• kdevel (36/36) Jan 04 2021 ~~~tsnf.d
• Paul Backus (13/16) Jan 04 2021 If you try to call .toString directly on a const(S), you get the
• kdevel (37/45) Jan 04 2021 Sure, but that's too late. IMHO the compiler should detect
• Jacob Carlborg (4/7) Jan 06 2021 DMD 2.095.0 now reports an error for this.

kdevel <kdevel vogtner.de> writes:

~~~tsnf.d
import std.stdio: writeln;

struct S {
    string s;
    string toString ()
    {
       return __PRETTY_FUNCTION__ ~ `: ` ~ s;
    }

    {
       return __PRETTY_FUNCTION__ ~ `: ` ~ s;
    }

    {
       return __PRETTY_FUNCTION__ ~ `: ` ~ s;
    }
}

void foo (T) (in ref T t) // fancy way of declaring t const
{
    writeln (t);
}

int main ()
{
    S s;
    writeln (s);
    foo (s);
    return 0;
}
~~~

$ dmd tsnf
$ ./tsnf
string tsnf.S.toString():
const(S)("")

expected output: none. The compiler should have rejected the code 

definitions instead. Is this a bug or a bug?

Paul Backus <snarwin gmail.com> writes:

On Tuesday, 5 January 2021 at 02:02:39 UTC, kdevel wrote:
 expected output: none. The compiler should have rejected the 

 both definitions instead. Is this a bug or a bug?

If you try to call .toString directly on a const(S), you get the 
following error:

onlineapp.d(27): Error: onlineapp.S.toString called with argument 
types () const matches both:
onlineapp.d(9):     onlineapp.S.toString() const
and:
onlineapp.d(13):     onlineapp.S.toString() const

The reason it is ignored when using writeln is that writeln uses 
__traits(compiles) to check whether it can call toString on a 
const(S). Since the call does not compile, it falls back to the 
default struct formatting code, which produces the output you saw 
in your example.

kdevel <kdevel vogtner.de> writes:

On Tuesday, 5 January 2021 at 03:36:47 UTC, Paul Backus wrote:
 On Tuesday, 5 January 2021 at 02:02:39 UTC, kdevel wrote:

[...]

 If you try to call .toString directly on a const(S), you get 
 the following error:

 onlineapp.d(27): Error: onlineapp.S.toString called with 
 argument types () const matches both:
 onlineapp.d(9):     onlineapp.S.toString() const
 and:
 onlineapp.d(13):     onlineapp.S.toString() const

Sure, but that's too late. IMHO the compiler should detect 
duplicate
definitions (probably concealing a program error) regardless 
whether
the method is called. There is a difference between calls which 
match
two or more functions and duplicate definitions. Isn't it?

Other compilers for other languages complain loud and clear:

```ddcpp.cc
template<typename T>
struct X {
    int foo ()
    {
       return 0;
    }
    int foo ()
    {
       return 0;
    }
};
```

result from clang++:

ddcpp.cc:7:8: error: class member cannot be redeclared
    int foo ()
        ^
ddcpp.cc:3:8: note: previous declaration is here
    int foo ()
        ^
ddcpp.cc:7:8: error: redefinition of 'foo'
    int foo ()
        ^
ddcpp.cc:3:8: note: previous definition is here
    int foo ()
        ^
2 errors generated.

Jacob Carlborg <doob me.com> writes:

On 2021-01-05 03:02, kdevel wrote:

 expected output: none. The compiler should have rejected the code after 

 instead. Is this a bug or a bug?

DMD 2.095.0 now reports an error for this.

-- 
/Jacob Carlborg