问题
gcov complains about one of my algorithms:
File 'Algorithm.h'
Lines executed:95.00% of 20
Algorithm.h:creating 'Algorithm.h.gcov'
17: 25:inline std::vector<std::string> starts_with(const std::vector<std::string>& input, const std::string& startsWith)
-: 26:{
17: 27: std::vector<std::string> output;
17: 28: std::remove_copy_if(input.begin(), input.end(), std::back_inserter(output), !boost::bind(&boost::starts_with<std::string,std::string>, _1, startsWith));
#####: 29: return output;
-: 30:}
My test looks like this, and it passes:
TEST (TestAlgorithm, starts_with)
{
std::vector<std::string> input = boost::assign::list_of("1")("2")("22")("33")("222");
EXPECT_TRUE(starts_with(input,"22") == boost::assign::list_of("22")("222"));
}
What might the problem be? I'm not using optimization.
UPDATE:
My CMakeList.txt contains:
if(CMAKE_COMPILER_IS_GNUCXX)
set(CMAKE_CXX_FLAGS "-O0") ## Optimize
endif()
回答1:
Try using the -fno-elide-constructors
switch in g++
From The Definitive Guide to GCC:
-fno-elide-constructors: This option when compiling C++ options causes GCC not to omit creating temporary objects when initializing objects of the same type, as permitted by the C++ standard. Specifying this option causes GCC to explicitly call the copy constructor in all cases.
Some discussions here: How can I get more accurate results from gcov? and here: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12076
来源:https://stackoverflow.com/questions/13624819/gcov-reporting-that-return-statement-is-not-hit