Swiping downward in perspective makes the map move in the opposite directions (upwards)

[mattgraham1]

Platform:
Android
Mapbox SDK version:
4.x

Steps to trigger behavior

1. Download your demo app from the play store
2. Open a map and put it into perspective mode
3. Swipe in a downward direction

Expected behavior

Map pans in a downward direction

Actual behavior

Map pans in a upward direction

[friedbunny]

@cammace and I couldn’t reproduce this with the current Android demo app on his 5X, though this sounds remarkably similar to the behavior on iOS (#5292).

[mattgraham1]

I'm able to reproduce it easily on my nexus 6p. Are you trying this in perspective view (two finger swipe up)? Then one finger swipe down will bounce the map up ward. See my attached video (need to unzip first).
device-2016-06-09-075356.mp4.zip

[HeyLookItsBrandon]

I've run into this as well on my LG G3. I've found that the map pans as expected when panning slowly. The issue appears to be specific to the fling gesture. When I pan with some velocity and release, the map "bounces" the opposite direction when I lift my finger.

[cammace]

After looking more into this, I have been able to reproduce on both 4.0.1 and 4.1.0 betas. This looks to be the same issue happening on iOS @friedbunny linked to.

cc: @zugaldia

[1ec5]

This effect reminds me of #3655, which affected untilted maps near the antimeridian. (It was fixed in #4214.)

[mdakram]

It seems its still there in v4.2.0 beta 4.

[tobrun]

Been debugging this issue in more depth and seeing a couple of scroll gestures occurring before we hit the fling gesture. Both scroll as fling hook into the Transform::moveBy method. The values outputted by the scroll gesture are based around x,y values while the values outputted by the fling gesture are provided as velocityX and velocityY. We now perform some calculations on these velocity values to create a x,y value from it.

Something I noticed while playing around with these calculations is that if we lower the velocity by dividing it with 6 instead of 4, the issue stops showing but this also results in the fling not feeling like a fling anymore.

An idea I have been playing around with is taking in account the current tilt value. What I mean by this is that when the map is not tilted the value of 4 remains, when we tilt the map the value will gradually increase.

Side note here that a related fix provided by @frederoni to handle the horizontal fling issue has made reproducing this a bit harder (related PR in #7105).

[tobrun]

Without taking tilt in account:

With taking tilt in account, and thus lowering velocity based on it:

• this uses value 7 instead of 4 that was mentioned above.

[tobrun]

Next steps to actually resolve the underlying issue, above is just a workaround, is to start writing some core tests to identify where this behaviour is coming from.

cc @brunoabinader