diff mbox series

[C++] Fix cp_parser_unqualified_id typo

Message ID c15d762d-67cb-fde5-baea-5553fc2f7b74@oracle.com
State New
Headers show
Series [C++] Fix cp_parser_unqualified_id typo | expand

Commit Message

Paolo Carlini May 31, 2019, 9:05 a.m. UTC
Hi,

noticed this typo: we are passing the location to build_nt as third 
argument instead of passing it to cp_expr as second argument. Tested 
x86_64-linux.

By the way, we have the option of using build_min_nt_loc instead of 
build_nt in all these places for BIT_NOT_EXPR in 
cp_parser_unqualified_id, Certainly it has the advantage that the 
location information survives in the tree node when cp_expr is 
eventually converted to a plain tree... (below tests fine) What do you 
think?

Thanks, Paolo.

////////////////////
2019-05-31  Paolo Carlini  <paolo.carlini@oracle.com>

      * parser.c (cp_parser_unqualified_id): Fix typo, properly call
      build_nt.

Index: parser.c
===================================================================
--- parser.c	(revision 271761)
+++ parser.c	(working copy)
@@ -6052,7 +6052,7 @@ cp_parser_unqualified_id (cp_parser* parser,
 		    && constructor_name_p (token->u.value, scope))))
 	  {
 	    cp_lexer_consume_token (parser->lexer);
-	    return cp_expr (build_nt (BIT_NOT_EXPR, scope), loc);
+	    return build_min_nt_loc (loc, BIT_NOT_EXPR, scope);
 	  }
 
 	/* ~auto means the destructor of whatever the object is.  */
@@ -6063,7 +6063,7 @@ cp_parser_unqualified_id (cp_parser* parser,
 		       "%<~auto%> only available with "
 		       "%<-std=c++14%> or %<-std=gnu++14%>");
 	    cp_lexer_consume_token (parser->lexer);
-	    return cp_expr (build_nt (BIT_NOT_EXPR, make_auto (), loc));
+	    return build_min_nt_loc (loc, BIT_NOT_EXPR, make_auto ());
 	  }
 
 	/* If there was an explicit qualification (S::~T), first look
@@ -6153,8 +6153,8 @@ cp_parser_unqualified_id (cp_parser* parser,
 		   time.  */
 		type_decl = cp_parser_identifier (parser);
 		if (type_decl != error_mark_node)
-		  type_decl = build_nt (BIT_NOT_EXPR, type_decl);
-		return cp_expr (type_decl, loc);
+		  type_decl = build_min_nt_loc (loc, BIT_NOT_EXPR, type_decl);
+		return type_decl;
 	      }
 	  }
 	/* If an error occurred, assume that the name of the
@@ -6162,7 +6162,7 @@ cp_parser_unqualified_id (cp_parser* parser,
 	   class.  That allows us to keep parsing after running
 	   into ill-formed destructor names.  */
 	if (type_decl == error_mark_node && scope)
-	  return build_nt (BIT_NOT_EXPR, scope);
+	  return build_min_nt_loc (loc, BIT_NOT_EXPR, scope);
 	else if (type_decl == error_mark_node)
 	  return error_mark_node;
 
@@ -6189,7 +6189,7 @@ cp_parser_unqualified_id (cp_parser* parser,
 		    "typedef-name %qD used as destructor declarator",
 		    type_decl);
 
-	return cp_expr (build_nt (BIT_NOT_EXPR, TREE_TYPE (type_decl), loc));
+	return build_min_nt_loc (loc, BIT_NOT_EXPR, TREE_TYPE (type_decl));
       }
 
     case CPP_KEYWORD:

Comments

Jason Merrill June 3, 2019, 5:34 p.m. UTC | #1
On 5/31/19 5:05 AM, Paolo Carlini wrote:
> 
> By the way, we have the option of using build_min_nt_loc instead of 
> build_nt in all these places for BIT_NOT_EXPR in 
> cp_parser_unqualified_id, Certainly it has the advantage that the 
> location information survives in the tree node when cp_expr is 
> eventually converted to a plain tree... (below tests fine) What do you 
> think?

This patch is OK.

Jason
diff mbox series

Patch

Index: parser.c
===================================================================
--- parser.c	(revision 271785)
+++ parser.c	(working copy)
@@ -6063,7 +6063,7 @@  cp_parser_unqualified_id (cp_parser* parser,
 		       "%<~auto%> only available with "
 		       "%<-std=c++14%> or %<-std=gnu++14%>");
 	    cp_lexer_consume_token (parser->lexer);
-	    return cp_expr (build_nt (BIT_NOT_EXPR, make_auto (), loc));
+	    return cp_expr (build_nt (BIT_NOT_EXPR, make_auto ()), loc);
 	  }
 
 	/* If there was an explicit qualification (S::~T), first look