bpo-36876: Moved Parser/listnode.c statics to interpreter state.

[vsajip]

https://bugs.python.org/issue36876

[ericsnowcurrently]

mostly LGTM

I have left just a few small suggestions to consider.

[ericsnowcurrently]

If you think there will be any other parser/compiler-related state to move to PyInterpreterState then consider making this struct stand-alone and perhaps even creating pycore_compiler.h (or pycore_parser.h) for it. That way the code doesn't get too cluttered.

[ericsnowcurrently]

I was going to point to the advantage of having stand-alone structs for places like this (so there isn't so much going on with each line). However, I actually kind of like how the line captures the indirection clearly. :)

[vstinner]

interp = _PyInterpreterState_GET_UNSAFE(); 

You can add listnode or parser variable to avoid having to write "interp->parser.listnode". But for that, you need to name the structure in pycore_pystate.h.

[vsajip]

I'm not quite sure what you mean. The change was specifically to move those variables into the interpreter state rather than stand-alone variables, as they were previously.

[vstinner]

I suggest to add parser = &interp->parser; and then use parser instead of interp. Or something similar with listnode.

[vsajip]

Ah, OK, I understand now :-)

Not sure what an additional variable buys in this case. I agree, if readability were adversely affected, it would be worth doing. As it is, it seems readable enough.

[vstinner]

It was just a minor remark for readability. You're free to ignore it ;-)