Paul Fultz II

> This does not work for tests. How so? > The tests are not libraries, but bcm_tests. There is no library involved. A `bcm_test` is both a cmake test and...

> My entire point is that I do not want to have to repeat the exact same requirement for every bloody test. Yes, thats why you use `bcm_test_link_libraries` so you...

I am not sure how that helps. For example, in `date_time` the definition of `BOOST_DATE_TIME_POSIX_TIME_STD_CONFIG` is added to every test using `bcm_test_link_libraries` like this: ```cmake add_library(data_time_tests_properties INTERFACE) target_compile_definitions(data_time_tests_properties INTERFACE BOOST_DATE_TIME_POSIX_TIME_STD_CONFIG)...

> What about setting the same properties for the directory itself using set_directory_properties ? But these properties dont set the properties on the target themselves, which can cause some problems(like...

> Are you saying that setting directory properties do not set the appropriate properties when creating targets from the files in that directory ( and sub-directories ) ? Nope, when...

I think its a lot cleaner to use `values()` instead of `mImpl->mValues`. However, `!mImp->mValues` is not the equivalent of `!values.empty()`. If we use `values()` then the checks for null can...

This seems to make the function a lot more complicated. How much actual time improvement in terms of seconds do we get for this? We should just be using NRVO...

Gcc 5.1 has lots of problems/regressions. It's best to declare it unsupported, and use gcc 5.2.

> Alternate approach to find clang path directly (independent of hip install and relative path) is addressed in https://github.com/ROCm-Developer-Tools/HIP/pull/2964. That is the wrong approach. The hip clang path should be...

> Why would it be beneficial to search for it relative to the compiler These are private headers used by the compiler. There is no guarantee of compatibility between different...