I cannot produce this bug on ChromeOS (Pixel Chromebook) latest stable (M58).
I get a pair of mouse down and mouse up regardless of how I interleave the touch and mouse.

Also, you mentioned you can repro this on linux with M30. Can you reproduce with M58?

I also tried Chrome M59 on Windows, the bug doesn't reproduce.

Hi, i re-tested on M57, it shows 2 mouseup events, but both on "Div 2", --
which i don't know is ok or wrong

2017-05-26 3:55 GMT+08:00 mus… via monorail <
monorail+v2.2341872843@chromium.org>:

Thank you for providing more feedback. Adding requester "majidvp@chromium.org" to the cc list and removing "Needs-Feedback" label.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Re-test on M58 stable & M59 beta, problem still there.

1, First, press mouse on div 1, and hold, you'll see mousedown on div 1
triggered;
2, Then, put finger on touch screen quickly on div 2, console log outputs
mousedown & up on div 2;
3, Then, release the mouse

No mouseup on div 1 shows. And what is interesting to me, the mouse
cursor's position wierdly moved to div 2, it is previously on div 1.

2017-05-26 19:33 GMT+08:00 小鱼儿 <ctengctsh@gmail.com>:

Sorry, typo. My chrome on windows is previously M58, and today i just
installed M59 beta to verify. There are currently 2 results:

result #1: 2 mouseup events triggered, but both event.target element seem
to the second one

or result #2: only 1 mouseup event triggered,the first element didn't
receive the mouseup

2017-05-26 19:40 GMT+08:00 小鱼儿 <ctengctsh@gmail.com>:

If i touch 2 div's using 2 fingers on the multi-touch screen, then no
synthetic mouse events will be triggered. -- Which i can't say is right or
wrong, but at least it's consistent.....

2017-05-26 19:46 GMT+08:00 小鱼儿 <ctengctsh@gmail.com>:

We send mousedown/up events for touch tap to make legacy websites work, and have stopped sending them for long-tap etc recently. I suspect we never sent them for two-finger taps.

There is no perfect solution here: legacy mouse behavior assumes a single mouse pointer, and we are trying to preserve the old behavior for multiple pointers here (mouse+touch or 2-touches) which is hard.

Also: we can never guarantee mousedown/up pairing even with a single mouse. You can drag out from a page and release the button outside, for example.

For multiple pointer cases you mentioned, we recommend using PointerEvents where each pointerdown/up/move/etc event has a unique pointerId so it's easy to differentiate between pointers (mouse/touch/pen).

I replaced mousedown/up by pointerdown/up, the same problem, so "unique
pointerId" doesn't apply.

Also, you mean "longtap doesn't trigger mouseup reliably"? I thought here
was a usability miss: mutlti-touch is basically used for "gesture
detection", but the W3C guys didn't consider the old plain legacy use.

Also, i can't agree with you. "legacy mouse behavior" can be extended in
multi-touch mode, & we should consider the final user special case, to
maximize the usability, not just falling adhered to an incomplete spec.

If there is mousemove after intial mousedown, the i thought mouseup doesn't
trigger is a good design: since if there is a pair of mousedown/up, there
will be a "click", the usees usually apply this principle to cancel the
click in UX ops. But in this case, there is no mousemove, only a "longtap"
gesture, so theoretically it's differentiatable?

2017-05-27 4:43 GMT+08:00 mus… via monorail <
monorail+v2.2341872843@chromium.org>:

pointerup/down: I meant pointerIds infer which pointer went down/up, which is an improvement over legacy MouseEvent.

mouseup on longtap: a longtap fires contextmenu event, so should never fire a click. Moreover, we don't know of any browsers doing it.  (Firing auxclick when contextmenu event gets prevented is a possibility though.)


Legacy mouse on multitap: This is a harder problem than it seems.  PointerEvent spec addresses this but only as a recommendation:
https://w3c.github.io/pointerevents/#tracking-the-effective-position-of-the-legacy-mouse-pointer

For example, what should happen to mouseenter/leave events (which must bracket every mousedown/mouseup) if two fingers are dragged on two different divs?  What the PointerEvent spec suggests is a long sequence of enter/leave/move sequence (leave@div1, enter@div2, move@div2, leave@div2, enter@div1, move@div1, ...) which is not great but there seems to be no good solution.

A related problem with multitouch/multipointers is that the visible mouse cursor movement is OS dependent, and to the best of our knowledge, only new Windows apps (those using WM_POINTER handlers) handle them perfectly: mouse cursor hides without losing its position when a new pointer (touch/pen) is active.  The single-mouse-pointer model in the legacy case is broken in this case.  In Linux/Mac, the mouse cursor always jumps to last active pointer so the legacy model makes sense.  But it has other problems:  e.g. Mac doesn't even have a way to identify a pointer at the OS level when e.g., a pen pointer leaves the digitizer range.

For all of this, tweaking legacy mouse events for multitouch is not a priority for us.