tr1/tests/thread is sporadically failing

[StephanTLavavej]

#1627 encountered a sporadic failure in the legacy tr1 test suite. The failure log was:

Build setup steps:
Build steps:
Command: "C:\Program Files (x86)\Microsoft Visual Studio\2019\Preview\VC\Tools\MSVC\14.28.29812\bin\HostX64\x64\cl.EXE" "C:\a\1\s\tests\tr1\tests\thread\test.cpp" "-ID:\build\out\inc" "-IC:\a\1\s\tests\tr1\include" "-IC:\a\1\s\tests\std\include" "/nologo" "/Od" "/W4" "/w14061" "/w14242" "/w14265" "/w14582" "/w14583" "/w14587" "/w14588" "/w14749" "/w14841" "/w14842" "/w15038" "/w15214" "/w15215" "/w15216" "/w15217" "/sdl" "/WX" "/Zc:strictStrings" "/D_ENABLE_STL_INTERNAL_CHECK" "/bigobj" "/FIforce_include.hpp" "/D_ENFORCE_FACET_SPECIALIZATIONS=1" "/D_CRT_SECURE_NO_WARNINGS" "/Za" "/EHsc" "/MD" "/std:c++latest" "/permissive-" "-FeD:\build\tests\tr1\tests\thread\Output\14\thread.exe" "-link" "-LIBPATH:D:\build\out\lib\amd64" "-LIBPATH:C:\Program Files (x86)\Microsoft Visual Studio\2019\Preview\VC\Tools\MSVC\14.28.29812\lib\x64" "/MANIFEST:EMBED"
Exit Code: 0
Standard Output:
--
test.cpp

--

Intellisense response file steps:
Test setup steps:
Test steps:
Test step failed unexpectedly.
Command: "D:\build\tests\tr1\tests\thread\Output\14\thread.exe"
Exit Code: 1
Standard Output:
--
__STDCPP_THREADS__ = <1>

 GOT 0 != 1

 FAIL test 113 at line 336 in C:\a\1\s\tests\tr1\tests\thread\test.cpp: tgt <= STD chrono::system_clock::now() == true

FINISHED testing <thread>

***** 1 erroneous test cases in <thread> *****

***** 113 successful test cases in <thread> *****

#FAILED: <thread>

--

Code:

STL/tests/tr1/tests/thread/test.cpp

Lines 332 to 342 in 47cd703

	{
	STD chrono::system_clock::duration dur = STD chrono::system_clock::duration(100);
	STD chrono::system_clock::time_point tgt = STD chrono::system_clock::now() + dur;
	STD this_thread::sleep_for(dur);
	CHECK_INT(tgt <= STD chrono::system_clock::now(), true);
	dur = -dur;
	tgt = STD chrono::system_clock::now() + dur;
	STD this_thread::sleep_for(dur);
	CHECK_INT(tgt <= STD chrono::system_clock::now(), true);
	}

Sporadic failures are doom. We should either fix this test to behave reliably, or skip it.

[CaseyCarter]

Is this not simply a consequence of #718?

[StephanTLavavej]

Failed again today, build 20210220.2.

[StephanTLavavej]

And again in 20210220.6. 😾

[StephanTLavavej]

And again in 20210222.7. 🙀

[StephanTLavavej]

The test also contains data races / timing assumptions:

STL/tests/tr1/tests/thread/test.cpp

Lines 16 to 20 in b52c379

	#define DELAY_VALUE 2 // milliseconds for delay()
	namespace {
	int called = 0;

STL/tests/tr1/tests/thread/test.cpp

Lines 110 to 113 in b52c379

	void delay() { // pause, then mark that function was called
	STD this_thread::sleep_for(STD chrono::milliseconds(DELAY_VALUE));
	called = 1;
	}

STL/tests/tr1/tests/thread/test.cpp

Lines 136 to 140 in b52c379

	called = 0;
	STD thread thr3(delay);
	CHECK_INT(thr3.joinable(), true);
	thr3.detach();
	CHECK_INT(called, 0);

called is an ordinary int (not atomic, not protected by a lock), and this code assumes that the main thread can't possibly stall for 2 milliseconds between spawning thr3 and verifying that called is still 0.

[frederick-vs-ja]

Is this issue fixed by #4457?

[StephanTLavavej]

This was actually fixed by #2595 deleting the test completely.