Re: [Darwin] Patch c++/15428

[Matt Austern <austern at apple dot com>]

On May 20, 2004, at 2:04 PM, Jason Merrill wrote:

> I still don't see why.  You explained above why this is necessary if
> there's a previous extern instantiation, but in that case we already do
> explicit instantiations on all targets.

If so, then I missed something!  (Always possible.)  When I did my 
original linkonce work for Darwin, I didn't find any logic that treated 
this case specially.  Here's what we had in 3.4, for example:

 /*       The explicit instantiation of a class template specialization
	 implies the instantiation of all of its members not
	 previously explicitly specialized in the translation unit
	 containing the explicit instantiation.

       Of course, we can't instantiate member template classes, since
       we don't have any arguments for them.  Note that the standard
       is unclear on whether the instantiation of the members are
       *explicit* instantiations or not.  We choose to be generous,
       and not set DECL_EXPLICIT_INSTANTIATION.  Therefore, we allow
       the explicit instantiation of a class where some of the members
       have no definition in the current translation unit.  */

    if (! static_p)
      for (tmp = TYPE_METHODS (t); tmp; tmp = TREE_CHAIN (tmp))
	if (TREE_CODE (tmp) == FUNCTION_DECL
	    && DECL_TEMPLATE_INSTANTIATION (tmp))
	  {
	    mark_decl_instantiated (tmp, extern_p);
	    repo_template_instantiated (tmp, extern_p);
	    if (! extern_p)
	      instantiate_decl (tmp, /*defer_ok=*/1);

In mainline, in do_type_instantiation, we have the 
explicitly_instantiate_members test.  But that's something that I added 
for Darwin, and what we're discussing now is whether I got it right.  I 
think what I'm hearing you say is that some other part of the compiler 
does the same test, so all that logic in  do_type_instantiation never 
had to go in in the first place, and it can be removed, not fixed.  Can 
you point me to that other place that does the same test?

			--Matt