compile error when initializing descendant struct

compile error when initializing descendant struct

[Dave Bender]

Dear list,
  When I attempt to compile the following code (test2.cpp):


struct A {
  int x;
};

struct B : public A {
  int y;
};

const B b = {1, 2};


g++ gives the following error:

test2.cpp:9: error: scalar object 'b' requires one element in initializer

My goal is to incorporate a large set of structs in a DSO (so that
they reside in the read only section). I don't see why this code
should not compile.

Thanks,
-Dave

RE: compile error when initializing descendant struct

[Young, Michael]

Maybe I'm wrong, but this doesn't appear to be valid C++ to me... you're
using a construct that's normally used for array initialization - I
don't think it's valid for initializing an object (instead, use a ctor).
Object layout is not guaranteed to be the same as an "array" of the data
members (even if the members are the same datatype, except for perhaps
PODs), but even if it the same, I don't believe you can use array
initialization syntax just because the memory layout is the same.

Here's what I think you want to do:

struct A {
  int x;
};

struct B : public A {

  B( int arg_x, int arg_y ) : y( arg_y ) { x = arg_x ; }

  int y;
};

const B b( 1, 2 );

I don't have g++ working on this machine at the moment, but this at
least compiles on another C++ compiler.
Note that you could also give A a ctor that initializes x, and then
initialize the base class in the initialization list of B ctor (instead
of doing the assignment of x in the body of B's ctor).

HTH,
  Mike

-----Original Message-----
From: gcc-help-owner@gcc.gnu.org [mailto:gcc-help-owner@gcc.gnu.org] On
Behalf Of Dave Bender
Sent: Wednesday, April 09, 2008 6:07 PM
To: gcc-help@gcc.gnu.org
Subject: compile error when initializing descendant struct

Dear list,
  When I attempt to compile the following code (test2.cpp):


struct A {
  int x;
};

struct B : public A {
  int y;
};

const B b = {1, 2};


g++ gives the following error:

test2.cpp:9: error: scalar object 'b' requires one element in
initializer

My goal is to incorporate a large set of structs in a DSO (so that they
reside in the read only section). I don't see why this code should not
compile.

Thanks,
-Dave

Re: compile error when initializing descendant struct

[Dave Bender]

Mike,
   In C++, structs can be initialized in such a way: please see
section 5.7 in TC++PL.

 Although your solution is correct, I want to instantiate the structs
in reside in the .rodata section of the DSO,
which cannot be done by using the output of a function call.

Dave

On Wed, Apr 9, 2008 at 6:34 PM, Young, Michael <Michael.Young@paetec.com> wrote:
> Maybe I'm wrong, but this doesn't appear to be valid C++ to me... you're
>  using a construct that's normally used for array initialization - I
>  don't think it's valid for initializing an object (instead, use a ctor).
>  Object layout is not guaranteed to be the same as an "array" of the data
>  members (even if the members are the same datatype, except for perhaps
>  PODs), but even if it the same, I don't believe you can use array
>  initialization syntax just because the memory layout is the same.
>
>  Here's what I think you want to do:
>
>
>  struct A {
>   int x;
>  };
>
>  struct B : public A {
>
>   B( int arg_x, int arg_y ) : y( arg_y ) { x = arg_x ; }
>
>   int y;
>  };
>
>  const B b( 1, 2 );
>
>  I don't have g++ working on this machine at the moment, but this at
>  least compiles on another C++ compiler.
>  Note that you could also give A a ctor that initializes x, and then
>  initialize the base class in the initialization list of B ctor (instead
>  of doing the assignment of x in the body of B's ctor).
>
>  HTH,
>   Mike
>
>
>
>  -----Original Message-----
>  From: gcc-help-owner@gcc.gnu.org [mailto:gcc-help-owner@gcc.gnu.org] On
>  Behalf Of Dave Bender
>  Sent: Wednesday, April 09, 2008 6:07 PM
>  To: gcc-help@gcc.gnu.org
>  Subject: compile error when initializing descendant struct
>
>  Dear list,
>   When I attempt to compile the following code (test2.cpp):
>
>
>  struct A {
>   int x;
>  };
>
>  struct B : public A {
>   int y;
>  };
>
>  const B b = {1, 2};
>
>
>  g++ gives the following error:
>
>  test2.cpp:9: error: scalar object 'b' requires one element in
>  initializer
>
>  My goal is to incorporate a large set of structs in a DSO (so that they
>  reside in the read only section). I don't see why this code should not
>  compile.
>
>  Thanks,
>  -Dave
>

Re: compile error when initializing descendant struct

[corey taylor]

On Wed, Apr 9, 2008 at 6:06 PM, Dave Bender <codehero@gmail.com> wrote:
> struct B : public A {
>  int y;
> };
>
> My goal is to incorporate a large set of structs in a DSO (so that
> they reside in the read only section). I don't see why this code
> should not compile.

Have you considered using composition and then initializing A then y
in your aggregate?

corey

Re: compile error when initializing descendant struct

[Brian Dessent]

Dave Bender wrote:

>  Although your solution is correct, I want to instantiate the structs
> in reside in the .rodata section of the DSO,
> which cannot be done by using the output of a function call.

I don't think it's possible, since having a base class makes struct B
non-POD.  There might be hope in C++0x, where the POD type rules have
been revisited, but that's still a draft and from the status page it's
not implemented in any gcc yet:
<http://gcc.gnu.org/projects/cxx0x.html>.

Brian

Re: compile error when initializing descendant struct

[Dave Bender]

Corey,
  That is a good suggestion and likely what I will end up doing. I
wanted to be able to cast between B and A but the read only storage is
more important.

Dave

On Wed, Apr 9, 2008 at 9:53 PM, corey taylor <corey.taylor@gmail.com> wrote:
> On Wed, Apr 9, 2008 at 6:06 PM, Dave Bender <codehero@gmail.com> wrote:
>  > struct B : public A {
>  >  int y;
>  > };
>  >
>
> > My goal is to incorporate a large set of structs in a DSO (so that
>  > they reside in the read only section). I don't see why this code
>  > should not compile.
>
>  Have you considered using composition and then initializing A then y
>  in your aggregate?
>
>  corey
>