On 04/29/2017 04:23 PM, Volker Reichelt wrote:
Hi,the following patch adds fix-it hints to the C++ parser for two wrongly used keywords detected in cp_parser_decl_specifier_seq. Bootstrapped and regtested on x86_64-pc-linux-gnu. OK for trunk? Regards, Volker 2017-04-29 Volker Reichelt <v.reich...@netcologne.de> * parser.c (cp_parser_decl_specifier_seq): Add fix-it hints for friend outside class and obsolete auto as storage-class-specifier. Index: gcc/cp/parser.c =================================================================== --- gcc/cp/parser.c 2017-04-28 +++ gcc/cp/parser.c 2017-04-28 @@ -13213,7 +13213,9 @@ case RID_FRIEND: if (!at_class_scope_p ()) { - error_at (token->location, "%<friend%> used outside of class"); + gcc_rich_location richloc (token->location); + richloc.add_fixit_remove (); + error_at_rich_loc (&richloc, "%<friend%> used outside of class"); cp_lexer_purge_token (parser->lexer); } else @@ -13277,8 +13279,11 @@ /* Complain about `auto' as a storage specifier, if we're complaining about C++0x compatibility. */ - warning_at (token->location, OPT_Wc__11_compat, "%<auto%>" - " changes meaning in C++11; please remove it"); + gcc_rich_location richloc (token->location); + richloc.add_fixit_remove (); + warning_at_rich_loc (&richloc, OPT_Wc__11_compat, + "%<auto%> changes meaning in C++11; " + "please remove it");
What caught my eye here is the "please remove it" part :) Maybe it's me but it seems a little too forceful for a warning (despite the please). I would find a more conventional phrasing like "suggest to remove it" to be more suitable. That said, I wonder if removing the auto is actually the best suggestion. Wouldn't it more in line with where C++ is headed to suggest to remove the type and keep the auto? Martin
