Gather demangled stack traces and report the same to console on crashes.

[chinmaygarde]

These should only be used on host binaries for more detailed crash reports.
Installing the handler on targets (iOS/Android) may cause use to break existing
crash reporting mechanisms users may have installed themselves in the process.

This should work on Darwin & Linux for now.

Doing something like int* a = nullptr; *a = 12; or abort or tripping an
assertion should print something the following before program termination. We
can tweak the report further if necessary.

[ERROR:flutter/fml/backtrace.cc(110)] Caught signal SIGSEGV during program execution.
Frame 0: 0x10658342c void testing::internal::HandleSehExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*)
Frame 1: 0x106555070 void testing::internal::HandleExceptionsInMethodIfSupported<testing::Test, void>(testing::Test*, void (testing::Test::*)(), char const*)
Frame 2: 0x106554f81 testing::Test::Run()
Frame 3: 0x106555dc3 testing::TestInfo::Run()
Frame 4: 0x1065570a1 testing::TestSuite::Run()
Frame 5: 0x106562a55 testing::internal::UnitTestImpl::RunAllTests()
Frame 6: 0x10658c22c bool testing::internal::HandleSehExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool>(testing::internal::UnitTestImpl*, bool (testing::internal::UnitTestImpl::*)(), char const*)
Frame 7: 0x1065625c3 bool testing::internal::HandleExceptionsInMethodIfSupported<testing::internal::UnitTestImpl, bool>(testing::internal::UnitTestImpl*, bool (testing::internal::UnitTestImpl::*)(), char const*)
Frame 8: 0x106562445 testing::UnitTest::Run()
Frame 9: 0x105c8dc33 RUN_ALL_TESTS()
Frame 10: 0x105c8dbe6 main
Frame 11: 0x7fff7c2dc3d5 start

Known issue: This routines that generate the stack trace are not signal safe.
But since we only use the same before the process is terminating, this ought to
be fine. I’ll work in a separate patch to convert all the internals to be signal
safe. In the meantime, this will help us better identify the causes of flakes on
our bots.

Fixes flutter/flutter#50244

[dnfield]

Lgtm

Do we have a big tracking making this signal safe?

[dnfield]

Have you tested this with flutter test --local-engine and hitting Ctrl + C while a test is running?

[dnfield]

Just manually tested it, seems fine.

[chinmaygarde]

I tested the following:

• Ctrl-C from the shell.
• kill -9 the process.
• Tripping a gtest assertions (it should NOT print a backtrace).

Do we have a big tracking making this signal safe?

No. I was waiting for approval on this in case folks objected to landing this as is with the caveat.

[dnfield]

I think it's worth doing even if it doesn't cover all scenarios safely - if it makes things worse we can revert it, but it seems like it wouldn't make things worse for users. And it would significantly help when CI tests fail, or if flutter_tester crashes with a segfault (Which can be very mysterious when it happens).

[chinmaygarde]

xref flutter/flutter#50993

[fluttergithubbot]

This pull request is not suitable for automatic merging in its current state.

• The status or check suite build_and_test_linux_unopt_debug has failed. Please fix the issues identified (or deflake) before re-applying this label.

[dnfield]

It seems like this is causing the framework tests to run more slowly and timeout. Not sure why.

[chinmaygarde]

It seems like this is causing the framework tests to run more slowly and timeout. Not sure why.

Unlikely to continue to investigate why.

[chinmaygarde]

As discussed with @dnfield. Enabling only on unit-test targets for now.

[chinmaygarde]

Done. PTAL.

[dnfield]

Still LGTM

[chinmaygarde]

LUCI is red because of an infra issue and the previously failing framework tests are fine. Landing.