[Bug c++/23840] New: [3.4/4.0/4.1 Regression] Bogus "invalid lvalue in unary '&'" diagnostic and ICE with va_arg

["jsm28 at gcc dot gnu dot org" <gcc-bugzilla at gcc dot gnu dot org>]

#include <stdarg.h>
struct S
{
  int f(int n) { return n; }
};
void f(int i, ...)
{
  va_list ap;
  va_start(ap, i);
  va_arg(ap, S).f(0);
}

yields a bogus diagnostic and ICE with current mainline:

t2.cc: In function 'void f(int, ...)':
t2.cc:10: error: invalid lvalue in unary '&'

tree check: expected class 'type', have 'exceptional' (error_mark) in
pp_c_type_qualifier_list, at c-pretty-print.c:232
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

4.0 also gives a bogus diagnostic:

t2.cc: In function 'void f(int, ...)':
t2.cc:10: error: invalid lvalue in unary '&'
t2.cc:10: error: no matching function for call to 'S::f(int)'
t2.cc:4: note: candidates are: int S::f(int)

as does 3.4, while 3.3 compiles the test OK.

-- 
           Summary: [3.4/4.0/4.1 Regression] Bogus "invalid lvalue in unary
                    '&'" diagnostic and ICE with va_arg
           Product: gcc
           Version: 4.1.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: jsm28 at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=23840