1238730 - Invalidation bug when scrolling

Status: RESOLVED DUPLICATE of bug 1238564

(Core :: Layout, defect)

Description

[Bill McCloskey [inactive unless it's an emergency] (:billm)]

Attachment: bug.png

STR:
1. Visit http://codepen.io/cgarnier/pen/zrzqaM
2. In the HTML pane, click on "View Compiled"
3. Scroll the HTML pane (mousewheel or scollbar)

Actual result:
There are black boxes on the left and right side of the scrolled region. I've attached a screenshot.

Expected result:
Text scrolls normally.

The problem happens to me on Linux and Windows, with and without e10s. So I think it's not caused by APZ, although it could be APZ-related I guess.

Comment 1

[Alice0775 White]

I can reproduce:
https://hg.mozilla.org/mozilla-central/rev/6020a4cb41a77a09484c24a5875bb221714c0e6a
Mozilla/5.0 (Windows NT 6.1; WOW64; rv:46.0) Gecko/20100101 Firefox/46.0 ID:20160111030207

Regression window:
https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=2ec54b38a33da939a8255612999d9e867eb11664&tochange=bb5becd378f40a9be14e4e635d7034f2835fc7b5

Regressed by: Bug 1147673

Comment 2

[Alice0775 White]

FWIW, setting layers.force-active = true seems to fix the problem.

Comment 3

[Markus Stange [:mstange]]

My patch for bug 1238564 seems to fix this bug as well.

Comment 4

[Milan Sreckovic [:milan] (needinfo for best results)]

This is a new regression in 46 - the patch to bug 1238564 looks "large", but I don't know if it's unsafe or not.  Would it be appropriate for an uplift to 46, in the next few days while on Aurora, or all the way to Beta if we wait until next week?  Or is there a smaller patch we can come up with that is appropriate for the uplift?

From comment 0, it isn't clear if this is a low frequency problem, but it still doesn't sound like it's something we should ship with.  Markus, how often do you think we'd hit this?

Comment 5

[Markus Stange [:mstange]]

(In reply to Milan Sreckovic [:milan] from comment #4)
> This is a new regression in 46 - the patch to bug 1238564 looks "large", but
> I don't know if it's unsafe or not.  Would it be appropriate for an uplift
> to 46, in the next few days while on Aurora, or all the way to Beta if we
> wait until next week?

Not really "safe", but I think it's the best option we have.

> Or is there a smaller patch we can come up with that
> is appropriate for the uplift?

I can't think of one, sorry. Basically, the patches in bug 1147673 were incomplete, and the ones in bug 1238564 add the missing parts. The only other option would be to back out bug 1147673 and live with blurry text when APZ is used.

> From comment 0, it isn't clear if this is a low frequency problem, but it
> still doesn't sound like it's something we should ship with.  Markus, how
> often do you think we'd hit this?

Frequently enough that we can't ship it. This bug here is basically a duplicate of bug 1238564, and bug 1246469 is another one.

Comment 6

[Markus Stange [:mstange]]

With that last sentence I was trying to say: The bug occurs frequently enough that two duplicates were filed.

Comment 7

[Liz Henry (:lizzard) (relman/hg->git project)]

Tracking, regression in 46

Comment 8

[Markus Stange [:mstange]]

This was fixed by bug 1238564, I'm marking it as a duplicate. Bug 1238564 has the same tracking flags as this one.