gaaclarke

I tried to implement what was written in skia's paper but it didn't work. I looked over the transcription a dozen times and I just suspect there is an error...

@domesticmouse, yep! It's on beta.

I'm stepping through this with a debugger. - The assert is failing not because `android_surface_` is null, but because `android_surface_->IsValid()` is returning false. - AndroidSurfaceGLImpeller is being built - AndroidSurfaceGLImpeller...

We looked at the device capabilities of the emulator and it looks like, for vulkan, multisample sizes of 1,4 show up for the emulator: http://vulkan.gpuinfo.org/displayreport.php?id=20115#properties (see `framebufferColorSampleCounts`) That means when...

@jmagman I'd have to play around with it but I'd check what happens when a UIView is placed on top of the FlutterView without a UIViewController, that might still be...

Another thought, an easy/correct fix might be to move the touches* methods to the FlutterView from FlutterViewController. Then it may interact more like we are expecting.

Yep, moving the touch handlers to the UIView fixes the issue. I'll test it a bit more and think about it a bit more then put it up for review...

Moving the touch events to the FlutterView worked at fixing the problem, but it broke gesture recognizers for platform views. I've been out for a couple of weeks and haven't...

> LGTM, but agree with `expect_death` if able. I'll give that another try today.

Yea looks like EXPECT_DEATH is unreliable in processes with more than one thread: ``` [ RUN ] ValidationTest.IsFatal [WARNING] ../../flutter/third_party/googletest/googletest/src/gtest-death-test.cc:1102:: Death tests use fork(), which is unsafe particularly in a...