These kernels should be added by the clad-feedstock if it detects xeus-cpp. They are added here:

https://github.com/conda-forge/clad-feedstock/blob/d5ca50e56bd1e853b4e9905f7b53c889e09ae95a/recipe/build.sh#L73-L90

Is there something that's not sufficient?

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 81.78%. Comparing base (51aeae3) to head (ec158cc).

@@           Coverage Diff           @@
##             main     #314   +/-   ##
=======================================
  Coverage   81.78%   81.78%           
=======================================
  Files          20       20           
  Lines         950      950           
  Branches       87       87           
=======================================
  Hits          777      777           
  Misses        173      173           

Hi,

Just trying to extrapolate this to the lite use case.

Can clad be built against emscripten -> and then used through -fplugin ?

I am not sure if clang plugins would work with lite (should they ? 🤔 )

Hi,

Just trying to extrapolate this to the lite use case.

Can clad be built against emscripten -> and then used through -fplugin ?

I am not sure if clang plugins would work with lite (should they ? 🤔 )

If dlopen works then they should.

I think we need to hold off on including plugins without defining a plugin API surface, or in the core of the package, as discussed with the Python magic.

We can mark it as complete because @kr-2003 solved it in the clad-feedstock. That’s not a xeus-cpp plugin but a clang plugin which is basically running the kernel with different flags.

Yeah can be closed I think. Thanks @kr-2003