Flaky crash of language/vm/allocation_sinking_vm_test

[mkustermann]

From this log:

/=============================================================================\
| language/vm/allocation_sinking_vm_test broke (Pass -> Crash, expected Pass) |
\=============================================================================/

--- Command "vm" (took 01:52.000158s):
DART_CONFIGURATION=DebugIA32 out/DebugIA32/dart --optimization-counter-threshold=10 --no-use-osr --no-background-compilation --sound-null-safety -Dtest_runner.configuration=dartk-strong-linux-debug-ia32 --ignore-unrecognized-flags --packages=/b/s/w/ir/.dart_tool/package_config.json /b/s/w/ir/tests/language/vm/allocation_sinking_vm_test.dart

exit code:
-6

stderr:
../../runtime/vm/raw_object.cc: 356: error: Invalid cid: 0, obj: 0xf354e6f0, tags: 78. Corrupt heap?
version=2.19.0-edge.e70dec4e82efdf843c76cedbcbaaa29be3777b7d (be) (Mon Oct 24 23:00:47 2022 +0000) on "linux_ia32"
pid=11105, thread=11295, isolate_group=kernel-service(0x5abc7f80), isolate=(nil)((nil))
os=linux, arch=ia32, comp=no, sim=no
isolate_instructions=0, vm_instructions=585855e0
  pc 0x589c81ab fp 0xf6d7dc48 dart::Profiler::DumpStackTrace(void*)+0x9b
  pc 0x58f97d81 fp 0xf6d7dc68 Dart_DumpNativeStackTrace+0x21
  pc 0x585857d6 fp 0xf6d7dc98 dart::Assert::Fail(char const*, ...) const+0x36
  pc 0x589e5e61 fp 0xf6d7df78 dart::UntaggedObject::VisitPointersPredefined(dart::ObjectPointerVisitor*, int)+0xce1
  pc 0x58ac8286 fp 0xf6d7dfd8 out/DebugIA32/dart+0x24a2286
  pc 0x58ac5aeb fp 0xf6d7e038 out/DebugIA32/dart+0x249faeb
  pc 0x58ac92fe fp 0xf6d7e0a8 out/DebugIA32/dart+0x24a32fe
  pc 0x58a91375 fp 0xf6d7e128 dart::ThreadPool::WorkerLoop(dart::ThreadPool::Worker*)+0x205
  pc 0x58a91cee fp 0xf6d7e178 dart::ThreadPool::Worker::Main(unsigned int)+0x14e
  pc 0x589c0ecb fp 0xf6d7e2d8 out/DebugIA32/dart+0x239aecb
  pc 0xf7f903bd fp 0xf6d7e3a8 /lib/i386-linux-gnu/libpthread.so.0+0x63bd
-- End of DumpStackTrace

--- Re-run this test:
python3 tools/test.py -n dartk-strong-linux-debug-ia32 language/vm/allocation_sinking_vm_test

In fact there seem to be many daily flaky crashes on ia32. They seemingly started around Oct 05 - which is when incremental marking support for ia32 landed (see 1bff4be)

@rmacnak-google Could you take a look (possibly revert the change if it takes longer to figure out)?

[mkustermann]

It does seem since 2399265 landed (Oct 27), the ia32 crash rate has reduced.

So let's close this issue for now.

[mkustermann]

Thanks for landing a fix, @rmacnak-google !