Can you clarify the motivation? I don't think Matz asked for it https://bugs.ruby-lang.org/issues/20182#note-5:

If the user intentionally redefines the fundamental method, it should be his/her own responsibility.

Integer#< and Integer#succ are two things that are already optimized well on the interpreter and YJIT, and replacing them with a random cexpr! would only complicate YJIT so much to maintain the current performance. The redefinition concern seems much less meaningful than Array#each's race condition one, unless it's something like Integer#/.

I'm fine with replacing block_yield? with defined?(yield) since it's not Primitive though.

This was just an experiment, but I found that builtin-indexes were not working with while.
#9533 worked because it moved i.succ into ary_fetch_next by chance.
In that sense, this might be better to focus that bug and be earlier 2 commits only.

This was just an experiment, but I found that builtin-indexes were not working with while.
In that sense, this might be better to focus that bug and be earlier 2 commits only.

It'd be nice to have those builtin index fixes 👍

It'd be nice to have those builtin index fixes 👍

Are you OK with the restriction, cexpr! must be up to one per line?

Yeah, I'm fine with it as long as it gives an error on conflict. RJIT is an exception, but it'd usually mean that you should merge such primitives.