Foo<int,(char*)&string>

Foo<int,(char*)&string>

[Michael Elizabeth Chastain]

Here's a test program, derived from gdb.cp/templates.cc:

  #include <iostream>
  #include <string.h>

  template <class T, char * name>
  class Foo
  {
    public:
      int method ();
  };

  template <class T, char * name> int Foo<T, name>::method ()
  {
    return strlen (name);
  }

  char string [10] = "hello";

  int main ()
  {
    Foo <int, string> f1;
    std::cout << f1.method () << std::endl;
    return 0;
  }

The second template parameter is a "char *"; a value, not a type.

gcc 3.3.2 -gstabs+ emits this:

  .stabs "Foo<int,&string>:Tt(1,1)=..."

gcc HEAD -gstabs+ just changed to emit this:

  .stabs "Foo<int,(char*)(&string)>:Tt(0,25)=..."

Do we care?  I think the "(char*)(&string)" looks ugly but is legal.
I want to change templates.exp to accept this as a PASS.  I've found
enough serious stabs+ bugs with gcc that I don't want to make a fuss
over this difference.

What do you think?

Michael C

Re: Foo<int,(char*)&string>

[Daniel Jacobowitz]

On Mon, Feb 16, 2004 at 09:20:08PM -0500, Michael Chastain wrote:
> The second template parameter is a "char *"; a value, not a type.
> 
> gcc 3.3.2 -gstabs+ emits this:
> 
>   .stabs "Foo<int,&string>:Tt(1,1)=..."
> 
> gcc HEAD -gstabs+ just changed to emit this:
> 
>   .stabs "Foo<int,(char*)(&string)>:Tt(0,25)=..."
> 
> Do we care?  I think the "(char*)(&string)" looks ugly but is legal.
> I want to change templates.exp to accept this as a PASS.  I've found
> enough serious stabs+ bugs with gcc that I don't want to make a fuss
> over this difference.
> 
> What do you think?

I don't care.  Eventually hopefully we will canonicalize this.  Code to
do it is on drow-cplus-branch (but not used, even there; it can't be
until some other startup performance issues are resolved...).

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer

Re: Foo<int,(char*)&string>

[David Carlton]

On Mon, 16 Feb 2004 21:20:08 -0500 (EST), mec.gnu@mindspring.com (Michael Elizabeth Chastain) said:

> The second template parameter is a "char *"; a value, not a type.

> gcc 3.3.2 -gstabs+ emits this:

>   .stabs "Foo<int,&string>:Tt(1,1)=..."

> gcc HEAD -gstabs+ just changed to emit this:

>   .stabs "Foo<int,(char*)(&string)>:Tt(0,25)=..."

> Do we care?  I think the "(char*)(&string)" looks ugly but is legal.
> I want to change templates.exp to accept this as a PASS.

Sounds good to me.

David Carlton
carlton@kealia.com