Re: [C++ PATCH] Fix up DECL_ARG_TYPE (PR c++/36631)

[Jason Merrill <jason at redhat dot com>]

Mark Mitchell wrote:
> That may be -- but I think the basic point remains true.  DECL_ARG_TYPE
> is not a front-end concept.  It's an implementation detail, describing
> how exactly the argument is passed.  As such, it really should be left
> to the middle-end to compute as needed.

Well, it depends on the ABI, which can be language-specific.

The issue is much like the one that motivated complete_vars: we can 
treat a class as complete while we're in the middle of the definition, 
but when we're done defining the class we need to go back and fix up the 
places which really need bits that we don't know until after 
finish_struct.  That can be done either toward the end of finish_struct, 
as with complete_vars, or as needed later on.  For functions being 
defined, we update DECL_ARG_TYPE as part of start_function 
(require_complete_type_for_parms).  Gimplification seems to be a fine 
time to do this for called functions.

I don't think this issue is specific to templates, however; the example 
in one of Jakub's mails

struct C
{
  C ();
  void foo (C, C);
};

void bar (void)
{
  C c;
  c.foo (c, c);
}

I would expect to have the same problem.

Jason