Treat warnings as errors in build of cppjit

[mcbarton]

If you look at the nightly builds (here for example https://github.com/compiler-research/cppjit/actions/runs/32926841974/job/98051372290#step:8:695), you'll see that the single_module is labelled as obsolete, so this PR removes the flag.

[aaronj0]

LGTM, thanks. Can you check if there is anything else thats stale in this MakeFile?

[mcbarton]

@aaronj0 I cannot see anything taking a quick look at the ci. If you update the reusable action here https://github.com/compiler-research/ci-workflows/blob/main/actions/build-and-test-cppjit/action.yml to do the cmake build with CMAKE_COMPILE_WARNING_AS_ERROR=ON you should catch any leftover warnings (if there is any) when building cppjit. To catch them when building the tests you'll want to add CXXFLAGS="-Werror" to the make command (and maybe a linker flag since I don't think this catches linker warnings).

To match Clad, xeus-cpp, CppInterOp I think it would make sense to remove the handwritten makefile and replace it with a cmake file which the main cmake can call, but that is just personal choice. You could also change it so the tests are built with Wall like the cppjit library for some extra checks.

You will need to merge this PR if happy with it, since I don't have permissions to do anything in this repo.

[aaronj0]

@aaronj0 I cannot see anything taking a quick look at the ci. If you update the reusable action here https://github.com/compiler-research/ci-workflows/blob/main/actions/build-and-test-cppjit/action.yml to do the cmake build with CMAKE_COMPILE_WARNING_AS_ERROR=ON you should catch any leftover warnings (if there is any) when building cppjit. To catch them when building the tests you'll want to add CXXFLAGS="-Werror" to the make command (and maybe a linker flag since I don't think this catches linker warnings).

That sounds like a good idea, but we should probably first inject the flag in the CMakeLists.txt and test MakeFile here and see what fails. We can then address the remaining issues as a part of this PR, before making that default in the ci-workflows yml.

To match Clad, xeus-cpp, CppInterOp I think it would make sense to remove the handwritten makefile and replace it with a cmake file which the main cmake can call, but that is just personal choice. You could also change it so the tests are built with Wall like the cppjit library for some extra checks.

Regarding moving to a CMake file, that sounds like a good idea, but perhaps the simplicity of this MakeFile is desired :)
One point is we cannot make building these dictionaries and test-time artifacts part of the higher level scikit-driven CMake, rather driven by the pytest driver when a user runs the Python tests on their system. An improved pytest driver that automates this instead of a user having to run make is a part of my incoming changes referred to in #46

[mcbarton]

I have added Werror and Wextra to the build of cppjit, and Wall, Werror and Wextra to the tests. These have raised some new warnings. I will make my way through them, but I might not start until next week.

[aaronj0]

I have added Werror and Wextra to the build of cppjit, and Wall, Werror and Wextra to the tests. These have raised some new warnings. I will make my way through them, but I might not start until next week.

Sounds good, thank you!

[mcbarton]

@Vipul-Cariappa @vgvassilev @aaronj0 How would you advise me proceeding with this PR now? I worked my way through the warnings outputted by adding -Wextra -Werror to the library build. About 99% of the errors were about PyCFunction and PyObject being incompatible when it comes to casting (I've 90% sure I didn't fix this error in the correct way, and probably the reason tests are failing). There was a few unused variable errors taken care of in another PR recently upstreamed from root, and one about fallthrough in interop_wrapper.cxx