834437 - [layers-refactoring] Make layers refactor not regress android

Status: RESOLVED FIXED

(Core :: Graphics: Layers, defect)

Comment 1

[Nick Cameron [:nrc]]

So I think this mostly works. Some artefacts to hunt down, but I think we'll leave that until we finish more of the big things because they will only get regressed again, or if we're lucky fixed by accident.

Comment 2

[Jeff Muizelaar [:jrmuizel]]

https://tbpl.mozilla.org/?tree=Try&rev=f2d4523dced8

Comment 3

[Nick Cameron [:nrc]]

I can repro the bad result in M8, but not that crash in M2. M1 and R2 look like webgl, which is busted for now. I'll try to fix M8 for now.

Comment 4

[Jeff Muizelaar [:jrmuizel]]

I fixed M8:
https://tbpl.mozilla.org/?tree=Try&rev=c851f7919ff3

Comment 5

[Nick Cameron [:nrc]]

(In reply to Jeff Muizelaar [:jrmuizel] from comment #4)
> I fixed M8:
> https://tbpl.mozilla.org/?tree=Try&rev=c851f7919ff3

The R1 break is aa on mask layers. I have a lot of bitterness with such bugs. We haven't changed the mask layers code as far as I know, other than moving it, so I don't see how this could happen. Probably we should fuzz it.

The R3 fails look like a tile misloading or something. I'll look into that.

Worryingly, jrmuizel is still reporting flickering and rendering glitches. That R3 test seems to be the only evidence in our tests, that is a bit scary.

Comment 6

[Jeff Muizelaar [:jrmuizel]]

We have landed