1123288 - [Mac] Horizontal scrolling triggers navigation (back/fwd) even when the page is scrollable

Status: RESOLVED WORKSFORME

(Firefox :: General, defect)

Description

[(Currently slow to respond) Philipp Sackl [:phlsa] (Firefox UX) please use needinfo]

Since around Jan. 12, using the trackpad to scroll left and right on a page always triggers history navigation. That means it is almost impossible to scroll horizontally on a page.

Expected behavior: Navigation should only be triggered if no more scrolling is possible (i.e. if the page is scrolled all the way to the left or the right).

It works as expected in Developer Edition (37).

Comment 1

[:Gijs (he/him)]

Philipp, do you have time to run a mozregression test? ( http://mozilla.github.io/mozregression/ )

Something like:

mozregression --good 2015-01-01

should find it in only a few (about 5) downloads+runs.

Comment 2

[Markus Stange [:mstange]]

Isn't this the e10s bug? I think we have a bug somewhere about it.

Comment 3

[Markus Stange [:mstange]]

So there's bug 1115532, which was duped to bug 863514. Not sure if that was a good idea.

Comment 4

[Steven Michaud [:smichaud] (Retired)]

I agree this is probably a dup of bug 1115532, and I'm quite sure that's not a dup of bug 863514.

Comment 5

[(Currently slow to respond) Philipp Sackl [:phlsa] (Firefox UX) please use needinfo]

Right, it seems to be an e10s bug.
The first »bad« version is 2015-01-18 as far as I can tell.

Comment 6

[Steven Michaud [:smichaud] (Retired)]

Bug 1115532 has a very different regression range.  See bug 1115532 comment #5 and bug 1115532 comment #9.

Are you sure you didn't get your regression range wrong, Philipp?  Note that you need to ensure e10s mode is on when you run your tests.

Comment 7

[Steven Michaud [:smichaud] (Retired)]

Philipp, you may need to look for the regression range "by hand", and not use the mozregression tool.

Comment 8

[(Currently slow to respond) Philipp Sackl [:phlsa] (Firefox UX) please use needinfo]

What do you mean by "by hand"?
I'm pretty sure I had e10s enabled in all test sessions, but perhaps I made some other mistake...

Comment 9

[Steven Michaud [:smichaud] (Retired)]

> What do you mean by "by hand"?

Download individual m-c nightlies and test them individually.  Make the ones you download (if you don't already have them) fit the pattern of a binary search.

I *always* search for regression ranges "by hand", and keep (and properly label) the individual nightlies that I downloaded.  That's all the fewer I have to download the next time I'm searching for a regression range (for a different bug).

But before you do that, please try this.  Download just the two following m-c nightlies and test with them.  First set browser.tabs.remote.autostart to true in about:config -- these nightlies date to before we turned e10s on by default, and provided a user-visible UI to turn it off and on.

firefox-2014-02-21-03-02-02-mozilla-central
firefox-2014-02-22-03-02-04-mozilla-central

This is the regression range for bug 1115532.  So I expect you'll see your bug with the second of these two nightlies, but not the first one.

Comment 10

[(Currently slow to respond) Philipp Sackl [:phlsa] (Firefox UX) please use needinfo]

I just got around to finally test this, and you're right: the regression happened between those two builds.

Comment 11

[David Lawrence [:dkl]]

Mistakenly filed against Firefox 38 and should be instead 38 Branch. Sorry for the spam. dkl

Comment 12

[Ryan VanderMeulen [:RyanVM]]

Looks like bug 863514 is fixed in 43+. Does this work for you now, Philipp?

Comment 13

[(Currently slow to respond) Philipp Sackl [:phlsa] (Firefox UX) please use needinfo]

Yep, seems to work now!