Here's a screenshot

Deleted: Screenshot from 2016-07-04 22-30-32.png 311 KB

	Screenshot from 2016-07-04 22-30-32.png 311 KB View Download

What I noticed is that the gesture appears to be enabled only after a short delay during which two fingers are both on the touchpad. That may explain some of the "inconsistencies" in behavior you noticed.

Additionally, when scrolling a page up/down with two fingers, the whole page tends to wobble ever so slightly along the horizontal axis, if you do not follow a perfectly vertical line: indeed, Chromium is starting the "maybe-I'm-gonna-go-back-in-history" animation while you're scrolling up/down if there is a bit of lateral motion in the initial vector of each swipe motion. And of course, there is nearly always some. It's really visually disturbing.

> What I noticed is that the gesture appears to be enabled only after a short
> delay during which two fingers are both on the touchpad. 

Do you mean it is a necessary, but not sufficient conditions? Or can you reproduce it systematically by just having both fingers on the touchpad for a time and then swiping? If that is the case, I can't.

> Additionally, when scrolling a page up/down with two fingers, the whole page
> tends to wobble ever so slightly along the horizontal axis, if you do not
> follow a perfectly vertical line

Never observed this either.

Are you on Mac OSX perhaps? I am on Linux.


Anyway, this behavior should not exist at all, certainly not on Linux and Windows.

Nor on OSX, unless it's a common native convention on that OS. In which case it is not automatically the right thing to do but it's worth considering.

This behavior is not limited to trackpads.

I've discovered this happens with a regular mouse that has a side-scrolling mouse-wheel. I'm using a basic Logitech mouse (M510). I can reproduce this  issue 100 % of the time. The trick to reproducing the issue is to NOT move the mouse cursor while scrolling. 

If I move the cursor while in the middle of this transition, the page will snap back. 

This is a video of this behavior on my computer.
https://youtu.be/8n-gpahfy8M

I'm running Chrome 52.0.2743.82 64bit on Ubuntu 16.04.

I'm observing this on Chrome 52.0.2743.116 64 bit on Ubuntu 14.04.1 on a laptop with a trackpad.

Unlike the video in comment 4, when I scroll horizontally, the page jumps horizontally by seemingly random amounts, but never more than about 50 pixels from the edge of the window, and if I release the scroll the page stays partially offset and does not snap back until I scroll again (horizontally or vertically). The video looks like something working as intended. What I'm seeing absolutely does not.

This happens on some pages and not on others. For example, it happens on GMail and this bug page, but it doesn't happen on www.google.com, www.youtube.com, or on really simple pages like chrome://version.

> The video looks like something working as intended.

Well that's not exactly true (unless we assume it is intended to be demential) for several reasons:
- navigating back and forth in history by horizontal scrolling should not exist as a feature in the first place
- the video seems to show (we can't tell for sure) the case where the user slowly does the whole scrolling from beginning to end. It doesn't show what happens if the user stops scrolling and releases the "sliding scroll wheel" (whatever that is) in the middle of the scrolling. The expected behavior in this case would be to either snap back or quickly continue and finish the animation (depending on whether a certain threshold was passed). To me the video suggests that the animation just remains half-way, though we would need to see what the user is doing with their finger to tell for sure.

> What I'm seeing absolutely does not.

Now that's for sure. Of course the behavior you are observing and describing does seem to be even crazier than the one in the video, as is the one that I observe.


This thing is pathetic, it looks like it's some work-in-progress incomplete feature (which shouldn't have been conceived in the first place because it's idiotic even if it was implemented in the most perfect way) that has accidentally slipped in to production by mistake.

I wonder how it is possible that nobody from chromium has commented yet.

> This thing is pathetic

There's no need to be rude. Bugs happen, and we should respect the hard work of the Chromium developers.

This clearly looks like the back/forth gesture on OS X, except that on OS X it's triggered by (I think) a three finger drag to distinguish it from a two finger scroll drag. I would guess the event handlers for that gesture unintentionally got entangled with the event handlers for regular horizontal scroll events on Linux.

I am also experiencing this issue, in Chromium 52.0.2743.116 on NixOS (Linux) 16.03. I've also experienced the same issue for a few months in Google Chrome (stable) on openSUSE 13.1, right up until I switched to NixOS.

In both cases I am using the sidescroll feature of a desktop mouse - the Trust GXT33 (also known under many other names: http://cryto.net/~joepie91/areson/). To my knowledge, this mouse does *not* simulate a trackpad.

The issue was not 100% reproducible for me - it did not work on all websites. This might be due to custom scroll event handlers on those sites, I'm not really sure.