Bug#35311585 merge_large_tests fails with clang in optimized mode

We were testing undefined behaviour, and there are now too many
exceptions (compilers, flags, platforms) for when it is supposed to
work. Just remove the test.

Change-Id: I701043a028738aa5157171f21ec91888f33ccbec
(cherry picked from commit e79855a83162d9bd14ba7ec47f9130a4b8200e98)

Author: Tor Didriksen

unittest/gunit/segfault-t.cc

@@ -142,36 +142,6 @@ TEST_F(FatalSignalDeathTest, CrashOnParallelAbort) {
   EXPECT_TRUE(contains_cached_result);
 }
 
-TEST_F(FatalSignalDeathTest, Segfault) {
-#if defined(_WIN32)
-  int *pint = NULL;
-  /*
-   After upgrading from gtest 1.5 to 1.6 this segfault is no longer
-   caught by handle_fatal_signal(). We get an empty error message from the
-   gtest library instead.
-  */
-  EXPECT_DEATH_IF_SUPPORTED(*pint = 42, "");
-#elif defined(HAVE_ASAN)
-/* gcc 4.8.1 with '-fsanitize=address -O1' */
-/* Newer versions of ASAN give other error message, disable it */
-// EXPECT_DEATH_IF_SUPPORTED(*pint= 42, ".*ASAN:SIGSEGV.*");
-#elif defined(__APPLE__) && defined(__aarch64__) && defined(NDEBUG)
-  // Disable also in non-debug mode on MacOS 11 arm, with -O1 or above, we get
-  // Result: died but not with expected error.
-  // Expected: contains regular expression ".* UTC - mysqld got signal .*"
-  // Actual msg:
-  // We do get: "Trace/BPT trap: 5" but not as part of the matcher input in
-  // EXPECT_DEATH(statement, matcher);
-#elif defined(HANDLE_FATAL_SIGNALS)
-  int *pint = nullptr;
-  /*
-   On most platforms we get SIGSEGV == 11, but SIGBUS == 10 is also possible.
-   And on Mac OsX we can get SIGILL == 4 (but only in optimized mode).
-  */
-  EXPECT_DEATH_IF_SUPPORTED(*pint = 42, ".* UTC - mysqld got signal .*");
-#endif
-}
-
 // Verifies that my_safe_utoa behaves like sprintf(_, "%llu", _)
 TEST(PrintUtilities, Utoa) {
   char buff[22];