Nicolás Ojeda Bär

There are some failures in the testsuite that seem wrong; I'll take a look tomorrow.

> I have not investigated what caused this behavior change. I think you are missing an `-alert` flag :) ``` nojebar@PERVERSESHEAF:~/tmp$ ../ocaml/local/bin/ocamlc -c a.ml nojebar@PERVERSESHEAF:~/tmp$ ../ocaml/local/bin/ocamlc -c b.ml nojebar@PERVERSESHEAF:~/tmp$ ../ocaml/local/bin/ocamlc...

I pushed a different (and this time, hopefully correect) fix. The key point as I understand it is that floating alerts `[@@@alert ...]` should only appear at the "top of...

> Do you know if this change was intentional? Looking quickly, I don't think the manual says one way or another whether we should expect alerts to be enabled by...

> I think your new test test shows the behavior of w53 is not quite right yet. Thanks for the quick review! You are right, I was also puzzled as...

I pushed a new version of the PR which following some of the suggestions of @ccasin. Rather than trying to propagate enough information to the typechecking routine to be able...

> Thanks for the changes, I think this came out very well! Thanks for the review. I pushed an additional test checking that warning is correctly triggered across `.cmi`s both...

This is ready to go from my side. In order to respect the protocol, this PR would need an approval from another core dev before it can be merged.

Thanks for the review @Octachron! Am planning to merge the the PR once the CI passes.

What is the status of this issue? Were the performance issues fixed by #11743?