Add C parser support for "restrict" and "_Atomic"

A user noticed that "watch -location" would fail with a "restrict"
pointer.  The issue here is that if the DWARF mentions "restrict", gdb
will put this into the type name -- but then the C parser will not be
able to parse this type.

This patch adds support for "restrict" and "_Atomic" to the C parser.
C++ doesn't have "restrict", but does have some GCC extensions.  The
type printer is changed to handle this difference as well, so that
watch expressions will work properly.

gdb/ChangeLog
2020-03-14  Tom Tromey  <tom@tromey.com>

	* c-typeprint.c (cp_type_print_method_args): Print "__restrict__"
	for C++.
	(c_type_print_modifier): Likewise.  Add "language" parameter.
	(c_type_print_varspec_prefix, c_type_print_base_struct_union)
	(c_type_print_base_1): Update.
	* type-stack.h (enum type_pieces) <tp_atomic, tp_restrict>: New
	constants.
	* type-stack.c (type_stack::insert): Handle tp_atomic and
	tp_restrict.
	(type_stack::follow_type_instance_flags): Likewise.
	(type_stack::follow_types): Likewise.  Merge type-following code.
	* c-exp.y (RESTRICT, ATOMIC): New tokens.
	(space_identifier, cv_with_space_id)
	(const_or_volatile_or_space_identifier_noopt)
	(const_or_volatile_or_space_identifier): Remove.
	(single_qualifier, qualifier_seq_noopt, qualifier_seq): New
	rules.
	(ptr_operator, typebase): Update.
	(enum token_flag) <FLAG_C>: New constant.
	(ident_tokens): Add "restrict", "__restrict__", "__restrict", and
	"_Atomic".
	(lex_one_token): Handle FLAG_C.

gdb/testsuite/ChangeLog
2020-03-14  Tom Tromey  <tom@tromey.com>

	* gdb.base/cvexpr.exp: Add test for _Atomic and restrict.

1 files changed, 39 insertions, 17 deletions

diff --git a/gdb/c-exp.y b/gdb/c-exp.y
index 3403a85..50a2eef 100644
--- a/gdb/c-exp.y
+++ b/gdb/c-exp.y
@@ -237,6 +237,7 @@ static void c_print_token (FILE *file, int type, YYSTYPE value);
 /* Special type cases, put in to allow the parser to distinguish different
    legal basetypes.  */
 %token SIGNED_KEYWORD LONG SHORT INT_KEYWORD CONST_KEYWORD VOLATILE_KEYWORD DOUBLE_KEYWORD
+%token RESTRICT ATOMIC
 
 %token <sval> DOLLAR_VARIABLE
 
@@ -1169,36 +1170,43 @@ variable:	name_not_typename
 			}
 	;
 
-space_identifier : '@' NAME
-		{
-		  cpstate->type_stack.insert (pstate,
-					      copy_name ($2.stoken).c_str ());
-		}
-	;
-
 const_or_volatile: const_or_volatile_noopt
 	|
 	;
 
-cv_with_space_id : const_or_volatile space_identifier const_or_volatile
+single_qualifier:
+		CONST_KEYWORD
+			{ cpstate->type_stack.insert (tp_const); }
+	| 	VOLATILE_KEYWORD
+			{ cpstate->type_stack.insert (tp_volatile); }
+	| 	ATOMIC
+			{ cpstate->type_stack.insert (tp_atomic); }
+	| 	RESTRICT
+			{ cpstate->type_stack.insert (tp_restrict); }
+	|	'@' NAME
+		{
+		  cpstate->type_stack.insert (pstate,
+					      copy_name ($2.stoken).c_str ());
+		}
 	;
 
-const_or_volatile_or_space_identifier_noopt: cv_with_space_id
-	| const_or_volatile_noopt
+qualifier_seq_noopt:
+		single_qualifier
+	| 	qualifier_seq single_qualifier
 	;
 
-const_or_volatile_or_space_identifier:
-		const_or_volatile_or_space_identifier_noopt
+qualifier_seq:
+		qualifier_seq_noopt
 	|
 	;
 
 ptr_operator:
 		ptr_operator '*'
 			{ cpstate->type_stack.insert (tp_pointer); }
-		const_or_volatile_or_space_identifier
+		qualifier_seq
 	|	'*'
 			{ cpstate->type_stack.insert (tp_pointer); }
-		const_or_volatile_or_space_identifier
+		qualifier_seq
 	|	'&'
 			{ cpstate->type_stack.insert (tp_reference); }
 	|	'&' ptr_operator
@@ -1472,9 +1480,9 @@ typebase
 			    (copy_name($2).c_str (), $4,
 			     pstate->expression_context_block);
 			}
-	| const_or_volatile_or_space_identifier_noopt typebase
+	|	qualifier_seq_noopt typebase
 			{ $$ = cpstate->type_stack.follow_types ($2); }
-	| typebase const_or_volatile_or_space_identifier_noopt
+	|	typebase qualifier_seq_noopt
 			{ $$ = cpstate->type_stack.follow_types ($1); }
 	;
 
@@ -2345,11 +2353,15 @@ enum token_flag
 
   FLAG_CXX = 1,
 
+  /* If this bit is set, the token is C-only.  */
+
+  FLAG_C = 2,
+
   /* If this bit is set, the token is conditional: if there is a
      symbol of the same name, then the token is a symbol; otherwise,
      the token is a keyword.  */
 
-  FLAG_SHADOW = 2
+  FLAG_SHADOW = 4
 };
 DEF_ENUM_FLAGS_TYPE (enum token_flag, token_flags);
 
@@ -2416,6 +2428,10 @@ static const struct token ident_tokens[] =
     {"union", UNION, OP_NULL, 0},
     {"short", SHORT, OP_NULL, 0},
     {"const", CONST_KEYWORD, OP_NULL, 0},
+    {"restrict", RESTRICT, OP_NULL, FLAG_C | FLAG_SHADOW},
+    {"__restrict__", RESTRICT, OP_NULL, 0},
+    {"__restrict", RESTRICT, OP_NULL, 0},
+    {"_Atomic", ATOMIC, OP_NULL, 0},
     {"enum", ENUM, OP_NULL, 0},
     {"long", LONG, OP_NULL, 0},
     {"true", TRUEKEYWORD, OP_NULL, FLAG_CXX},
@@ -2550,6 +2566,7 @@ lex_one_token (struct parser_state *par_state, bool *is_quoted_name)
 	if ((tokentab3[i].flags & FLAG_CXX) != 0
 	    && par_state->language ()->la_language != language_cplus)
 	  break;
+	gdb_assert ((tokentab3[i].flags & FLAG_C) == 0);
 
 	pstate->lexptr += 3;
 	yylval.opcode = tokentab3[i].opcode;
@@ -2563,6 +2580,7 @@ lex_one_token (struct parser_state *par_state, bool *is_quoted_name)
 	if ((tokentab2[i].flags & FLAG_CXX) != 0
 	    && par_state->language ()->la_language != language_cplus)
 	  break;
+	gdb_assert ((tokentab3[i].flags & FLAG_C) == 0);
 
 	pstate->lexptr += 2;
 	yylval.opcode = tokentab2[i].opcode;
@@ -2857,6 +2875,10 @@ lex_one_token (struct parser_state *par_state, bool *is_quoted_name)
 	if ((ident_tokens[i].flags & FLAG_CXX) != 0
 	    && par_state->language ()->la_language != language_cplus)
 	  break;
+	if ((ident_tokens[i].flags & FLAG_C) != 0
+	    && par_state->language ()->la_language != language_c
+	    && par_state->language ()->la_language != language_objc)
+	  break;
 
 	if ((ident_tokens[i].flags & FLAG_SHADOW) != 0)
 	  {