1626080 - Clicking area is not aligned properly on the button when touch simulation is activated in RDM mode

Status: RESOLVED DUPLICATE of bug 1556556

(DevTools :: Responsive Design Mode, defect, P2)

Description

[Clara Guerrero ( Need Info Brindusa Tot Please)]

Steps to reproduce
This happens in Mac OS
1-visit aliexpress.com
2- open RDM
3- start a search (any ite)
4- click on the filters at the left of the page
5- The clicking area is not aligned properly on the button
6- go to login page
7- "show all/hide" buttons, and "quick access with" section are not working properly.

Comment 1

[Clara Guerrero ( Need Info Brindusa Tot Please)]

Attachment: click.MOV

Comment 2

[Clara Guerrero ( Need Info Brindusa Tot Please)]

Attachment: cick 2.MOV

Comment 3

[Brad Werth [:bradwerth]]

I can replicate with a clean profile. Thanks for finding this.

Comment 4

[Brad Werth [:bradwerth]]

As I work on this, I notice a similar effect to Bug 1619402, where with Touch Simulation / Meta Viewport on, clicking on a checkbox also causes the page to scroll. The first part of the patch will attempt to address that issue. It's possible that Bug 1619402 should be marked as a blocker here, but for now I'll just make it a See Also.

Comment 5

[Brad Werth [:bradwerth]]

For me, this is also manifest the offset tooltips in Bug 1342555. There's a lot going on here.

Comment 6

[Brad Werth [:bradwerth]]

Revised Steps to Reproduce:

1. Navigate to https://www.aliexpress.com/wholesale?catId=0&initiative_id=SB_20200403134306&SearchText=bracelet
2. Open the devtools Inspector, and turn on element picker. This is not required to get the problem to occur, but it makes it very easy to confirm that the problem is occurring.
3. Enter Responsive Design Mode; I'm using an 800 x 600 window. Ensure that Touch Simulation is on.
4. Close RDM and re-open RDM. The bug only appears consistently if RDM opens with Touch Simulation is on, though you can sometimes get it to occur by toggling Touch Simulation off and back on again.
5. Confirm that the cursor is offset by moving the cursor over elements on the page, which will show the highlighters for elements above and to the left the element under the cursor. This is easiest to do with the checkbox items on the left sidebar.

The originally reported bug is now occurring. If you now follow these steps...

6. Click in the margin area of the RDM pane -- a light grey color.

The highlighters now appear correctly. The cursor is no longer offset. To experience the bug again, you have to return to Step 4.

As I tried to debug this, I tried to build an offline testcase by making a complete copy of the page locally. That page does not replicate the bug. That indicates to me that this is triggered somehow by JavaScript running after the page is loaded. That's a similar case to Bug 1598145, though the patches I've built for that bug do not resolve this bug.

I'll keep working on it. Needinfo'ing Botond in case Step 6 raises alarm bells. What in APZ or layer transforms is reset by a click outside the viewport area, and what might cause it to get screwed up in the first place (probably by script)?

Comment 7

[Botond Ballo [:botond]]

The reason clicking fixes many problems is that it forces a repaint, which has the effect of synchronizing scrolling state between the main thread and the compositor (among other things).

As for what might cause things to get screwed up in the first place -- I'm not sure off the top of my head, but bug 1556556 fixes many hit testing issues involving scenarios where there is a resolution, including bug 1619402, so a good first step would be to see if it fixes this bug as well.

Comment 8

[Botond Ballo [:botond]]

(In reply to Botond Ballo [:botond] from comment #7)

bug 1556556 fixes many hit testing issues involving scenarios where there is a resolution, including bug 1619402, so a good first step would be to see if it fixes this bug as well.

I can't reproduce this bug on Linux -- possibly it's specific to HiDPI or something like that -- and I don't have access to a Mac.

If you need a Mac build with the bug 1556556 patches applied, here is one (from this push).

Comment 9

[Brad Werth [:bradwerth]]

(In reply to Botond Ballo [:botond] from comment #7)
bug 1556556 fixes many hit testing issues involving scenarios where there is a resolution, including bug 1619402, so a good first step would be to see if it fixes this bug as well.

Thank you, Botond. The artifact build you supplied in comment 8 worked for me and demonstrated correct behavior; I'm also updating my local build to double-check.

Comment 10

[Brad Werth [:bradwerth]]

(In reply to Brad Werth [:bradwerth] from comment #9)

Thank you, Botond. The artifact build you supplied in comment 8 worked for me and demonstrated correct behavior; I'm also updating my local build to double-check.

After solving some rebasing issues, I've confirmed that my locally-built Nightly with patches for Bug 1556556 applied no longer exhibits this bug. Wonderful!