Vincent Laviron

After a bit more investigating, I believe that we cannot cause the compiler to produce wrong code, but the reasoning is subtle and I think it would help to have...

@gasche I think you're missing my point. You mention in your message substitutions and freshening, but I think you're not noticing that not only are we using the same code...

Note for reviewers: this is about `OCAMLRUNPARAM`, passing runtime parameters to a program, and not `OCAMLPARAM`. which deals with compiler flags.

In OCaml, it's fine. Even though this `setfield` is never executed, the handler for `exit` also does `setfield` so in the end everything is fine. (This happens to break the...

Why wouldn't this be valid ? It's similar to `1 + raise Not_found`, the `1 +` part is dead code but it still works as intended.

> @lthls right, it should work, but this is an Lstaticraise, so in general that's exactly one of the goals of simplif right? I know that it is semantically valid,...

In `Asmpackager.package_files`, there is a piece of code like that: ```ocaml let cmi = Unit_info.(companion_cmi @@ Artifact.from_filename targetcmx) in let obj = Unit_info.companion_obj cmi in ``` I think this should...

There's no `-for-pack` involved ? I'm very much convinced that we shouldn't support the sequence of compilation commands shown above. I'll try to look at ocamlbuild's code and see if...

Yes, I had already looked at the new commit and I plan to redo a review (I've tried to find a way to mark my previous review as stale, the...

Could you add a comment on how this compares to #12678 ?