Re: [Caml-list] typechecking in 4.05+beta2

[Gabriel Scherer <gabriel.scherer@gmail.com>]

Thanks for the bisection work. The related code change reproduces, at
the level of functors, a change in handling of non-generalized
inference variables that was done at the module level in 4.04, and for
which there were a couple similar known regressions. I believe that
the conclusion would be the same here : it is unfortunate that your
code breaks, but we don't expect the new behavior of 4.05 (which plugs
a small soundness hole in the type system) to change to become more
permissive again.

( I posted in essence the same comment on the relevant pull request
discussion: https://github.com/ocaml/ocaml/pull/929#issuecomment-282283191
)

On Fri, Feb 24, 2017 at 3:47 AM, Virgile Prevosto
<virgile.prevosto@m4x.org> wrote:
> Dear list,
>
> 2017-02-22 19:54 GMT+01:00 Virgile Prevosto <virgile.prevosto@m4x.org>:
>> while playing with the new beta, I stumbled upon the following issue.
>> The program below is now rejected by the compiler, while this was not
>> the case with 4.04.
>> [snip]
>
> I've bisected the issue and it is apparently caused by commit
> 2bd22967b that fixes mantis issue 7414, which is about weak type
> variables and first-class modules. I don't know the type-checker well
> enough to understand whether it is possible to accept my example while
> rejecting the one from MPR#7414.
>
> Best regards,
> --
> E tutto per oggi, a la prossima volta
> Virgile
>
> --
> Caml-list mailing list.  Subscription management and archives:
> https://sympa.inria.fr/sympa/arc/caml-list
> Beginner's list: http://groups.yahoo.com/group/ocaml_beginners
> Bug reports: http://caml.inria.fr/bin/caml-bugs