db/fixup_kernel.sh: fix clear_user() handling
[smatch.git] / parse.c
blobe7c19c042195ee5c9d73454080c4a3ff0a44b71a
1 /*
2 * Stupid C parser, version 1e-6.
4 * Let's see how hard this is to do.
6 * Copyright (C) 2003 Transmeta Corp.
7 * 2003-2004 Linus Torvalds
8 * Copyright (C) 2004 Christopher Li
10 * Permission is hereby granted, free of charge, to any person obtaining a copy
11 * of this software and associated documentation files (the "Software"), to deal
12 * in the Software without restriction, including without limitation the rights
13 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
14 * copies of the Software, and to permit persons to whom the Software is
15 * furnished to do so, subject to the following conditions:
17 * The above copyright notice and this permission notice shall be included in
18 * all copies or substantial portions of the Software.
20 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
23 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
24 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
25 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
26 * THE SOFTWARE.
29 #include <stdarg.h>
30 #include <stdlib.h>
31 #include <stdio.h>
32 #include <string.h>
33 #include <ctype.h>
34 #include <unistd.h>
35 #include <fcntl.h>
36 #include <limits.h>
38 #include "lib.h"
39 #include "allocate.h"
40 #include "token.h"
41 #include "parse.h"
42 #include "symbol.h"
43 #include "scope.h"
44 #include "expression.h"
45 #include "target.h"
47 static struct symbol_list **function_symbol_list;
48 struct symbol_list *function_computed_target_list;
49 struct statement_list *function_computed_goto_list;
51 static struct token *statement(struct token *token, struct statement **tree);
52 static struct token *handle_attributes(struct token *token, struct decl_state *ctx);
54 typedef struct token *declarator_t(struct token *, struct symbol *, struct decl_state *);
55 static declarator_t
56 struct_specifier, union_specifier, enum_specifier,
57 attribute_specifier, typeof_specifier,
58 storage_specifier, thread_specifier;
59 static declarator_t generic_qualifier;
60 static declarator_t autotype_specifier;
62 static struct token *parse_if_statement(struct token *token, struct statement *stmt);
63 static struct token *parse_return_statement(struct token *token, struct statement *stmt);
64 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt);
65 static struct token *parse_default_statement(struct token *token, struct statement *stmt);
66 static struct token *parse_case_statement(struct token *token, struct statement *stmt);
67 static struct token *parse_switch_statement(struct token *token, struct statement *stmt);
68 static struct token *parse_for_statement(struct token *token, struct statement *stmt);
69 static struct token *parse_while_statement(struct token *token, struct statement *stmt);
70 static struct token *parse_do_statement(struct token *token, struct statement *stmt);
71 static struct token *parse_goto_statement(struct token *token, struct statement *stmt);
72 static struct token *parse_context_statement(struct token *token, struct statement *stmt);
73 static struct token *parse_range_statement(struct token *token, struct statement *stmt);
74 static struct token *parse_asm_statement(struct token *token, struct statement *stmt);
75 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list);
76 static struct token *parse_static_assert(struct token *token, struct symbol_list **unused);
77 static struct token *ignore_seg_gs(struct token *token, struct symbol_list **unused);
79 typedef struct token *attr_t(struct token *, struct symbol *,
80 struct decl_state *);
82 static attr_t
83 attribute_packed, attribute_aligned, attribute_modifier,
84 attribute_function,
85 attribute_bitwise,
86 attribute_address_space, attribute_context,
87 attribute_cleanup,
88 attribute_designated_init,
89 attribute_transparent_union, ignore_attribute,
90 attribute_mode, attribute_force;
92 typedef struct symbol *to_mode_t(struct symbol *);
94 static to_mode_t
95 to_QI_mode, to_HI_mode, to_SI_mode, to_DI_mode, to_TI_mode;
96 static to_mode_t to_pointer_mode, to_word_mode;
98 enum {
99 Set_T = 1,
100 Set_S = 2,
101 Set_Char = 4,
102 Set_Int = 8,
103 Set_Double = 16,
104 Set_Float = 32,
105 Set_Signed = 64,
106 Set_Unsigned = 128,
107 Set_Short = 256,
108 Set_Long = 512,
109 Set_Vlong = 1024,
110 Set_Int128 = 2048,
111 Set_Any = Set_T | Set_Short | Set_Long | Set_Signed | Set_Unsigned
114 enum {
115 CInt = 0, CSInt, CUInt, CReal,
118 static void asm_modifier(struct token *token, unsigned long *mods, unsigned long mod)
120 if (*mods & mod)
121 warning(token->pos, "duplicated asm modifier");
122 *mods |= mod;
125 static struct symbol_op typedef_op = {
126 .type = KW_MODIFIER,
127 .declarator = storage_specifier,
130 static struct symbol_op inline_op = {
131 .type = KW_MODIFIER,
132 .declarator = generic_qualifier,
133 .asm_modifier = asm_modifier,
136 static struct symbol_op noreturn_op = {
137 .type = KW_MODIFIER,
138 .declarator = generic_qualifier,
141 static declarator_t alignas_specifier;
142 static struct symbol_op alignas_op = {
143 .type = KW_MODIFIER,
144 .declarator = alignas_specifier,
147 static struct symbol_op auto_op = {
148 .type = KW_MODIFIER,
149 .declarator = storage_specifier,
152 static struct symbol_op register_op = {
153 .type = KW_MODIFIER,
154 .declarator = storage_specifier,
157 static struct symbol_op static_op = {
158 .type = KW_MODIFIER|KW_STATIC,
159 .declarator = storage_specifier,
162 static struct symbol_op extern_op = {
163 .type = KW_MODIFIER,
164 .declarator = storage_specifier,
167 static struct symbol_op thread_op = {
168 .type = KW_MODIFIER,
169 .declarator = thread_specifier,
172 static struct symbol_op const_op = {
173 .type = KW_QUALIFIER,
174 .declarator = generic_qualifier,
177 static struct symbol_op volatile_op = {
178 .type = KW_QUALIFIER,
179 .declarator = generic_qualifier,
180 .asm_modifier = asm_modifier,
183 static struct symbol_op restrict_op = {
184 .type = KW_QUALIFIER,
185 .declarator = generic_qualifier,
188 static struct symbol_op atomic_op = {
189 .type = KW_QUALIFIER,
190 .declarator = generic_qualifier,
193 static struct symbol_op typeof_op = {
194 .type = KW_SPECIFIER,
195 .declarator = typeof_specifier,
196 .test = Set_Any,
197 .set = Set_S|Set_T,
200 static struct symbol_op autotype_op = {
201 .type = KW_SPECIFIER,
202 .declarator = autotype_specifier,
203 .test = Set_Any,
204 .set = Set_S|Set_T,
207 static struct symbol_op attribute_op = {
208 .type = KW_ATTRIBUTE,
209 .declarator = attribute_specifier,
212 static struct symbol_op struct_op = {
213 .type = KW_SPECIFIER,
214 .declarator = struct_specifier,
215 .test = Set_Any,
216 .set = Set_S|Set_T,
219 static struct symbol_op union_op = {
220 .type = KW_SPECIFIER,
221 .declarator = union_specifier,
222 .test = Set_Any,
223 .set = Set_S|Set_T,
226 static struct symbol_op enum_op = {
227 .type = KW_SPECIFIER,
228 .declarator = enum_specifier,
229 .test = Set_Any,
230 .set = Set_S|Set_T,
233 static struct symbol_op spec_op = {
234 .type = KW_SPECIFIER | KW_EXACT,
235 .test = Set_Any,
236 .set = Set_S|Set_T,
239 static struct symbol_op char_op = {
240 .type = KW_SPECIFIER,
241 .test = Set_T|Set_Long|Set_Short,
242 .set = Set_T|Set_Char,
243 .class = CInt,
246 static struct symbol_op int_op = {
247 .type = KW_SPECIFIER,
248 .test = Set_T,
249 .set = Set_T|Set_Int,
252 static struct symbol_op double_op = {
253 .type = KW_SPECIFIER,
254 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Vlong,
255 .set = Set_T|Set_Double,
256 .class = CReal,
259 static struct symbol_op float_op = {
260 .type = KW_SPECIFIER,
261 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Long,
262 .set = Set_T|Set_Float,
263 .class = CReal,
266 static struct symbol_op short_op = {
267 .type = KW_SPECIFIER,
268 .test = Set_S|Set_Char|Set_Float|Set_Double|Set_Long|Set_Short,
269 .set = Set_Short,
272 static struct symbol_op signed_op = {
273 .type = KW_SPECIFIER,
274 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
275 .set = Set_Signed,
276 .class = CSInt,
279 static struct symbol_op unsigned_op = {
280 .type = KW_SPECIFIER,
281 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
282 .set = Set_Unsigned,
283 .class = CUInt,
286 static struct symbol_op long_op = {
287 .type = KW_SPECIFIER,
288 .test = Set_S|Set_Char|Set_Float|Set_Short|Set_Vlong,
289 .set = Set_Long,
292 static struct symbol_op int128_op = {
293 .type = KW_SPECIFIER,
294 .test = Set_S|Set_T|Set_Char|Set_Short|Set_Int|Set_Float|Set_Double|Set_Long|Set_Vlong|Set_Int128,
295 .set = Set_T|Set_Int128|Set_Vlong,
296 .class = CInt,
299 static struct symbol_op if_op = {
300 .statement = parse_if_statement,
303 static struct symbol_op return_op = {
304 .statement = parse_return_statement,
307 static struct symbol_op loop_iter_op = {
308 .statement = parse_loop_iterator,
311 static struct symbol_op default_op = {
312 .statement = parse_default_statement,
315 static struct symbol_op case_op = {
316 .statement = parse_case_statement,
319 static struct symbol_op switch_op = {
320 .statement = parse_switch_statement,
323 static struct symbol_op for_op = {
324 .statement = parse_for_statement,
327 static struct symbol_op while_op = {
328 .statement = parse_while_statement,
331 static struct symbol_op do_op = {
332 .statement = parse_do_statement,
335 static struct symbol_op goto_op = {
336 .statement = parse_goto_statement,
339 static struct symbol_op __context___op = {
340 .statement = parse_context_statement,
341 .attribute = attribute_context,
344 static struct symbol_op range_op = {
345 .statement = parse_range_statement,
348 static struct symbol_op asm_op = {
349 .type = KW_ASM,
350 .statement = parse_asm_statement,
351 .toplevel = toplevel_asm_declaration,
354 static struct symbol_op static_assert_op = {
355 .toplevel = parse_static_assert,
358 static struct symbol_op seg_gs = {
359 .toplevel = ignore_seg_gs,
362 static struct symbol_op packed_op = {
363 .attribute = attribute_packed,
366 static struct symbol_op aligned_op = {
367 .attribute = attribute_aligned,
370 static struct symbol_op cleanup_op = {
371 .attribute = attribute_cleanup,
374 static struct symbol_op attr_mod_op = {
375 .attribute = attribute_modifier,
378 static struct symbol_op attr_fun_op = {
379 .attribute = attribute_function,
382 static struct symbol_op attr_bitwise_op = {
383 .attribute = attribute_bitwise,
386 static struct symbol_op attr_force_op = {
387 .attribute = attribute_force,
390 static struct symbol_op address_space_op = {
391 .attribute = attribute_address_space,
394 static struct symbol_op mode_op = {
395 .attribute = attribute_mode,
398 static struct symbol_op context_op = {
399 .attribute = attribute_context,
402 static struct symbol_op designated_init_op = {
403 .attribute = attribute_designated_init,
406 static struct symbol_op transparent_union_op = {
407 .attribute = attribute_transparent_union,
410 static struct symbol_op ignore_attr_op = {
411 .attribute = ignore_attribute,
414 static struct symbol_op mode_QI_op = {
415 .type = KW_MODE,
416 .to_mode = to_QI_mode
419 static struct symbol_op mode_HI_op = {
420 .type = KW_MODE,
421 .to_mode = to_HI_mode
424 static struct symbol_op mode_SI_op = {
425 .type = KW_MODE,
426 .to_mode = to_SI_mode
429 static struct symbol_op mode_DI_op = {
430 .type = KW_MODE,
431 .to_mode = to_DI_mode
434 static struct symbol_op mode_TI_op = {
435 .type = KW_MODE,
436 .to_mode = to_TI_mode
439 static struct symbol_op mode_pointer_op = {
440 .type = KW_MODE,
441 .to_mode = to_pointer_mode
444 static struct symbol_op mode_word_op = {
445 .type = KW_MODE,
446 .to_mode = to_word_mode
450 * Define the keyword and their effects.
451 * The entries in the 'typedef' and put in NS_TYPEDEF and
452 * are automatically set as reserved keyword while the ones
453 * in the 'keyword' table are just put in NS_KEYWORD.
455 * The entries are added via the 3 macros:
456 * N() for entries with "name" only,
457 * D() for entries with "name" & "__name__",
458 * A() for entries with "name", "__name" & "__name__",
459 * U() for entries with "__name" & "__name__".
461 static struct init_keyword {
462 const char *name;
463 struct symbol_op *op;
464 struct symbol *type;
465 unsigned long mods;
466 } typedefs[] = {
467 #define N(I, O,...) { I, O,##__VA_ARGS__ }
468 #define D(I, O,...) N(I,O,##__VA_ARGS__ ), \
469 N("__" I "__",O,##__VA_ARGS__)
470 #define A(I, O,...) N(I,O,##__VA_ARGS__ ), \
471 N("__" I,O,##__VA_ARGS__), \
472 N("__" I "__",O,##__VA_ARGS__)
473 #define U(I, O,...) N("__" I,O,##__VA_ARGS__), \
474 N("__" I "__",O,##__VA_ARGS__)
475 /* Storage classes */
476 N("auto", &auto_op, .mods = MOD_AUTO),
477 N("register", &register_op, .mods = MOD_REGISTER),
478 N("static", &static_op, .mods = MOD_STATIC),
479 N("extern", &extern_op, .mods = MOD_EXTERN),
480 N("__thread", &thread_op),
481 N("_Thread_local", &thread_op),
483 A("inline", &inline_op, .mods = MOD_INLINE),
485 /* Typedef ... */
486 N("typedef", &typedef_op, .mods = MOD_USERTYPE),
487 A("typeof", &typeof_op),
488 N("__auto_type", &autotype_op),
490 /* Type qualifiers */
491 A("const", &const_op, .mods = MOD_CONST),
492 A("volatile", &volatile_op, .mods = MOD_VOLATILE),
493 A("restrict", &restrict_op, .mods = MOD_RESTRICT),
495 N("_Atomic", &atomic_op, .mods = MOD_ATOMIC),
496 N("_Noreturn", &noreturn_op, .mods = MOD_NORETURN),
497 N("_Alignas", &alignas_op),
499 U("attribute", &attribute_op),
501 /* Type specifiers */
502 N("struct", &struct_op),
503 N("union", &union_op),
504 N("enum", &enum_op),
506 N("void", &spec_op, .type = &void_ctype),
507 N("char", &char_op),
508 N("short", &short_op),
509 N("int", &int_op),
510 N("long", &long_op),
511 N("float", &float_op),
512 N("double", &double_op),
513 A("signed", &signed_op),
514 N("unsigned", &unsigned_op),
515 N("__int128", &int128_op),
516 N("_Bool", &spec_op, .type = &bool_ctype),
518 /* Predeclared types */
519 N("__builtin_va_list", &spec_op, .type = &ptr_ctype),
520 N("__builtin_ms_va_list",&spec_op, .type = &ptr_ctype),
521 N("__int128_t", &spec_op, .type = &sint128_ctype),
522 N("__uint128_t", &spec_op, .type = &uint128_ctype),
523 N("_Float32", &spec_op, .type = &float32_ctype),
524 N("_Float32x", &spec_op, .type = &float32x_ctype),
525 N("_Float64", &spec_op, .type = &float64_ctype),
526 N("_Float64x", &spec_op, .type = &float64x_ctype),
527 N("_Float128", &spec_op, .type = &float128_ctype),
528 N("__float128", &spec_op, .type = &float128_ctype),
530 N("__seg_gs", &seg_gs),
532 }, keywords[] = {
533 /* Statements */
534 N("if", &if_op),
535 N("return", &return_op),
536 N("break", &loop_iter_op),
537 N("continue", &loop_iter_op),
538 N("default", &default_op),
539 N("case", &case_op),
540 N("switch", &switch_op),
541 N("for", &for_op),
542 N("while", &while_op),
543 N("do", &do_op),
544 N("goto", &goto_op),
545 A("asm", &asm_op),
546 N("context", &context_op),
547 N("__context__", &__context___op),
548 N("__range__", &range_op),
549 N("_Static_assert", &static_assert_op),
551 /* Attributes */
552 D("packed", &packed_op),
553 D("aligned", &aligned_op),
554 D("cleanup", &cleanup_op),
555 D("nocast", &attr_mod_op, .mods = MOD_NOCAST),
556 D("noderef", &attr_mod_op, .mods = MOD_NODEREF),
557 D("safe", &attr_mod_op, .mods = MOD_SAFE),
558 D("unused", &attr_mod_op, .mods = MOD_UNUSED),
559 D("externally_visible", &attr_mod_op, .mods = MOD_EXT_VISIBLE),
560 D("force", &attr_force_op),
561 D("bitwise", &attr_bitwise_op, .mods = MOD_BITWISE),
562 D("address_space", &address_space_op),
563 D("designated_init", &designated_init_op),
564 D("transparent_union", &transparent_union_op),
565 D("noreturn", &attr_fun_op, .mods = MOD_NORETURN),
566 D("pure", &attr_fun_op, .mods = MOD_PURE),
567 A("const", &attr_fun_op, .mods = MOD_PURE),
568 D("gnu_inline", &attr_fun_op, .mods = MOD_GNU_INLINE),
570 /* Modes */
571 D("mode", &mode_op),
572 D("QI", &mode_QI_op),
573 D("HI", &mode_HI_op),
574 D("SI", &mode_SI_op),
575 D("DI", &mode_DI_op),
576 D("TI", &mode_TI_op),
577 D("byte", &mode_QI_op),
578 D("pointer", &mode_pointer_op),
579 D("word", &mode_word_op),
583 static const char *ignored_attributes[] = {
585 #define GCC_ATTR(x) \
586 STRINGIFY(x), \
587 STRINGIFY(__##x##__),
589 #include "gcc-attr-list.h"
591 #undef GCC_ATTR
593 "bounded",
594 "__bounded__",
595 "__noclone",
596 "__nonnull",
597 "__nothrow",
601 static void init_keyword(int stream, struct init_keyword *kw, enum namespace ns)
603 struct symbol *sym = create_symbol(stream, kw->name, SYM_KEYWORD, ns);
604 sym->ident->keyword = 1;
605 sym->ident->reserved |= (ns == NS_TYPEDEF);
606 sym->ctype.modifiers = kw->mods;
607 sym->ctype.base_type = kw->type;
608 sym->op = kw->op;
611 void init_parser(int stream)
613 int i;
615 for (i = 0; i < ARRAY_SIZE(typedefs); i++)
616 init_keyword(stream, &typedefs[i], NS_TYPEDEF);
617 for (i = 0; i < ARRAY_SIZE(keywords); i++)
618 init_keyword(stream, &keywords[i], NS_KEYWORD);
620 for (i = 0; i < ARRAY_SIZE(ignored_attributes); i++) {
621 const char * name = ignored_attributes[i];
622 struct symbol *sym = create_symbol(stream, name, SYM_KEYWORD,
623 NS_KEYWORD);
624 if (!sym->op) {
625 sym->ident->keyword = 1;
626 sym->op = &ignore_attr_op;
632 static struct token *skip_to(struct token *token, int op)
634 while (!match_op(token, op) && !eof_token(token))
635 token = token->next;
636 return token;
639 static struct token bad_token = { .pos.type = TOKEN_BAD };
640 struct token *expect(struct token *token, int op, const char *where)
642 if (!match_op(token, op)) {
643 if (token != &bad_token) {
644 bad_token.next = token;
645 sparse_error(token->pos, "Expected %s %s", show_special(op), where);
646 sparse_error(token->pos, "got %s", show_token(token));
648 if (op == ';')
649 return skip_to(token, op);
650 return &bad_token;
652 return token->next;
656 // issue an error message on new parsing errors
657 // @token: the current token
658 // @errmsg: the error message
659 // If the current token is from a previous error, an error message
660 // has already been issued, so nothing more is done.
661 // Otherwise, @errmsg is displayed followed by the current token.
662 static void unexpected(struct token *token, const char *errmsg)
664 if (token == &bad_token)
665 return;
666 sparse_error(token->pos, "%s", errmsg);
667 sparse_error(token->pos, "got %s", show_token(token));
670 // Add a symbol to the list of function-local symbols
671 static void fn_local_symbol(struct symbol *sym)
673 if (function_symbol_list)
674 add_symbol(function_symbol_list, sym);
677 struct statement *alloc_statement(struct position pos, int type)
679 struct statement *stmt = __alloc_statement(0);
680 stmt->type = type;
681 stmt->pos = pos;
682 return stmt;
685 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list);
687 static void apply_ctype(struct position pos, struct ctype *dst, struct ctype *src);
689 static void apply_modifiers(struct position pos, struct decl_state *ctx)
691 struct symbol *ctype;
692 if (!ctx->mode)
693 return;
694 ctype = ctx->mode->to_mode(ctx->ctype.base_type);
695 if (!ctype)
696 sparse_error(pos, "don't know how to apply mode to %s",
697 show_typename(ctx->ctype.base_type));
698 else
699 ctx->ctype.base_type = ctype;
703 static struct symbol * alloc_indirect_symbol(struct position pos, struct ctype *ctype, int type)
705 struct symbol *sym = alloc_symbol(pos, type);
707 sym->ctype.base_type = ctype->base_type;
708 sym->ctype.modifiers = ctype->modifiers;
710 ctype->base_type = sym;
711 ctype->modifiers = 0;
712 return sym;
716 * NOTE! NS_LABEL is not just a different namespace,
717 * it also ends up using function scope instead of the
718 * regular symbol scope.
720 struct symbol *label_symbol(struct token *token, int used)
722 struct symbol *sym = lookup_symbol(token->ident, NS_LABEL);
723 if (!sym) {
724 sym = alloc_symbol(token->pos, SYM_LABEL);
725 bind_symbol(sym, token->ident, NS_LABEL);
726 if (used)
727 sym->used = 1;
728 fn_local_symbol(sym);
730 return sym;
733 static struct token *struct_union_enum_specifier(enum type type,
734 struct token *token, struct decl_state *ctx,
735 struct token *(*parse)(struct token *, struct symbol *))
737 struct decl_state attr = { };
738 struct symbol *sym;
739 struct position *repos;
741 token = handle_attributes(token, &attr);
742 if (token_type(token) == TOKEN_IDENT) {
743 sym = lookup_symbol(token->ident, NS_STRUCT);
744 if (!sym ||
745 (is_outer_scope(sym->scope) &&
746 (match_op(token->next,';') || match_op(token->next,'{')))) {
747 // Either a new symbol, or else an out-of-scope
748 // symbol being redefined.
749 sym = alloc_symbol(token->pos, type);
750 bind_symbol(sym, token->ident, NS_STRUCT);
752 if (sym->type != type)
753 error_die(token->pos, "invalid tag applied to %s", show_typename (sym));
754 ctx->ctype.base_type = sym;
755 repos = &token->pos;
756 token = token->next;
757 if (!match_op(token, '{'))
758 return token;
760 // The following test is actually wrong for empty
761 // structs, but (1) they are not C99, (2) gcc does
762 // the same thing, and (3) it's easier.
763 if (sym->symbol_list)
764 error_die(token->pos, "redefinition of %s", show_typename (sym));
765 sym->pos = *repos;
767 // Mark the structure as needing re-examination
768 sym->examined = 0;
769 } else if (match_op(token, '{')) {
770 // private struct/union/enum type
771 sym = alloc_symbol(token->pos, type);
772 set_current_scope(sym); // used by dissect
773 ctx->ctype.base_type = sym;
774 } else {
775 sparse_error(token->pos, "expected declaration");
776 ctx->ctype.base_type = &bad_ctype;
777 return token;
780 token = parse(token->next, sym);
781 token = expect(token, '}', "at end of specifier");
782 attr.ctype.base_type = sym;
783 token = handle_attributes(token, &attr);
784 apply_ctype(token->pos, &sym->ctype, &attr.ctype);
785 sym->packed = attr.packed;
787 sym->endpos = token->pos;
789 return token;
792 static struct token *parse_struct_declaration(struct token *token, struct symbol *sym)
794 struct symbol *field, *last = NULL;
795 struct token *res;
796 res = struct_declaration_list(token, &sym->symbol_list);
797 FOR_EACH_PTR(sym->symbol_list, field) {
798 if (!field->ident) {
799 struct symbol *base = field->ctype.base_type;
800 if (base && base->type == SYM_BITFIELD)
801 continue;
803 if (last)
804 last->next_subobject = field;
805 last = field;
806 } END_FOR_EACH_PTR(field);
807 return res;
810 static struct token *parse_union_declaration(struct token *token, struct symbol *sym)
812 return struct_declaration_list(token, &sym->symbol_list);
815 static struct token *struct_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
817 return struct_union_enum_specifier(SYM_STRUCT, token, ctx, parse_struct_declaration);
820 static struct token *union_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
822 return struct_union_enum_specifier(SYM_UNION, token, ctx, parse_union_declaration);
826 // safe right shift
828 // This allow to use a shift amount as big (or bigger)
829 // than the width of the value to be shifted, in which case
830 // the result is, of course, 0.
831 static unsigned long long rshift(unsigned long long val, unsigned int n)
833 if (n >= (sizeof(val) * 8))
834 return 0;
835 return val >> n;
838 struct range {
839 long long neg;
840 unsigned long long pos;
843 static void update_range(struct range *range, unsigned long long uval, struct symbol *vtype)
845 long long sval = uval;
847 if (is_signed_type(vtype) && (sval < 0)) {
848 if (sval < range->neg)
849 range->neg = sval;
850 } else {
851 if (uval > range->pos)
852 range->pos = uval;
856 static int type_is_ok(struct symbol *type, struct range range)
858 int shift = type->bit_size;
859 int is_unsigned = type->ctype.modifiers & MOD_UNSIGNED;
861 if (!is_unsigned)
862 shift--;
863 if (rshift(range.pos, shift))
864 return 0;
865 if (range.neg == 0)
866 return 1;
867 if (is_unsigned)
868 return 0;
869 if (rshift(~range.neg, shift))
870 return 0;
871 return 1;
874 static struct range type_range(struct symbol *type)
876 struct range range;
877 unsigned int size = type->bit_size;
878 unsigned long long max;
879 long long min;
881 if (is_signed_type(type)) {
882 min = sign_bit(size);
883 max = min - 1;
884 } else {
885 min = 0;
886 max = bits_mask(size);
889 range.pos = max;
890 range.neg = min;
891 return range;
894 static int val_in_range(struct range *range, long long sval, struct symbol *vtype)
896 unsigned long long uval = sval;
898 if (is_signed_type(vtype) && (sval < 0))
899 return range->neg <= sval;
900 else
901 return uval <= range->pos;
904 static void cast_enum_list(struct symbol_list *list, struct symbol *base_type)
906 struct range irange = type_range(&int_ctype);
907 struct symbol *sym;
909 FOR_EACH_PTR(list, sym) {
910 struct expression *expr = sym->initializer;
911 struct symbol *ctype;
912 long long val;
913 if (expr->type != EXPR_VALUE)
914 continue;
915 ctype = expr->ctype;
916 val = get_expression_value(expr);
917 if (is_int_type(ctype) && val_in_range(&irange, val, ctype)) {
918 expr->ctype = &int_ctype;
919 continue;
921 cast_value(expr, base_type, expr);
922 } END_FOR_EACH_PTR(sym);
925 static struct token *parse_enum_declaration(struct token *token, struct symbol *parent)
927 unsigned long long lastval = 0;
928 struct symbol *ctype = NULL, *base_type = NULL;
929 struct range range = { };
930 int mix_bitwise = 0;
932 parent->examined = 1;
933 parent->ctype.base_type = &int_ctype;
934 while (token_type(token) == TOKEN_IDENT) {
935 struct expression *expr = NULL;
936 struct token *next = token->next;
937 struct decl_state ctx = { };
938 struct symbol *sym;
940 // FIXME: only 'deprecated' should be accepted
941 next = handle_attributes(next, &ctx);
943 if (match_op(next, '=')) {
944 next = constant_expression(next->next, &expr);
945 lastval = get_expression_value(expr);
946 ctype = &void_ctype;
947 if (expr && expr->ctype)
948 ctype = expr->ctype;
949 } else if (!ctype) {
950 ctype = &int_ctype;
951 } else if (is_int_type(ctype)) {
952 lastval++;
953 } else {
954 error_die(token->pos, "can't increment the last enum member");
957 if (!expr) {
958 expr = alloc_expression(token->pos, EXPR_VALUE);
959 expr->value = lastval;
960 expr->ctype = ctype;
963 sym = alloc_symbol(token->pos, SYM_NODE);
964 bind_symbol(sym, token->ident, NS_SYMBOL);
965 sym->ctype.modifiers &= ~MOD_ADDRESSABLE;
966 sym->initializer = expr;
967 sym->enum_member = 1;
968 sym->ctype.base_type = parent;
969 add_ptr_list(&parent->symbol_list, sym);
971 if (base_type != &bad_ctype) {
972 if (ctype->type == SYM_NODE)
973 ctype = ctype->ctype.base_type;
974 if (ctype->type == SYM_ENUM) {
975 if (ctype == parent)
976 ctype = base_type;
977 else
978 ctype = ctype->ctype.base_type;
981 * base_type rules:
982 * - if all enums are of the same type, then
983 * the base_type is that type (two first
984 * cases)
985 * - if enums are of different types, they
986 * all have to be integer types, and the
987 * base type is at least "int_ctype".
988 * - otherwise the base_type is "bad_ctype".
990 if (!base_type || ctype == &bad_ctype) {
991 base_type = ctype;
992 } else if (ctype == base_type) {
993 /* nothing */
994 } else if (is_int_type(base_type) && is_int_type(ctype)) {
995 base_type = &int_ctype;
996 } else if (is_restricted_type(base_type) != is_restricted_type(ctype)) {
997 if (!mix_bitwise++) {
998 warning(expr->pos, "mixed bitwiseness");
1000 } else if (is_restricted_type(base_type) && base_type != ctype) {
1001 sparse_error(expr->pos, "incompatible restricted type");
1002 info(expr->pos, " expected: %s", show_typename(base_type));
1003 info(expr->pos, " got: %s", show_typename(ctype));
1004 base_type = &bad_ctype;
1005 } else if (base_type != &bad_ctype) {
1006 sparse_error(token->pos, "bad enum definition");
1007 base_type = &bad_ctype;
1009 parent->ctype.base_type = base_type;
1011 if (is_int_type(base_type)) {
1012 update_range(&range, lastval, ctype);
1014 token = next;
1016 sym->endpos = token->pos;
1018 if (!match_op(token, ','))
1019 break;
1020 token = token->next;
1022 if (!base_type) {
1023 sparse_error(token->pos, "empty enum definition");
1024 base_type = &bad_ctype;
1026 else if (!is_int_type(base_type))
1028 else if (type_is_ok(&uint_ctype, range))
1029 base_type = &uint_ctype;
1030 else if (type_is_ok(&int_ctype, range))
1031 base_type = &int_ctype;
1032 else if (type_is_ok(&ulong_ctype, range))
1033 base_type = &ulong_ctype;
1034 else if (type_is_ok(&long_ctype, range))
1035 base_type = &long_ctype;
1036 else if (type_is_ok(&ullong_ctype, range))
1037 base_type = &ullong_ctype;
1038 else if (type_is_ok(&llong_ctype, range))
1039 base_type = &llong_ctype;
1040 else
1041 base_type = &bad_ctype;
1042 parent->ctype.base_type = base_type;
1043 parent->ctype.modifiers |= (base_type->ctype.modifiers & MOD_UNSIGNED);
1044 parent->examined = 0;
1046 if (mix_bitwise)
1047 return token;
1048 cast_enum_list(parent->symbol_list, base_type);
1050 return token;
1053 static struct token *enum_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1055 struct token *ret = struct_union_enum_specifier(SYM_ENUM, token, ctx, parse_enum_declaration);
1056 struct ctype *ctype = &ctx->ctype.base_type->ctype;
1058 if (!ctype->base_type)
1059 ctype->base_type = &incomplete_ctype;
1061 return ret;
1064 static struct token *typeof_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1067 if (!match_op(token, '(')) {
1068 sparse_error(token->pos, "expected '(' after typeof");
1069 return token;
1071 if (lookup_type(token->next)) {
1072 struct symbol *sym;
1073 token = typename(token->next, &sym, NULL);
1074 ctx->ctype.base_type = sym->ctype.base_type;
1075 apply_ctype(token->pos, &ctx->ctype, &sym->ctype);
1076 } else {
1077 struct symbol *typeof_sym = alloc_symbol(token->pos, SYM_TYPEOF);
1078 token = parse_expression(token->next, &typeof_sym->initializer);
1080 typeof_sym->endpos = token->pos;
1081 if (!typeof_sym->initializer) {
1082 sparse_error(token->pos, "expected expression after the '(' token");
1083 typeof_sym = &bad_ctype;
1085 ctx->ctype.base_type = typeof_sym;
1087 return expect(token, ')', "after typeof");
1090 static struct token *autotype_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1092 ctx->ctype.base_type = &autotype_ctype;
1093 ctx->autotype = 1;
1094 return token;
1097 static struct token *ignore_attribute(struct token *token, struct symbol *attr, struct decl_state *ctx)
1099 struct expression *expr = NULL;
1100 if (match_op(token, '('))
1101 token = parens_expression(token, &expr, "in attribute");
1102 return token;
1105 static struct token *attribute_packed(struct token *token, struct symbol *attr, struct decl_state *ctx)
1107 if (!ctx->ctype.alignment) {
1108 ctx->ctype.alignment = 1;
1109 ctx->packed = 1;
1111 return token;
1114 static struct token *attribute_aligned(struct token *token, struct symbol *attr, struct decl_state *ctx)
1116 int alignment = max_alignment;
1117 struct expression *expr = NULL;
1119 if (match_op(token, '(')) {
1120 token = parens_expression(token, &expr, "in attribute");
1121 if (expr)
1122 alignment = const_expression_value(expr);
1124 if (alignment & (alignment-1)) {
1125 warning(token->pos, "I don't like non-power-of-2 alignments");
1126 return token;
1127 } else if (alignment > ctx->ctype.alignment)
1128 ctx->ctype.alignment = alignment;
1129 return token;
1132 static struct token *attribute_cleanup(struct token *token, struct symbol *attr, struct decl_state *ctx)
1134 struct expression *expr = NULL;
1136 if (match_op(token, '(')) {
1137 token = token->next;
1138 if (match_op(token, ')'))
1139 sparse_error(token->pos, "an argument is expected for attribute 'cleanup'");
1140 else if (token_type(token) != TOKEN_IDENT)
1141 sparse_error(token->pos, "argument is not an identifier");
1142 token = primary_expression(token, &expr);
1143 if (expr && expr->type == EXPR_SYMBOL)
1144 ctx->cleanup = expr;
1145 return expect(token, ')', "after attribute's argument'");
1148 sparse_error(token->pos, "an argument is expected for attribute 'cleanup'");
1149 return token;
1152 static void apply_mod(struct position *pos, unsigned long *mods, unsigned long mod)
1154 if (*mods & mod & ~MOD_DUP_OK)
1155 warning(*pos, "duplicate %s", modifier_name(mod));
1156 *mods |= mod;
1159 static void apply_qualifier(struct position *pos, struct ctype *ctx, unsigned long qual)
1161 apply_mod(pos, &ctx->modifiers, qual);
1164 static struct token *attribute_modifier(struct token *token, struct symbol *attr, struct decl_state *ctx)
1166 apply_mod(&token->pos, &ctx->ctype.modifiers, attr->ctype.modifiers);
1167 return token;
1170 static struct token *attribute_function(struct token *token, struct symbol *attr, struct decl_state *ctx)
1172 apply_mod(&token->pos, &ctx->f_modifiers, attr->ctype.modifiers);
1173 return token;
1176 static struct token *attribute_bitwise(struct token *token, struct symbol *attr, struct decl_state *ctx)
1178 if (Wbitwise)
1179 attribute_modifier(token, attr, ctx);
1180 return token;
1183 static struct ident *numerical_address_space(int asn)
1185 char buff[32];
1187 if (!asn)
1188 return NULL;
1189 sprintf(buff, "<asn:%d>", asn);
1190 return built_in_ident(buff);
1193 static struct token *attribute_address_space(struct token *token, struct symbol *attr, struct decl_state *ctx)
1195 struct expression *expr = NULL;
1196 struct ident *as = NULL;
1197 struct token *next;
1199 token = expect(token, '(', "after address_space attribute");
1200 switch (token_type(token)) {
1201 case TOKEN_NUMBER:
1202 next = primary_expression(token, &expr);
1203 if (expr->type != EXPR_VALUE)
1204 goto invalid;
1205 as = numerical_address_space(expr->value);
1206 break;
1207 case TOKEN_IDENT:
1208 next = token->next;
1209 as = token->ident;
1210 break;
1211 default:
1212 next = token->next;
1213 invalid:
1214 as = NULL;
1215 warning(token->pos, "invalid address space name");
1218 if (Waddress_space && as) {
1219 if (ctx->ctype.as)
1220 sparse_error(token->pos,
1221 "multiple address spaces given: %s & %s",
1222 show_as(ctx->ctype.as), show_as(as));
1223 ctx->ctype.as = as;
1225 token = expect(next, ')', "after address_space attribute");
1226 return token;
1229 static struct symbol *to_QI_mode(struct symbol *ctype)
1231 if (ctype->ctype.base_type != &int_type)
1232 return NULL;
1233 if (ctype == &char_ctype)
1234 return ctype;
1235 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uchar_ctype
1236 : &schar_ctype;
1239 static struct symbol *to_HI_mode(struct symbol *ctype)
1241 if (ctype->ctype.base_type != &int_type)
1242 return NULL;
1243 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ushort_ctype
1244 : &sshort_ctype;
1247 static struct symbol *to_SI_mode(struct symbol *ctype)
1249 if (ctype->ctype.base_type != &int_type)
1250 return NULL;
1251 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint_ctype
1252 : &sint_ctype;
1255 static struct symbol *to_DI_mode(struct symbol *ctype)
1257 if (ctype->ctype.base_type != &int_type)
1258 return NULL;
1259 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ullong_ctype
1260 : &sllong_ctype;
1263 static struct symbol *to_TI_mode(struct symbol *ctype)
1265 if (ctype->ctype.base_type != &int_type)
1266 return NULL;
1267 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint128_ctype
1268 : &sint128_ctype;
1271 static struct symbol *to_pointer_mode(struct symbol *ctype)
1273 if (ctype->ctype.base_type != &int_type)
1274 return NULL;
1275 return ctype->ctype.modifiers & MOD_UNSIGNED ? uintptr_ctype
1276 : intptr_ctype;
1279 static struct symbol *to_word_mode(struct symbol *ctype)
1281 if (ctype->ctype.base_type != &int_type)
1282 return NULL;
1283 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulong_ctype
1284 : &slong_ctype;
1287 static struct token *attribute_mode(struct token *token, struct symbol *attr, struct decl_state *ctx)
1289 token = expect(token, '(', "after mode attribute");
1290 if (token_type(token) == TOKEN_IDENT) {
1291 struct symbol *mode = lookup_keyword(token->ident, NS_KEYWORD);
1292 if (mode && mode->op->type & KW_MODE)
1293 ctx->mode = mode->op;
1294 else
1295 sparse_error(token->pos, "unknown mode attribute %s", show_ident(token->ident));
1296 token = token->next;
1297 } else
1298 sparse_error(token->pos, "expect attribute mode symbol\n");
1299 token = expect(token, ')', "after mode attribute");
1300 return token;
1303 static struct token *attribute_context(struct token *token, struct symbol *attr, struct decl_state *ctx)
1305 struct context *context = alloc_context();
1306 struct expression *args[3];
1307 int idx = 0;
1309 token = expect(token, '(', "after context attribute");
1310 token = conditional_expression(token, &args[0]);
1311 token = expect(token, ',', "after context 1st argument");
1312 token = conditional_expression(token, &args[1]);
1313 if (match_op(token, ',')) {
1314 token = token->next;
1315 token = conditional_expression(token, &args[2]);
1316 token = expect(token, ')', "after context 3rd argument");
1317 context->context = args[0];
1318 idx++;
1319 } else {
1320 token = expect(token, ')', "after context 2nd argument");
1322 context->in = get_expression_value(args[idx++]);
1323 context->out = get_expression_value(args[idx++]);
1324 add_ptr_list(&ctx->ctype.contexts, context);
1325 return token;
1328 static struct token *attribute_designated_init(struct token *token, struct symbol *attr, struct decl_state *ctx)
1330 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_STRUCT)
1331 ctx->ctype.base_type->designated_init = 1;
1332 else
1333 warning(token->pos, "attribute designated_init applied to non-structure type");
1334 return token;
1337 static struct token *attribute_transparent_union(struct token *token, struct symbol *attr, struct decl_state *ctx)
1339 if (Wtransparent_union)
1340 warning(token->pos, "attribute __transparent_union__");
1342 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_UNION)
1343 ctx->ctype.base_type->transparent_union = 1;
1344 else
1345 warning(token->pos, "attribute __transparent_union__ applied to non-union type");
1346 return token;
1349 static struct token *recover_unknown_attribute(struct token *token)
1351 struct expression *expr = NULL;
1353 if (Wunknown_attribute)
1354 warning(token->pos, "unknown attribute '%s'", show_ident(token->ident));
1355 token = token->next;
1356 if (match_op(token, '('))
1357 token = parens_expression(token, &expr, "in attribute");
1358 return token;
1361 static struct token *attribute_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1363 token = expect(token, '(', "after attribute");
1364 token = expect(token, '(', "after attribute");
1366 while (token_type(token) == TOKEN_IDENT) {
1367 struct symbol *attr = lookup_keyword(token->ident, NS_KEYWORD);
1368 if (attr && attr->op->attribute)
1369 token = attr->op->attribute(token->next, attr, ctx);
1370 else
1371 token = recover_unknown_attribute(token);
1373 if (!match_op(token, ','))
1374 break;
1375 token = token->next;
1378 token = expect(token, ')', "after attribute");
1379 token = expect(token, ')', "after attribute");
1380 return token;
1383 static unsigned long decl_modifiers(struct decl_state *ctx)
1385 unsigned long mods = ctx->ctype.modifiers & MOD_DECLARE;
1386 ctx->ctype.modifiers &= ~MOD_DECLARE;
1387 return ctx->storage_class | mods;
1390 static struct token *storage_specifier(struct token *next, struct symbol *sym, struct decl_state *ctx)
1392 int is_tls = ctx->ctype.modifiers & MOD_TLS;
1393 unsigned long class = sym->ctype.modifiers;
1394 const char *storage = modifier_name(class);
1396 /* __thread can be used alone, or with extern or static */
1397 if (is_tls && (class & ~(MOD_STATIC|MOD_EXTERN)))
1398 sparse_error(next->pos, "__thread cannot be used with '%s'", storage);
1399 else if (!ctx->storage_class)
1400 ctx->storage_class = class;
1401 else if (ctx->storage_class == class)
1402 sparse_error(next->pos, "duplicate %s", storage);
1403 else
1404 sparse_error(next->pos, "multiple storage classes");
1405 return next;
1408 static struct token *thread_specifier(struct token *next, struct symbol *sym, struct decl_state *ctx)
1410 /* This GCC extension can be used alone, or with extern or static */
1411 if (!(ctx->storage_class & ~(MOD_STATIC|MOD_EXTERN))) {
1412 apply_qualifier(&next->pos, &ctx->ctype, MOD_TLS);
1413 } else {
1414 sparse_error(next->pos, "__thread cannot be used with '%s'",
1415 modifier_name(ctx->storage_class));
1418 return next;
1421 static struct token *attribute_force(struct token *token, struct symbol *attr, struct decl_state *ctx)
1423 ctx->forced = 1;
1424 return token;
1427 static struct token *alignas_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1429 int alignment = 0;
1431 if (!match_op(token, '(')) {
1432 sparse_error(token->pos, "expected '(' after _Alignas");
1433 return token;
1435 if (lookup_type(token->next)) {
1436 struct symbol *sym = NULL;
1437 token = typename(token->next, &sym, NULL);
1438 sym = examine_symbol_type(sym);
1439 alignment = sym->ctype.alignment;
1440 token = expect(token, ')', "after _Alignas(...");
1441 } else {
1442 struct expression *expr = NULL;
1443 token = parens_expression(token, &expr, "after _Alignas");
1444 if (!expr)
1445 return token;
1446 alignment = const_expression_value(expr);
1449 if (alignment < 0) {
1450 warning(token->pos, "non-positive alignment");
1451 return token;
1453 if (alignment & (alignment-1)) {
1454 warning(token->pos, "non-power-of-2 alignment");
1455 return token;
1457 if (alignment > ctx->ctype.alignment)
1458 ctx->ctype.alignment = alignment;
1459 return token;
1462 static struct token *generic_qualifier(struct token *next, struct symbol *sym, struct decl_state *ctx)
1464 apply_qualifier(&next->pos, &ctx->ctype, sym->ctype.modifiers);
1465 return next;
1468 static void apply_ctype(struct position pos, struct ctype *dst, struct ctype *src)
1470 unsigned long mod = src->modifiers;
1472 if (mod)
1473 apply_qualifier(&pos, dst, mod);
1475 /* Context */
1476 concat_ptr_list((struct ptr_list *)src->contexts,
1477 (struct ptr_list **)&dst->contexts);
1479 /* Alignment */
1480 if (src->alignment > dst->alignment)
1481 dst->alignment = src->alignment;
1483 /* Address space */
1484 if (src->as)
1485 dst->as = src->as;
1488 static void specifier_conflict(struct position pos, int what, struct ident *new)
1490 const char *old;
1491 if (what & (Set_S | Set_T))
1492 goto Catch_all;
1493 if (what & Set_Char)
1494 old = "char";
1495 else if (what & Set_Double)
1496 old = "double";
1497 else if (what & Set_Float)
1498 old = "float";
1499 else if (what & Set_Signed)
1500 old = "signed";
1501 else if (what & Set_Unsigned)
1502 old = "unsigned";
1503 else if (what & Set_Short)
1504 old = "short";
1505 else if (what & Set_Long)
1506 old = "long";
1507 else
1508 old = "long long";
1509 sparse_error(pos, "impossible combination of type specifiers: %s %s",
1510 old, show_ident(new));
1511 return;
1513 Catch_all:
1514 sparse_error(pos, "two or more data types in declaration specifiers");
1517 static struct symbol * const int_types[] =
1518 {&char_ctype, &short_ctype, &int_ctype, &long_ctype, &llong_ctype, &int128_ctype};
1519 static struct symbol * const signed_types[] =
1520 {&schar_ctype, &sshort_ctype, &sint_ctype, &slong_ctype, &sllong_ctype,
1521 &sint128_ctype};
1522 static struct symbol * const unsigned_types[] =
1523 {&uchar_ctype, &ushort_ctype, &uint_ctype, &ulong_ctype, &ullong_ctype,
1524 &uint128_ctype};
1525 static struct symbol * const real_types[] =
1526 {&float_ctype, &double_ctype, &ldouble_ctype};
1527 static struct symbol * const * const types[] = {
1528 [CInt] = int_types + 2,
1529 [CSInt] = signed_types + 2,
1530 [CUInt] = unsigned_types + 2,
1531 [CReal] = real_types + 1,
1534 struct symbol *ctype_integer(int size, int want_unsigned)
1536 return types[want_unsigned ? CUInt : CInt][size];
1539 static struct token *handle_qualifiers(struct token *t, struct decl_state *ctx)
1541 while (token_type(t) == TOKEN_IDENT) {
1542 struct symbol *s = lookup_keyword(t->ident, NS_TYPEDEF);
1543 if (!s)
1544 break;
1545 if (!(s->op->type & (KW_ATTRIBUTE | KW_QUALIFIER)))
1546 break;
1547 t = t->next;
1548 if (s->op->declarator)
1549 t = s->op->declarator(t, s, ctx);
1551 return t;
1554 static struct token *declaration_specifiers(struct token *token, struct decl_state *ctx)
1556 int seen = 0;
1557 int class = CInt;
1558 int rank = 0;
1560 while (token_type(token) == TOKEN_IDENT) {
1561 struct symbol *s = lookup_symbol(token->ident,
1562 NS_TYPEDEF | NS_SYMBOL);
1563 if (!s || !(s->namespace & NS_TYPEDEF))
1564 break;
1565 if (s->type != SYM_KEYWORD) {
1566 if (seen & Set_Any)
1567 break;
1568 seen |= Set_S | Set_T;
1569 ctx->ctype.base_type = s->ctype.base_type;
1570 apply_ctype(token->pos, &ctx->ctype, &s->ctype);
1571 token = token->next;
1572 continue;
1574 if (s->op->type & KW_SPECIFIER) {
1575 if (seen & s->op->test) {
1576 specifier_conflict(token->pos,
1577 seen & s->op->test,
1578 token->ident);
1579 break;
1581 seen |= s->op->set;
1582 class += s->op->class;
1583 if (s->op->set & Set_Int128)
1584 rank = 3;
1585 else if (s->op->set & Set_Char)
1586 rank = -2;
1587 if (s->op->set & (Set_Short|Set_Float)) {
1588 rank = -1;
1589 } else if (s->op->set & Set_Long && rank++) {
1590 if (class == CReal) {
1591 specifier_conflict(token->pos,
1592 Set_Vlong,
1593 &double_ident);
1594 break;
1596 seen |= Set_Vlong;
1599 token = token->next;
1600 if (s->op->declarator) // Note: this eats attributes
1601 token = s->op->declarator(token, s, ctx);
1602 if (s->op->type & KW_EXACT) {
1603 ctx->ctype.base_type = s->ctype.base_type;
1604 ctx->ctype.modifiers |= s->ctype.modifiers;
1608 if (!(seen & Set_S)) { /* not set explicitly? */
1609 struct symbol *base = &incomplete_ctype;
1610 if (seen & Set_Any)
1611 base = types[class][rank];
1612 ctx->ctype.base_type = base;
1615 if (ctx->ctype.modifiers & MOD_BITWISE) {
1616 struct symbol *type;
1617 ctx->ctype.modifiers &= ~MOD_BITWISE;
1618 if (!is_int_type(ctx->ctype.base_type)) {
1619 sparse_error(token->pos, "invalid modifier");
1620 return token;
1622 type = alloc_symbol(token->pos, SYM_BASETYPE);
1623 *type = *ctx->ctype.base_type;
1624 type->ctype.modifiers &= ~MOD_SPECIFIER;
1625 type->ctype.base_type = ctx->ctype.base_type;
1626 type->type = SYM_RESTRICT;
1627 ctx->ctype.base_type = type;
1628 create_fouled(type);
1630 return token;
1633 static struct token *abstract_array_declarator(struct token *token, struct symbol *sym)
1635 struct expression *expr = NULL;
1636 int has_static = 0;
1638 while (token_type(token) == TOKEN_IDENT) {
1639 struct symbol *sym = lookup_keyword(token->ident, NS_TYPEDEF);
1640 if (!sym || !(sym->op->type & (KW_STATIC|KW_QUALIFIER)))
1641 break;
1642 if (has_static && (sym->op->type & KW_STATIC))
1643 sparse_error(token->pos, "duplicate array static declarator");
1644 has_static |= (sym->op->type & KW_STATIC);
1645 token = token->next;
1647 if (match_op(token, '*') && match_op(token->next, ']')) {
1648 // FIXME: '[*]' is treated like '[]'
1649 token = token->next;
1650 } else {
1651 token = assignment_expression(token, &expr);
1653 sym->array_size = expr;
1654 return token;
1657 static struct token *parameter_type_list(struct token *, struct symbol *);
1658 static struct token *identifier_list(struct token *, struct symbol *);
1659 static struct token *declarator(struct token *token, struct decl_state *ctx);
1661 static struct token *handle_asm_name(struct token *token, struct decl_state *ctx)
1663 struct expression *expr;
1664 struct symbol *keyword;
1666 if (token_type(token) != TOKEN_IDENT)
1667 return token;
1668 keyword = lookup_keyword(token->ident, NS_KEYWORD);
1669 if (!keyword)
1670 return token;
1671 if (!(keyword->op->type & KW_ASM))
1672 return token;
1674 token = token->next;
1675 token = expect(token, '(', "after asm");
1676 token = string_expression(token, &expr, "asm name");
1677 token = expect(token, ')', "after asm");
1678 return token;
1682 // test if @token is '__attribute__' (or one of its variant)
1683 static bool match_attribute(struct token *token)
1685 struct symbol *sym;
1687 if (token_type(token) != TOKEN_IDENT)
1688 return false;
1689 sym = lookup_keyword(token->ident, NS_TYPEDEF);
1690 if (!sym || !sym->op)
1691 return false;
1692 return sym->op->type & KW_ATTRIBUTE;
1695 static struct token *skip_attribute(struct token *token)
1697 token = token->next;
1698 if (match_op(token, '(')) {
1699 int depth = 1;
1700 token = token->next;
1701 while (depth && !eof_token(token)) {
1702 if (token_type(token) == TOKEN_SPECIAL) {
1703 if (token->special == '(')
1704 depth++;
1705 else if (token->special == ')')
1706 depth--;
1708 token = token->next;
1711 return token;
1714 static struct token *skip_attributes(struct token *token)
1716 while (match_attribute(token)) {
1717 token = expect(token->next, '(', "after attribute");
1718 token = expect(token, '(', "after attribute");
1719 while (token_type(token) == TOKEN_IDENT) {
1720 token = skip_attribute(token);
1721 if (!match_op(token, ','))
1722 break;
1723 token = token->next;
1725 token = expect(token, ')', "after attribute");
1726 token = expect(token, ')', "after attribute");
1728 return token;
1731 static struct token *handle_attributes(struct token *token, struct decl_state *ctx)
1733 while (match_attribute(token))
1734 token = attribute_specifier(token->next, NULL, ctx);
1735 return token;
1738 static int is_nested(struct token *token, struct token **p,
1739 int prefer_abstract)
1742 * This can be either a parameter list or a grouping.
1743 * For the direct (non-abstract) case, we know if must be
1744 * a parameter list if we already saw the identifier.
1745 * For the abstract case, we know if must be a parameter
1746 * list if it is empty or starts with a type.
1748 struct token *next = token->next;
1750 *p = next = skip_attributes(next);
1752 if (token_type(next) == TOKEN_IDENT) {
1753 if (lookup_type(next))
1754 return !prefer_abstract;
1755 return 1;
1758 if (match_op(next, ')') || match_op(next, SPECIAL_ELLIPSIS))
1759 return 0;
1761 return 1;
1764 enum kind {
1765 Empty, K_R, Proto, Bad_Func,
1768 static enum kind which_func(struct token *token,
1769 struct ident **n,
1770 int prefer_abstract)
1772 struct token *next = token->next;
1774 if (token_type(next) == TOKEN_IDENT) {
1775 if (lookup_type(next))
1776 return Proto;
1777 /* identifier list not in definition; complain */
1778 if (prefer_abstract)
1779 warning(token->pos,
1780 "identifier list not in definition");
1781 return K_R;
1784 if (token_type(next) != TOKEN_SPECIAL)
1785 return Bad_Func;
1787 if (next->special == ')') {
1788 /* don't complain about those */
1789 if (!n || match_op(next->next, ';') || match_op(next->next, ','))
1790 return Empty;
1791 if (Wstrict_prototypes)
1792 warning(next->pos,
1793 "non-ANSI function declaration of function '%s'",
1794 show_ident(*n));
1795 return Empty;
1798 if (next->special == SPECIAL_ELLIPSIS) {
1799 warning(next->pos,
1800 "variadic functions must have one named argument");
1801 return Proto;
1804 return Bad_Func;
1807 static struct token *direct_declarator(struct token *token, struct decl_state *ctx)
1809 struct ctype *ctype = &ctx->ctype;
1810 struct token *next;
1811 struct ident **p = ctx->ident;
1813 if (ctx->ident && token_type(token) == TOKEN_IDENT) {
1814 *ctx->ident = token->ident;
1815 token = token->next;
1816 } else if (match_op(token, '(') &&
1817 is_nested(token, &next, ctx->prefer_abstract)) {
1818 struct symbol *base_type = ctype->base_type;
1819 if (token->next != next)
1820 next = handle_attributes(token->next, ctx);
1821 token = declarator(next, ctx);
1822 token = expect(token, ')', "in nested declarator");
1823 while (ctype->base_type != base_type)
1824 ctype = &ctype->base_type->ctype;
1825 p = NULL;
1828 if (match_op(token, '(')) {
1829 enum kind kind = which_func(token, p, ctx->prefer_abstract);
1830 struct symbol *fn;
1831 fn = alloc_indirect_symbol(token->pos, ctype, SYM_FN);
1832 ctype->modifiers |= ctx->f_modifiers;
1833 token = token->next;
1834 if (kind == K_R)
1835 token = identifier_list(token, fn);
1836 else if (kind == Proto)
1837 token = parameter_type_list(token, fn);
1838 token = expect(token, ')', "in function declarator");
1839 fn->endpos = token->pos;
1840 return token;
1843 while (match_op(token, '[')) {
1844 struct symbol *array;
1845 array = alloc_indirect_symbol(token->pos, ctype, SYM_ARRAY);
1846 token = abstract_array_declarator(token->next, array);
1847 token = expect(token, ']', "in abstract_array_declarator");
1848 array->endpos = token->pos;
1849 ctype = &array->ctype;
1851 return token;
1854 static struct token *pointer(struct token *token, struct decl_state *ctx)
1856 while (match_op(token,'*')) {
1857 struct symbol *ptr = alloc_symbol(token->pos, SYM_PTR);
1858 ptr->ctype.modifiers = ctx->ctype.modifiers;
1859 ptr->ctype.base_type = ctx->ctype.base_type;
1860 ptr->ctype.as = ctx->ctype.as;
1861 ptr->ctype.contexts = ctx->ctype.contexts;
1862 ctx->ctype.modifiers = 0;
1863 ctx->ctype.base_type = ptr;
1864 ctx->ctype.as = NULL;
1865 ctx->ctype.contexts = NULL;
1866 ctx->ctype.alignment = 0;
1868 token = handle_qualifiers(token->next, ctx);
1869 ctx->ctype.base_type->endpos = token->pos;
1871 return token;
1874 static struct token *declarator(struct token *token, struct decl_state *ctx)
1876 token = pointer(token, ctx);
1877 return direct_declarator(token, ctx);
1880 static struct token *handle_bitfield(struct token *token, struct decl_state *ctx)
1882 struct ctype *ctype = &ctx->ctype;
1883 struct expression *expr;
1884 struct symbol *bitfield;
1885 long long width;
1887 if (ctype->base_type != &int_type && !is_int_type(ctype->base_type)) {
1888 sparse_error(token->pos, "invalid bitfield specifier for type %s.",
1889 show_typename(ctype->base_type));
1890 // Parse this to recover gracefully.
1891 return conditional_expression(token->next, &expr);
1894 bitfield = alloc_indirect_symbol(token->pos, ctype, SYM_BITFIELD);
1895 token = conditional_expression(token->next, &expr);
1896 width = const_expression_value(expr);
1897 bitfield->bit_size = width;
1899 if (width < 0 || width > INT_MAX || (*ctx->ident && width == 0)) {
1900 sparse_error(token->pos, "bitfield '%s' has invalid width (%lld)",
1901 show_ident(*ctx->ident), width);
1902 width = -1;
1903 } else if (*ctx->ident) {
1904 struct symbol *base_type = bitfield->ctype.base_type;
1905 struct symbol *bitfield_type = base_type == &int_type ? bitfield : base_type;
1906 int is_signed = !(bitfield_type->ctype.modifiers & MOD_UNSIGNED);
1907 if (Wone_bit_signed_bitfield && width == 1 && is_signed) {
1908 // Valid values are either {-1;0} or {0}, depending on integer
1909 // representation. The latter makes for very efficient code...
1910 sparse_error(token->pos, "dubious one-bit signed bitfield");
1912 if (Wdefault_bitfield_sign &&
1913 bitfield_type->type != SYM_ENUM &&
1914 !(bitfield_type->ctype.modifiers & MOD_EXPLICITLY_SIGNED) &&
1915 is_signed) {
1916 // The sign of bitfields is unspecified by default.
1917 warning(token->pos, "dubious bitfield without explicit `signed' or `unsigned'");
1920 bitfield->bit_size = width;
1921 bitfield->endpos = token->pos;
1922 bitfield->ident = *ctx->ident;
1923 return token;
1926 static struct token *declaration_list(struct token *token, struct symbol_list **list)
1928 struct decl_state ctx = {.prefer_abstract = 0};
1929 struct ctype saved;
1930 unsigned long mod;
1932 token = declaration_specifiers(token, &ctx);
1933 mod = decl_modifiers(&ctx);
1934 saved = ctx.ctype;
1935 for (;;) {
1936 struct symbol *decl = alloc_symbol(token->pos, SYM_NODE);
1937 ctx.cleanup = NULL;
1938 ctx.ident = &decl->ident;
1940 token = declarator(token, &ctx);
1941 if (match_op(token, ':'))
1942 token = handle_bitfield(token, &ctx);
1944 token = handle_attributes(token, &ctx);
1945 apply_modifiers(token->pos, &ctx);
1947 decl->ctype = ctx.ctype;
1948 decl->ctype.modifiers |= mod;
1949 decl->cleanup = ctx.cleanup;
1950 decl->endpos = token->pos;
1951 add_symbol(list, decl);
1952 if (!match_op(token, ','))
1953 break;
1954 token = token->next;
1955 ctx.ctype = saved;
1957 return token;
1960 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list)
1962 while (!match_op(token, '}')) {
1963 if (match_ident(token, &_Static_assert_ident)) {
1964 token = parse_static_assert(token, NULL);
1965 continue;
1967 if (!match_op(token, ';'))
1968 token = declaration_list(token, list);
1969 if (!match_op(token, ';')) {
1970 sparse_error(token->pos, "expected ; at end of declaration");
1971 break;
1973 token = token->next;
1975 return token;
1978 static struct token *parameter_declaration(struct token *token, struct symbol *sym)
1980 struct decl_state ctx = {.prefer_abstract = 1};
1982 token = declaration_specifiers(token, &ctx);
1983 ctx.ident = &sym->ident;
1984 token = declarator(token, &ctx);
1985 token = handle_attributes(token, &ctx);
1986 apply_modifiers(token->pos, &ctx);
1987 sym->ctype = ctx.ctype;
1988 sym->ctype.modifiers |= decl_modifiers(&ctx);
1989 sym->endpos = token->pos;
1990 sym->forced_arg = ctx.forced;
1991 return token;
1994 struct token *typename(struct token *token, struct symbol **p, int *forced)
1996 struct decl_state ctx = {.prefer_abstract = 1};
1997 unsigned long class;
1998 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
1999 *p = sym;
2000 token = declaration_specifiers(token, &ctx);
2001 token = declarator(token, &ctx);
2002 apply_modifiers(token->pos, &ctx);
2003 sym->ctype = ctx.ctype;
2004 sym->cleanup = ctx.cleanup;
2005 sym->endpos = token->pos;
2006 class = ctx.storage_class;
2007 if (forced)
2008 *forced = ctx.forced;
2009 if (class)
2010 warning(sym->pos, "storage class in typename (%s%s)",
2011 modifier_string(class), show_typename(sym));
2012 return token;
2015 static struct token *parse_underscore_Pragma(struct token *token)
2017 struct token *next;
2019 next = token->next;
2020 if (!match_op(next, '('))
2021 return next;
2022 next = next->next;
2023 if (next->pos.type != TOKEN_STRING)
2024 return next;
2025 next = next->next;
2026 if (!match_op(next, ')'))
2027 return next;
2028 return next->next;
2031 static struct token *expression_statement(struct token *token, struct expression **tree)
2033 if (match_ident(token, &_Pragma_ident))
2034 return parse_underscore_Pragma(token);
2036 token = parse_expression(token, tree);
2037 return expect(token, ';', "at end of statement");
2040 static struct token *parse_asm_operands(struct token *token, struct statement *stmt,
2041 struct asm_operand_list **inout)
2043 /* Allow empty operands */
2044 if (match_op(token->next, ':') || match_op(token->next, ')'))
2045 return token->next;
2046 do {
2047 struct asm_operand *op = __alloc_asm_operand(0);
2048 if (match_op(token->next, '[') &&
2049 token_type(token->next->next) == TOKEN_IDENT &&
2050 match_op(token->next->next->next, ']')) {
2051 op->name = token->next->next->ident;
2052 token = token->next->next->next;
2054 token = token->next;
2055 token = string_expression(token, &op->constraint, "asm constraint");
2056 token = parens_expression(token, &op->expr, "in asm parameter");
2057 add_ptr_list(inout, op);
2058 } while (match_op(token, ','));
2059 return token;
2062 static struct token *parse_asm_clobbers(struct token *token, struct statement *stmt,
2063 struct expression_list **clobbers)
2065 struct expression *expr;
2067 do {
2068 token = primary_expression(token->next, &expr);
2069 if (expr)
2070 add_expression(clobbers, expr);
2071 } while (match_op(token, ','));
2072 return token;
2075 static struct token *parse_asm_labels(struct token *token, struct statement *stmt,
2076 struct symbol_list **labels)
2078 struct symbol *label;
2080 do {
2081 token = token->next; /* skip ':' and ',' */
2082 if (token_type(token) != TOKEN_IDENT)
2083 return token;
2084 label = label_symbol(token, 1);
2085 add_symbol(labels, label);
2086 token = token->next;
2087 } while (match_op(token, ','));
2088 return token;
2091 static struct token *parse_asm_statement(struct token *token, struct statement *stmt)
2093 unsigned long mods = 0;
2095 token = token->next;
2096 stmt->type = STMT_ASM;
2097 while (token_type(token) == TOKEN_IDENT) {
2098 struct symbol *s = lookup_keyword(token->ident, NS_TYPEDEF);
2099 if (s && s->op->asm_modifier)
2100 s->op->asm_modifier(token, &mods, s->ctype.modifiers);
2101 else if (token->ident == &goto_ident)
2102 asm_modifier(token, &mods, MOD_ASM_GOTO);
2103 token = token->next;
2105 token = expect(token, '(', "after asm");
2106 token = string_expression(token, &stmt->asm_string, "inline asm");
2107 if (match_op(token, ':'))
2108 token = parse_asm_operands(token, stmt, &stmt->asm_outputs);
2109 if (match_op(token, ':'))
2110 token = parse_asm_operands(token, stmt, &stmt->asm_inputs);
2111 if (match_op(token, ':'))
2112 token = parse_asm_clobbers(token, stmt, &stmt->asm_clobbers);
2113 if (match_op(token, ':') && (mods & MOD_ASM_GOTO))
2114 token = parse_asm_labels(token, stmt, &stmt->asm_labels);
2115 token = expect(token, ')', "after asm");
2116 return expect(token, ';', "at end of asm-statement");
2119 static struct token *parse_static_assert(struct token *token, struct symbol_list **unused)
2121 struct expression *cond = NULL, *message = NULL;
2123 token = expect(token->next, '(', "after _Static_assert");
2124 token = constant_expression(token, &cond);
2125 if (!cond)
2126 sparse_error(token->pos, "Expected constant expression");
2127 if (match_op(token, ',')) {
2128 token = token->next;
2129 token = string_expression(token, &message, "_Static_assert()");
2130 if (!message)
2131 cond = NULL;
2133 token = expect(token, ')', "after diagnostic message in _Static_assert");
2134 token = expect(token, ';', "after _Static_assert()");
2136 if (cond && !const_expression_value(cond) && cond->type == EXPR_VALUE) {
2137 const char *sep = "", *msg = "";
2139 if (message) {
2140 sep = ": ";
2141 msg = show_string(message->string);
2143 sparse_error(cond->pos, "static assertion failed%s%s", sep, msg);
2146 return token;
2149 static struct token *ignore_seg_gs(struct token *token, struct symbol_list **unused)
2151 return token->next;
2154 /* Make a statement out of an expression */
2155 static struct statement *make_statement(struct expression *expr)
2157 struct statement *stmt;
2159 if (!expr)
2160 return NULL;
2161 stmt = alloc_statement(expr->pos, STMT_EXPRESSION);
2162 stmt->expression = expr;
2163 return stmt;
2167 * All iterators have two symbols associated with them:
2168 * the "continue" and "break" symbols, which are targets
2169 * for continue and break statements respectively.
2171 * They are in a special name-space, but they follow
2172 * all the normal visibility rules, so nested iterators
2173 * automatically work right.
2175 static void start_iterator(struct statement *stmt)
2177 struct symbol *cont, *brk;
2179 start_block_scope(stmt->pos);
2180 cont = alloc_symbol(stmt->pos, SYM_NODE);
2181 bind_symbol(cont, &continue_ident, NS_ITERATOR);
2182 brk = alloc_symbol(stmt->pos, SYM_NODE);
2183 bind_symbol(brk, &break_ident, NS_ITERATOR);
2185 stmt->type = STMT_ITERATOR;
2186 stmt->iterator_break = brk;
2187 stmt->iterator_continue = cont;
2188 fn_local_symbol(brk);
2189 fn_local_symbol(cont);
2192 static void end_iterator(struct statement *stmt)
2194 end_block_scope();
2197 static struct statement *start_function(struct symbol *sym)
2199 struct symbol *ret;
2200 struct statement *stmt = alloc_statement(sym->pos, STMT_COMPOUND);
2202 start_function_scope(sym->pos);
2203 ret = alloc_symbol(sym->pos, SYM_NODE);
2204 ret->ctype = sym->ctype.base_type->ctype;
2205 ret->ctype.modifiers &= ~(MOD_STORAGE | MOD_QUALIFIER | MOD_TLS | MOD_ACCESS | MOD_NOCAST | MOD_NODEREF);
2206 ret->ctype.modifiers |= (MOD_AUTO | MOD_REGISTER);
2207 bind_symbol(ret, &return_ident, NS_ITERATOR);
2208 stmt->ret = ret;
2209 fn_local_symbol(ret);
2211 // Currently parsed symbol for __func__/__FUNCTION__/__PRETTY_FUNCTION__
2212 current_fn = sym;
2214 return stmt;
2217 static void end_function(struct symbol *sym)
2219 current_fn = NULL;
2220 end_function_scope();
2224 * A "switch()" statement, like an iterator, has a
2225 * the "break" symbol associated with it. It works
2226 * exactly like the iterator break - it's the target
2227 * for any break-statements in scope, and means that
2228 * "break" handling doesn't even need to know whether
2229 * it's breaking out of an iterator or a switch.
2231 * In addition, the "case" symbol is a marker for the
2232 * case/default statements to find the switch statement
2233 * that they are associated with.
2235 static void start_switch(struct statement *stmt)
2237 struct symbol *brk, *switch_case;
2239 start_block_scope(stmt->pos);
2240 brk = alloc_symbol(stmt->pos, SYM_NODE);
2241 bind_symbol(brk, &break_ident, NS_ITERATOR);
2243 switch_case = alloc_symbol(stmt->pos, SYM_NODE);
2244 bind_symbol(switch_case, &case_ident, NS_ITERATOR);
2245 switch_case->stmt = stmt;
2247 stmt->type = STMT_SWITCH;
2248 stmt->switch_break = brk;
2249 stmt->switch_case = switch_case;
2251 fn_local_symbol(brk);
2252 fn_local_symbol(switch_case);
2255 static void end_switch(struct statement *stmt)
2257 if (!stmt->switch_case->symbol_list)
2258 warning(stmt->pos, "switch with no cases");
2259 end_block_scope();
2262 static void add_case_statement(struct statement *stmt)
2264 struct symbol *target = lookup_symbol(&case_ident, NS_ITERATOR);
2265 struct symbol *sym;
2267 if (!target) {
2268 sparse_error(stmt->pos, "not in switch scope");
2269 stmt->type = STMT_NONE;
2270 return;
2272 sym = alloc_symbol(stmt->pos, SYM_NODE);
2273 add_symbol(&target->symbol_list, sym);
2274 sym->stmt = stmt;
2275 stmt->case_label = sym;
2276 fn_local_symbol(sym);
2279 static struct token *parse_return_statement(struct token *token, struct statement *stmt)
2281 struct symbol *target = lookup_symbol(&return_ident, NS_ITERATOR);
2283 if (!target)
2284 error_die(token->pos, "internal error: return without a function target");
2285 stmt->type = STMT_RETURN;
2286 stmt->ret_target = target;
2287 return expression_statement(token->next, &stmt->ret_value);
2290 static void validate_for_loop_decl(struct symbol *sym)
2292 unsigned long storage = sym->ctype.modifiers & MOD_STORAGE;
2294 if (storage & ~(MOD_AUTO | MOD_REGISTER)) {
2295 const char *name = show_ident(sym->ident);
2296 sparse_error(sym->pos, "non-local var '%s' in for-loop initializer", name);
2297 sym->ctype.modifiers &= ~MOD_STORAGE;
2301 static struct token *parse_for_statement(struct token *token, struct statement *stmt)
2303 struct symbol_list *syms;
2304 struct expression *e1, *e2, *e3;
2305 struct statement *iterator;
2307 start_iterator(stmt);
2308 token = expect(token->next, '(', "after 'for'");
2310 syms = NULL;
2311 e1 = NULL;
2312 /* C99 variable declaration? */
2313 if (lookup_type(token)) {
2314 token = external_declaration(token, &syms, validate_for_loop_decl);
2315 } else {
2316 token = parse_expression(token, &e1);
2317 token = expect(token, ';', "in 'for'");
2319 token = parse_expression(token, &e2);
2320 token = expect(token, ';', "in 'for'");
2321 token = parse_expression(token, &e3);
2322 token = expect(token, ')', "in 'for'");
2323 token = statement(token, &iterator);
2325 stmt->iterator_syms = syms;
2326 stmt->iterator_pre_statement = make_statement(e1);
2327 stmt->iterator_pre_condition = e2;
2328 stmt->iterator_post_statement = make_statement(e3);
2329 stmt->iterator_post_condition = NULL;
2330 stmt->iterator_statement = iterator;
2331 end_iterator(stmt);
2333 return token;
2336 static struct token *parse_while_statement(struct token *token, struct statement *stmt)
2338 struct expression *expr;
2339 struct statement *iterator;
2341 start_iterator(stmt);
2342 token = parens_expression(token->next, &expr, "after 'while'");
2343 token = statement(token, &iterator);
2345 stmt->iterator_pre_condition = expr;
2346 stmt->iterator_post_condition = NULL;
2347 stmt->iterator_statement = iterator;
2348 end_iterator(stmt);
2350 return token;
2353 static struct token *parse_do_statement(struct token *token, struct statement *stmt)
2355 struct expression *expr;
2356 struct statement *iterator;
2358 start_iterator(stmt);
2359 token = statement(token->next, &iterator);
2360 if (token_type(token) == TOKEN_IDENT && token->ident == &while_ident)
2361 token = token->next;
2362 else
2363 sparse_error(token->pos, "expected 'while' after 'do'");
2364 token = parens_expression(token, &expr, "after 'do-while'");
2366 stmt->iterator_post_condition = expr;
2367 stmt->iterator_statement = iterator;
2368 end_iterator(stmt);
2370 if (iterator && iterator->type != STMT_COMPOUND && Wdo_while)
2371 warning(iterator->pos, "do-while statement is not a compound statement");
2373 return expect(token, ';', "after statement");
2376 static struct token *parse_if_statement(struct token *token, struct statement *stmt)
2378 stmt->type = STMT_IF;
2379 token = parens_expression(token->next, &stmt->if_conditional, "after if");
2380 token = statement(token, &stmt->if_true);
2381 if (token_type(token) != TOKEN_IDENT)
2382 return token;
2383 if (token->ident != &else_ident)
2384 return token;
2385 return statement(token->next, &stmt->if_false);
2388 static inline struct token *case_statement(struct token *token, struct statement *stmt)
2390 stmt->type = STMT_CASE;
2391 token = expect(token, ':', "after default/case");
2392 add_case_statement(stmt);
2393 if (match_op(token, '}')) {
2394 warning(token->pos, "statement expected after case label");
2395 stmt->case_statement = alloc_statement(token->pos, STMT_NONE);
2396 return token;
2398 return statement(token, &stmt->case_statement);
2401 static struct token *parse_case_statement(struct token *token, struct statement *stmt)
2403 token = parse_expression(token->next, &stmt->case_expression);
2404 if (match_op(token, SPECIAL_ELLIPSIS))
2405 token = parse_expression(token->next, &stmt->case_to);
2406 return case_statement(token, stmt);
2409 static struct token *parse_default_statement(struct token *token, struct statement *stmt)
2411 return case_statement(token->next, stmt);
2414 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt)
2416 struct symbol *target = lookup_symbol(token->ident, NS_ITERATOR);
2417 stmt->type = STMT_GOTO;
2418 stmt->goto_label = target;
2419 if (!target)
2420 sparse_error(stmt->pos, "break/continue not in iterator scope");
2421 return expect(token->next, ';', "at end of statement");
2424 static struct token *parse_switch_statement(struct token *token, struct statement *stmt)
2426 stmt->type = STMT_SWITCH;
2427 start_switch(stmt);
2428 token = parens_expression(token->next, &stmt->switch_expression, "after 'switch'");
2429 token = statement(token, &stmt->switch_statement);
2430 end_switch(stmt);
2431 return token;
2434 static void warn_label_usage(struct position def, struct position use, struct ident *ident)
2436 const char *id = show_ident(ident);
2437 sparse_error(use, "label '%s' used outside statement expression", id);
2438 info(def, " label '%s' defined here", id);
2439 current_fn->bogus_linear = 1;
2442 void check_label_usage(struct symbol *label, struct position use_pos)
2444 struct statement *def = label->stmt;
2446 if (def) {
2447 if (!is_in_scope(def->label_scope, label_scope))
2448 warn_label_usage(def->pos, use_pos, label->ident);
2449 } else if (!label->label_scope) {
2450 label->label_scope = label_scope;
2451 label->label_pos = use_pos;
2455 static struct token *parse_goto_statement(struct token *token, struct statement *stmt)
2457 stmt->type = STMT_GOTO;
2458 token = token->next;
2459 if (match_op(token, '*')) {
2460 token = parse_expression(token->next, &stmt->goto_expression);
2461 add_statement(&function_computed_goto_list, stmt);
2462 } else if (token_type(token) == TOKEN_IDENT) {
2463 struct symbol *label = label_symbol(token, 1);
2464 stmt->goto_label = label;
2465 check_label_usage(label, stmt->pos);
2466 token = token->next;
2467 } else {
2468 sparse_error(token->pos, "Expected identifier or goto expression");
2470 return expect(token, ';', "at end of statement");
2473 static struct token *parse_context_statement(struct token *token, struct statement *stmt)
2475 stmt->type = STMT_CONTEXT;
2476 token = token->next;
2477 token = expect(token, '(', "after __context__ statement");
2478 token = assignment_expression(token, &stmt->expression);
2479 if (!stmt->expression)
2480 unexpected(token, "expression expected after '('");
2481 if (match_op(token, ',')) {
2482 token = token->next;
2483 stmt->context = stmt->expression;
2484 token = assignment_expression(token, &stmt->expression);
2485 if (!stmt->expression)
2486 unexpected(token, "expression expected after ','");
2488 token = expect(token, ')', "at end of __context__ statement");
2489 return expect(token, ';', "at end of statement");
2492 static struct token *parse_range_statement(struct token *token, struct statement *stmt)
2494 stmt->type = STMT_RANGE;
2495 token = token->next;
2496 token = expect(token, '(', "after __range__ statement");
2497 token = assignment_expression(token, &stmt->range_expression);
2498 token = expect(token, ',', "after range expression");
2499 token = assignment_expression(token, &stmt->range_low);
2500 token = expect(token, ',', "after low range");
2501 token = assignment_expression(token, &stmt->range_high);
2502 token = expect(token, ')', "after range statement");
2503 return expect(token, ';', "after range statement");
2506 static struct token *handle_label_attributes(struct token *token, struct symbol *label)
2508 struct decl_state ctx = { };
2510 token = handle_attributes(token, &ctx);
2511 label->label_modifiers = ctx.ctype.modifiers;
2512 return token;
2515 static struct token *statement(struct token *token, struct statement **tree)
2517 struct statement *stmt = alloc_statement(token->pos, STMT_NONE);
2519 *tree = stmt;
2520 if (token_type(token) == TOKEN_IDENT) {
2521 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2522 if (s && s->op->statement)
2523 return s->op->statement(token, stmt);
2525 if (match_op(token->next, ':')) {
2526 struct symbol *s = label_symbol(token, 0);
2527 token = handle_label_attributes(token->next->next, s);
2528 if (s->stmt) {
2529 sparse_error(stmt->pos, "label '%s' redefined", show_ident(s->ident));
2530 // skip the label to avoid multiple definitions
2531 return statement(token, tree);
2533 stmt->type = STMT_LABEL;
2534 stmt->label_identifier = s;
2535 stmt->label_scope = label_scope;
2536 if (s->label_scope) {
2537 if (!is_in_scope(label_scope, s->label_scope))
2538 warn_label_usage(stmt->pos, s->label_pos, s->ident);
2540 s->stmt = stmt;
2541 if (match_op(token, '}')) {
2542 warning(token->pos, "statement expected after label");
2543 stmt->label_statement = alloc_statement(token->pos, STMT_NONE);
2544 return token;
2546 return statement(token, &stmt->label_statement);
2550 if (match_op(token, '{')) {
2551 token = compound_statement(token->next, stmt);
2552 return expect(token, '}', "at end of compound statement");
2555 stmt->type = STMT_EXPRESSION;
2556 return expression_statement(token, &stmt->expression);
2559 /* gcc extension - __label__ ident-list; in the beginning of compound stmt */
2560 static struct token *label_statement(struct token *token)
2562 while (token_type(token) == TOKEN_IDENT) {
2563 struct symbol *sym = alloc_symbol(token->pos, SYM_LABEL);
2564 /* it's block-scope, but we want label namespace */
2565 bind_symbol_with_scope(sym, token->ident, NS_LABEL, block_scope);
2566 fn_local_symbol(sym);
2567 token = token->next;
2568 if (!match_op(token, ','))
2569 break;
2570 token = token->next;
2572 return expect(token, ';', "at end of label declaration");
2575 static struct token * statement_list(struct token *token, struct statement_list **list)
2577 int seen_statement = 0;
2578 while (token_type(token) == TOKEN_IDENT &&
2579 token->ident == &__label___ident)
2580 token = label_statement(token->next);
2581 for (;;) {
2582 struct statement * stmt;
2583 if (eof_token(token))
2584 break;
2585 if (match_op(token, '}'))
2586 break;
2587 if (match_ident(token, &_Static_assert_ident)) {
2588 token = parse_static_assert(token, NULL);
2589 continue;
2591 if (lookup_type(token)) {
2592 if (seen_statement) {
2593 warning(token->pos, "mixing declarations and code");
2594 seen_statement = 0;
2596 stmt = alloc_statement(token->pos, STMT_DECLARATION);
2597 token = external_declaration(token, &stmt->declaration, NULL);
2598 } else {
2599 seen_statement = Wdeclarationafterstatement;
2600 token = statement(token, &stmt);
2602 add_statement(list, stmt);
2604 return token;
2607 static struct token *identifier_list(struct token *token, struct symbol *fn)
2609 struct symbol_list **list = &fn->arguments;
2610 for (;;) {
2611 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2612 sym->ident = token->ident;
2613 token = token->next;
2614 sym->endpos = token->pos;
2615 sym->ctype.base_type = &incomplete_ctype;
2616 add_symbol(list, sym);
2617 if (!match_op(token, ',') ||
2618 token_type(token->next) != TOKEN_IDENT ||
2619 lookup_type(token->next))
2620 break;
2621 token = token->next;
2623 return token;
2626 static struct token *parameter_type_list(struct token *token, struct symbol *fn)
2628 struct symbol_list **list = &fn->arguments;
2630 for (;;) {
2631 struct symbol *sym;
2633 if (match_op(token, SPECIAL_ELLIPSIS)) {
2634 fn->variadic = 1;
2635 token = token->next;
2636 break;
2639 sym = alloc_symbol(token->pos, SYM_NODE);
2640 token = parameter_declaration(token, sym);
2641 if (sym->ctype.base_type == &void_ctype) {
2642 /* Special case: (void) */
2643 if (!*list && !sym->ident)
2644 break;
2645 warning(token->pos, "void parameter");
2647 add_symbol(list, sym);
2648 if (!match_op(token, ','))
2649 break;
2650 token = token->next;
2652 return token;
2655 struct token *compound_statement(struct token *token, struct statement *stmt)
2657 stmt->type = STMT_COMPOUND;
2658 start_block_scope(token->pos);
2659 token = statement_list(token, &stmt->stmts);
2660 end_block_scope();
2661 return token;
2664 static struct expression *identifier_expression(struct token *token)
2666 struct expression *expr = alloc_expression(token->pos, EXPR_IDENTIFIER);
2667 expr->expr_ident = token->ident;
2668 return expr;
2671 static struct expression *index_expression(struct expression *from, struct expression *to)
2673 int idx_from, idx_to;
2674 struct expression *expr = alloc_expression(from->pos, EXPR_INDEX);
2676 idx_from = const_expression_value(from);
2677 idx_to = idx_from;
2678 if (to) {
2679 idx_to = const_expression_value(to);
2680 if (idx_to < idx_from || idx_from < 0)
2681 warning(from->pos, "nonsense array initializer index range");
2683 expr->idx_from = idx_from;
2684 expr->idx_to = idx_to;
2685 return expr;
2688 static struct token *single_initializer(struct expression **ep, struct token *token)
2690 int expect_equal = 0;
2691 struct token *next = token->next;
2692 struct expression **tail = ep;
2693 int nested;
2695 *ep = NULL;
2697 if ((token_type(token) == TOKEN_IDENT) && match_op(next, ':')) {
2698 struct expression *expr = identifier_expression(token);
2699 if (Wold_initializer)
2700 warning(token->pos, "obsolete struct initializer, use C99 syntax");
2701 token = initializer(&expr->ident_expression, next->next);
2702 if (expr->ident_expression)
2703 *ep = expr;
2704 return token;
2707 for (tail = ep, nested = 0; ; nested++, next = token->next) {
2708 if (match_op(token, '.') && (token_type(next) == TOKEN_IDENT)) {
2709 struct expression *expr = identifier_expression(next);
2710 *tail = expr;
2711 tail = &expr->ident_expression;
2712 expect_equal = 1;
2713 token = next->next;
2714 } else if (match_op(token, '[')) {
2715 struct expression *from = NULL, *to = NULL, *expr;
2716 token = constant_expression(token->next, &from);
2717 if (!from) {
2718 sparse_error(token->pos, "Expected constant expression");
2719 break;
2721 if (match_op(token, SPECIAL_ELLIPSIS))
2722 token = constant_expression(token->next, &to);
2723 expr = index_expression(from, to);
2724 *tail = expr;
2725 tail = &expr->idx_expression;
2726 token = expect(token, ']', "at end of initializer index");
2727 if (nested)
2728 expect_equal = 1;
2729 } else {
2730 break;
2733 if (nested && !expect_equal) {
2734 if (!match_op(token, '='))
2735 warning(token->pos, "obsolete array initializer, use C99 syntax");
2736 else
2737 expect_equal = 1;
2739 if (expect_equal)
2740 token = expect(token, '=', "at end of initializer index");
2742 token = initializer(tail, token);
2743 if (!*tail)
2744 *ep = NULL;
2745 return token;
2748 static struct token *initializer_list(struct expression_list **list, struct token *token)
2750 struct expression *expr;
2752 for (;;) {
2753 token = single_initializer(&expr, token);
2754 if (!expr)
2755 break;
2756 add_expression(list, expr);
2757 if (!match_op(token, ','))
2758 break;
2759 token = token->next;
2761 return token;
2764 struct token *initializer(struct expression **tree, struct token *token)
2766 if (match_op(token, '{')) {
2767 struct expression *expr = alloc_expression(token->pos, EXPR_INITIALIZER);
2768 *tree = expr;
2769 if (!Wuniversal_initializer) {
2770 struct token *next = token->next;
2771 // '{ 0 }' is equivalent to '{ }' except for some
2772 // warnings, like using 0 to initialize a null-pointer.
2773 if (match_token_zero(next)) {
2774 if (match_op(next->next, '}'))
2775 expr->zero_init = 1;
2779 token = initializer_list(&expr->expr_list, token->next);
2780 return expect(token, '}', "at end of initializer");
2782 return assignment_expression(token, tree);
2785 static void declare_argument(struct symbol *sym, struct symbol *fn)
2787 if (!sym->ident) {
2788 sparse_error(sym->pos, "no identifier for function argument");
2789 return;
2791 if (sym->ctype.base_type == &incomplete_ctype) {
2792 sym->ctype.base_type = &int_ctype;
2794 if (Wimplicit_int) {
2795 sparse_error(sym->pos, "missing type declaration for parameter '%s'",
2796 show_ident(sym->ident));
2799 bind_symbol(sym, sym->ident, NS_SYMBOL);
2802 static int is_syscall(struct symbol *sym)
2804 char *macro;
2805 char *name;
2806 int is_syscall = 0;
2808 macro = get_macro_name(sym->pos);
2809 if (macro &&
2810 (strncmp("SYSCALL_DEFINE", macro, strlen("SYSCALL_DEFINE")) == 0 ||
2811 strncmp("COMPAT_SYSCALL_DEFINE", macro, strlen("COMPAT_SYSCALL_DEFINE")) == 0))
2812 is_syscall = 1;
2814 name = sym->ident->name;
2816 if (name && strncmp(name, "sys_", 4) ==0)
2817 is_syscall = 1;
2819 if (name && strncmp(name, "compat_sys_", 11) == 0)
2820 is_syscall = 1;
2822 return is_syscall;
2826 static struct token *parse_function_body(struct token *token, struct symbol *decl,
2827 struct symbol_list **list)
2829 struct symbol_list **old_symbol_list;
2830 struct symbol *base_type = decl->ctype.base_type;
2831 struct statement *stmt, **p;
2832 struct symbol *prev;
2833 struct symbol *arg;
2835 old_symbol_list = function_symbol_list;
2836 if (decl->ctype.modifiers & MOD_INLINE) {
2837 function_symbol_list = &decl->inline_symbol_list;
2838 p = &base_type->inline_stmt;
2839 } else {
2840 function_symbol_list = &decl->symbol_list;
2841 p = &base_type->stmt;
2843 function_computed_target_list = NULL;
2844 function_computed_goto_list = NULL;
2846 if ((decl->ctype.modifiers & (MOD_EXTERN|MOD_INLINE)) == MOD_EXTERN) {
2847 if (Wexternal_function_has_definition)
2848 warning(decl->pos, "function '%s' with external linkage has definition", show_ident(decl->ident));
2850 if (!(decl->ctype.modifiers & MOD_STATIC))
2851 decl->ctype.modifiers |= MOD_EXTERN;
2853 stmt = start_function(decl);
2854 *p = stmt;
2856 FOR_EACH_PTR (base_type->arguments, arg) {
2857 declare_argument(arg, base_type);
2858 } END_FOR_EACH_PTR(arg);
2860 token = statement_list(token->next, &stmt->stmts);
2861 end_function(decl);
2863 if (!(decl->ctype.modifiers & MOD_INLINE))
2864 add_symbol(list, decl);
2865 else if (is_syscall(decl)) {
2866 add_symbol(list, decl);
2868 printf("parse.c decl: %s\n", decl->ident->name);
2869 char *macro = get_macro_name(decl->pos);
2870 printf("decl macro: %s\n", macro);
2873 check_declaration(decl);
2874 decl->definition = decl;
2875 prev = decl->same_symbol;
2876 if (prev && prev->definition) {
2877 warning(decl->pos, "multiple definitions for function '%s'",
2878 show_ident(decl->ident));
2879 info(prev->definition->pos, " the previous one is here");
2880 } else {
2881 while (prev) {
2882 rebind_scope(prev, decl->scope);
2883 prev->definition = decl;
2884 prev = prev->same_symbol;
2887 function_symbol_list = old_symbol_list;
2888 if (function_computed_goto_list) {
2889 if (!function_computed_target_list)
2890 warning(decl->pos, "function '%s' has computed goto but no targets?", show_ident(decl->ident));
2891 else {
2892 FOR_EACH_PTR(function_computed_goto_list, stmt) {
2893 stmt->target_list = function_computed_target_list;
2894 } END_FOR_EACH_PTR(stmt);
2897 return expect(token, '}', "at end of function");
2900 static void promote_k_r_types(struct symbol *arg)
2902 struct symbol *base = arg->ctype.base_type;
2903 if (base && base->ctype.base_type == &int_type && base->rank < 0) {
2904 arg->ctype.base_type = &int_ctype;
2908 static void apply_k_r_types(struct symbol_list *argtypes, struct symbol *fn)
2910 struct symbol_list *real_args = fn->ctype.base_type->arguments;
2911 struct symbol *arg;
2913 FOR_EACH_PTR(real_args, arg) {
2914 struct symbol *type;
2916 /* This is quadratic in the number of arguments. We _really_ don't care */
2917 FOR_EACH_PTR(argtypes, type) {
2918 if (type->ident == arg->ident)
2919 goto match;
2920 } END_FOR_EACH_PTR(type);
2921 if (Wimplicit_int) {
2922 warning(arg->pos, "missing type declaration for parameter '%s'",
2923 show_ident(arg->ident));
2925 type = alloc_symbol(arg->pos, SYM_NODE);
2926 type->ident = arg->ident;
2927 type->ctype.base_type = &int_ctype;
2928 match:
2929 type->used = 1;
2930 /* "char" and "short" promote to "int" */
2931 promote_k_r_types(type);
2933 arg->ctype = type->ctype;
2934 } END_FOR_EACH_PTR(arg);
2936 FOR_EACH_PTR(argtypes, arg) {
2937 if (!arg->used)
2938 warning(arg->pos, "nonsensical parameter declaration '%s'", show_ident(arg->ident));
2939 } END_FOR_EACH_PTR(arg);
2943 static struct token *parse_k_r_arguments(struct token *token, struct symbol *decl,
2944 struct symbol_list **list)
2946 struct symbol_list *args = NULL;
2948 if (Wold_style_definition)
2949 warning(token->pos, "non-ANSI definition of function '%s'", show_ident(decl->ident));
2951 do {
2952 token = declaration_list(token, &args);
2953 if (!match_op(token, ';')) {
2954 sparse_error(token->pos, "expected ';' at end of parameter declaration");
2955 break;
2957 token = token->next;
2958 } while (lookup_type(token));
2960 apply_k_r_types(args, decl);
2962 if (!match_op(token, '{')) {
2963 sparse_error(token->pos, "expected function body");
2964 return token;
2966 return parse_function_body(token, decl, list);
2969 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list)
2971 struct symbol *anon = alloc_symbol(token->pos, SYM_NODE);
2972 struct symbol *fn = alloc_symbol(token->pos, SYM_FN);
2973 struct statement *stmt;
2975 anon->ctype.base_type = fn;
2976 stmt = alloc_statement(token->pos, STMT_NONE);
2977 fn->stmt = stmt;
2979 token = parse_asm_statement(token, stmt);
2981 // FIXME: add_symbol(list, anon);
2982 return token;
2985 struct token *external_declaration(struct token *token, struct symbol_list **list,
2986 validate_decl_t validate_decl)
2988 struct ident *ident = NULL;
2989 struct symbol *decl;
2990 struct decl_state ctx = { .ident = &ident };
2991 struct ctype saved;
2992 struct symbol *base_type;
2993 unsigned long mod;
2994 int is_typedef;
2996 if (match_ident(token, &_Pragma_ident))
2997 return parse_underscore_Pragma(token);
2999 /* Top-level inline asm or static assertion? */
3000 if (token_type(token) == TOKEN_IDENT) {
3001 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
3002 if (s && s->op->toplevel)
3003 return s->op->toplevel(token, list);
3006 /* Parse declaration-specifiers, if any */
3007 token = declaration_specifiers(token, &ctx);
3008 mod = decl_modifiers(&ctx);
3009 decl = alloc_symbol(token->pos, SYM_NODE);
3010 /* Just a type declaration? */
3011 if (match_op(token, ';')) {
3012 apply_modifiers(token->pos, &ctx);
3013 return token->next;
3016 saved = ctx.ctype;
3017 token = declarator(token, &ctx);
3018 token = handle_asm_name(token, &ctx);
3019 token = handle_attributes(token, &ctx);
3020 apply_modifiers(token->pos, &ctx);
3022 decl->ctype = ctx.ctype;
3023 decl->ctype.modifiers |= mod;
3024 decl->cleanup = ctx.cleanup;
3025 decl->endpos = token->pos;
3027 /* Just a type declaration? */
3028 if (!ident) {
3029 warning(token->pos, "missing identifier in declaration");
3030 return expect(token, ';', "at the end of type declaration");
3033 /* type define declaration? */
3034 is_typedef = ctx.storage_class == MOD_USERTYPE;
3036 /* Typedefs don't have meaningful storage */
3037 if (is_typedef)
3038 decl->ctype.modifiers |= MOD_USERTYPE;
3040 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
3042 base_type = decl->ctype.base_type;
3044 if (is_typedef) {
3045 if (base_type && !base_type->ident) {
3046 switch (base_type->type) {
3047 case SYM_STRUCT:
3048 case SYM_UNION:
3049 case SYM_ENUM:
3050 case SYM_RESTRICT:
3051 base_type->ident = ident;
3052 break;
3053 default:
3054 break;
3057 } else if (base_type && base_type->type == SYM_FN) {
3058 if (base_type->ctype.base_type == &autotype_ctype) {
3059 sparse_error(decl->pos, "'%s()' has __auto_type return type",
3060 show_ident(decl->ident));
3061 base_type->ctype.base_type = &int_ctype;
3063 if (base_type->ctype.base_type == &incomplete_ctype) {
3064 warning(decl->pos, "'%s()' has implicit return type",
3065 show_ident(decl->ident));
3066 base_type->ctype.base_type = &int_ctype;
3068 /* apply attributes placed after the declarator */
3069 decl->ctype.modifiers |= ctx.f_modifiers;
3071 /* K&R argument declaration? */
3072 if (lookup_type(token))
3073 return parse_k_r_arguments(token, decl, list);
3074 if (match_op(token, '{'))
3075 return parse_function_body(token, decl, list);
3077 if (!(decl->ctype.modifiers & MOD_STATIC))
3078 decl->ctype.modifiers |= MOD_EXTERN;
3079 } else if (base_type == &void_ctype && !(decl->ctype.modifiers & MOD_EXTERN)) {
3080 sparse_error(token->pos, "void declaration");
3082 if (base_type == &incomplete_ctype) {
3083 warning(decl->pos, "'%s' has implicit type", show_ident(decl->ident));
3084 decl->ctype.base_type = &int_ctype;;
3087 for (;;) {
3088 if (!is_typedef && match_op(token, '=')) {
3089 struct token *next = token->next;
3090 token = initializer(&decl->initializer, next);
3091 if (token == next)
3092 sparse_error(token->pos, "expression expected before '%s'", show_token(token));
3094 if (!is_typedef) {
3095 if (validate_decl)
3096 validate_decl(decl);
3098 if (decl->initializer && decl->ctype.modifiers & MOD_EXTERN) {
3099 warning(decl->pos, "symbol with external linkage has initializer");
3100 decl->ctype.modifiers &= ~MOD_EXTERN;
3103 if (!(decl->ctype.modifiers & (MOD_EXTERN | MOD_INLINE))) {
3104 add_symbol(list, decl);
3105 fn_local_symbol(decl);
3108 check_declaration(decl);
3109 if (decl->same_symbol) {
3110 decl->definition = decl->same_symbol->definition;
3111 decl->op = decl->same_symbol->op;
3112 if (is_typedef) {
3113 // TODO: handle -std=c89 --pedantic
3114 check_duplicates(decl);
3118 if (ctx.autotype) {
3119 const char *msg = NULL;
3120 if (decl->ctype.base_type != &autotype_ctype)
3121 msg = "on non-identifier";
3122 else if (match_op(token, ','))
3123 msg = "on declaration list";
3124 else if (!decl->initializer)
3125 msg = "without initializer";
3126 else if (decl->initializer->type == EXPR_SYMBOL &&
3127 decl->initializer->symbol == decl)
3128 msg = "on self-init var";
3129 if (msg) {
3130 sparse_error(decl->pos, "__auto_type %s", msg);
3131 decl->ctype.base_type = &bad_ctype;
3135 if (!match_op(token, ','))
3136 break;
3138 token = token->next;
3139 ident = NULL;
3140 decl = alloc_symbol(token->pos, SYM_NODE);
3141 ctx.ctype = saved;
3142 ctx.cleanup = NULL;
3143 token = handle_attributes(token, &ctx);
3144 token = declarator(token, &ctx);
3145 token = handle_asm_name(token, &ctx);
3146 token = handle_attributes(token, &ctx);
3147 apply_modifiers(token->pos, &ctx);
3148 decl->ctype = ctx.ctype;
3149 decl->ctype.modifiers |= mod;
3150 decl->cleanup = ctx.cleanup;
3151 decl->endpos = token->pos;
3152 if (!ident) {
3153 sparse_error(token->pos, "expected identifier name in type definition");
3154 return token;
3157 if (is_typedef)
3158 decl->ctype.modifiers |= MOD_USERTYPE;
3160 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
3162 /* Function declarations are automatically extern unless specifically static */
3163 base_type = decl->ctype.base_type;
3164 if (!is_typedef && base_type && base_type->type == SYM_FN) {
3165 if (!(decl->ctype.modifiers & MOD_STATIC))
3166 decl->ctype.modifiers |= MOD_EXTERN;
3169 return expect(token, ';', "at end of declaration");