* expression.h (OP_LABELED): New operator, for Chill

	labeled structre tuples.
	* ch-exp.y (tuple_element, named_record_element, tuple_elements):
	New non-terminals, to handle labeled structure tuples.
	(tuple):  Re-define using tuple_elements.
	* eval.c (evaluate_labeled_field_init):  New function, to handle
	initialization of structure fields, possibly using OP_LABELED.
	(evaluate_subexp):  Use it.
	* expprint.c (print_subexp case):  For OP_ARRAY, use Chill syntax
	for Chill.  Handled OP_LABELED.
	* parse.c (length_of_subexp, prefixify_subexp):  Handle OP_LABELED.

	* eval.c (evaluate_subexp):  Handle Chill Powerset tuples.
	* valarith.c (value_bit_index):  Just treat bitstring as represented
	by an array of bytes.  Alignment is handled by compiler.

1 files changed, 162 insertions, 115 deletions

diff --git a/gdb/valarith.c b/gdb/valarith.c
index 3848029..79c0943 100644
--- a/gdb/valarith.c
+++ b/gdb/valarith.c
@@ -1,5 +1,6 @@
 /* Perform arithmetic and other operations on values, for GDB.
-   Copyright 1986, 1989, 1991, 1992 Free Software Foundation, Inc.
+   Copyright 1986, 1989, 1991, 1992, 1993, 1994
+   Free Software Foundation, Inc.
 
 This file is part of GDB.
 
@@ -34,15 +35,14 @@ Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  */
 #define TRUNCATION_TOWARDS_ZERO ((-5 / 2) == -2)
 #endif
 
-static value
-value_subscripted_rvalue PARAMS ((value, value));
+static value_ptr value_subscripted_rvalue PARAMS ((value_ptr, value_ptr));
 
 
-value
+value_ptr
 value_add (arg1, arg2)
-	value arg1, arg2;
+     value_ptr arg1, arg2;
 {
-  register value valint, valptr;
+  register value_ptr valint, valptr;
   register int len;
 
   COERCE_ARRAY (arg1);
@@ -75,9 +75,9 @@ value_add (arg1, arg2)
   return value_binop (arg1, arg2, BINOP_ADD);
 }
 
-value
+value_ptr
 value_sub (arg1, arg2)
-	value arg1, arg2;
+     value_ptr arg1, arg2;
 {
 
   COERCE_ARRAY (arg1);
@@ -94,13 +94,15 @@ value_sub (arg1, arg2)
 	     - (TYPE_LENGTH (TYPE_TARGET_TYPE (VALUE_TYPE (arg1)))
 		* value_as_long (arg2)));
 	}
-      else if (VALUE_TYPE (arg1) == VALUE_TYPE (arg2))
+      else if (TYPE_CODE (VALUE_TYPE (arg2)) == TYPE_CODE_PTR
+	       && TYPE_LENGTH (TYPE_TARGET_TYPE (VALUE_TYPE (arg1)))
+		  == TYPE_LENGTH (TYPE_TARGET_TYPE (VALUE_TYPE (arg2))))
 	{
 	  /* pointer to <type x> - pointer to <type x>.  */
 	  return value_from_longest
 	    (builtin_type_long,		/* FIXME -- should be ptrdiff_t */
 	     (value_as_long (arg1) - value_as_long (arg2))
-	     / TYPE_LENGTH (TYPE_TARGET_TYPE (VALUE_TYPE (arg1))));
+	     / (LONGEST) (TYPE_LENGTH (TYPE_TARGET_TYPE (VALUE_TYPE (arg1)))));
 	}
       else
 	{
@@ -119,12 +121,12 @@ an integer nor a pointer of the same type.");
    FIXME:  Perhaps we should validate that the index is valid and if
    verbosity is set, warn about invalid indices (but still use them). */
 
-value
+value_ptr
 value_subscript (array, idx)
-     value array, idx;
+     value_ptr array, idx;
 {
   int lowerbound;
-  value bound;
+  value_ptr bound;
   struct type *range_type;
 
   COERCE_REF (array);
@@ -152,14 +154,14 @@ value_subscript (array, idx)
    (eg, a vector register).  This routine used to promote floats
    to doubles, but no longer does.  */
 
-static value
+static value_ptr
 value_subscripted_rvalue (array, idx)
-     value array, idx;
+     value_ptr array, idx;
 {
   struct type *elt_type = TYPE_TARGET_TYPE (VALUE_TYPE (array));
   int elt_size = TYPE_LENGTH (elt_type);
   int elt_offs = elt_size * longest_to_int (value_as_long (idx));
-  value v;
+  value_ptr v;
 
   if (elt_offs >= TYPE_LENGTH (VALUE_TYPE (array)))
     error ("no such vector element");
@@ -186,7 +188,7 @@ value_subscripted_rvalue (array, idx)
 int
 binop_user_defined_p (op, arg1, arg2)
      enum exp_opcode op;
-     value arg1, arg2;
+     value_ptr arg1, arg2;
 {
   if (op == BINOP_ASSIGN)
     return 0;
@@ -206,7 +208,7 @@ binop_user_defined_p (op, arg1, arg2)
 
 int unop_user_defined_p (op, arg1)
      enum exp_opcode op;
-     value arg1;
+     value_ptr arg1;
 {
   if (op == UNOP_ADDR)
     return 0;
@@ -224,14 +226,14 @@ int unop_user_defined_p (op, arg1)
    is the opcode saying how to modify it.  Otherwise, OTHEROP is
    unused.  */
 
-value
+value_ptr
 value_x_binop (arg1, arg2, op, otherop)
-     value arg1, arg2;
+     value_ptr arg1, arg2;
      enum exp_opcode op, otherop;
 {
-  value * argvec;
-  char *ptr, *mangle_ptr;
-  char tstr[13], mangle_tstr[13];
+  value_ptr * argvec;
+  char *ptr;
+  char tstr[13];
   int static_memfuncp;
 
   COERCE_REF (arg1);
@@ -245,7 +247,7 @@ value_x_binop (arg1, arg2, op, otherop)
   if (TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_STRUCT)
     error ("Can't do that binary op on that type");  /* FIXME be explicit */
 
-  argvec = (value *) alloca (sizeof (value) * 4);
+  argvec = (value_ptr *) alloca (sizeof (value_ptr) * 4);
   argvec[1] = value_addr (arg1);
   argvec[2] = arg2;
   argvec[3] = 0;
@@ -321,12 +323,12 @@ value_x_binop (arg1, arg2, op, otherop)
    and return that value (where '@' is (almost) any unary operator which
    is legal for GNU C++).  */
 
-value
+value_ptr
 value_x_unop (arg1, op)
-     value arg1;
+     value_ptr arg1;
      enum exp_opcode op;
 {
-  value * argvec;
+  value_ptr * argvec;
   char *ptr, *mangle_ptr;
   char tstr[13], mangle_tstr[13];
   int static_memfuncp;
@@ -339,7 +341,7 @@ value_x_unop (arg1, op)
   if (TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_STRUCT)
     error ("Can't do that unary op on that type");  /* FIXME be explicit */
 
-  argvec = (value *) alloca (sizeof (value) * 3);
+  argvec = (value_ptr *) alloca (sizeof (value_ptr) * 3);
   argvec[1] = value_addr (arg1);
   argvec[2] = 0;
 
@@ -398,11 +400,11 @@ value_x_unop (arg1, op)
     	string values of length 1.
 */
 
-value
+value_ptr
 value_concat (arg1, arg2)
-     value arg1, arg2;
+     value_ptr arg1, arg2;
 {
-  register value inval1, inval2, outval;
+  register value_ptr inval1, inval2, outval;
   int inval1len, inval2len;
   int count, idx;
   char *ptr;
@@ -517,14 +519,6 @@ value_concat (arg1, arg2)
 }
 
 
-/* The type we give to value_binop results.  This is a kludge to get around
-   the fact that we don't know how to determine the result type from
-   the types of the operands.  (I'm not really sure how much we feel
-   the need to duplicate the exact rules of the current language.
-   They can get really hairy.  But not to do so makes it hard to document
-   just what we *do* do).  */
-static struct type *signed_operation_result;
-static struct type *unsigned_operation_result;
 
 /* Perform a binary operation on two operands which have reasonable
    representations as integers or floats.  This includes booleans,
@@ -532,31 +526,27 @@ static struct type *unsigned_operation_result;
    Does not support addition and subtraction on pointers;
    use value_add or value_sub if you want to handle those possibilities.  */
 
-value
+value_ptr
 value_binop (arg1, arg2, op)
-     value arg1, arg2;
+     value_ptr arg1, arg2;
      enum exp_opcode op;
 {
-  register value val;
+  register value_ptr val;
 
   COERCE_ENUM (arg1);
   COERCE_ENUM (arg2);
 
   if ((TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_FLT
-       &&
-       TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_CHAR
-       &&
-       TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_INT
-       &&
-       TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_BOOL)
+       && TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_CHAR
+       && TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_INT
+       && TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_BOOL
+       && TYPE_CODE (VALUE_TYPE (arg1)) != TYPE_CODE_RANGE)
       ||
       (TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_FLT
-       &&
-       TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_CHAR
-       &&
-       TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_INT
-       &&
-       TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_BOOL))
+       && TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_CHAR
+       && TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_INT
+       && TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_BOOL
+       && TYPE_CODE (VALUE_TYPE (arg2)) != TYPE_CODE_RANGE))
     error ("Argument to arithmetic operation not a number or boolean.");
 
   if (TYPE_CODE (VALUE_TYPE (arg1)) == TYPE_CODE_FLT
@@ -629,16 +619,64 @@ value_binop (arg1, arg2, op)
   else
     /* Integral operations here.  */
     /* FIXME:  Also mixed integral/booleans, with result an integer. */
+    /* FIXME: This implements ANSI C rules (also correct for C++).
+       What about FORTRAN and chill?  */
     {
-      /* Should we promote to unsigned longest?  */
-      if ((TYPE_UNSIGNED (VALUE_TYPE (arg1))
-	   || TYPE_UNSIGNED (VALUE_TYPE (arg2)))
-	  && (TYPE_LENGTH (VALUE_TYPE (arg1)) >= sizeof (unsigned LONGEST)
-	      || TYPE_LENGTH (VALUE_TYPE (arg2)) >= sizeof (unsigned LONGEST)))
+      struct type *type1 = VALUE_TYPE (arg1);
+      struct type *type2 = VALUE_TYPE (arg2);
+      int promoted_len1 = TYPE_LENGTH (type1);
+      int promoted_len2 = TYPE_LENGTH (type2);
+      int is_unsigned1 = TYPE_UNSIGNED (type1);
+      int is_unsigned2 = TYPE_UNSIGNED (type2);
+      int result_len;
+      int unsigned_operation;
+
+      /* Determine type length and signedness after promotion for
+	 both operands.  */
+      if (promoted_len1 < TYPE_LENGTH (builtin_type_int))
+	{
+	  is_unsigned1 = 0;
+	  promoted_len1 = TYPE_LENGTH (builtin_type_int);
+	}
+      if (promoted_len2 < TYPE_LENGTH (builtin_type_int))
+	{
+	  is_unsigned2 = 0;
+	  promoted_len2 = TYPE_LENGTH (builtin_type_int);
+	}
+
+      /* Determine type length of the result, and if the operation should
+	 be done unsigned.
+	 Use the signedness of the operand with the greater length.
+	 If both operands are of equal length, use unsigned operation
+	 if one of the operands is unsigned.  */
+      if (promoted_len1 > promoted_len2)
+	{
+	  unsigned_operation = is_unsigned1;
+	  result_len = promoted_len1;
+	}
+      else if (promoted_len2 > promoted_len1)
+	{
+	  unsigned_operation = is_unsigned2;
+	  result_len = promoted_len2;
+	}
+      else
+	{
+	  unsigned_operation = is_unsigned1 || is_unsigned2;
+	  result_len = promoted_len1;
+	}
+
+      if (unsigned_operation)
 	{
 	  unsigned LONGEST v1, v2, v;
 	  v1 = (unsigned LONGEST) value_as_long (arg1);
 	  v2 = (unsigned LONGEST) value_as_long (arg2);
+
+	  /* Truncate values to the type length of the result.  */
+	  if (result_len < sizeof (unsigned LONGEST))
+	    {
+	      v1 &= ((LONGEST) 1 << HOST_CHAR_BIT * result_len) - 1;
+	      v2 &= ((LONGEST) 1 << HOST_CHAR_BIT * result_len) - 1;
+	    }
 	  
 	  switch (op)
 	    {
@@ -718,12 +756,32 @@ value_binop (arg1, arg2, op)
 	    case BINOP_MAX:
 	      v = v1 > v2 ? v1 : v2;
 	      break;
+
+	    case BINOP_EQUAL:
+	      v = v1 == v2;
+	      break;
+
+	    case BINOP_LESS:
+	      v = v1 < v2;
+	      break;
 	      
 	    default:
 	      error ("Invalid binary operation on numbers.");
 	    }
 
-	  val = allocate_value (unsigned_operation_result);
+	  /* This is a kludge to get around the fact that we don't
+	     know how to determine the result type from the types of
+	     the operands.  (I'm not really sure how much we feel the
+	     need to duplicate the exact rules of the current
+	     language.  They can get really hairy.  But not to do so
+	     makes it hard to document just what we *do* do).  */
+
+	  /* Can't just call init_type because we wouldn't know what
+	     name to give the type.  */
+	  val = allocate_value
+	    (result_len > TARGET_LONG_BIT / HOST_CHAR_BIT
+	     ? builtin_type_unsigned_long_long
+	     : builtin_type_unsigned_long);
 	  store_unsigned_integer (VALUE_CONTENTS_RAW (val),
 				  TYPE_LENGTH (VALUE_TYPE (val)),
 				  v);
@@ -816,12 +874,32 @@ value_binop (arg1, arg2, op)
 	    case BINOP_MAX:
 	      v = v1 > v2 ? v1 : v2;
 	      break;
+
+	    case BINOP_EQUAL:
+	      v = v1 == v2;
+	      break;
+
+	    case BINOP_LESS:
+	      v = v1 < v2;
+	      break;
 	      
 	    default:
 	      error ("Invalid binary operation on numbers.");
 	    }
-	  
-	  val = allocate_value (signed_operation_result);
+
+	  /* This is a kludge to get around the fact that we don't
+	     know how to determine the result type from the types of
+	     the operands.  (I'm not really sure how much we feel the
+	     need to duplicate the exact rules of the current
+	     language.  They can get really hairy.  But not to do so
+	     makes it hard to document just what we *do* do).  */
+
+	  /* Can't just call init_type because we wouldn't know what
+	     name to give the type.  */
+	  val = allocate_value
+	    (result_len > TARGET_LONG_BIT / HOST_CHAR_BIT
+	     ? builtin_type_long_long
+	     : builtin_type_long);
 	  store_signed_integer (VALUE_CONTENTS_RAW (val),
 				TYPE_LENGTH (VALUE_TYPE (val)),
 				v);
@@ -835,7 +913,7 @@ value_binop (arg1, arg2, op)
 
 int
 value_logical_not (arg1)
-     value arg1;
+     value_ptr arg1;
 {
   register int len;
   register char *p;
@@ -862,7 +940,7 @@ value_logical_not (arg1)
 
 int
 value_equal (arg1, arg2)
-     register value arg1, arg2;
+     register value_ptr arg1, arg2;
 
 {
   register int len;
@@ -877,7 +955,8 @@ value_equal (arg1, arg2)
   code2 = TYPE_CODE (VALUE_TYPE (arg2));
 
   if (code1 == TYPE_CODE_INT && code2 == TYPE_CODE_INT)
-    return value_as_long (arg1) == value_as_long (arg2);
+    return longest_to_int (value_as_long (value_binop (arg1, arg2,
+						       BINOP_EQUAL)));
   else if ((code1 == TYPE_CODE_FLT || code1 == TYPE_CODE_INT)
 	   && (code2 == TYPE_CODE_FLT || code2 == TYPE_CODE_INT))
     return value_as_double (arg1) == value_as_double (arg2);
@@ -914,7 +993,7 @@ value_equal (arg1, arg2)
 
 int
 value_less (arg1, arg2)
-     register value arg1, arg2;
+     register value_ptr arg1, arg2;
 {
   register enum type_code code1;
   register enum type_code code2;
@@ -926,14 +1005,8 @@ value_less (arg1, arg2)
   code2 = TYPE_CODE (VALUE_TYPE (arg2));
 
   if (code1 == TYPE_CODE_INT && code2 == TYPE_CODE_INT)
-    {
-      if (TYPE_UNSIGNED (VALUE_TYPE (arg1))
-       || TYPE_UNSIGNED (VALUE_TYPE (arg2)))
-	return ((unsigned LONGEST) value_as_long (arg1)
-		< (unsigned LONGEST) value_as_long (arg2));
-      else
-	return value_as_long (arg1) < value_as_long (arg2);
-    }
+    return longest_to_int (value_as_long (value_binop (arg1, arg2,
+						       BINOP_LESS)));
   else if ((code1 == TYPE_CODE_FLT || code1 == TYPE_CODE_INT)
 	   && (code2 == TYPE_CODE_FLT || code2 == TYPE_CODE_INT))
     return value_as_double (arg1) < value_as_double (arg2);
@@ -956,9 +1029,9 @@ value_less (arg1, arg2)
 
 /* The unary operators - and ~.  Both free the argument ARG1.  */
 
-value
+value_ptr
 value_neg (arg1)
-     register value arg1;
+     register value_ptr arg1;
 {
   register struct type *type;
 
@@ -976,9 +1049,9 @@ value_neg (arg1)
   }
 }
 
-value
+value_ptr
 value_complement (arg1)
-     register value arg1;
+     register value_ptr arg1;
 {
   COERCE_ENUM (arg1);
 
@@ -999,8 +1072,9 @@ value_bit_index (type, valaddr, index)
      int index;
 {
   struct type *range;
-  int low_bound, high_bound, bit_length;
+  int low_bound, high_bound;
   LONGEST word;
+  unsigned rel_index;
   range = TYPE_FIELD_TYPE (type, 0);
   if (TYPE_CODE (range) != TYPE_CODE_RANGE)
     return -2;
@@ -1008,33 +1082,18 @@ value_bit_index (type, valaddr, index)
   high_bound = TYPE_HIGH_BOUND (range);
   if (index < low_bound || index > high_bound)
     return -1;
-  bit_length = high_bound - low_bound + 1;
-  index -= low_bound;
-  if (bit_length <= TARGET_CHAR_BIT)
-    word = unpack_long (builtin_type_unsigned_char, valaddr);
-  else if (bit_length <= TARGET_SHORT_BIT)
-    word = unpack_long (builtin_type_unsigned_short, valaddr);
-  else
-    {
-      int word_start_index = (index / TARGET_INT_BIT) * TARGET_INT_BIT;
-      index -= word_start_index;
-      word = unpack_long (builtin_type_unsigned_int,
-			  valaddr + (word_start_index / HOST_CHAR_BIT));
-    }
-#if BITS_BIG_ENDIAN
-  if (bit_length <= TARGET_CHAR_BIT)
-    index = TARGET_CHAR_BIT - 1 - index;
-  else if (bit_length <= TARGET_SHORT_BIT)
-    index = TARGET_SHORT_BIT - 1 - index;
-  else
-    index = TARGET_INT_BIT - 1 - index;
-#endif
-  return (word >> index) & 1;
+  rel_index = index - low_bound;
+  word = unpack_long (builtin_type_unsigned_char,
+		      valaddr + (rel_index / TARGET_CHAR_BIT));
+  rel_index %= TARGET_CHAR_BIT;
+  if (BITS_BIG_ENDIAN)
+    rel_index = TARGET_CHAR_BIT - 1 - rel_index;
+  return (word >> rel_index) & 1;
 }
 
-value
+value_ptr
 value_in (element, set)
-     value element, set;
+     value_ptr element, set;
 {
   int member;
   if (TYPE_CODE (VALUE_TYPE (set)) != TYPE_CODE_SET)
@@ -1054,16 +1113,4 @@ value_in (element, set)
 void
 _initialize_valarith ()
 {
-  /* Can't just call init_type because we wouldn't know what names to give
-     them.  */
-  if (sizeof (LONGEST) > TARGET_LONG_BIT / HOST_CHAR_BIT)
-    {
-      unsigned_operation_result = builtin_type_unsigned_long_long;
-      signed_operation_result = builtin_type_long_long;
-    }
-  else
-    {
-      unsigned_operation_result = builtin_type_unsigned_long;
-      signed_operation_result = builtin_type_long;
-    }
 }