gaaclarke

Results 267 comments of gaaclarke

> Oh, is the render pass that's being eliminated not what the second blur kernel pass writes to? The thing that the second blur pass was rendering to was typically...

The following patch seems to render things correctish. Here is the render versus the golden before the change. I have no idea where this 2.0 factor is coming from though....

I chatted with @bdero. The fudge factor is the downscale amount. I wasn't thinking. The inline blur pass will now have to be scaled to match the destination size we...

That fudge scalar isn't directly the downsample scalar. I tried that and it doesn't work. It makes sense we have to blur more when drawing directly in the parent pass...

Here are the measurements before and after with the 2.0 fudge factor that seems to render things correctly: |before|after| |-|-| ||| Given the fact that we have to scale the...

Note: [31b7bdc](https://github.com/flutter/engine/pull/55092/commits/31b7bdcfbc607cb78ced9f683a855a2f82ab2af6) is where this was merged last.

It can be reproduced locally by just running the one test: ``` $ ./flutter_desktop_darwin_unittests --gtest_filter="FlutterViewControllerMockEngineTest.TestTrackpadGesturesAreSentToFramework" [INFO:flutter/testing/test_timeout_listener.cc(76)] Test timeout of 300 seconds per test case will be enforced. Note: Google Test...

Here's the stack trace for the crash: ``` (lldb) bt * thread #1, queue = 'com.apple.main-thread', stop reason = EXC_BREAKPOINT (code=1, subcode=0x18c0915a4) * frame #0: 0x000000018c0915a4 libobjc.A.dylib`object_getClass + 48 frame...

Here is invocation in the crash: ``` (lldb) up frame #2: 0x0000000100b99624 flutter_desktop_darwin_unittests`-[OCMInvocationStub invokeArgActionsForInvocation:](self=0x00006000023e5780, _cmd=, anInvocation=0x00006000036d4fc0) at OCMInvocationStub.m:87:27 [opt] 84 continue; 85 86 if([recordedArg isKindOfClass:[NSValue class]]) -> 87 recordedArg =...

I reverted the change to run_tests.py since there appears to be some other issues with that test runner, not just the one that I fixed. I still think this is...