If someones achieve to build them (or can explain how it should be done) I would be super happy

Thank you everyone

Guillaume

Thanks for your answer

Guillaume

if(WIN32)
    message("windows support limited")
    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DTEST -fPIC" )
    set(PD_CMAKE_PATH "C:/Program Files/Pd") 
    set(EXECUTABLE_NAME mi4pd)
endif()

also, to force cmake to use Unix Makefile generator instead of default MSVC
I added file next to CMakeList.txt called PreLoad.cmake with following:

set(CMAKE_GENERATOR "Unix Makefiles" CACHE INTERNAL "" FORCE)

So,

mkdir build && cd build && cmake .. && make

produces DLLs in ./release/bin/ with some minor warnings, but all audio object causes segmentation fault and crashes Pd straight after enabling DSP. Only grids is usable.

The original hardware modules have different and very varied sample rates. In short, I think clouds uses 32k, rings 48k, braids 96k, etc...

They are great externals, I managed to compile them in Linux after looking for some information and since then I haven't stopped using them.

Does that sound about right? Are samples in the actual hardware, therefore subject to copyright stuff etc..?

Of course can make our own impulse..but just struck (no pun intended) me as a little odd.

I made some changes to the code and built another version:

lmnts~.pd_linux

There do seem to be some subtle variations between the versions' responses to strike_timbre but this could just be my imagination.

Not sure what happened here, perhaps a sample-rate issue? In any case, I have closed the GitHub issue for now. I am still wondering if any other users have experienced this behavior.