Leo White

This will definitely silently break plenty of programs. I believe that the reasoning for this change is flawed because it assumes that the compiler's view of the language -- that...

I really think that these links should always go to the documentation pages rather than to mli files. I agree that bouncing between interface and implementation is valuable, but the...

This all looks great to me. One thing that isn’t made clear in the description is the linking behaviour of explicit static aliases when `-no-alias-deps` *isn’t* passed to the compiler....

I don’t have any further feedback, and the changes I proposed above are small things that can easily be adjusted after the PR is written.

@damiendoligez Implemented various approaches to aging in the past, which is what your first suggestion is. They always turned out to perform worse than the current approach because the extra...

Even if you do 2 you might as well define the types like: ```ocaml type ('r, 'v) field type 'v loc = Loc : 'r * ('r, 'v) field ->...

> If I remember correctly, it was done this way on flambda backend because they wanted to minimize the diff with the upstream compiler. It is actually done that way...

> I think that makes a lot of sense, I will give your suggestion a try. I would hold off on trying to make that change. I think it will...

I'm not sure why they are in this patch -- probably I just fixed them as I was going past -- but they both look like correct bug fixes to...

I didn’t mean something in the testsuite would fail. I meant that, with some effort, it should be possible to produce an example of misbehaviour that they fix. Feel free...