RE: possible parser bug?

RE: possible parser bug?

[Wegher, Jeff]

Your line:

  B(GetA()->f());

looks like a function declaration rather than a constructor, and it's
invalid syntax for a function declaration.

Try B aB(GetA()->f());

Works fine.

Jeff


> -----Original Message-----
> From:	Michael Iber [SMTP:michael.iber@siemens.at]
> Sent:	Friday, November 06, 1998 5:25 AM
> To:	egcs@cygnus.com
> Subject:	possible parser bug?
> 
> I have a problem with the egcs parser. 
> It raises an error message on code that seems syntactically correct to me.
> It seems the egcs parser knows, it is inside a constructor parameter list.
> 
> In a function call parameter list there is no parse error.
> 
> the code:
> 
> class A
> {
>   public:
>     A() {};
>     int f() { return 1; };
> };
> 
> class B
> {
>   public:
>     B(int) {};
> };
>     
> A* GetA() { return new A(); };
> 
> int main()
> {
>   B(GetA()->f());
> }
> 
> the error message:
> 
> test1.cxx: In function `int main()':
> test1.cxx:18: parse error before `->'
> 
> egcs-1.0.3, egcs-1.1 and egcs-ss981101 show the same behaviour.
> 
> mike
> 

Re: possible parser bug?

[Alexandre Oliva]

On Nov  6, 1998, "Wegher, Jeff" <JWegher@neonsoft.com> wrote:

> Your line:

>   B(GetA()->f());

> looks like a function declaration rather than a constructor, and it's
> invalid syntax for a function declaration.

I don't think this is one of the cases of ambiguity.  This can't
possibly be a declaration, so it must be interpreted as an
expression-statement.

-- 
Alexandre Oliva  http://www.dcc.unicamp.br/~oliva  aoliva@{acm.org}
oliva@{dcc.unicamp.br,gnu.org,egcs.cygnus.com,samba.anu.edu.au}
Universidade Estadual de Campinas, SP, Brasil

Re: possible parser bug?

[Martin von Loewis]

> Your line:
> 
>   B(GetA()->f());
> 
> looks like a function declaration rather than a constructor, and it's
> invalid syntax for a function declaration.

Although this explanation is correct with regard to the
implementation, it is not with regard to the standard.

There is no way this could be a function declaration: The return type
is missing, so this is clearly a function-style. Of course, g++
assumes "int" here, which is an error.

In addition, this has a class member access in it, so it can be
clearly disambiguated as an expression, according to
[stmt.ambig]. It's just that g++ does not get it right.

Regards,
Martin

RE: possible parser bug?

[Michael Iber]

>Your line:
>
>  B(GetA()->f());
>
>looks like a function declaration rather than a constructor, and it's
>invalid syntax for a function declaration.
>
>Try B aB(GetA()->f());
>
>Works fine.

It should be a constructor, but I think the example was too short.
I use code like this

  if (flag)
    B(GetA()->f()).Execute();

e.g to display a modal window on screen, depending on some condition.
An instance of B is created only to call Execute() and destroyed
automatically after the call.
There the parse error occurs. It works fine for function calls.

mike

possible parser bug?

[Michael Iber]

I have a problem with the egcs parser. 
It raises an error message on code that seems syntactically correct to me.
It seems the egcs parser knows, it is inside a constructor parameter list. 
In a function call parameter list there is no parse error.

the code:

class A
{
  public:
    A() {};
    int f() { return 1; };
};

class B
{
  public:
    B(int) {};
};
    
A* GetA() { return new A(); };

int main()
{
  B(GetA()->f());
}

the error message:

test1.cxx: In function `int main()':
test1.cxx:18: parse error before `->'

egcs-1.0.3, egcs-1.1 and egcs-ss981101 show the same behaviour.

mike