fold_build1 (NOP_EXPR, ...) vs. fold_build1 (CONVERT_EXPR, ...)

[Richard Guenther]

It looks like it is safe to exchange both of them (the first one for sure)
to fold_convert (...) due to the fact that fold_unary handles NOP_EXPR
the same way than CONVERT_EXPR apart from cases that look like oversights,
f.i.

      /* Convert (T1)((T2)X op Y) into (T1)X op Y, for pointer types T1 
and
         T2 being pointers to types of the same size.  */
      if (POINTER_TYPE_P (type)
          && BINARY_CLASS_P (arg0)
          && TREE_CODE (TREE_OPERAND (arg0, 0)) == NOP_EXPR
          && POINTER_TYPE_P (TREE_TYPE (TREE_OPERAND (arg0, 0))))
        {

or

    case ABS_EXPR:
...
      else if (TREE_CODE (arg0) == NOP_EXPR
               && TREE_CODE (type) == REAL_TYPE)
        {
          tree targ0 = strip_float_extensions (arg0);
          if (targ0 != arg0)
            return fold_convert (type, fold_build1 (ABS_EXPR,
                                                    TREE_TYPE (targ0),
                                                    targ0));
        }

or

      /* Simplify ((int)c & 0377) into (int)c, if c is unsigned char.  */
      if (TREE_CODE (arg1) == INTEGER_CST && TREE_CODE (arg0) == NOP_EXPR
          && TYPE_UNSIGNED (TREE_TYPE (TREE_OPERAND (arg0, 0))))
        {

etc.

In fact, I remember a plan of merging NOP_EXPR and CONVERT_EXPR.

Is this correct?

Thanks,
Richard.