06_warnings

Exercise

First part

Edit the implementation of the Sphere class in the cxx/Sphere.cpp file and have a look at the code of the method Sphere::volume().

double Sphere::volume() const
{
  return 4 * M_PI * pow (this->radius_, 3.) / 3.;
}

This is the method which is tested in the unique test of the library, and the test is implemented in the file /cxx/tests/testSphere.cpp.

Let's imagine we want to improve the efficiency of the Sphere::volume() method by pre-computing the value 4 * M_PI / 3

TODO :

• Extract this value into a class data member.
• Run the benchmark. This may show only a very, very minimal improvement.
• Run the test.

The test may still be successful : this may depend on your hardware and compiler.

Let's assume it is successful : do not forget to commit your modifications.

git commit -a -m "precomputation of 4pi/3"
git push

The test should now fail on GitLab CI.

Here is the output you can see on GitLab CI job result console:

$ ctest --output-on-failure
Test project /home/user/gitlabciintroduction/cxx/build
    Start 1: TestVolume
1/1 Test #1: TestVolume .......................***Failed    0.00 sec
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from TestSphere
[ RUN      ] TestSphere.testVolume
/home/user/gitlabciintroduction/cxx/tests/TestSphere.cpp:14: Failure
Value of: 14.137166941154069 == sphere.volume()
  Actual: false
Expected: true
[  FAILED  ] TestSphere.testVolume (0 ms)
[----------] 1 test from TestSphere (0 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (0 ms total)
[  PASSED  ] 0 tests.
[  FAILED  ] 1 test, listed below:
[  FAILED  ] TestSphere.testVolume

 1 FAILED TEST


0% tests passed, 1 tests failed out of 1

Total Test time (real) =   0.00 sec

The following tests FAILED:
	  1 - TestVolume (Failed)
Errors while running CTest

Second part

The -funsafe-math-optimizations implied by -ffast-math allows for reordering of floating points operations and this may lead to different results.

The direct check of equality of floating point numbers with the == operator (cf. tests/TestSphere.cpp file) should be avoided. So we can add a warning (only if understood by the compiler) in order not to reproduce this kind of error.

Since version 4.6, GNU C++ provides this warning -Wfloat-equal. With portability in mind, before adding the flag to the compiler command, we need to check that the compiler accepts it.

A macro add_cxx_compiler_flag is provided in the cxx/cmake/utils.cmake module file.

To load this utils module, you need to modify the root CMakelists.txt file.

• Set the CMAKE_MODULE_PATH to this directory (using CMAKE_CURRENT_SOURCE_DIR ):

set(CMAKE_MODULE_PATH ${CMAKE_CURRENT_SOURCE_DIR}/cmake)

• Load the module utils. Beware: this should be done before using the macro:

include(utils)

TODO :

• In the CMake configuration file, CMakeLists.txt, set the CMAKE_MODULE_PATH and include the utils module.
• With the provided CMake add_cxx_compiler_flag macro, add the compiler warning -Wfloat-equal to the CMake configuration.
• Verify that the flag -Wfloat-equal is used for the compilation.
• Verify that the warnings is printed during compilation.

Home | C++ Home | << C++ Previous - GitLab CI setup | >> C++ Next - GitLab CI pipelines