> The inexorable conclusion is there is a fundamental bug in the structure of > the language which causes this error to occur.
Are you sure about this? I found at least one difference between the C++ and the Nim implementation (the first loop in `next_pinit` starts with 1 in C++ and and with 0 in Nim. While this doesn't seem to change the behavior, it means that you can't rule out that there are other subtle differences. (I only browsed the code briefly, I didn't check it for equivalence.) It is of course entirely possible that there's a compiler bug (there've been others before, after all), but I'd like somewhat stronger evidence that this isn't just a subtle semantic difference between the two implementations before going down that rabbit hole.
