Re: Type inference with String_Literal.

["Randy Brukardt" <randy@rrsoftware.com>]

Well, the rules for conditional expressions are separate from the rules for 
literals, so it certainly is possible for them to get different results.

However, it appears to me that 4.5.7(13/3) applies to your examples on line 
18 (the expression is "expected" to be of type My_Type1) and on line 19 
(here the expression "shall resolve to" My_Type1). If that rule applies, 
then the dependent expressions should be resolved individually using 
My_Type1 - which obviously should work. So I think this is a bug in GNAT, 
although I'm not 100% certain (if the rule doesn't apply, then there isn't 
enough context to resolve the expression -- but I don't know why the rule 
wouldn't apply).

                              Randy.

"Blady" <p.p11@orange.fr> wrote in message 
news:rs6uql$uj1$1@gioia.aioe.org...
> Hello,
>
> In the following test program, line 16 the string literal is correctly 
> inferred to Wide_Wide_String and line 17 the string literal is correctly 
> inferred to My_Type1 as aspect String_Literal is defined line 7 for this 
> type.
> But line 18, GNAT reports a type match error in a "if expression" with 
> string literals.
> String literal in the "if expression" has to qualified by My_Type1 as in 
> line 20:
>
>     1. with Ada.Wide_Wide_Text_IO;
>      2. procedure Test_20201225_str_lit is
>      3.
>      4.    type My_Type1 (Length : Natural) is record
>      5.       Value : Wide_Wide_String (1 .. Length);
>      6.    end record with
>      7.       String_Literal => From_String1;
>      8.    function From_String1 (Value : Wide_Wide_String) return 
> My_Type1 is ((Length => Value'Length, Value => Value));
>      9.    procedure Print1 (Self : My_Type1) is
>     10.    begin
>     11.       Ada.Wide_Wide_Text_IO.Put_Line (Self.Value);
>     12.    end Print1;
>     13.
>     14.    function Test return Boolean is (True);
>     15.
>     16.    S : Wide_Wide_String := (if Test then "test0" else "");
>     17.    MV0 : My_Type1 := "test0";
>     18.    MV1 : My_Type1 := (if Test then "test1" else "");
>                               |
>         >>> expected type "My_Type1" defined at line 4
>         >>> found a string type
>
>     19.    MV2 : My_Type1 := My_Type1'(if Test then "test2" else "");
>                                        |
>         >>> expected type "My_Type1" defined at line 4
>         >>> found a string type
>
>     20.    MV3 : My_Type1 := (if Test then My_Type1'("test3") else "");
>     21.
>     22. begin
>     23.    Print1 ("Test ""string""");
>     24.    Print1 ("88");
>     25.    Print1 (MV1);
>     26.    Print1 (MV2);
>     27. end Test_20201225_str_lit;
>
> Shouldn't GNAT set the type My_Type1 for the string literal in line 18 as 
> it does line 17?
>
> Thanks, Pascal.
>