Re: [Caml-list] Bug or useless error message

[Jacques Garrigue <garrigue@math.nagoya-u.ac.jp>]

You should submit a bug report, with all the source files required.
Impossible to tell you anything without more information.

Jacques Garrigue

On 2015/08/27 22:52, Christoph Höger wrote:
> 
> Dear all,
> 
> right now I am unable to produce a working example of that problem, so
> please bear with me.
> 
> Basically, I just patched ppx_deriving_yojson to generate polymorphic
> functions for recursive types, i.e.
> 
> type 'a foo = ...
> and bar = int foo [@@deriving yojson]
> 
> becomes
> 
> let rec foo_of_yojson : 'a. (json -> [`Ok of 'a, `Error of string]) ->
> json -> [`Ok of 'a, `Error of string] = ...
> and bar_of_yojson = ...
> 
> this does, indeed, work in the test cases.
> 
> If I apply this method to my (large) real world module, I get a strange
> error:
> 
> Error: Signature mismatch:
>       ...
>       Values do not match:
>         val typedef_struct_of_yojson :
>           (Yojson.Safe.json -> [ `Error of string | `Ok of 'a ]) ->
>           Yojson.Safe.json ->
>           [ `Error of string | `Ok of 'a typedef_struct ]
>       is not included in
>         val typedef_struct_of_yojson :
>           (Yojson.Safe.json -> [ `Error of string | `Ok of 'a ]) ->
>           Yojson.Safe.json ->
>           [ `Error of string | `Ok of 'a typedef_struct ]
> 
> Syntactically, the types are the same, but obviously they are not equal
> in their internal representation.
> 
> I can workaround this problem by hiding this particular function from
> the interface (yes, it only occurs in one function out of ~300).
> However, this message bugs me: Is something like this expected when
> using explicit generalization?
> 
> regards,
> 
> Christoph