direct_notype_declarator and constructors

[Clinton Popetz <cpopetz at cygnus dot com>]

Removing the precedence declaration from this rule in cp/parse.y:

direct_notype_declarator:
          complex_direct_notype_declarator
        /* This precedence declaration is to prefer this reduce
           to the Koenig lookup shift in primary, below.  I hate yacc.  */
        | notype_unqualified_id %prec '('

Causes g++.brendan/parse4.C to XPASS, and introduces no new failures in
check-g++.

The test looks like:

class B
{
public:
    B( int t ) {}
    void f() {}
};

int g() { return 0; } // gets bogus error - referenced below XFAIL *-*-*

int main()
{
    B( g() ).f();    // gets bogus error - treated as decl XFAIL *-*-*
}

Can someone post code that exhibits the need for the %prec?  If not, is the
patch below ok?  FWIW, this introduces two shift/reduce conflicts ('(' and
LEFT_RIGHT) which bison resolves as reduce.  

(Disclaimer: backend mole attempts to pull weeds above ground in the C++
parser, is nearly trampled by a bison, and dives back underground.)

				-Clint


2000-02-25  Clinton Popetz  <cpopetz@cygnus.com>

	* parse.y (direct_notype_declarator): Don't use precedence 
	declaration.

Index: parse.y
===================================================================
RCS file: /cvs/gcc/egcs/gcc/cp/parse.y,v
retrieving revision 1.167
diff -c -2 -p -r1.167 parse.y
*** parse.y	2000/02/25 01:36:03	1.167
--- parse.y	2000/02/25 21:20:32
*************** notype_template_declarator:
*** 1428,1434 ****
  direct_notype_declarator:
  	  complex_direct_notype_declarator
! 	/* This precedence declaration is to prefer this reduce
! 	   to the Koenig lookup shift in primary, below.  I hate yacc.  */
! 	| notype_unqualified_id %prec '('
  	| notype_template_declarator
  	| '(' expr_or_declarator_intern ')'
--- 1428,1432 ----
  direct_notype_declarator:
  	  complex_direct_notype_declarator
! 	| notype_unqualified_id 
  	| notype_template_declarator
  	| '(' expr_or_declarator_intern ')'