add support for FILE_SET HEADERS

[ClausKlein]

to support new CMake feature for interface-libraries

this would fix #32

[ClausKlein]

The use case is that a header only project may check that every header is self contented and compiles.

Too it is possible to run-clang-tidy for each header, because CMake generate an object target for each header file in this set.

see https://discourse.cmake.org/t/verifying-header-sets-not-supported-for-header-only-interface-libraries/6139/8

and https://github.com/ClausKlein/asio/pull/3/commits/7701d95f38a09b196670fd70ef2e7d98e864c513

[TheLartians]

So if I understand correctly from the docs, the advantage is that we don't need to specify the include destination explicitly but can now use the BASE_DIRS flag in the library definition? How about adding a test that relies on this feature?

[ClausKlein]

So if I understand correctly from the docs, the advantage is that we don't need to specify the include destination explicitly but can now use the BASE_DIRS flag in the library definition? How about adding a test that relies on this feature?

No, the main advantages are

• for an interface library you may check, that every header is self-contained
• you may also check them with run-clang-tidy without writing tests.

see too https://github.com/aminya/project_options/issues/143 and CMAKE_VERIFY_INTERFACE_HEADER_SETS

[TheLartians]

I still don't really understand the feature then, do you think you could add an independent test case that would fail with the current version, but pass using the suggested changes?

[ClausKlein]

I still don't really understand the feature then, do you think you could add an independent test case that would fail with the current version, but pass using the suggested changes?

@TheLartians I push a version that is not usable with CMake v3.24.2 (pip install cmake)

How will you support to install a header only library using this new CMake feature?

[ClausKlein]

@TheLartians may you think about this MR again?

[ClausKlein]

@TheLartians Q: Is the project not maintained anymore?

[TheLartians]

Hey @ClausKlein I apologise for not being able to respond so far. I'm still planning to maintain this project, however I don't have any experience with FILE_SET HEADERS and don't yet understand what problem it solves. From a quick look at the readme it allows omitting the target_include_directories?

As before it would be great if you could avoid changing the test behaviour based on the CMake version. You can assume that tests will be performed with CMake > 3.23.0, but I still want to keep the existing test code that doesn't use the new feature. Perhaps you instead create another dependency test/file_set_dependency/CMakeLists.txt using file set instead?

[ClausKlein]

FYI: from import-cmake-c20-modules

Once you have built a compiler that supports p1689 and a version of CMake that supports C++20 modules and have run the tests to make sure things are set up correctly. To use modules in CMake you need to use target_sources and FILE_SETS. https://cmake.org/cmake/help/latest/command/target_sources.html

[TheLartians]

I see, so the files sets are required to support modules? In that case how about adding a separate test case that demonstrates this functionality by using modules instead of modifying the existing tests?

[ClausKlein]

I can not longer use this cmake module because it has no support for moderern CMake file sets!