Vlad Lucaci, Desktop QA (:vlucaci) Reporter
	Description • 6 years ago

[Affected versions]:

- 64.0b7
- 65.0a1 (2018-11-06)
- 63.0

[Affected platforms]:

- Ubuntu 16.04 x64
- Ubuntu 16.04 x86
- Ubuntu 18.04 x64

[Steps to reproduce]:

1. Launch FF with a fresh profile.
2. Set the window to be either full screen or windowed.
3. Restart the browser either via Shift+F2 or via CTRL+Shift+J/CTRL+ALT+R.

[Expected result]:

- The browser is restarted without any corrupted UI elements.

[Actual result]:

- Upon restarting the browser, the right side and footer of the window is corrupted.

[Regression range]:

- https://hg.mozilla.org/mozilla-central/pushloghtml?fromchange=25aad10380b10b6efa50c2b4d97245f078d870a0&tochange=a31334a65a1c75638efae4452ecd271450df2ad0

-Attachments of the issue:
(Video of the issue, Images and about:support information)
https://drive.google.com/drive/folders/1hODl0F2H5zItOuvemY8GaqOanrYHPERP?usp=sharing

[Additional notes]:

- This issue does not reproduce on any other work station except on the following hardware configuration:

Processor: Intel Core i3-5005U CPU @ 2.00GHz x 4
Memory: 4 GB RAM
Graphics: Intel @ HD Graphics 5500 (Broadwell GT2)

-This issue occurs even while in full screen or in windowed mode. Issue can also occur if restarting the browser as a user would. Above methods were used as an example.

	Andrew Osmond [:aosmond] (he/him)
	Comment 1 • 6 years ago

This is happening with the basic compositor, so it is theoretically platform agnostic. OMTP wasn't enabled on Linux by default until 2018 so I think we can rule out that.

Nothing jumps out at me from the regression range as obviously related, so it is probably a second order effect from another change. Maybe one of these?
	e0c53e24ab2a	sotaro — Bug 1391262 - Create TabChild::CreateRemoteLayerManager() r=dvander
	38330dd0d9e5	sotaro — Bug 1391262 - Do not use remote LayerManager when its initialization fails r=dvander 
	d6fdc1d3b070	Timothy Nikkel — Bug 1405397. Part 2. ScrollFrameHelper::BuildDisplayList should only have one way to determine if we are using a displayport/building a async scrollable layer. r=mstange
	bda0ae08740f	Timothy Nikkel — Bug 1405397. Part 1. Only add scrollbars in the "ignore scroll frame" case if we are painting to the window. r=mstange

	Timothy Nikkel (:tnikkel)
	Comment 2 • 6 years ago

If my layout change had an effect (even invisible in the vast majority of cases) I'd expect to have heard about some sort of fallout from it by now.

	Timothy Nikkel (:tnikkel)
	Comment 3 • 6 years ago

(In reply to Timothy Nikkel (:tnikkel) from comment #2)
> If my layout change had an effect (even invisible in the vast majority of
> cases) I'd expect to have heard about some sort of fallout from it by now.

An effect on the default UI that is, not some weird random webpage with special edge cases.

	Sotaro Ikeda [:sotaro]
	Comment 4 • 6 years ago

From symptom, it seems that Widget seems not have correct size and pos info. Bug 1391262 does not change how widget size and pos are calculated. Then Bug 1391262 seems not related to the bug.

I tried to reproduce the problem on Ubuntu 18.04 x64 with Firefox 63. I could reproduce the problem once, but I failed to reproduce the problem since then.

	Sotaro Ikeda [:sotaro]
	Comment 5 • 6 years ago

:vlucaci, does the problem happen with latest nightly?

	Sotaro Ikeda [:sotaro]
	Comment 6 • 6 years ago

From the symptom, I wonder if gfk's nsWindow::mBounds is not updated correctly.

	Sotaro Ikeda [:sotaro]
	Comment 7 • 6 years ago

If nsWindow::OnSizeAllocate() worked as expected, it invalidate the new part of the window.
  https://dxr.mozilla.org/mozilla-central/source/widget/gtk/nsWindow.cpp#2442

	Vlad Lucaci, Desktop QA (:vlucaci) Reporter
	Comment 8 • 6 years ago

Hello, 

To answer the question from Comment 5:

Yes, this issue occurs with the exact same repro steps on the latest Nightly 65.0a1(2018-11-22) and the latest Beta as well 64.0b11.

	Liz Henry (:lizzard) (relman/hg->git project)
	Comment 9 • 5 years ago

Since this is triaged and has a priority set, marking this fix-optional to remove it from regression triage. 
Happy to still take a patch in nightly.