Add touch scrolling modes experimental flags (DEPRECATED)

Add touch scrolling modes experimental flags

Superceded by https://codereview.chromium.org/166923002/.

In particular, adds --touch-scrolling-mode=absorb-touchmove

TODO: work in progress
Appears to work properly on Android, but I've only run some basic tests and done some manual testing.  I haven't tested Aura recently so it may be broken (and tests probably need to be updated for it).

BUG=328503

[jdduke (slow), 2014-01-31 20:01:16]

I don't have any serious problems with this addition.  I'd love to get a more
descriptive term than |is_independent()|, though I can't immediately come up
with a single better word or phrase for the semantics =/

https://codereview.chromium.org/140253005/diff/60001/content/browser/renderer...
File content/browser/renderer_host/input/touch_event_queue.cc (right):

https://codereview.chromium.org/140253005/diff/60001/content/browser/renderer...
content/browser/renderer_host/input/touch_event_queue.cc:380: if
(touch_scrolling_mode_ != TOUCH_SCROLLING_MODE_ABSORB_TOUCHMOVE)
Could this be racy? If the GSU ack is delayed?  Or is the only consequence that
there might be some lag time when transitioning between absorption states?

https://codereview.chromium.org/140253005/diff/60001/content/browser/renderer...
content/browser/renderer_host/input/touch_event_queue.cc:549:
TouchEventQueue::TouchScrollingMode TouchEventQueue::GetTouchScrollingMode()
Any reason not to just put this in the anon namespace above and remove the
static member?  Or even better, have the mode be an arg to the TEQ and move
commandline parsing to the InputRouter?

[Rick Byers, 2014-02-01 23:09:33]

Thanks Jared.

I realized I wasn't properly handling GestureScrollEnd/GesturePinchEnd (we can't
throw them away when their corresponding touchend was consumed).  I've updated
the patch to handle this in the most direct way I see, but I'm not entirely
happy with it - it's getting quite complex and I fear we're lacking some general
principles here.

I've tried to expand on what I think we need in the eager GR touch dispositions
doc:
https://docs.google.com/a/chromium.org/document/d/176xYUC3WbiSl7qd08SBW4NiI4_....
 Perhaps we should just switch this code to use a table like I define in that
doc?

Or perhaps this isn't so bad and we should stick with it for now as we think
about how to evolve this further.

Tim/Jared, feel free to take over this CL while I'm on vacation and do whatever
you think is best here.  Otherwise I can pick it back up when I return on Feb
13th.

Thanks!
   Rick

https://codereview.chromium.org/140253005/diff/60001/content/browser/renderer...
File content/browser/renderer_host/input/touch_event_queue.cc (right):

https://codereview.chromium.org/140253005/diff/60001/content/browser/renderer...
content/browser/renderer_host/input/touch_event_queue.cc:380: if
(touch_scrolling_mode_ != TOUCH_SCROLLING_MODE_ABSORB_TOUCHMOVE)
On 2014/01/31 20:01:16, jdduke wrote:
> Could this be racy? If the GSU ack is delayed?  Or is the only consequence
that
> there might be some lag time when transitioning between absorption states?

Yes, this could be racy.  I was thinking it was OK, but upon further reflection
I'm a little worried about the impact of unavoidably triggering overscroll. 
I've add some thoughts on this to the design doc here:
https://docs.google.com/a/chromium.org/document/d/1MGuCb0Lb7UfaDmH44bp9YKVG-C...
and will leave a TODO in the code for now.  I think this is fine for
experimentation purposes (worst case should cause some minor UI glitches when
under load), but we should resolve it before deciding whether to ship this or
not.

https://codereview.chromium.org/140253005/diff/60001/content/browser/renderer...
content/browser/renderer_host/input/touch_event_queue.cc:549:
TouchEventQueue::TouchScrollingMode TouchEventQueue::GetTouchScrollingMode()
On 2014/01/31 20:01:16, jdduke wrote:
> Any reason not to just put this in the anon namespace above and remove the
> static member?  Or even better, have the mode be an arg to the TEQ and move
> commandline parsing to the InputRouter?

Done.

[tdresser, 2014-02-03 15:52:57]

I'm leaning towards re-implementing as a more explicit table. I think long term
that will be more flexible and easier to reason about.

Other than that, I just have a few minor nits.

jjduke@, do you think making the table more explicit would be beneficial? If so,
I'll give it a whirl.

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
File content/browser/renderer_host/input/gesture_event_packet.cc (right):

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_packet.cc:40:
is_independent_(true) {}
It feels odd to me that we default this to true, and then switch it to false if
necessary. It would feel more natural to default to false and switch to true if
necessary.

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
File content/browser/renderer_host/input/gesture_event_queue.cc (right):

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_queue.cc:115: bool send =
false;
Need {} in this conditional.

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
File content/browser/renderer_host/input/gesture_event_queue_unittest.cc
(right):

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_queue_unittest.cc:210: // An
consumed touch's gesture should be sent.
I'm not clear on the purpose of this comment modification.
Aren't we still testing that consumed a consumed touch's gesture should not be
sent?

Also, "An" -> "A".

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_queue_unittest.cc:294: // End
events dispatched when their start events were independent
"when their start events were" -> "when their start events were,"

[tdresser, 2014-02-03 18:55:14]

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
File content/browser/renderer_host/input/gesture_event_packet.cc (right):

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_packet.cc:40:
is_independent_(true) {}
On 2014/02/03 15:52:58, tdresser wrote:
> It feels odd to me that we default this to true, and then switch it to false
if
> necessary. It would feel more natural to default to false and switch to true
if
> necessary.

Nevermind, this makes sense.

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
File content/browser/renderer_host/input/gesture_event_queue.cc (right):

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_queue.cc:115: bool send =
false;
On 2014/02/03 15:52:58, tdresser wrote:
> Need {} in this conditional.

Done.

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
File content/browser/renderer_host/input/gesture_event_queue_unittest.cc
(right):

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_queue_unittest.cc:210: // An
consumed touch's gesture should be sent.
On 2014/02/03 15:52:58, tdresser wrote:
> I'm not clear on the purpose of this comment modification.
> Aren't we still testing that consumed a consumed touch's gesture should not be
> sent?
> 
> Also, "An" -> "A". 

Done.

https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
content/browser/renderer_host/input/gesture_event_queue_unittest.cc:294: // End
events dispatched when their start events were independent
On 2014/02/03 15:52:58, tdresser wrote:
> "when their start events were" -> "when their start events were,"

Done.

[tdresser, 2014-02-10 18:35:02]

On 2014/02/03 18:55:14, tdresser wrote:
>
https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
> File content/browser/renderer_host/input/gesture_event_packet.cc (right):
> 
>
https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
> content/browser/renderer_host/input/gesture_event_packet.cc:40:
> is_independent_(true) {}
> On 2014/02/03 15:52:58, tdresser wrote:
> > It feels odd to me that we default this to true, and then switch it to false
> if
> > necessary. It would feel more natural to default to false and switch to true
> if
> > necessary.
> 
> Nevermind, this makes sense.
> 
>
https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
> File content/browser/renderer_host/input/gesture_event_queue.cc (right):
> 
>
https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
> content/browser/renderer_host/input/gesture_event_queue.cc:115: bool send =
> false;
> On 2014/02/03 15:52:58, tdresser wrote:
> > Need {} in this conditional.
> 
> Done.
> 
>
https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
> File content/browser/renderer_host/input/gesture_event_queue_unittest.cc
> (right):
> 
>
https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
> content/browser/renderer_host/input/gesture_event_queue_unittest.cc:210: // An
> consumed touch's gesture should be sent.
> On 2014/02/03 15:52:58, tdresser wrote:
> > I'm not clear on the purpose of this comment modification.
> > Aren't we still testing that consumed a consumed touch's gesture should not
be
> > sent?
> > 
> > Also, "An" -> "A". 
> 
> Done.
> 
>
https://codereview.chromium.org/140253005/diff/120001/content/browser/rendere...
> content/browser/renderer_host/input/gesture_event_queue_unittest.cc:294: //
End
> events dispatched when their start events were independent
> On 2014/02/03 15:52:58, tdresser wrote:
> > "when their start events were" -> "when their start events were,"
> 
> Done.

I've rewritten using the table approach here
(https://codereview.chromium.org/156783006).
There are a few issues with the current table which I've commented on in the
doc.

I'll add some unit tests, and get Aura behaving reasonably.