Hi @Vizonex, can you check this PR?

Looks like using uv_pipe instead of uv_socketpair makes the bad fds disappear. In any case, test_dealloc passes this way, and also the bad fds are gone for test_context and test_signals (but these latter tests still have other issues).

Also, uv_pipe looks like a reasonable alternative for AF_UNIX sockets on Windows.

I'll let this one through let me make sure my version is up to date with this one by testing this on my computer

Feel free to let me know when I can release 0.1.4

Sure @Vizonex you can release 0.1.4 now. More updates will follow, but it's good to have this version tested already.
Make sure to include the Python 3.8 version and compile everything with cython==0.29.37 for best results.

Managed to fix the publish workflow by adding the use of submodules, v0.1.4 is now on PyPI.