How to suppress problems detected by Valgrind with EXPECT_DEBUG_DEATH Google test macro?

[fajlip]

I’ve run to the problem with Google unit testing framework and valgrind interaction. When I use EXPECT_DEBUG_DEATH macro and memcheck tests code with valgrind I get massive log output.

valgrind --error-exitcode=1 --leak-check=full --show-leak-kinds=all ./tests

Output starts with:

[ RUN      ] TestCase.MyTest
==15==
==15== Process terminating with default action of signal 6 (SIGABRT): dumping core
==15==    at 0x4C47B94: __pthread_kill_implementation (in /usr/lib64/libc.so.6)
==15==    by 0x4BF6AED: raise (in /usr/lib64/libc.so.6)
==15==    by 0x4BDF87E: abort (in /usr/lib64/libc.so.6)
==15==    by 0x4BDF79A: __assert_fail_base.cold (in /usr/lib64/libc.so.6)
==15==    by 0x4BEF176: __assert_fail (in /usr/lib64/libc.so.6)

And a lot of stuff like this for every record (there is 690 of them):

59 bytes in 1 blocks are still reachable in loss record 1 of 690

Valgrind does not summarize result as the one with warnings or errors, but as I said it generates tons of output messages.

I found that google test uses clone to spawn child process and if test crashes (in this case on purpose), valgrind thinks that test was not exited correctly. There is a flag GTEST_DEATH_TEST_USE_FORK that should fix this. [Link 1]

Solutions that did not work for me:

Tried to run cmake with -DGTEST_DEATH_TEST_USE_FORK=1, did not have any effect on valgrind output.

I’ve also used GTEST_FLAG_SET(death_test_use_fork, true), which gives error: ‘FLAGS_gtest_death_test_use_fork’ is not a member of ‘testing’; yet when I use GTEST_FLAG_SET(death_test_style, "threadsafe") it compiles – this flag is in the same file and just above death_test_use_fork [Link 1 line 86].. did it just to test that it can be done and compiled.

Valgrind version: 3.21.0

Google test version: release 1.12.1

What else could I please do to make valgrind in peace with EXPECT_DEBUG_DEATH?

Link 1: death_test_use_fork