Re: V3 static data in classes vs AIX

[Gabriel Dos Reis <gdr at codesourcery dot com>]

David Edelsohn <dje@watson.ibm.com> writes:

| 	IBM xlC 3.6.6.0 actually produces a common symbol as well (for
| class T instead of typename T).  Your latest example does have a
| definition, not a reference:
| 
|      template <typename T> int S<T>::i;
| 
| 	This latest commentary really doesn't make sense to me.  Why is
| the compiler wrong and the library header right?

I don't fully understand all your message but I will try to answer
this and don't hesitate to correct if I misunderstand the issue.

With Mark's example 

  template <typename T>
  struct S {
    static int i;		// this is -not- a definition.
  };

  template <typename T>
  int S<T>::i;			// this is a template definition
				// and need not be instantiated.


  int j = S<double>::i;		// We need to "instantiate"
				// S<double>::i from the template
				// definition.


| 	Why are the multiple definitions correct?  Yes, you can fake out
| the compiler, but why is it correct for the compiler to *NOT* create
| storage for a definition?  Why should these multiple storage definitions
| be merged?

For implicit instantiations, the compiler is mandated to take
appropriate steps to merge possible multiple definitions.

| 	Do weak symbols specifically allow this type of use?  Or is this
| specific to late binding?  I still do not understand the differences
| between the semantics of COMMON and WEAK very well.

I would like to be educated on this issue as well :-)

-- Gaby
CodeSourcery, LLC                       http://www.codesourcery.com