> (I think it would be good if a C expert compared it to the *current* C code.)

That would be very beneficial but is quite a big task, specially given that codeop design is quite surprising. I personally find quite confusing the way codeop is handling the decisions on continuing or not (is based on compiling the code with different levels of new lines and then checking the repr() of the exceptions). It would be good if someone familiar with "codeop" design can take a look on what's the best way to proceed.