Nicolás Ojeda Bär
Nicolás Ojeda Bär
Just out of curiosity, there is another approach that could be considered: version the `ocamlformat` executable itself. So you could have `ocamlformat-0.26.1` and also `ocamlformat-0.24.0` installed at the same time....
> It would be reasonable to restrict the check to only git-versioned .ml files. I run into this problem in the past, and I also thought of this solution. However:...
> 3\. A new user-exposed stdlib module `Thread_local_storage` that exports TLS, reused from the previous PRs. Why not `Thread.TLS` ? I find the asymmetry with `Domain.DLS` rather jarring.
This PR (or something very close to it) has already beed proposed in the past several times and rejected (#652, #7252, #724, #9910), mostly due to a strong objection to...
Meta-point: this is not just an optimization because `{a with foo}` is currently guaranteed to make a copy of `a`, and this behaviour cannot be changed without potentially breaking code,...
> It's not valid for mutable records, of course. Yes, sorry, I had this case in mind.
This alert is generated by the compiler; there isn't much Dune can do about it. It may be worth reporting it upstream over at https://github.com/ocaml/ocaml.
Thanks, this looks like a bug indeed!
Thanks @v-gb for the review! I will look at your comments in detail a bit later today, but here are some initial responses: > It'd be slightly better to have...
> I think there's no chance that this can work. Yes, indeed, sorry, I have forgotten some of the details since I first wrote the code :)