This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: varargs fix for array 'va_list'.


> Date: Thu, 23 Mar 2000 23:49:18 -0800
> From: Richard Henderson <rth@cygnus.com>
> Cc: gcc-patches@gcc.gnu.org
> 
> > ! stabilize_va_list (valist, needs_lvalue)
> >        tree valist;
> > !      int needs_lvalue ATTRIBUTE_UNUSED;
> 
> You don't enforce needs_lvalue anywhere?  What are you
> going to do with
> 
> 	__builtin_stdarg_start(args+0)
> 	__builtin_stdarg_start(foo())
> 
> If you can give an error, perfect, but I'm thinking it
> might be as bad as a crash.

Hmmm.  I'll look at it.

> You waid you tried the reference thing and it failed?
> Was that just for ppc's array type, or in general?  Cause
> there's nothing saying the internal prototype can't
> depend on the class of va_list_type.

It turns out that

struct foo;
typedef struct foo t[1];

extern void bar (t &);

void foo (t s)
{
  bar (s);
}

which is what we want to do, is not legal C++:

t.cc: In function `void foo (foo *)':
t.cc:8: could not convert `s' to `foo (&)[1]'
t.cc:4: in passing argument 1 of `bar (foo (&)[1])'

Sigh.  But now that you mention it, perhaps we _could_ do it for the
non-array case...   I'll look into that too.

-- 
- Geoffrey Keating <geoffk@cygnus.com>

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]