Coverage analysis is typically achieved by hooking the code that looks up methods or functions. A log is then created if which areas have been used, doing this across languages isn’t really possible unless you are prepared to coerce them into the same runtime with something like parrot. I would not recommend this as it will be such an unusual situation that the test is unlikely to be meaningful.
You could make a simple version of this by having a log of calls between the two codebases and comparing it to as list of possible calls, this could be done with a protocol specific spy. I’m not aware of generic tools for this as this is a very unusual requirement.
Tests in a compiled language for code in interpreted language is very strange.