Investigate adding address sanitizer

[fcooper8472]

This seems to be the current best practice for detecting memory errors in C++ code, and it seems very straightforward to use.

AddressSanitizer is a fast memory error detector. It consists of a compiler instrumentation module and a run-time library. The tool can detect the following types of bugs:

Out-of-bounds accesses to heap, stack and globals Use-after-free Use-after-return (runtime flag ASAN_OPTIONS=detect_stack_use_after_return=1) Use-after-scope (clang flag -fsanitize-address-use-after-scope) Double-free, invalid free Memory leaks (experimental) Typical slowdown introduced by AddressSanitizer is 2x.

@martinjrobins what's the modern-CMake-approved method of passing

-O1 -g -fsanitize=address -fno-omit-frame-pointer

to the (clang) compiler, and

-g -fsanitize=address

to the (clang) linker?

[martinjrobins]

have a look at how I did the openmp compile and link flags using target_compile_options and target_link_libraries:

# OpenMP as an optional component for tests
option(Template_USE_OPENMP "Use OpenMP for shared memory parallism" OFF)
if (Template_USE_OPENMP)
    find_package(OpenMP REQUIRED)
    target_compile_options(mytest PRIVATE "${OpenMP_CXX_FLAGS}")
    target_link_libraries(mytest PRIVATE "${OpenMP_EXE_LINKER_FLAGS}")
endif()

You will also need to filter out only the clang compiler as well If you need these flags for both the library and test executables, best way is to add them to mylib as PUBLIC, then they will automatically be transferred to mytest

[fcooper8472]

I have an implementation ready to go, just want your opinion @martinjrobins:

If someone runs cmake -DTemplate_MEMCHECK=ON and they are using gcc rather than clang, would you prefer a CMake WARNING or a CMake FATAL_ERROR?

I think if the user specifically requests functionality it's OK to make it a FATAL_ERROR?

[martinjrobins]

I agree, make it a FATAL_ERROR. If I request a feature and it can't be satisfied, that is an error.

[fcooper8472]

Happy with that changeset @martinjrobins ?

[martinjrobins]

looks good. We might also want to change the way that the coverage option is handled so that it is consistent with the memcheck

[fcooper8472]

Do we just want coverage on GCC, or Clang too?

[martinjrobins]

well, probably want the ability to check coverage on both gcc and clang, but travis only do it for one of them.

On 9 May 2018 at 16:24, Fergus Cooper notifications@github.com wrote:

Do we just want coverage on GCC, or Clang too?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/OxfordRSE/template-project-cpp/issues/8#issuecomment-387777102, or mute the thread https://github.com/notifications/unsubscribe-auth/ABGF9LDTe_mNRnlNwYSQLoj-zXOTMrggks5twwowgaJpZM4T4Lv9 .

[fcooper8472]

OK - I'll close this issue and open a new one for sorting out coverage.