Oddness concerning private variants in 3.10.0

Oddness concerning private variants in 3.10.0

[Chris King]

Hi,

I just took the leap to 3.10 and encountered some strangeness
compiling my project with it:

The implementation input_test.ml
does not match the interface (inferred signature):
Type declarations do not match:
  type button = FrGui.button
is not included in
  type button = FrGui.button#row

First, I don't understand how a module doesn't match its inferred
signature.  Strangely, creating a blank input_test.mli file gets
around this error.

Second, the type button (which is a private variant) isn't even
defined in input_test.ml, it's defined in a module opened by it.

I tried coming up with some minimal test cases and found even more
interesting results:

Compiling this:

module A: sig
  type t = [ `A | `B ]
end = struct
  type t = private [> `A ]
end

results in the error:

Type declarations do not match:
  type t = private [> `A ]
is not included in
  type t = [ `A | `B ]

But compiling this:

type t = [ `A | `B ]

with this .mli file:

type t = private [> `A ]

works fine.  What is going on here?

I haven't yet been able to reproduce the error in my project using a
minimal test case but I will post it when I do.

Thanks,
Chris

Re: [Caml-list] Oddness concerning private variants in 3.10.0

[Jacques GARRIGUE]

From: "Chris King" <colanderman@gmail.com>
> I just took the leap to 3.10 and encountered some strangeness
> compiling my project with it:
> 
> The implementation input_test.ml
> does not match the interface (inferred signature):
> Type declarations do not match:
>   type button = FrGui.button
> is not included in
>   type button = FrGui.button#row
> 
> First, I don't understand how a module doesn't match its inferred
> signature.  Strangely, creating a blank input_test.mli file gets
> around this error.

This looks like a bug report, but I clearly miss information to solve it.

> Compiling this:
> 
> module A: sig
>   type t = [ `A | `B ]
> end = struct
>   type t = private [> `A ]
> end
> 
> results in the error:

It certainly should: here the signature is written first!
If you write it the other way round, it is accepted without problem.

Concerning your first problem, could you submit a bu report, attaching
relevant information?

Jacques Garrigue

Re: [Caml-list] Oddness concerning private variants in 3.10.0

[Chris King]

On 9/7/07, Jacques GARRIGUE <garrigue@math.nagoya-u.ac.jp> wrote:
> From: "Chris King" <colanderman@gmail.com>
> > Compiling this:
> >
> > module A: sig
> >   type t = [ `A | `B ]
> > end = struct
> >   type t = private [> `A ]
> > end
> >
> > results in the error:
>
> It certainly should: here the signature is written first!
> If you write it the other way round, it is accepted without problem.

D'oh!  Braino on my part.  You are right, I did mean the other way around.

> Concerning your first problem, could you submit a bu report, attaching
> relevant information?

Sure... I will try to come up with a smaller program to reproduce it
first though.

- Chris

Re: Oddness concerning private variants in 3.10.0

[Chris King]

On 9/7/07, Chris King <colanderman@gmail.com> wrote:
> Second, the type button (which is a private variant) isn't even
> defined in input_test.ml, it's defined in a module opened by it.

Another braino on my part... input_test.ml does actually include the
module defining type button.

The core of the bug remains however, and a bug report (#4383) has been filed.

- Chris