Closed Bug 1258426 Opened 8 years ago Closed 8 years ago

Failures to paint when scrolling on Mac nightly

Categories

(Core :: Graphics, defect)

defect
Not set
normal

Tracking

()

RESOLVED DUPLICATE of bug 1257491
Tracking Status
firefox48 --- affected

People

(Reporter: bzbarsky, Unassigned)

References

Details

Since updating to the 2016-03-20 nightly, I fairly commonly see us fail to paint when scrolling.  That is, I will scroll and will just see blank space where the page content should be.  Attempting to drag-select, say, fixes the problem.

I'm not sure what nightly I was on before this; I'm pretty sure it was after 2016-02-22, and looking at my about:crashes may well have been 2016-03-14 or 2016-03-17 (the last times I crashed).  

Have we recently made changes to Mac painting (esp. on e10s) in some way?
Flags: needinfo?(nical.bugzilla)
Summary: Failures to paint when scrolling on Mac nightly` → Failures to paint when scrolling on Mac nightly
Just to clarify, this is different from APZ checkerboarding?
Yes, it is.

I just ran into this on https://github.com/whatwg/html/issues/517#issuecomment-198460384 and it behaved as follows:

1)  I loaded that URI.
2)  I used the touchpad (two-finger drag) to scroll all the way to the top of the page.

Now what my browser showed was an all-white viewport _except_ for the box on the right with "Labels", "Milestone", etc.  The one that's "position: sticky" and has a z-index and hence is probably getting its own layer or something.

I let it sit for about 10 seconds and nothing changed.  As soon as I clicked and dragged in the middle of the viewport (to start a drag-selection), all the other stuff that should have been showing popped in.
Oh, and if I'm compiling at the same time, this happens nearly 100% of the time...  Echoes of the "our IPC message timed out" stuff we were having before.
Can you try setting apz.peek_messages.enabled to false (no restart required) and see if you still run into this problem?
Flags: needinfo?(bzbarsky)
FWIW, I'm seeing this on win10 after updating today as well.
As far as I can tell, that makes the problem go away; I didn't see in the several hours of browsing I did after flipping the pref.  I still see the blankness for a bit (a second or two at most) if the system is under load, but then stuff paints in.  I just tried flipping the pref back to true and reproduced almost immediately with my STR while compiling.  Flipped it to false again and have failed to reproduce again after trying a number of times (still while compiling).
Flags: needinfo?(bzbarsky)
Looks like that pref was added in bug 1242609.

So it's possible that this is a dup of bug 1257491...
Depends on: 1257491
Yeah I strongly suspect it is a dupe. I'll land the patch on that bug and we'll see if this goes away in the next nightly.
Flags: needinfo?(nical.bugzilla)
Optimistically duping, but please reopen if you still see this issue. Note that bug 1257491 missed the March 22 nightly, it should be in the March 23 nightly.
Status: NEW → RESOLVED
Closed: 8 years ago
No longer depends on: 1257491
Resolution: --- → DUPLICATE
You need to log in before you can comment on or make changes to this bug.