touch is broken on this browser: it listens to pointer events that are never emitted.

You can test that with the two following dojo mobile tests, that are failing on Google Chrome 35:

• test_Overlay.html: click the popup button, and then click the "Done" button of the overlay: nothing happens (expected: the overlay is closed)
• test_ScrollableView-v-ah-af.html: you cannot scroll the view, nor start a transition by clicking a list item.

[bill]

Thanks for tracking that down. As described in #18019, it also breaks dropdowns in dijit (see for example dijit/tests/_HasDropDown.html).

I guess the test should be "onpointerdown" in document, similar to the check for touch-events: "ontouchstart" in document?

[bill]

#18019 is a duplicate of this ticket.

[Sebastien Pereira]

Chrome Desktop M35 introduce a partial support of the pointer events specification: it supports the touch-ction CSS property ("touchAction" in document.body.style is true) and also defines "maxTouchPoints" in navigator, but it doesn't support the events (pointerdown...), so we need another way to detect the support for pointer events.

Chrome Mobile M35 is not impacted, but Chrome M36 plan to add the same partial support (1).

Firefox is not impacted yet, but also plan to provide pointer-event support. An experimental support is available since FF 29 behind flags dom.w3c_pointer_events.enabled and layout.css.touch_action.enabled (2).

(1) ​https://code.google.com/p/chromium/issues/detail?id=372357 (2) ​https://bugzilla.mozilla.org/show_bug.cgi?id=960316

[Sebastien Pereira]

Replying to bill:

I guess the test should be "onpointerdown" in document, similar to the check for touch-events: "ontouchstart" in document?

Yes, according the from the Pointer Events spec we can do that (1). I pushed the change in 36a44e063bfd1fde48de1944fdcf89a316dddc9a

(2) ​http://www.w3.org/TR/pointerevents/#extensions-to-the-globaleventhandlers-interface

[Sebastien Pereira]

c2f382cd194f147582e03c01dc41463eb978fd4a fixes has(touch) that caused #18019.

[bill]

This is fixed, right? Marking as enhancement so we remember to doc this in the release notes.