Closed Bug 1507770 Opened 6 years ago Closed 5 years ago

Enabling wayland and WebRender results in blank window

Categories

(Core :: Widget: Gtk, defect)

65 Branch
defect
Not set
normal

Tracking

()

RESOLVED DUPLICATE of bug 1514156

People

(Reporter: D.J, Unassigned)

Details

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:65.0) Gecko/20100101 Firefox/65.0

Steps to reproduce:

Using up to date Fedora 29:
 - Enable WebRender using gfx.webrender.all=true config option.
 - Start Firefox Nightly (20181116100115) with wayland backend by running GDK_BACKEND=wayland ~/bin/firefox/firefox



Actual results:

The browser starts up but only displays the title bar. The rest of the window is grey. Resizing the window to force a repaint does nothing. The browser appears to be functioning in that Ctrl-Tab flips through the tabs which should be open. Right-clicking the window displays the expected context menu for the area clicked.


Expected results:

Browser chrome and content should be displayed.
Disabling WebRender but keeping wayland enabled allows the browser to function completely. Wayland is functioning according to about:support. Disabling wayland and re-enabling WebRender also allows the browser to function fully.
(In reply to David J. Fiddes from comment #0)
> The browser starts up but only displays the title bar. The rest of the window is grey.

Do you still see a grey window with current Nightly?

For me it is completely transparent except for the title bar, so I am wondering if that's a different bug, maybe depending on the hardware/driver.
Yes. That is how it behaves for me now. I think it changed around the time the fix for bug 1482350 landed. I've not been testing it regularly though. Apologies for not updating the bug when I noticed this.
I started Firefox with a fresh profile several times in a row and I observed that sometimes Firefox started without this problem. So I had the following idea: I booted with the parameter "maxcpus=1" to only enable one core of my dual core CPU. Then I started Firefox again several (~10) times in a row and the window content was always displayed normally. So I think there is a race condition somewhere, but this is just a guess.

David, could you try to confirm my observation, i.e. by booting with "maxcpus=1" and starting Firefox several times in a row?
Just in case that it's not clear, I started Firefox *and* closed it right after a successful start, and repeated this several times in a row.
It seems that bug 1514156 is a duplicate. I will ask for confirmation.
I've tried this with the kernel configured to have maxcpus=1. I didn't observe any difference over 4-5 restarts. Limiting the CPUs like this would only make a startup/initialization race less likely rather than impossible.

David, can you please verify if this bug is fixed for you now that bug 1514156 is fixed?

This looks like it is a straight duplicate of bug #1514156 and should have been marked as such.

I've re-tested and it now appears to function as a browser though it does crash if I open the About box (this appears an unrelated and new bug).

Status: UNCONFIRMED → RESOLVED
Closed: 5 years ago
Resolution: --- → DUPLICATE
You need to log in before you can comment on or make changes to this bug.