- add -f[no-]short-wchar
[cparser] / parser.c
index 8a79feb..baa943a 100644 (file)
--- a/parser.c
+++ b/parser.c
 #include "adt/error.h"
 #include "adt/array.h"
 
+/** if wchar_t is equal to unsigned short. */
+bool opt_short_wchar_t =
+#ifdef _WIN32
+       true;
+#else
+       false;
+#endif
+
 //#define PRINT_TOKENS
 #define MAX_LOOKAHEAD 2
 
@@ -840,7 +848,7 @@ static void label_pop_to(size_t new_top)
 }
 
 
-static int get_rank(const type_t *type)
+static atomic_type_kind_t get_rank(const type_t *type)
 {
        assert(!is_typeref(type));
        /* The C-standard allows promoting enums to int or unsigned int (see ยง 7.2.2
@@ -959,9 +967,9 @@ static void report_assign_error(assign_error_t error, type_t *orig_type_left,
                /* the left type has all qualifiers from the right type */
                unsigned missing_qualifiers
                        = points_to_right->base.qualifiers & ~points_to_left->base.qualifiers;
-               errorf(source_position,
-                      "destination type '%T' in %s from type '%T' lacks qualifiers '%Q' in pointed-to type",
-                      orig_type_left, context, orig_type_right, missing_qualifiers);
+               warningf(source_position,
+                        "destination type '%T' in %s from type '%T' lacks qualifiers '%Q' in pointed-to type",
+                        orig_type_left, context, orig_type_right, missing_qualifiers);
                return;
        }
 
@@ -3054,6 +3062,7 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers)
        type_modifiers_t   modifiers       = TYPE_MODIFIER_NONE;
        unsigned           type_specifiers = 0;
        bool               newtype         = false;
+       bool               saw_error       = false;
 
        specifiers->source_position = token.source_position;
 
@@ -3207,13 +3216,50 @@ static void parse_declaration_specifiers(declaration_specifiers_t *specifiers)
 
                case T_IDENTIFIER: {
                        /* only parse identifier if we haven't found a type yet */
-                       if (type != NULL || type_specifiers != 0)
-                               goto finish_specifiers;
+                       if (type != NULL || type_specifiers != 0) {
+                               /* Be somewhat resilient to typos like 'unsigned lng* f()' in a
+                                * declaration, so it doesn't generate errors about expecting '(' or
+                                * '{' later on. */
+                               switch (look_ahead(1)->type) {
+                                       STORAGE_CLASSES
+                                       TYPE_SPECIFIERS
+                                       case T_const:
+                                       case T_restrict:
+                                       case T_volatile:
+                                       case T_inline:
+                                       case T__forceinline: /* ^ DECLARATION_START except for __attribute__ */
+                                       case T_IDENTIFIER:
+                                       case '*':
+                                               errorf(HERE, "discarding stray %K in declaration specifer", &token);
+                                               next_token();
+                                               continue;
 
-                       type_t *typedef_type = get_typedef_type(token.v.symbol);
+                                       default:
+                                               goto finish_specifiers;
+                               }
+                       }
+
+                       type_t *const typedef_type = get_typedef_type(token.v.symbol);
+                       if (typedef_type == NULL) {
+                               /* Be somewhat resilient to typos like 'vodi f()' at the beginning of a
+                                * declaration, so it doesn't generate 'implicit int' followed by more
+                                * errors later on. */
+                               token_type_t const la1_type = (token_type_t)look_ahead(1)->type;
+                               switch (la1_type) {
+                                       DECLARATION_START
+                                       case T_IDENTIFIER:
+                                       case '*':
+                                               errorf(HERE, "%K does not name a type", &token);
+                                               next_token();
+                                               saw_error = true;
+                                               if (la1_type == '*')
+                                                       goto finish_specifiers;
+                                               continue;
 
-                       if (typedef_type == NULL)
-                               goto finish_specifiers;
+                                       default:
+                                               goto finish_specifiers;
+                               }
+                       }
 
                        next_token();
                        type = typedef_type;
@@ -3366,7 +3412,12 @@ warn_about_long_long:
                default:
                        /* invalid specifier combination, give an error message */
                        if (type_specifiers == 0) {
-                               if (! strict_mode) {
+                               if (saw_error) {
+                                       specifiers->type = type_error_type;
+                                       return;
+                               }
+
+                               if (!strict_mode) {
                                        if (warning.implicit_int) {
                                                warningf(HERE, "no type specifiers in declaration, using 'int'");
                                        }
@@ -3399,10 +3450,8 @@ warn_about_long_long:
                        type->atomic.akind = atomic_type;
                }
                newtype = true;
-       } else {
-               if (type_specifiers != 0) {
-                       errorf(HERE, "multiple datatypes in declaration");
-               }
+       } else if (type_specifiers != 0) {
+               errorf(HERE, "multiple datatypes in declaration");
        }
 
        /* FIXME: check type qualifiers here */
@@ -4665,6 +4714,10 @@ static void check_declarations(void)
        if (warning.unused_parameter) {
                const scope_t *scope = &current_function->scope;
 
+               if (is_sym_main(current_function->symbol)) {
+                       /* do not issue unused warnings for main */
+                       return;
+               }
                const declaration_t *parameter = scope->declarations;
                for (; parameter != NULL; parameter = parameter->next) {
                        if (! parameter->used) {
@@ -4745,9 +4798,7 @@ static void check_reachable(statement_t *const stmt)
                                                continue;
                                        }
 
-                                       expression_t *const case_expr = i->expression;
-                                       if (is_constant_expression(case_expr) &&
-                                           fold_constant(case_expr) == val) {
+                                       if (i->first_case <= val && val <= i->last_case) {
                                                check_reachable((statement_t*)i);
                                                return;
                                        }
@@ -4900,6 +4951,7 @@ found_break_parent:
                        type_t *const ret  = skip_typeref(type->function.return_type);
                        if (warning.return_type                    &&
                            !is_type_atomic(ret, ATOMIC_TYPE_VOID) &&
+                           is_type_valid(ret)                     &&
                            !is_sym_main(current_function->symbol)) {
                                warningf(&stmt->base.source_position,
                                         "control reaches end of non-void function");
@@ -5472,7 +5524,7 @@ static expression_t *parse_string_const(void)
                        /* note: that we use type_char_ptr here, which is already the
                         * automatic converted type. revert_automatic_type_conversion
                         * will construct the array type */
-                       cnst->base.type    = type_char_ptr;
+                       cnst->base.type    = warning.write_strings ? type_const_char_ptr : type_char_ptr;
                        cnst->string.value = res;
                        return cnst;
                }
@@ -5495,7 +5547,7 @@ static expression_t *parse_string_const(void)
 
                        default: {
                                expression_t *const cnst = allocate_expression_zero(EXPR_WIDE_STRING_LITERAL);
-                               cnst->base.type         = type_wchar_t_ptr;
+                               cnst->base.type         = warning.write_strings ? type_const_wchar_t_ptr : type_wchar_t_ptr;
                                cnst->wide_string.value = wres;
                                return cnst;
                        }
@@ -5928,9 +5980,9 @@ end_error:
 }
 
 /**
- * Parse a braced expression.
+ * Parse a parenthesized expression.
  */
-static expression_t *parse_brace_expression(void)
+static expression_t *parse_parenthesized_expression(void)
 {
        eat('(');
        add_anchor_token(')');
@@ -6416,7 +6468,7 @@ static expression_t *parse_primary_expression(void)
                case T___builtin_prefetch:       return parse_builtin_prefetch();
                case T__assume:                  return parse_assume();
 
-               case '(':                        return parse_brace_expression();
+               case '(':                        return parse_parenthesized_expression();
                case T___noop:                   return parse_noop_expression();
        }
 
@@ -6957,18 +7009,28 @@ static bool check_pointer_arithmetic(const source_position_t *source_position,
        type_t *points_to = pointer_type->pointer.points_to;
        points_to = skip_typeref(points_to);
 
-       if (is_type_incomplete(points_to) &&
-                       (! (c_mode & _GNUC)
-                        || !is_type_atomic(points_to, ATOMIC_TYPE_VOID))) {
-               errorf(source_position,
-                          "arithmetic with pointer to incomplete type '%T' not allowed",
-                          orig_pointer_type);
-               return false;
+       if (is_type_incomplete(points_to)) {
+               if (!(c_mode & _GNUC) || !is_type_atomic(points_to, ATOMIC_TYPE_VOID)) {
+                       errorf(source_position,
+                              "arithmetic with pointer to incomplete type '%T' not allowed",
+                              orig_pointer_type);
+                       return false;
+               } else if (warning.pointer_arith) {
+                       warningf(source_position,
+                                "pointer of type '%T' used in arithmetic",
+                                orig_pointer_type);
+               }
        } else if (is_type_function(points_to)) {
-               errorf(source_position,
-                          "arithmetic with pointer to function type '%T' not allowed",
-                          orig_pointer_type);
-               return false;
+               if (!(c_mode && _GNUC)) {
+                       errorf(source_position,
+                              "arithmetic with pointer to function type '%T' not allowed",
+                              orig_pointer_type);
+                       return false;
+               } else if (warning.pointer_arith) {
+                       warningf(source_position,
+                                "pointer to a function '%T' used in arithmetic",
+                                orig_pointer_type);
+               }
        }
        return true;
 }
@@ -7164,16 +7226,16 @@ static type_t *semantic_arithmetic(type_t *type_left, type_t *type_right)
 
        bool const signed_left  = is_type_signed(type_left);
        bool const signed_right = is_type_signed(type_right);
-       int  const rank_left    = get_rank(type_left);
-       int  const rank_right   = get_rank(type_right);
+       atomic_type_kind_t const rank_left  = get_rank(type_left);
+       atomic_type_kind_t const rank_right = get_rank(type_right);
 
        if (signed_left == signed_right)
                return rank_left >= rank_right ? type_left : type_right;
 
-       int     s_rank;
-       int     u_rank;
-       type_t *s_type;
-       type_t *u_type;
+       atomic_type_kind_t  s_rank;
+       atomic_type_kind_t  u_rank;
+       type_t             *s_type;
+       type_t             *u_type;
        if (signed_left) {
                s_rank = rank_left;
                s_type = type_left;
@@ -8152,34 +8214,54 @@ static statement_t *parse_case_statement(void)
 
        *pos                             = token.source_position;
        statement->case_label.expression = parse_expression();
-
-       PUSH_PARENT(statement);
+       if (! is_constant_expression(statement->case_label.expression)) {
+               errorf(pos, "case label does not reduce to an integer constant");
+               statement->case_label.is_bad = true;
+       } else {
+               long const val = fold_constant(statement->case_label.expression);
+               statement->case_label.first_case = val;
+               statement->case_label.last_case  = val;
+       }
 
        if (c_mode & _GNUC) {
                if (token.type == T_DOTDOTDOT) {
                        next_token();
                        statement->case_label.end_range = parse_expression();
+                       if (! is_constant_expression(statement->case_label.end_range)) {
+                               errorf(pos, "case range does not reduce to an integer constant");
+                               statement->case_label.is_bad = true;
+                       } else {
+                               long const val = fold_constant(statement->case_label.end_range);
+                               statement->case_label.last_case = val;
+
+                               if (val < statement->case_label.first_case) {
+                                       statement->case_label.is_empty = true;
+                                       warningf(pos, "empty range specified");
+                               }
+                       }
                }
        }
 
+       PUSH_PARENT(statement);
+
        expect(':');
 
-       if (! is_constant_expression(statement->case_label.expression)) {
-               errorf(pos, "case label does not reduce to an integer constant");
-       } else if (current_switch != NULL) {
-               /* Check for duplicate case values */
-               /* FIXME slow */
-               long const val = fold_constant(statement->case_label.expression);
-               for (case_label_statement_t *l = current_switch->first_case; l != NULL; l = l->next) {
-                       expression_t const* const e = l->expression;
-                       if (e == NULL || !is_constant_expression(e) || fold_constant(e) != val)
-                               continue;
+       if (current_switch != NULL) {
+               if (! statement->case_label.is_bad) {
+                       /* Check for duplicate case values */
+                       case_label_statement_t *c = &statement->case_label;
+                       for (case_label_statement_t *l = current_switch->first_case; l != NULL; l = l->next) {
+                               if (l->is_bad || l->is_empty)
+                                       continue;
 
-                       errorf(pos, "duplicate case value (previously used %P)",
-                              &l->base.source_position);
-                       break;
-               }
+                               if (c->last_case < l->first_case || c->first_case > l->last_case)
+                                       continue;
 
+                               errorf(pos, "duplicate case value (previously used %P)",
+                                      &l->base.source_position);
+                               break;
+                       }
+               }
                /* link all cases into the switch statement */
                if (current_switch->last_case == NULL) {
                        current_switch->first_case      = &statement->case_label;
@@ -8204,20 +8286,6 @@ end_error:
        return create_invalid_statement();
 }
 
-/**
- * Finds an existing default label of a switch statement.
- */
-static case_label_statement_t *
-find_default_label(const switch_statement_t *statement)
-{
-       case_label_statement_t *label = statement->first_case;
-       for ( ; label != NULL; label = label->next) {
-               if (label->expression == NULL)
-                       return label;
-       }
-       return NULL;
-}
-
 /**
  * Parse a default statement.
  */
@@ -8232,11 +8300,13 @@ static statement_t *parse_default_statement(void)
 
        expect(':');
        if (current_switch != NULL) {
-               const case_label_statement_t *def_label = find_default_label(current_switch);
+               const case_label_statement_t *def_label = current_switch->default_label;
                if (def_label != NULL) {
                        errorf(HERE, "multiple default labels in one switch (previous declared %P)",
                               &def_label->base.source_position);
                } else {
+                       current_switch->default_label = &statement->case_label;
+
                        /* link all cases into the switch statement */
                        if (current_switch->last_case == NULL) {
                                current_switch->first_case      = &statement->case_label;
@@ -8387,6 +8457,46 @@ end_error:
        return create_invalid_statement();
 }
 
+/**
+ * Check that all enums are handled in a switch.
+ *
+ * @param statement  the switch statement to check
+ */
+static void check_enum_cases(const switch_statement_t *statement) {
+       const type_t *type = skip_typeref(statement->expression->base.type);
+       if (! is_type_enum(type))
+               return;
+       const enum_type_t *enumt = &type->enumt;
+
+       /* if we have a default, no warnings */
+       if (statement->default_label != NULL)
+               return;
+
+       /* FIXME: calculation of value should be done while parsing */
+       const declaration_t *declaration;
+       long last_value = -1;
+       for (declaration = enumt->declaration->next;
+            declaration != NULL && declaration->storage_class == STORAGE_CLASS_ENUM_ENTRY;
+                declaration = declaration->next) {
+               const expression_t *expression = declaration->init.enum_value;
+               long                value      = expression != NULL ? fold_constant(expression) : last_value + 1;
+               bool                found      = false;
+               for (const case_label_statement_t *l = statement->first_case; l != NULL; l = l->next) {
+                       if (l->expression == NULL)
+                               continue;
+                       if (l->first_case <= value && value <= l->last_case) {
+                               found = true;
+                               break;
+                       }
+               }
+               if (! found) {
+                       warningf(&statement->base.source_position,
+                               "enumeration value '%Y' not handled in switch", declaration->symbol);
+               }
+               last_value = value;
+       }
+}
+
 /**
  * Parse a switch statement.
  */
@@ -8400,6 +8510,7 @@ static statement_t *parse_switch(void)
        PUSH_PARENT(statement);
 
        expect('(');
+       add_anchor_token(')');
        expression_t *const expr = parse_expression();
        type_t       *      type = skip_typeref(expr->base.type);
        if (is_type_integer(type)) {
@@ -8411,6 +8522,7 @@ static statement_t *parse_switch(void)
        }
        statement->switchs.expression = create_implicit_cast(expr, type);
        expect(')');
+       rem_anchor_token(')');
 
        switch_statement_t *rem = current_switch;
        current_switch          = &statement->switchs;
@@ -8418,9 +8530,11 @@ static statement_t *parse_switch(void)
        current_switch          = rem;
 
        if (warning.switch_default &&
-          find_default_label(&statement->switchs) == NULL) {
+           statement->switchs.default_label == NULL) {
                warningf(&statement->base.source_position, "switch has no default case");
        }
+       if (warning.switch_enum)
+               check_enum_cases(&statement->switchs);
 
        POP_PARENT;
        return statement;
@@ -8881,8 +8995,9 @@ static statement_t *parse_empty_statement(void)
        if (warning.empty_statement) {
                warningf(HERE, "statement is empty");
        }
+       statement_t *const statement = create_empty_statement();
        eat(';');
-       return create_empty_statement();
+       return statement;
 }
 
 /**
@@ -8984,9 +9099,8 @@ static statement_t *parse_compound_statement(bool inside_expression_statement)
        scope_t *last_scope = scope;
        set_scope(&statement->compound.scope);
 
-       statement_t *last_statement = NULL;
-
-       bool only_decls_so_far = true;
+       statement_t **anchor            = &statement->compound.statements;
+       bool          only_decls_so_far = true;
        while (token.type != '}' && token.type != T_EOF) {
                statement_t *sub_statement = intern_parse_statement();
                if (is_invalid_statement(sub_statement)) {
@@ -9005,16 +9119,12 @@ static statement_t *parse_compound_statement(bool inside_expression_statement)
                        }
                }
 
-               if (last_statement != NULL) {
-                       last_statement->base.next = sub_statement;
-               } else {
-                       statement->compound.statements = sub_statement;
-               }
+               *anchor = sub_statement;
 
                while (sub_statement->base.next != NULL)
                        sub_statement = sub_statement->base.next;
 
-               last_statement = sub_statement;
+               anchor = &sub_statement->base.next;
        }
 
        if (token.type == '}') {
@@ -9064,13 +9174,20 @@ static void initialize_builtin_types(void)
        type_ptrdiff_t   = make_global_typedef("__PTRDIFF_TYPE__",  type_long);
        type_uintmax_t   = make_global_typedef("__uintmax_t__",     type_unsigned_long_long);
        type_uptrdiff_t  = make_global_typedef("__UPTRDIFF_TYPE__", type_unsigned_long);
-       type_wchar_t     = make_global_typedef("__WCHAR_TYPE__",    type_int);
+       type_wchar_t     = make_global_typedef("__WCHAR_TYPE__",    opt_short_wchar_t ? type_unsigned_short : type_int);
        type_wint_t      = make_global_typedef("__WINT_TYPE__",     type_int);
 
        type_intmax_t_ptr  = make_pointer_type(type_intmax_t,  TYPE_QUALIFIER_NONE);
        type_ptrdiff_t_ptr = make_pointer_type(type_ptrdiff_t, TYPE_QUALIFIER_NONE);
        type_ssize_t_ptr   = make_pointer_type(type_ssize_t,   TYPE_QUALIFIER_NONE);
        type_wchar_t_ptr   = make_pointer_type(type_wchar_t,   TYPE_QUALIFIER_NONE);
+
+       /* const version of wchar_t */
+       type_const_wchar_t = allocate_type_zero(TYPE_TYPEDEF, &builtin_source_position);
+       type_const_wchar_t->typedeft.declaration  = type_wchar_t->typedeft.declaration;
+       type_const_wchar_t->base.qualifiers      |= TYPE_QUALIFIER_CONST;
+
+       type_const_wchar_t_ptr = make_pointer_type(type_const_wchar_t, TYPE_QUALIFIER_NONE);
 }
 
 /**