sm_math1: update the validation test
[smatch.git] / parse.c
blob960d0606888278981d2d3c8d28d4dc6ab93e51e8
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, typeof_unqual_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 typeof_unqual_op = {
201 .type = KW_SPECIFIER,
202 .declarator = typeof_unqual_specifier,
203 .test = Set_Any,
204 .set = Set_S|Set_T,
207 static struct symbol_op autotype_op = {
208 .type = KW_SPECIFIER,
209 .declarator = autotype_specifier,
210 .test = Set_Any,
211 .set = Set_S|Set_T,
214 static struct symbol_op attribute_op = {
215 .type = KW_ATTRIBUTE,
216 .declarator = attribute_specifier,
219 static struct symbol_op struct_op = {
220 .type = KW_SPECIFIER,
221 .declarator = struct_specifier,
222 .test = Set_Any,
223 .set = Set_S|Set_T,
226 static struct symbol_op union_op = {
227 .type = KW_SPECIFIER,
228 .declarator = union_specifier,
229 .test = Set_Any,
230 .set = Set_S|Set_T,
233 static struct symbol_op enum_op = {
234 .type = KW_SPECIFIER,
235 .declarator = enum_specifier,
236 .test = Set_Any,
237 .set = Set_S|Set_T,
240 static struct symbol_op spec_op = {
241 .type = KW_SPECIFIER | KW_EXACT,
242 .test = Set_Any,
243 .set = Set_S|Set_T,
246 static struct symbol_op char_op = {
247 .type = KW_SPECIFIER,
248 .test = Set_T|Set_Long|Set_Short,
249 .set = Set_T|Set_Char,
250 .class = CInt,
253 static struct symbol_op int_op = {
254 .type = KW_SPECIFIER,
255 .test = Set_T,
256 .set = Set_T|Set_Int,
259 static struct symbol_op double_op = {
260 .type = KW_SPECIFIER,
261 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Vlong,
262 .set = Set_T|Set_Double,
263 .class = CReal,
266 static struct symbol_op float_op = {
267 .type = KW_SPECIFIER,
268 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Long,
269 .set = Set_T|Set_Float,
270 .class = CReal,
273 static struct symbol_op short_op = {
274 .type = KW_SPECIFIER,
275 .test = Set_S|Set_Char|Set_Float|Set_Double|Set_Long|Set_Short,
276 .set = Set_Short,
279 static struct symbol_op signed_op = {
280 .type = KW_SPECIFIER,
281 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
282 .set = Set_Signed,
283 .class = CSInt,
286 static struct symbol_op unsigned_op = {
287 .type = KW_SPECIFIER,
288 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
289 .set = Set_Unsigned,
290 .class = CUInt,
293 static struct symbol_op long_op = {
294 .type = KW_SPECIFIER,
295 .test = Set_S|Set_Char|Set_Float|Set_Short|Set_Vlong,
296 .set = Set_Long,
299 static struct symbol_op int128_op = {
300 .type = KW_SPECIFIER,
301 .test = Set_S|Set_T|Set_Char|Set_Short|Set_Int|Set_Float|Set_Double|Set_Long|Set_Vlong|Set_Int128,
302 .set = Set_T|Set_Int128|Set_Vlong,
303 .class = CInt,
306 static struct symbol_op if_op = {
307 .statement = parse_if_statement,
310 static struct symbol_op return_op = {
311 .statement = parse_return_statement,
314 static struct symbol_op loop_iter_op = {
315 .statement = parse_loop_iterator,
318 static struct symbol_op default_op = {
319 .statement = parse_default_statement,
322 static struct symbol_op case_op = {
323 .statement = parse_case_statement,
326 static struct symbol_op switch_op = {
327 .statement = parse_switch_statement,
330 static struct symbol_op for_op = {
331 .statement = parse_for_statement,
334 static struct symbol_op while_op = {
335 .statement = parse_while_statement,
338 static struct symbol_op do_op = {
339 .statement = parse_do_statement,
342 static struct symbol_op goto_op = {
343 .statement = parse_goto_statement,
346 static struct symbol_op __context___op = {
347 .statement = parse_context_statement,
348 .attribute = attribute_context,
351 static struct symbol_op range_op = {
352 .statement = parse_range_statement,
355 static struct symbol_op asm_op = {
356 .type = KW_ASM,
357 .statement = parse_asm_statement,
358 .toplevel = toplevel_asm_declaration,
361 static struct symbol_op static_assert_op = {
362 .toplevel = parse_static_assert,
365 static struct symbol_op seg_gs = {
366 .toplevel = ignore_seg_gs,
369 static struct symbol_op packed_op = {
370 .attribute = attribute_packed,
373 static struct symbol_op aligned_op = {
374 .attribute = attribute_aligned,
377 static struct symbol_op cleanup_op = {
378 .attribute = attribute_cleanup,
381 static struct symbol_op attr_mod_op = {
382 .attribute = attribute_modifier,
385 static struct symbol_op attr_fun_op = {
386 .attribute = attribute_function,
389 static struct symbol_op attr_bitwise_op = {
390 .attribute = attribute_bitwise,
393 static struct symbol_op attr_force_op = {
394 .attribute = attribute_force,
397 static struct symbol_op address_space_op = {
398 .attribute = attribute_address_space,
401 static struct symbol_op mode_op = {
402 .attribute = attribute_mode,
405 static struct symbol_op context_op = {
406 .attribute = attribute_context,
409 static struct symbol_op designated_init_op = {
410 .attribute = attribute_designated_init,
413 static struct symbol_op transparent_union_op = {
414 .attribute = attribute_transparent_union,
417 static struct symbol_op ignore_attr_op = {
418 .attribute = ignore_attribute,
421 static struct symbol_op mode_QI_op = {
422 .type = KW_MODE,
423 .to_mode = to_QI_mode
426 static struct symbol_op mode_HI_op = {
427 .type = KW_MODE,
428 .to_mode = to_HI_mode
431 static struct symbol_op mode_SI_op = {
432 .type = KW_MODE,
433 .to_mode = to_SI_mode
436 static struct symbol_op mode_DI_op = {
437 .type = KW_MODE,
438 .to_mode = to_DI_mode
441 static struct symbol_op mode_TI_op = {
442 .type = KW_MODE,
443 .to_mode = to_TI_mode
446 static struct symbol_op mode_pointer_op = {
447 .type = KW_MODE,
448 .to_mode = to_pointer_mode
451 static struct symbol_op mode_word_op = {
452 .type = KW_MODE,
453 .to_mode = to_word_mode
457 * Define the keyword and their effects.
458 * The entries in the 'typedef' and put in NS_TYPEDEF and
459 * are automatically set as reserved keyword while the ones
460 * in the 'keyword' table are just put in NS_KEYWORD.
462 * The entries are added via the 3 macros:
463 * N() for entries with "name" only,
464 * D() for entries with "name" & "__name__",
465 * A() for entries with "name", "__name" & "__name__",
466 * U() for entries with "__name" & "__name__".
468 static struct init_keyword {
469 const char *name;
470 struct symbol_op *op;
471 struct symbol *type;
472 unsigned long mods;
473 } typedefs[] = {
474 #define N(I, O,...) { I, O,##__VA_ARGS__ }
475 #define D(I, O,...) N(I,O,##__VA_ARGS__ ), \
476 N("__" I "__",O,##__VA_ARGS__)
477 #define A(I, O,...) N(I,O,##__VA_ARGS__ ), \
478 N("__" I,O,##__VA_ARGS__), \
479 N("__" I "__",O,##__VA_ARGS__)
480 #define U(I, O,...) N("__" I,O,##__VA_ARGS__), \
481 N("__" I "__",O,##__VA_ARGS__)
482 /* Storage classes */
483 N("auto", &auto_op, .mods = MOD_AUTO),
484 N("register", &register_op, .mods = MOD_REGISTER),
485 N("static", &static_op, .mods = MOD_STATIC),
486 N("extern", &extern_op, .mods = MOD_EXTERN),
487 N("__thread", &thread_op),
488 N("_Thread_local", &thread_op),
490 A("inline", &inline_op, .mods = MOD_INLINE),
492 /* Typedef ... */
493 N("typedef", &typedef_op, .mods = MOD_USERTYPE),
494 A("typeof", &typeof_op),
495 A("typeof_unqual", &typeof_unqual_op),
496 N("__auto_type", &autotype_op),
498 /* Type qualifiers */
499 A("const", &const_op, .mods = MOD_CONST),
500 A("volatile", &volatile_op, .mods = MOD_VOLATILE),
501 A("restrict", &restrict_op, .mods = MOD_RESTRICT),
503 N("_Atomic", &atomic_op, .mods = MOD_ATOMIC),
504 N("_Noreturn", &noreturn_op, .mods = MOD_NORETURN),
505 N("_Alignas", &alignas_op),
507 U("attribute", &attribute_op),
509 /* Type specifiers */
510 N("struct", &struct_op),
511 N("union", &union_op),
512 N("enum", &enum_op),
514 N("void", &spec_op, .type = &void_ctype),
515 N("char", &char_op),
516 N("short", &short_op),
517 N("int", &int_op),
518 N("long", &long_op),
519 N("float", &float_op),
520 N("double", &double_op),
521 A("signed", &signed_op),
522 N("unsigned", &unsigned_op),
523 N("__int128", &int128_op),
524 N("_Bool", &spec_op, .type = &bool_ctype),
526 /* Predeclared types */
527 N("__builtin_va_list", &spec_op, .type = &ptr_ctype),
528 N("__builtin_ms_va_list",&spec_op, .type = &ptr_ctype),
529 N("__int128_t", &spec_op, .type = &sint128_ctype),
530 N("__uint128_t", &spec_op, .type = &uint128_ctype),
531 N("_Float32", &spec_op, .type = &float32_ctype),
532 N("_Float32x", &spec_op, .type = &float32x_ctype),
533 N("_Float64", &spec_op, .type = &float64_ctype),
534 N("_Float64x", &spec_op, .type = &float64x_ctype),
535 N("_Float128", &spec_op, .type = &float128_ctype),
536 N("__float128", &spec_op, .type = &float128_ctype),
538 N("__seg_gs", &seg_gs),
540 }, keywords[] = {
541 /* Statements */
542 N("if", &if_op),
543 N("return", &return_op),
544 N("break", &loop_iter_op),
545 N("continue", &loop_iter_op),
546 N("default", &default_op),
547 N("case", &case_op),
548 N("switch", &switch_op),
549 N("for", &for_op),
550 N("while", &while_op),
551 N("do", &do_op),
552 N("goto", &goto_op),
553 A("asm", &asm_op),
554 N("context", &context_op),
555 N("__context__", &__context___op),
556 N("__range__", &range_op),
557 N("_Static_assert", &static_assert_op),
559 /* Attributes */
560 D("packed", &packed_op),
561 D("aligned", &aligned_op),
562 D("cleanup", &cleanup_op),
563 D("nocast", &attr_mod_op, .mods = MOD_NOCAST),
564 D("noderef", &attr_mod_op, .mods = MOD_NODEREF),
565 D("safe", &attr_mod_op, .mods = MOD_SAFE),
566 D("unused", &attr_mod_op, .mods = MOD_UNUSED),
567 D("externally_visible", &attr_mod_op, .mods = MOD_EXT_VISIBLE),
568 D("force", &attr_force_op),
569 D("bitwise", &attr_bitwise_op, .mods = MOD_BITWISE),
570 D("address_space", &address_space_op),
571 D("designated_init", &designated_init_op),
572 D("transparent_union", &transparent_union_op),
573 D("noreturn", &attr_fun_op, .mods = MOD_NORETURN),
574 D("pure", &attr_fun_op, .mods = MOD_PURE),
575 A("const", &attr_fun_op, .mods = MOD_PURE),
576 D("gnu_inline", &attr_fun_op, .mods = MOD_GNU_INLINE),
578 /* Modes */
579 D("mode", &mode_op),
580 D("QI", &mode_QI_op),
581 D("HI", &mode_HI_op),
582 D("SI", &mode_SI_op),
583 D("DI", &mode_DI_op),
584 D("TI", &mode_TI_op),
585 D("byte", &mode_QI_op),
586 D("pointer", &mode_pointer_op),
587 D("word", &mode_word_op),
591 static const char *ignored_attributes[] = {
593 #define GCC_ATTR(x) \
594 STRINGIFY(x), \
595 STRINGIFY(__##x##__),
597 #include "gcc-attr-list.h"
599 #undef GCC_ATTR
601 "bounded",
602 "__bounded__",
603 "__noclone",
604 "__nonnull",
605 "__nothrow",
609 static void init_keyword(int stream, struct init_keyword *kw, enum namespace ns)
611 struct symbol *sym = create_symbol(stream, kw->name, SYM_KEYWORD, ns);
612 sym->ident->keyword = 1;
613 sym->ident->reserved |= (ns == NS_TYPEDEF);
614 sym->ctype.modifiers = kw->mods;
615 sym->ctype.base_type = kw->type;
616 sym->op = kw->op;
619 void init_parser(int stream)
621 int i;
623 for (i = 0; i < ARRAY_SIZE(typedefs); i++)
624 init_keyword(stream, &typedefs[i], NS_TYPEDEF);
625 for (i = 0; i < ARRAY_SIZE(keywords); i++)
626 init_keyword(stream, &keywords[i], NS_KEYWORD);
628 for (i = 0; i < ARRAY_SIZE(ignored_attributes); i++) {
629 const char * name = ignored_attributes[i];
630 struct symbol *sym = create_symbol(stream, name, SYM_KEYWORD,
631 NS_KEYWORD);
632 if (!sym->op) {
633 sym->ident->keyword = 1;
634 sym->op = &ignore_attr_op;
640 static struct token *skip_to(struct token *token, int op)
642 while (!match_op(token, op) && !eof_token(token))
643 token = token->next;
644 return token;
647 static struct token bad_token = { .pos.type = TOKEN_BAD };
648 struct token *expect(struct token *token, int op, const char *where)
650 if (!match_op(token, op)) {
651 if (token != &bad_token) {
652 bad_token.next = token;
653 sparse_error(token->pos, "Expected %s %s", show_special(op), where);
654 sparse_error(token->pos, "got %s", show_token(token));
656 if (op == ';')
657 return skip_to(token, op);
658 return &bad_token;
660 return token->next;
664 // issue an error message on new parsing errors
665 // @token: the current token
666 // @errmsg: the error message
667 // If the current token is from a previous error, an error message
668 // has already been issued, so nothing more is done.
669 // Otherwise, @errmsg is displayed followed by the current token.
670 static void unexpected(struct token *token, const char *errmsg)
672 if (token == &bad_token)
673 return;
674 sparse_error(token->pos, "%s", errmsg);
675 sparse_error(token->pos, "got %s", show_token(token));
678 // Add a symbol to the list of function-local symbols
679 static void fn_local_symbol(struct symbol *sym)
681 if (function_symbol_list)
682 add_symbol(function_symbol_list, sym);
685 struct statement *alloc_statement(struct position pos, int type)
687 struct statement *stmt = __alloc_statement(0);
688 stmt->type = type;
689 stmt->pos = pos;
690 return stmt;
693 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list);
695 static void apply_ctype(struct position pos, struct ctype *dst, struct ctype *src);
697 static void apply_modifiers(struct position pos, struct decl_state *ctx)
699 struct symbol *ctype;
700 if (!ctx->mode)
701 return;
702 ctype = ctx->mode->to_mode(ctx->ctype.base_type);
703 if (!ctype)
704 sparse_error(pos, "don't know how to apply mode to %s",
705 show_typename(ctx->ctype.base_type));
706 else
707 ctx->ctype.base_type = ctype;
711 static struct symbol * alloc_indirect_symbol(struct position pos, struct ctype *ctype, int type)
713 struct symbol *sym = alloc_symbol(pos, type);
715 sym->ctype.base_type = ctype->base_type;
716 sym->ctype.modifiers = ctype->modifiers;
718 ctype->base_type = sym;
719 ctype->modifiers = 0;
720 return sym;
724 * NOTE! NS_LABEL is not just a different namespace,
725 * it also ends up using function scope instead of the
726 * regular symbol scope.
728 struct symbol *label_symbol(struct token *token, int used)
730 struct symbol *sym = lookup_symbol(token->ident, NS_LABEL);
731 if (!sym) {
732 sym = alloc_symbol(token->pos, SYM_LABEL);
733 bind_symbol(sym, token->ident, NS_LABEL);
734 if (used)
735 sym->used = 1;
736 fn_local_symbol(sym);
738 return sym;
741 static struct token *struct_union_enum_specifier(enum type type,
742 struct token *token, struct decl_state *ctx,
743 struct token *(*parse)(struct token *, struct symbol *))
745 struct decl_state attr = { };
746 struct symbol *sym;
747 struct position *repos;
749 token = handle_attributes(token, &attr);
750 if (token_type(token) == TOKEN_IDENT) {
751 sym = lookup_symbol(token->ident, NS_STRUCT);
752 if (!sym ||
753 (is_outer_scope(sym->scope) &&
754 (match_op(token->next,';') || match_op(token->next,'{')))) {
755 // Either a new symbol, or else an out-of-scope
756 // symbol being redefined.
757 sym = alloc_symbol(token->pos, type);
758 bind_symbol(sym, token->ident, NS_STRUCT);
760 if (sym->type != type)
761 error_die(token->pos, "invalid tag applied to %s", show_typename (sym));
762 ctx->ctype.base_type = sym;
763 repos = &token->pos;
764 token = token->next;
765 if (!match_op(token, '{'))
766 return token;
768 // The following test is actually wrong for empty
769 // structs, but (1) they are not C99, (2) gcc does
770 // the same thing, and (3) it's easier.
771 if (sym->symbol_list)
772 error_die(token->pos, "redefinition of %s", show_typename (sym));
773 sym->pos = *repos;
775 // Mark the structure as needing re-examination
776 sym->examined = 0;
777 } else if (match_op(token, '{')) {
778 // private struct/union/enum type
779 sym = alloc_symbol(token->pos, type);
780 set_current_scope(sym); // used by dissect
781 ctx->ctype.base_type = sym;
782 } else {
783 sparse_error(token->pos, "expected declaration");
784 ctx->ctype.base_type = &bad_ctype;
785 return token;
788 token = parse(token->next, sym);
789 token = expect(token, '}', "at end of specifier");
790 attr.ctype.base_type = sym;
791 token = handle_attributes(token, &attr);
792 apply_ctype(token->pos, &sym->ctype, &attr.ctype);
793 sym->packed = attr.packed;
795 sym->endpos = token->pos;
797 return token;
800 static struct token *parse_struct_declaration(struct token *token, struct symbol *sym)
802 struct symbol *field, *last = NULL;
803 struct token *res;
804 res = struct_declaration_list(token, &sym->symbol_list);
805 FOR_EACH_PTR(sym->symbol_list, field) {
806 if (!field->ident) {
807 struct symbol *base = field->ctype.base_type;
808 if (base && base->type == SYM_BITFIELD)
809 continue;
811 if (last)
812 last->next_subobject = field;
813 last = field;
814 } END_FOR_EACH_PTR(field);
815 return res;
818 static struct token *parse_union_declaration(struct token *token, struct symbol *sym)
820 return struct_declaration_list(token, &sym->symbol_list);
823 static struct token *struct_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
825 return struct_union_enum_specifier(SYM_STRUCT, token, ctx, parse_struct_declaration);
828 static struct token *union_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
830 return struct_union_enum_specifier(SYM_UNION, token, ctx, parse_union_declaration);
834 // safe right shift
836 // This allow to use a shift amount as big (or bigger)
837 // than the width of the value to be shifted, in which case
838 // the result is, of course, 0.
839 static unsigned long long rshift(unsigned long long val, unsigned int n)
841 if (n >= (sizeof(val) * 8))
842 return 0;
843 return val >> n;
846 struct range {
847 long long neg;
848 unsigned long long pos;
851 static void update_range(struct range *range, unsigned long long uval, struct symbol *vtype)
853 long long sval = uval;
855 if (is_signed_type(vtype) && (sval < 0)) {
856 if (sval < range->neg)
857 range->neg = sval;
858 } else {
859 if (uval > range->pos)
860 range->pos = uval;
864 static int type_is_ok(struct symbol *type, struct range range)
866 int shift = type->bit_size;
867 int is_unsigned = type->ctype.modifiers & MOD_UNSIGNED;
869 if (!is_unsigned)
870 shift--;
871 if (rshift(range.pos, shift))
872 return 0;
873 if (range.neg == 0)
874 return 1;
875 if (is_unsigned)
876 return 0;
877 if (rshift(~range.neg, shift))
878 return 0;
879 return 1;
882 static struct range type_range(struct symbol *type)
884 struct range range;
885 unsigned int size = type->bit_size;
886 unsigned long long max;
887 long long min;
889 if (is_signed_type(type)) {
890 min = sign_bit(size);
891 max = min - 1;
892 } else {
893 min = 0;
894 max = bits_mask(size);
897 range.pos = max;
898 range.neg = min;
899 return range;
902 static int val_in_range(struct range *range, long long sval, struct symbol *vtype)
904 unsigned long long uval = sval;
906 if (is_signed_type(vtype) && (sval < 0))
907 return range->neg <= sval;
908 else
909 return uval <= range->pos;
912 static void cast_enum_list(struct symbol_list *list, struct symbol *base_type)
914 struct range irange = type_range(&int_ctype);
915 struct symbol *sym;
917 FOR_EACH_PTR(list, sym) {
918 struct expression *expr = sym->initializer;
919 struct symbol *ctype;
920 long long val;
921 if (expr->type != EXPR_VALUE)
922 continue;
923 ctype = expr->ctype;
924 val = get_expression_value(expr);
925 if (is_int_type(ctype) && val_in_range(&irange, val, ctype)) {
926 expr->ctype = &int_ctype;
927 continue;
929 cast_value(expr, base_type, expr);
930 } END_FOR_EACH_PTR(sym);
933 static struct token *parse_enum_declaration(struct token *token, struct symbol *parent)
935 unsigned long long lastval = 0;
936 struct symbol *ctype = NULL, *base_type = NULL;
937 struct range range = { };
938 int mix_bitwise = 0;
940 parent->examined = 1;
941 parent->ctype.base_type = &int_ctype;
942 while (token_type(token) == TOKEN_IDENT) {
943 struct expression *expr = NULL;
944 struct token *next = token->next;
945 struct decl_state ctx = { };
946 struct symbol *sym;
948 // FIXME: only 'deprecated' should be accepted
949 next = handle_attributes(next, &ctx);
951 if (match_op(next, '=')) {
952 next = constant_expression(next->next, &expr);
953 lastval = get_expression_value(expr);
954 ctype = &void_ctype;
955 if (expr && expr->ctype)
956 ctype = expr->ctype;
957 } else if (!ctype) {
958 ctype = &int_ctype;
959 } else if (is_int_type(ctype)) {
960 lastval++;
961 } else {
962 error_die(token->pos, "can't increment the last enum member");
965 if (!expr) {
966 expr = alloc_expression(token->pos, EXPR_VALUE);
967 expr->value = lastval;
968 expr->ctype = ctype;
971 sym = alloc_symbol(token->pos, SYM_NODE);
972 bind_symbol(sym, token->ident, NS_SYMBOL);
973 sym->ctype.modifiers &= ~MOD_ADDRESSABLE;
974 sym->initializer = expr;
975 sym->enum_member = 1;
976 sym->ctype.base_type = parent;
977 add_ptr_list(&parent->symbol_list, sym);
979 if (base_type != &bad_ctype) {
980 if (ctype->type == SYM_NODE)
981 ctype = ctype->ctype.base_type;
982 if (ctype->type == SYM_ENUM) {
983 if (ctype == parent)
984 ctype = base_type;
985 else
986 ctype = ctype->ctype.base_type;
989 * base_type rules:
990 * - if all enums are of the same type, then
991 * the base_type is that type (two first
992 * cases)
993 * - if enums are of different types, they
994 * all have to be integer types, and the
995 * base type is at least "int_ctype".
996 * - otherwise the base_type is "bad_ctype".
998 if (!base_type || ctype == &bad_ctype) {
999 base_type = ctype;
1000 } else if (ctype == base_type) {
1001 /* nothing */
1002 } else if (is_int_type(base_type) && is_int_type(ctype)) {
1003 base_type = &int_ctype;
1004 } else if (is_restricted_type(base_type) != is_restricted_type(ctype)) {
1005 if (!mix_bitwise++) {
1006 warning(expr->pos, "mixed bitwiseness");
1008 } else if (is_restricted_type(base_type) && base_type != ctype) {
1009 sparse_error(expr->pos, "incompatible restricted type");
1010 info(expr->pos, " expected: %s", show_typename(base_type));
1011 info(expr->pos, " got: %s", show_typename(ctype));
1012 base_type = &bad_ctype;
1013 } else if (base_type != &bad_ctype) {
1014 sparse_error(token->pos, "bad enum definition");
1015 base_type = &bad_ctype;
1017 parent->ctype.base_type = base_type;
1019 if (is_int_type(base_type)) {
1020 update_range(&range, lastval, ctype);
1022 token = next;
1024 sym->endpos = token->pos;
1026 if (!match_op(token, ','))
1027 break;
1028 token = token->next;
1030 if (!base_type) {
1031 sparse_error(token->pos, "empty enum definition");
1032 base_type = &bad_ctype;
1034 else if (!is_int_type(base_type))
1036 else if (type_is_ok(&uint_ctype, range))
1037 base_type = &uint_ctype;
1038 else if (type_is_ok(&int_ctype, range))
1039 base_type = &int_ctype;
1040 else if (type_is_ok(&ulong_ctype, range))
1041 base_type = &ulong_ctype;
1042 else if (type_is_ok(&long_ctype, range))
1043 base_type = &long_ctype;
1044 else if (type_is_ok(&ullong_ctype, range))
1045 base_type = &ullong_ctype;
1046 else if (type_is_ok(&llong_ctype, range))
1047 base_type = &llong_ctype;
1048 else
1049 base_type = &bad_ctype;
1050 parent->ctype.base_type = base_type;
1051 parent->ctype.modifiers |= (base_type->ctype.modifiers & MOD_UNSIGNED);
1052 parent->examined = 0;
1054 if (mix_bitwise)
1055 return token;
1056 cast_enum_list(parent->symbol_list, base_type);
1058 return token;
1061 static struct token *enum_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1063 struct token *ret = struct_union_enum_specifier(SYM_ENUM, token, ctx, parse_enum_declaration);
1064 struct ctype *ctype = &ctx->ctype.base_type->ctype;
1066 if (!ctype->base_type)
1067 ctype->base_type = &incomplete_ctype;
1069 return ret;
1072 static struct token *typeof_specifier_helper(struct token *token, struct symbol *sym, struct decl_state *ctx, bool qual)
1075 if (!match_op(token, '(')) {
1076 sparse_error(token->pos, "expected '(' after typeof");
1077 return token;
1079 if (lookup_type(token->next)) {
1080 struct symbol *sym;
1081 token = typename(token->next, &sym, NULL);
1082 ctx->ctype.base_type = sym->ctype.base_type;
1083 apply_ctype(token->pos, &ctx->ctype, &sym->ctype);
1084 } else {
1085 struct symbol *typeof_sym = alloc_symbol(token->pos, qual? SYM_TYPEOF : SYM_TYPEOF_UNQUAL);
1086 token = parse_expression(token->next, &typeof_sym->initializer);
1088 typeof_sym->endpos = token->pos;
1089 if (!typeof_sym->initializer) {
1090 sparse_error(token->pos, "expected expression after the '(' token");
1091 typeof_sym = &bad_ctype;
1093 ctx->ctype.base_type = typeof_sym;
1095 return expect(token, ')', "after typeof");
1098 static struct token *typeof_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1100 return typeof_specifier_helper(token, sym, ctx, true);
1103 static struct token *typeof_unqual_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1105 return typeof_specifier_helper(token, sym, ctx, false);
1108 static struct token *autotype_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1110 ctx->ctype.base_type = &autotype_ctype;
1111 ctx->autotype = 1;
1112 return token;
1115 static struct token *ignore_attribute(struct token *token, struct symbol *attr, struct decl_state *ctx)
1117 struct expression *expr = NULL;
1118 if (match_op(token, '('))
1119 token = parens_expression(token, &expr, "in attribute");
1120 return token;
1123 static struct token *attribute_packed(struct token *token, struct symbol *attr, struct decl_state *ctx)
1125 if (!ctx->ctype.alignment) {
1126 ctx->ctype.alignment = 1;
1127 ctx->packed = 1;
1129 return token;
1132 static struct token *attribute_aligned(struct token *token, struct symbol *attr, struct decl_state *ctx)
1134 int alignment = max_alignment;
1135 struct expression *expr = NULL;
1137 if (match_op(token, '(')) {
1138 token = parens_expression(token, &expr, "in attribute");
1139 if (expr)
1140 alignment = const_expression_value(expr);
1142 if (alignment & (alignment-1)) {
1143 warning(token->pos, "I don't like non-power-of-2 alignments");
1144 return token;
1145 } else if (alignment > ctx->ctype.alignment)
1146 ctx->ctype.alignment = alignment;
1147 return token;
1150 static struct token *attribute_cleanup(struct token *token, struct symbol *attr, struct decl_state *ctx)
1152 struct expression *expr = NULL;
1154 if (match_op(token, '(')) {
1155 token = token->next;
1156 if (match_op(token, ')'))
1157 sparse_error(token->pos, "an argument is expected for attribute 'cleanup'");
1158 else if (token_type(token) != TOKEN_IDENT)
1159 sparse_error(token->pos, "argument is not an identifier");
1160 token = primary_expression(token, &expr);
1161 if (expr && expr->type == EXPR_SYMBOL)
1162 ctx->cleanup = expr;
1163 return expect(token, ')', "after attribute's argument'");
1166 sparse_error(token->pos, "an argument is expected for attribute 'cleanup'");
1167 return token;
1170 static void apply_mod(struct position *pos, unsigned long *mods, unsigned long mod)
1172 if (*mods & mod & ~MOD_DUP_OK)
1173 warning(*pos, "duplicate %s", modifier_name(mod));
1174 *mods |= mod;
1177 static void apply_qualifier(struct position *pos, struct ctype *ctx, unsigned long qual)
1179 apply_mod(pos, &ctx->modifiers, qual);
1182 static struct token *attribute_modifier(struct token *token, struct symbol *attr, struct decl_state *ctx)
1184 apply_mod(&token->pos, &ctx->ctype.modifiers, attr->ctype.modifiers);
1185 return token;
1188 static struct token *attribute_function(struct token *token, struct symbol *attr, struct decl_state *ctx)
1190 apply_mod(&token->pos, &ctx->f_modifiers, attr->ctype.modifiers);
1191 return token;
1194 static struct token *attribute_bitwise(struct token *token, struct symbol *attr, struct decl_state *ctx)
1196 if (Wbitwise)
1197 attribute_modifier(token, attr, ctx);
1198 return token;
1201 static struct ident *numerical_address_space(int asn)
1203 char buff[32];
1205 if (!asn)
1206 return NULL;
1207 sprintf(buff, "<asn:%d>", asn);
1208 return built_in_ident(buff);
1211 static struct token *attribute_address_space(struct token *token, struct symbol *attr, struct decl_state *ctx)
1213 struct expression *expr = NULL;
1214 struct ident *as = NULL;
1215 struct token *next;
1217 token = expect(token, '(', "after address_space attribute");
1218 switch (token_type(token)) {
1219 case TOKEN_NUMBER:
1220 next = primary_expression(token, &expr);
1221 if (expr->type != EXPR_VALUE)
1222 goto invalid;
1223 as = numerical_address_space(expr->value);
1224 break;
1225 case TOKEN_IDENT:
1226 next = token->next;
1227 as = token->ident;
1228 break;
1229 default:
1230 next = token->next;
1231 invalid:
1232 as = NULL;
1233 warning(token->pos, "invalid address space name");
1236 if (Waddress_space && as) {
1237 if (ctx->ctype.as)
1238 sparse_error(token->pos,
1239 "multiple address spaces given: %s & %s",
1240 show_as(ctx->ctype.as), show_as(as));
1241 ctx->ctype.as = as;
1243 token = expect(next, ')', "after address_space attribute");
1244 return token;
1247 static struct symbol *to_QI_mode(struct symbol *ctype)
1249 if (ctype->ctype.base_type != &int_type)
1250 return NULL;
1251 if (ctype == &char_ctype)
1252 return ctype;
1253 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uchar_ctype
1254 : &schar_ctype;
1257 static struct symbol *to_HI_mode(struct symbol *ctype)
1259 if (ctype->ctype.base_type != &int_type)
1260 return NULL;
1261 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ushort_ctype
1262 : &sshort_ctype;
1265 static struct symbol *to_SI_mode(struct symbol *ctype)
1267 if (ctype->ctype.base_type != &int_type)
1268 return NULL;
1269 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint_ctype
1270 : &sint_ctype;
1273 static struct symbol *to_DI_mode(struct symbol *ctype)
1275 if (ctype->ctype.base_type != &int_type)
1276 return NULL;
1277 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ullong_ctype
1278 : &sllong_ctype;
1281 static struct symbol *to_TI_mode(struct symbol *ctype)
1283 if (ctype->ctype.base_type != &int_type)
1284 return NULL;
1285 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint128_ctype
1286 : &sint128_ctype;
1289 static struct symbol *to_pointer_mode(struct symbol *ctype)
1291 if (ctype->ctype.base_type != &int_type)
1292 return NULL;
1293 return ctype->ctype.modifiers & MOD_UNSIGNED ? uintptr_ctype
1294 : intptr_ctype;
1297 static struct symbol *to_word_mode(struct symbol *ctype)
1299 if (ctype->ctype.base_type != &int_type)
1300 return NULL;
1301 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulong_ctype
1302 : &slong_ctype;
1305 static struct token *attribute_mode(struct token *token, struct symbol *attr, struct decl_state *ctx)
1307 token = expect(token, '(', "after mode attribute");
1308 if (token_type(token) == TOKEN_IDENT) {
1309 struct symbol *mode = lookup_keyword(token->ident, NS_KEYWORD);
1310 if (mode && mode->op->type & KW_MODE)
1311 ctx->mode = mode->op;
1312 else
1313 sparse_error(token->pos, "unknown mode attribute %s", show_ident(token->ident));
1314 token = token->next;
1315 } else
1316 sparse_error(token->pos, "expect attribute mode symbol\n");
1317 token = expect(token, ')', "after mode attribute");
1318 return token;
1321 static struct token *attribute_context(struct token *token, struct symbol *attr, struct decl_state *ctx)
1323 struct context *context = alloc_context();
1324 struct expression *args[3];
1325 int idx = 0;
1327 token = expect(token, '(', "after context attribute");
1328 token = conditional_expression(token, &args[0]);
1329 token = expect(token, ',', "after context 1st argument");
1330 token = conditional_expression(token, &args[1]);
1331 if (match_op(token, ',')) {
1332 token = token->next;
1333 token = conditional_expression(token, &args[2]);
1334 token = expect(token, ')', "after context 3rd argument");
1335 context->context = args[0];
1336 idx++;
1337 } else {
1338 token = expect(token, ')', "after context 2nd argument");
1340 context->in = get_expression_value(args[idx++]);
1341 context->out = get_expression_value(args[idx++]);
1342 add_ptr_list(&ctx->ctype.contexts, context);
1343 return token;
1346 static struct token *attribute_designated_init(struct token *token, struct symbol *attr, struct decl_state *ctx)
1348 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_STRUCT)
1349 ctx->ctype.base_type->designated_init = 1;
1350 else
1351 warning(token->pos, "attribute designated_init applied to non-structure type");
1352 return token;
1355 static struct token *attribute_transparent_union(struct token *token, struct symbol *attr, struct decl_state *ctx)
1357 if (Wtransparent_union)
1358 warning(token->pos, "attribute __transparent_union__");
1360 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_UNION)
1361 ctx->ctype.base_type->transparent_union = 1;
1362 else
1363 warning(token->pos, "attribute __transparent_union__ applied to non-union type");
1364 return token;
1367 static struct token *recover_unknown_attribute(struct token *token)
1369 struct expression *expr = NULL;
1371 if (Wunknown_attribute)
1372 warning(token->pos, "unknown attribute '%s'", show_ident(token->ident));
1373 token = token->next;
1374 if (match_op(token, '('))
1375 token = parens_expression(token, &expr, "in attribute");
1376 return token;
1379 static struct token *attribute_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1381 token = expect(token, '(', "after attribute");
1382 token = expect(token, '(', "after attribute");
1384 while (token_type(token) == TOKEN_IDENT) {
1385 struct symbol *attr = lookup_keyword(token->ident, NS_KEYWORD);
1386 if (attr && attr->op->attribute)
1387 token = attr->op->attribute(token->next, attr, ctx);
1388 else
1389 token = recover_unknown_attribute(token);
1391 if (!match_op(token, ','))
1392 break;
1393 token = token->next;
1396 token = expect(token, ')', "after attribute");
1397 token = expect(token, ')', "after attribute");
1398 return token;
1401 static unsigned long decl_modifiers(struct decl_state *ctx)
1403 unsigned long mods = ctx->ctype.modifiers & MOD_DECLARE;
1404 ctx->ctype.modifiers &= ~MOD_DECLARE;
1405 return ctx->storage_class | mods;
1408 static struct token *storage_specifier(struct token *next, struct symbol *sym, struct decl_state *ctx)
1410 int is_tls = ctx->ctype.modifiers & MOD_TLS;
1411 unsigned long class = sym->ctype.modifiers;
1412 const char *storage = modifier_name(class);
1414 /* __thread can be used alone, or with extern or static */
1415 if (is_tls && (class & ~(MOD_STATIC|MOD_EXTERN)))
1416 sparse_error(next->pos, "__thread cannot be used with '%s'", storage);
1417 else if (!ctx->storage_class)
1418 ctx->storage_class = class;
1419 else if (ctx->storage_class == class)
1420 sparse_error(next->pos, "duplicate %s", storage);
1421 else
1422 sparse_error(next->pos, "multiple storage classes");
1423 return next;
1426 static struct token *thread_specifier(struct token *next, struct symbol *sym, struct decl_state *ctx)
1428 /* This GCC extension can be used alone, or with extern or static */
1429 if (!(ctx->storage_class & ~(MOD_STATIC|MOD_EXTERN))) {
1430 apply_qualifier(&next->pos, &ctx->ctype, MOD_TLS);
1431 } else {
1432 sparse_error(next->pos, "__thread cannot be used with '%s'",
1433 modifier_name(ctx->storage_class));
1436 return next;
1439 static struct token *attribute_force(struct token *token, struct symbol *attr, struct decl_state *ctx)
1441 ctx->forced = 1;
1442 return token;
1445 static struct token *alignas_specifier(struct token *token, struct symbol *sym, struct decl_state *ctx)
1447 int alignment = 0;
1449 if (!match_op(token, '(')) {
1450 sparse_error(token->pos, "expected '(' after _Alignas");
1451 return token;
1453 if (lookup_type(token->next)) {
1454 struct symbol *sym = NULL;
1455 token = typename(token->next, &sym, NULL);
1456 sym = examine_symbol_type(sym);
1457 alignment = sym->ctype.alignment;
1458 token = expect(token, ')', "after _Alignas(...");
1459 } else {
1460 struct expression *expr = NULL;
1461 token = parens_expression(token, &expr, "after _Alignas");
1462 if (!expr)
1463 return token;
1464 alignment = const_expression_value(expr);
1467 if (alignment < 0) {
1468 warning(token->pos, "non-positive alignment");
1469 return token;
1471 if (alignment & (alignment-1)) {
1472 warning(token->pos, "non-power-of-2 alignment");
1473 return token;
1475 if (alignment > ctx->ctype.alignment)
1476 ctx->ctype.alignment = alignment;
1477 return token;
1480 static struct token *generic_qualifier(struct token *next, struct symbol *sym, struct decl_state *ctx)
1482 apply_qualifier(&next->pos, &ctx->ctype, sym->ctype.modifiers);
1483 return next;
1486 static void apply_ctype(struct position pos, struct ctype *dst, struct ctype *src)
1488 unsigned long mod = src->modifiers;
1490 if (mod)
1491 apply_qualifier(&pos, dst, mod);
1493 /* Context */
1494 concat_ptr_list((struct ptr_list *)src->contexts,
1495 (struct ptr_list **)&dst->contexts);
1497 /* Alignment */
1498 if (src->alignment > dst->alignment)
1499 dst->alignment = src->alignment;
1501 /* Address space */
1502 if (src->as)
1503 dst->as = src->as;
1506 static void specifier_conflict(struct position pos, int what, struct ident *new)
1508 const char *old;
1509 if (what & (Set_S | Set_T))
1510 goto Catch_all;
1511 if (what & Set_Char)
1512 old = "char";
1513 else if (what & Set_Double)
1514 old = "double";
1515 else if (what & Set_Float)
1516 old = "float";
1517 else if (what & Set_Signed)
1518 old = "signed";
1519 else if (what & Set_Unsigned)
1520 old = "unsigned";
1521 else if (what & Set_Short)
1522 old = "short";
1523 else if (what & Set_Long)
1524 old = "long";
1525 else
1526 old = "long long";
1527 sparse_error(pos, "impossible combination of type specifiers: %s %s",
1528 old, show_ident(new));
1529 return;
1531 Catch_all:
1532 sparse_error(pos, "two or more data types in declaration specifiers");
1535 static struct symbol * const int_types[] =
1536 {&char_ctype, &short_ctype, &int_ctype, &long_ctype, &llong_ctype, &int128_ctype};
1537 static struct symbol * const signed_types[] =
1538 {&schar_ctype, &sshort_ctype, &sint_ctype, &slong_ctype, &sllong_ctype,
1539 &sint128_ctype};
1540 static struct symbol * const unsigned_types[] =
1541 {&uchar_ctype, &ushort_ctype, &uint_ctype, &ulong_ctype, &ullong_ctype,
1542 &uint128_ctype};
1543 static struct symbol * const real_types[] =
1544 {&float_ctype, &double_ctype, &ldouble_ctype};
1545 static struct symbol * const * const types[] = {
1546 [CInt] = int_types + 2,
1547 [CSInt] = signed_types + 2,
1548 [CUInt] = unsigned_types + 2,
1549 [CReal] = real_types + 1,
1552 struct symbol *ctype_integer(int size, int want_unsigned)
1554 return types[want_unsigned ? CUInt : CInt][size];
1557 static struct token *handle_qualifiers(struct token *t, struct decl_state *ctx)
1559 while (token_type(t) == TOKEN_IDENT) {
1560 struct symbol *s = lookup_keyword(t->ident, NS_TYPEDEF);
1561 if (!s)
1562 break;
1563 if (!(s->op->type & (KW_ATTRIBUTE | KW_QUALIFIER)))
1564 break;
1565 t = t->next;
1566 if (s->op->declarator)
1567 t = s->op->declarator(t, s, ctx);
1569 return t;
1572 static struct token *declaration_specifiers(struct token *token, struct decl_state *ctx)
1574 int seen = 0;
1575 int class = CInt;
1576 int rank = 0;
1578 while (token_type(token) == TOKEN_IDENT) {
1579 struct symbol *s = lookup_symbol(token->ident,
1580 NS_TYPEDEF | NS_SYMBOL);
1581 if (!s || !(s->namespace & NS_TYPEDEF))
1582 break;
1583 if (s->type != SYM_KEYWORD) {
1584 if (seen & Set_Any)
1585 break;
1586 seen |= Set_S | Set_T;
1587 ctx->ctype.base_type = s->ctype.base_type;
1588 apply_ctype(token->pos, &ctx->ctype, &s->ctype);
1589 token = token->next;
1590 continue;
1592 if (s->op->type & KW_SPECIFIER) {
1593 if (seen & s->op->test) {
1594 specifier_conflict(token->pos,
1595 seen & s->op->test,
1596 token->ident);
1597 break;
1599 seen |= s->op->set;
1600 class += s->op->class;
1601 if (s->op->set & Set_Int128)
1602 rank = 3;
1603 else if (s->op->set & Set_Char)
1604 rank = -2;
1605 if (s->op->set & (Set_Short|Set_Float)) {
1606 rank = -1;
1607 } else if (s->op->set & Set_Long && rank++) {
1608 if (class == CReal) {
1609 specifier_conflict(token->pos,
1610 Set_Vlong,
1611 &double_ident);
1612 break;
1614 seen |= Set_Vlong;
1617 token = token->next;
1618 if (s->op->declarator) // Note: this eats attributes
1619 token = s->op->declarator(token, s, ctx);
1620 if (s->op->type & KW_EXACT) {
1621 ctx->ctype.base_type = s->ctype.base_type;
1622 ctx->ctype.modifiers |= s->ctype.modifiers;
1626 if (!(seen & Set_S)) { /* not set explicitly? */
1627 struct symbol *base = &incomplete_ctype;
1628 if (seen & Set_Any)
1629 base = types[class][rank];
1630 ctx->ctype.base_type = base;
1633 if (ctx->ctype.modifiers & MOD_BITWISE) {
1634 struct symbol *type;
1635 ctx->ctype.modifiers &= ~MOD_BITWISE;
1636 if (!is_int_type(ctx->ctype.base_type)) {
1637 sparse_error(token->pos, "invalid modifier");
1638 return token;
1640 type = alloc_symbol(token->pos, SYM_BASETYPE);
1641 *type = *ctx->ctype.base_type;
1642 type->ctype.modifiers &= ~MOD_SPECIFIER;
1643 type->ctype.base_type = ctx->ctype.base_type;
1644 type->type = SYM_RESTRICT;
1645 ctx->ctype.base_type = type;
1646 create_fouled(type);
1648 return token;
1651 static struct token *abstract_array_declarator(struct token *token, struct symbol *sym)
1653 struct expression *expr = NULL;
1654 int has_static = 0;
1656 while (token_type(token) == TOKEN_IDENT) {
1657 struct symbol *sym = lookup_keyword(token->ident, NS_TYPEDEF);
1658 if (!sym || !(sym->op->type & (KW_STATIC|KW_QUALIFIER)))
1659 break;
1660 if (has_static && (sym->op->type & KW_STATIC))
1661 sparse_error(token->pos, "duplicate array static declarator");
1662 has_static |= (sym->op->type & KW_STATIC);
1663 token = token->next;
1665 if (match_op(token, '*') && match_op(token->next, ']')) {
1666 // FIXME: '[*]' is treated like '[]'
1667 token = token->next;
1668 } else {
1669 token = assignment_expression(token, &expr);
1671 sym->array_size = expr;
1672 return token;
1675 static struct token *parameter_type_list(struct token *, struct symbol *);
1676 static struct token *identifier_list(struct token *, struct symbol *);
1677 static struct token *declarator(struct token *token, struct decl_state *ctx);
1679 static struct token *handle_asm_name(struct token *token, struct decl_state *ctx)
1681 struct expression *expr;
1682 struct symbol *keyword;
1684 if (token_type(token) != TOKEN_IDENT)
1685 return token;
1686 keyword = lookup_keyword(token->ident, NS_KEYWORD);
1687 if (!keyword)
1688 return token;
1689 if (!(keyword->op->type & KW_ASM))
1690 return token;
1692 token = token->next;
1693 token = expect(token, '(', "after asm");
1694 token = string_expression(token, &expr, "asm name");
1695 token = expect(token, ')', "after asm");
1696 return token;
1700 // test if @token is '__attribute__' (or one of its variant)
1701 static bool match_attribute(struct token *token)
1703 struct symbol *sym;
1705 if (token_type(token) != TOKEN_IDENT)
1706 return false;
1707 sym = lookup_keyword(token->ident, NS_TYPEDEF);
1708 if (!sym || !sym->op)
1709 return false;
1710 return sym->op->type & KW_ATTRIBUTE;
1713 static struct token *skip_attribute(struct token *token)
1715 token = token->next;
1716 if (match_op(token, '(')) {
1717 int depth = 1;
1718 token = token->next;
1719 while (depth && !eof_token(token)) {
1720 if (token_type(token) == TOKEN_SPECIAL) {
1721 if (token->special == '(')
1722 depth++;
1723 else if (token->special == ')')
1724 depth--;
1726 token = token->next;
1729 return token;
1732 static struct token *skip_attributes(struct token *token)
1734 while (match_attribute(token)) {
1735 token = expect(token->next, '(', "after attribute");
1736 token = expect(token, '(', "after attribute");
1737 while (token_type(token) == TOKEN_IDENT) {
1738 token = skip_attribute(token);
1739 if (!match_op(token, ','))
1740 break;
1741 token = token->next;
1743 token = expect(token, ')', "after attribute");
1744 token = expect(token, ')', "after attribute");
1746 return token;
1749 static struct token *handle_attributes(struct token *token, struct decl_state *ctx)
1751 while (match_attribute(token))
1752 token = attribute_specifier(token->next, NULL, ctx);
1753 return token;
1756 static int is_nested(struct token *token, struct token **p,
1757 int prefer_abstract)
1760 * This can be either a parameter list or a grouping.
1761 * For the direct (non-abstract) case, we know if must be
1762 * a parameter list if we already saw the identifier.
1763 * For the abstract case, we know if must be a parameter
1764 * list if it is empty or starts with a type.
1766 struct token *next = token->next;
1768 *p = next = skip_attributes(next);
1770 if (token_type(next) == TOKEN_IDENT) {
1771 if (lookup_type(next))
1772 return !prefer_abstract;
1773 return 1;
1776 if (match_op(next, ')') || match_op(next, SPECIAL_ELLIPSIS))
1777 return 0;
1779 return 1;
1782 enum kind {
1783 Empty, K_R, Proto, Bad_Func,
1786 static enum kind which_func(struct token *token,
1787 struct ident **n,
1788 int prefer_abstract)
1790 struct token *next = token->next;
1792 if (token_type(next) == TOKEN_IDENT) {
1793 if (lookup_type(next))
1794 return Proto;
1795 /* identifier list not in definition; complain */
1796 if (prefer_abstract)
1797 warning(token->pos,
1798 "identifier list not in definition");
1799 return K_R;
1802 if (token_type(next) != TOKEN_SPECIAL)
1803 return Bad_Func;
1805 if (next->special == ')') {
1806 /* don't complain about those */
1807 if (!n || match_op(next->next, ';') || match_op(next->next, ','))
1808 return Empty;
1809 if (Wstrict_prototypes)
1810 warning(next->pos,
1811 "non-ANSI function declaration of function '%s'",
1812 show_ident(*n));
1813 return Empty;
1816 if (next->special == SPECIAL_ELLIPSIS) {
1817 warning(next->pos,
1818 "variadic functions must have one named argument");
1819 return Proto;
1822 return Bad_Func;
1825 static struct token *direct_declarator(struct token *token, struct decl_state *ctx)
1827 struct ctype *ctype = &ctx->ctype;
1828 struct token *next;
1829 struct ident **p = ctx->ident;
1831 if (ctx->ident && token_type(token) == TOKEN_IDENT) {
1832 *ctx->ident = token->ident;
1833 token = token->next;
1834 } else if (match_op(token, '(') &&
1835 is_nested(token, &next, ctx->prefer_abstract)) {
1836 struct symbol *base_type = ctype->base_type;
1837 if (token->next != next)
1838 next = handle_attributes(token->next, ctx);
1839 token = declarator(next, ctx);
1840 token = expect(token, ')', "in nested declarator");
1841 while (ctype->base_type != base_type)
1842 ctype = &ctype->base_type->ctype;
1843 p = NULL;
1846 if (match_op(token, '(')) {
1847 enum kind kind = which_func(token, p, ctx->prefer_abstract);
1848 struct symbol *fn;
1849 fn = alloc_indirect_symbol(token->pos, ctype, SYM_FN);
1850 ctype->modifiers |= ctx->f_modifiers;
1851 token = token->next;
1852 if (kind == K_R)
1853 token = identifier_list(token, fn);
1854 else if (kind == Proto)
1855 token = parameter_type_list(token, fn);
1856 token = expect(token, ')', "in function declarator");
1857 fn->endpos = token->pos;
1858 return token;
1861 while (match_op(token, '[')) {
1862 struct symbol *array;
1863 array = alloc_indirect_symbol(token->pos, ctype, SYM_ARRAY);
1864 token = abstract_array_declarator(token->next, array);
1865 token = expect(token, ']', "in abstract_array_declarator");
1866 array->endpos = token->pos;
1867 ctype = &array->ctype;
1869 return token;
1872 static struct token *pointer(struct token *token, struct decl_state *ctx)
1874 while (match_op(token,'*')) {
1875 struct symbol *ptr = alloc_symbol(token->pos, SYM_PTR);
1876 ptr->ctype.modifiers = ctx->ctype.modifiers;
1877 ptr->ctype.base_type = ctx->ctype.base_type;
1878 ptr->ctype.as = ctx->ctype.as;
1879 ptr->ctype.contexts = ctx->ctype.contexts;
1880 ctx->ctype.modifiers = 0;
1881 ctx->ctype.base_type = ptr;
1882 ctx->ctype.as = NULL;
1883 ctx->ctype.contexts = NULL;
1884 ctx->ctype.alignment = 0;
1886 token = handle_qualifiers(token->next, ctx);
1887 ctx->ctype.base_type->endpos = token->pos;
1889 return token;
1892 static struct token *declarator(struct token *token, struct decl_state *ctx)
1894 token = pointer(token, ctx);
1895 return direct_declarator(token, ctx);
1898 static struct token *handle_bitfield(struct token *token, struct decl_state *ctx)
1900 struct ctype *ctype = &ctx->ctype;
1901 struct expression *expr;
1902 struct symbol *bitfield;
1903 long long width;
1905 if (ctype->base_type != &int_type && !is_int_type(ctype->base_type)) {
1906 sparse_error(token->pos, "invalid bitfield specifier for type %s.",
1907 show_typename(ctype->base_type));
1908 // Parse this to recover gracefully.
1909 return conditional_expression(token->next, &expr);
1912 bitfield = alloc_indirect_symbol(token->pos, ctype, SYM_BITFIELD);
1913 token = conditional_expression(token->next, &expr);
1914 width = const_expression_value(expr);
1915 bitfield->bit_size = width;
1917 if (width < 0 || width > INT_MAX || (*ctx->ident && width == 0)) {
1918 sparse_error(token->pos, "bitfield '%s' has invalid width (%lld)",
1919 show_ident(*ctx->ident), width);
1920 width = -1;
1921 } else if (*ctx->ident) {
1922 struct symbol *base_type = bitfield->ctype.base_type;
1923 struct symbol *bitfield_type = base_type == &int_type ? bitfield : base_type;
1924 int is_signed = !(bitfield_type->ctype.modifiers & MOD_UNSIGNED);
1925 if (Wone_bit_signed_bitfield && width == 1 && is_signed) {
1926 // Valid values are either {-1;0} or {0}, depending on integer
1927 // representation. The latter makes for very efficient code...
1928 sparse_error(token->pos, "dubious one-bit signed bitfield");
1930 if (Wdefault_bitfield_sign &&
1931 bitfield_type->type != SYM_ENUM &&
1932 !(bitfield_type->ctype.modifiers & MOD_EXPLICITLY_SIGNED) &&
1933 is_signed) {
1934 // The sign of bitfields is unspecified by default.
1935 warning(token->pos, "dubious bitfield without explicit `signed' or `unsigned'");
1938 bitfield->bit_size = width;
1939 bitfield->endpos = token->pos;
1940 bitfield->ident = *ctx->ident;
1941 return token;
1944 static struct token *declaration_list(struct token *token, struct symbol_list **list)
1946 struct decl_state ctx = {.prefer_abstract = 0};
1947 struct ctype saved;
1948 unsigned long mod;
1950 token = declaration_specifiers(token, &ctx);
1951 mod = decl_modifiers(&ctx);
1952 saved = ctx.ctype;
1953 for (;;) {
1954 struct symbol *decl = alloc_symbol(token->pos, SYM_NODE);
1955 ctx.cleanup = NULL;
1956 ctx.ident = &decl->ident;
1958 token = declarator(token, &ctx);
1959 if (match_op(token, ':'))
1960 token = handle_bitfield(token, &ctx);
1962 token = handle_attributes(token, &ctx);
1963 apply_modifiers(token->pos, &ctx);
1965 decl->ctype = ctx.ctype;
1966 decl->ctype.modifiers |= mod;
1967 decl->cleanup = ctx.cleanup;
1968 decl->endpos = token->pos;
1969 add_symbol(list, decl);
1970 if (!match_op(token, ','))
1971 break;
1972 token = token->next;
1973 ctx.ctype = saved;
1975 return token;
1978 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list)
1980 while (!match_op(token, '}')) {
1981 if (match_ident(token, &_Static_assert_ident)) {
1982 token = parse_static_assert(token, NULL);
1983 continue;
1985 if (!match_op(token, ';'))
1986 token = declaration_list(token, list);
1987 if (!match_op(token, ';')) {
1988 sparse_error(token->pos, "expected ; at end of declaration");
1989 break;
1991 token = token->next;
1993 return token;
1996 static struct token *parameter_declaration(struct token *token, struct symbol *sym)
1998 struct decl_state ctx = {.prefer_abstract = 1};
2000 token = declaration_specifiers(token, &ctx);
2001 ctx.ident = &sym->ident;
2002 token = declarator(token, &ctx);
2003 token = handle_attributes(token, &ctx);
2004 apply_modifiers(token->pos, &ctx);
2005 sym->ctype = ctx.ctype;
2006 sym->ctype.modifiers |= decl_modifiers(&ctx);
2007 sym->endpos = token->pos;
2008 sym->forced_arg = ctx.forced;
2009 return token;
2012 struct token *typename(struct token *token, struct symbol **p, int *forced)
2014 struct decl_state ctx = {.prefer_abstract = 1};
2015 unsigned long class;
2016 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2017 *p = sym;
2018 token = declaration_specifiers(token, &ctx);
2019 token = declarator(token, &ctx);
2020 apply_modifiers(token->pos, &ctx);
2021 sym->ctype = ctx.ctype;
2022 sym->cleanup = ctx.cleanup;
2023 sym->endpos = token->pos;
2024 class = ctx.storage_class;
2025 if (forced)
2026 *forced = ctx.forced;
2027 if (class)
2028 warning(sym->pos, "storage class in typename (%s%s)",
2029 modifier_string(class), show_typename(sym));
2030 return token;
2033 static struct token *parse_underscore_Pragma(struct token *token)
2035 struct token *next;
2037 next = token->next;
2038 if (!match_op(next, '('))
2039 return next;
2040 next = next->next;
2041 if (next->pos.type != TOKEN_STRING)
2042 return next;
2043 next = next->next;
2044 if (!match_op(next, ')'))
2045 return next;
2046 return next->next;
2049 static struct token *expression_statement(struct token *token, struct expression **tree)
2051 if (match_ident(token, &_Pragma_ident))
2052 return parse_underscore_Pragma(token);
2054 token = parse_expression(token, tree);
2055 return expect(token, ';', "at end of statement");
2058 static struct token *parse_asm_operands(struct token *token, struct statement *stmt,
2059 struct asm_operand_list **inout)
2061 /* Allow empty operands */
2062 if (match_op(token->next, ':') || match_op(token->next, ')'))
2063 return token->next;
2064 do {
2065 struct asm_operand *op = __alloc_asm_operand(0);
2066 if (match_op(token->next, '[') &&
2067 token_type(token->next->next) == TOKEN_IDENT &&
2068 match_op(token->next->next->next, ']')) {
2069 op->name = token->next->next->ident;
2070 token = token->next->next->next;
2072 token = token->next;
2073 token = string_expression(token, &op->constraint, "asm constraint");
2074 token = parens_expression(token, &op->expr, "in asm parameter");
2075 add_ptr_list(inout, op);
2076 } while (match_op(token, ','));
2077 return token;
2080 static struct token *parse_asm_clobbers(struct token *token, struct statement *stmt,
2081 struct expression_list **clobbers)
2083 struct expression *expr;
2085 do {
2086 token = primary_expression(token->next, &expr);
2087 if (expr)
2088 add_expression(clobbers, expr);
2089 } while (match_op(token, ','));
2090 return token;
2093 static struct token *parse_asm_labels(struct token *token, struct statement *stmt,
2094 struct symbol_list **labels)
2096 struct symbol *label;
2098 do {
2099 token = token->next; /* skip ':' and ',' */
2100 if (token_type(token) != TOKEN_IDENT)
2101 return token;
2102 label = label_symbol(token, 1);
2103 add_symbol(labels, label);
2104 token = token->next;
2105 } while (match_op(token, ','));
2106 return token;
2109 static struct token *parse_asm_statement(struct token *token, struct statement *stmt)
2111 unsigned long mods = 0;
2113 token = token->next;
2114 stmt->type = STMT_ASM;
2115 while (token_type(token) == TOKEN_IDENT) {
2116 struct symbol *s = lookup_keyword(token->ident, NS_TYPEDEF);
2117 if (s && s->op->asm_modifier)
2118 s->op->asm_modifier(token, &mods, s->ctype.modifiers);
2119 else if (token->ident == &goto_ident)
2120 asm_modifier(token, &mods, MOD_ASM_GOTO);
2121 token = token->next;
2123 token = expect(token, '(', "after asm");
2124 token = string_expression(token, &stmt->asm_string, "inline asm");
2125 if (match_op(token, ':'))
2126 token = parse_asm_operands(token, stmt, &stmt->asm_outputs);
2127 if (match_op(token, ':'))
2128 token = parse_asm_operands(token, stmt, &stmt->asm_inputs);
2129 if (match_op(token, ':'))
2130 token = parse_asm_clobbers(token, stmt, &stmt->asm_clobbers);
2131 if (match_op(token, ':') && (mods & MOD_ASM_GOTO))
2132 token = parse_asm_labels(token, stmt, &stmt->asm_labels);
2133 token = expect(token, ')', "after asm");
2134 return expect(token, ';', "at end of asm-statement");
2137 static struct token *parse_static_assert(struct token *token, struct symbol_list **unused)
2139 struct expression *cond = NULL, *message = NULL;
2141 token = expect(token->next, '(', "after _Static_assert");
2142 token = constant_expression(token, &cond);
2143 if (!cond)
2144 sparse_error(token->pos, "Expected constant expression");
2145 if (match_op(token, ',')) {
2146 token = token->next;
2147 token = string_expression(token, &message, "_Static_assert()");
2148 if (!message)
2149 cond = NULL;
2151 token = expect(token, ')', "after diagnostic message in _Static_assert");
2152 token = expect(token, ';', "after _Static_assert()");
2154 if (cond && !const_expression_value(cond) && cond->type == EXPR_VALUE) {
2155 const char *sep = "", *msg = "";
2157 if (message) {
2158 sep = ": ";
2159 msg = show_string(message->string);
2161 sparse_error(cond->pos, "static assertion failed%s%s", sep, msg);
2164 return token;
2167 static struct token *ignore_seg_gs(struct token *token, struct symbol_list **unused)
2169 return token->next;
2172 /* Make a statement out of an expression */
2173 static struct statement *make_statement(struct expression *expr)
2175 struct statement *stmt;
2177 if (!expr)
2178 return NULL;
2179 stmt = alloc_statement(expr->pos, STMT_EXPRESSION);
2180 stmt->expression = expr;
2181 return stmt;
2185 * All iterators have two symbols associated with them:
2186 * the "continue" and "break" symbols, which are targets
2187 * for continue and break statements respectively.
2189 * They are in a special name-space, but they follow
2190 * all the normal visibility rules, so nested iterators
2191 * automatically work right.
2193 static void start_iterator(struct statement *stmt)
2195 struct symbol *cont, *brk;
2197 start_block_scope(stmt->pos);
2198 cont = alloc_symbol(stmt->pos, SYM_NODE);
2199 bind_symbol(cont, &continue_ident, NS_ITERATOR);
2200 brk = alloc_symbol(stmt->pos, SYM_NODE);
2201 bind_symbol(brk, &break_ident, NS_ITERATOR);
2203 stmt->type = STMT_ITERATOR;
2204 stmt->iterator_break = brk;
2205 stmt->iterator_continue = cont;
2206 fn_local_symbol(brk);
2207 fn_local_symbol(cont);
2210 static void end_iterator(struct statement *stmt)
2212 end_block_scope();
2215 static struct statement *start_function(struct symbol *sym)
2217 struct symbol *ret;
2218 struct statement *stmt = alloc_statement(sym->pos, STMT_COMPOUND);
2220 start_function_scope(sym->pos);
2221 ret = alloc_symbol(sym->pos, SYM_NODE);
2222 ret->ctype = sym->ctype.base_type->ctype;
2223 ret->ctype.modifiers &= ~(MOD_STORAGE | MOD_QUALIFIER | MOD_TLS | MOD_ACCESS | MOD_NOCAST | MOD_NODEREF);
2224 ret->ctype.modifiers |= (MOD_AUTO | MOD_REGISTER);
2225 bind_symbol(ret, &return_ident, NS_ITERATOR);
2226 stmt->ret = ret;
2227 fn_local_symbol(ret);
2229 // Currently parsed symbol for __func__/__FUNCTION__/__PRETTY_FUNCTION__
2230 current_fn = sym;
2232 return stmt;
2235 static void end_function(struct symbol *sym)
2237 current_fn = NULL;
2238 end_function_scope();
2242 * A "switch()" statement, like an iterator, has a
2243 * the "break" symbol associated with it. It works
2244 * exactly like the iterator break - it's the target
2245 * for any break-statements in scope, and means that
2246 * "break" handling doesn't even need to know whether
2247 * it's breaking out of an iterator or a switch.
2249 * In addition, the "case" symbol is a marker for the
2250 * case/default statements to find the switch statement
2251 * that they are associated with.
2253 static void start_switch(struct statement *stmt)
2255 struct symbol *brk, *switch_case;
2257 start_block_scope(stmt->pos);
2258 brk = alloc_symbol(stmt->pos, SYM_NODE);
2259 bind_symbol(brk, &break_ident, NS_ITERATOR);
2261 switch_case = alloc_symbol(stmt->pos, SYM_NODE);
2262 bind_symbol(switch_case, &case_ident, NS_ITERATOR);
2263 switch_case->stmt = stmt;
2265 stmt->type = STMT_SWITCH;
2266 stmt->switch_break = brk;
2267 stmt->switch_case = switch_case;
2269 fn_local_symbol(brk);
2270 fn_local_symbol(switch_case);
2273 static void end_switch(struct statement *stmt)
2275 if (!stmt->switch_case->symbol_list)
2276 warning(stmt->pos, "switch with no cases");
2277 end_block_scope();
2280 static void add_case_statement(struct statement *stmt)
2282 struct symbol *target = lookup_symbol(&case_ident, NS_ITERATOR);
2283 struct symbol *sym;
2285 if (!target) {
2286 sparse_error(stmt->pos, "not in switch scope");
2287 stmt->type = STMT_NONE;
2288 return;
2290 sym = alloc_symbol(stmt->pos, SYM_NODE);
2291 add_symbol(&target->symbol_list, sym);
2292 sym->stmt = stmt;
2293 stmt->case_label = sym;
2294 fn_local_symbol(sym);
2297 static struct token *parse_return_statement(struct token *token, struct statement *stmt)
2299 struct symbol *target = lookup_symbol(&return_ident, NS_ITERATOR);
2301 if (!target)
2302 error_die(token->pos, "internal error: return without a function target");
2303 stmt->type = STMT_RETURN;
2304 stmt->ret_target = target;
2305 return expression_statement(token->next, &stmt->ret_value);
2308 static void validate_for_loop_decl(struct symbol *sym)
2310 unsigned long storage = sym->ctype.modifiers & MOD_STORAGE;
2312 if (storage & ~(MOD_AUTO | MOD_REGISTER)) {
2313 const char *name = show_ident(sym->ident);
2314 sparse_error(sym->pos, "non-local var '%s' in for-loop initializer", name);
2315 sym->ctype.modifiers &= ~MOD_STORAGE;
2319 static struct token *parse_for_statement(struct token *token, struct statement *stmt)
2321 struct symbol_list *syms;
2322 struct expression *e1, *e2, *e3;
2323 struct statement *iterator;
2325 start_iterator(stmt);
2326 token = expect(token->next, '(', "after 'for'");
2328 syms = NULL;
2329 e1 = NULL;
2330 /* C99 variable declaration? */
2331 if (lookup_type(token)) {
2332 token = external_declaration(token, &syms, validate_for_loop_decl);
2333 } else {
2334 token = parse_expression(token, &e1);
2335 token = expect(token, ';', "in 'for'");
2337 token = parse_expression(token, &e2);
2338 token = expect(token, ';', "in 'for'");
2339 token = parse_expression(token, &e3);
2340 token = expect(token, ')', "in 'for'");
2341 token = statement(token, &iterator);
2343 stmt->iterator_syms = syms;
2344 stmt->iterator_pre_statement = make_statement(e1);
2345 stmt->iterator_pre_condition = e2;
2346 stmt->iterator_post_statement = make_statement(e3);
2347 stmt->iterator_post_condition = NULL;
2348 stmt->iterator_statement = iterator;
2349 end_iterator(stmt);
2351 return token;
2354 static struct token *parse_while_statement(struct token *token, struct statement *stmt)
2356 struct expression *expr;
2357 struct statement *iterator;
2359 start_iterator(stmt);
2360 token = parens_expression(token->next, &expr, "after 'while'");
2361 token = statement(token, &iterator);
2363 stmt->iterator_pre_condition = expr;
2364 stmt->iterator_post_condition = NULL;
2365 stmt->iterator_statement = iterator;
2366 end_iterator(stmt);
2368 return token;
2371 static struct token *parse_do_statement(struct token *token, struct statement *stmt)
2373 struct expression *expr;
2374 struct statement *iterator;
2376 start_iterator(stmt);
2377 token = statement(token->next, &iterator);
2378 if (token_type(token) == TOKEN_IDENT && token->ident == &while_ident)
2379 token = token->next;
2380 else
2381 sparse_error(token->pos, "expected 'while' after 'do'");
2382 token = parens_expression(token, &expr, "after 'do-while'");
2384 stmt->iterator_post_condition = expr;
2385 stmt->iterator_statement = iterator;
2386 end_iterator(stmt);
2388 if (iterator && iterator->type != STMT_COMPOUND && Wdo_while)
2389 warning(iterator->pos, "do-while statement is not a compound statement");
2391 return expect(token, ';', "after statement");
2394 static struct token *parse_if_statement(struct token *token, struct statement *stmt)
2396 stmt->type = STMT_IF;
2397 token = parens_expression(token->next, &stmt->if_conditional, "after if");
2398 token = statement(token, &stmt->if_true);
2399 if (token_type(token) != TOKEN_IDENT)
2400 return token;
2401 if (token->ident != &else_ident)
2402 return token;
2403 return statement(token->next, &stmt->if_false);
2406 static inline struct token *case_statement(struct token *token, struct statement *stmt)
2408 stmt->type = STMT_CASE;
2409 token = expect(token, ':', "after default/case");
2410 add_case_statement(stmt);
2411 if (match_op(token, '}')) {
2412 warning(token->pos, "statement expected after case label");
2413 stmt->case_statement = alloc_statement(token->pos, STMT_NONE);
2414 return token;
2416 return statement(token, &stmt->case_statement);
2419 static struct token *parse_case_statement(struct token *token, struct statement *stmt)
2421 token = parse_expression(token->next, &stmt->case_expression);
2422 if (match_op(token, SPECIAL_ELLIPSIS))
2423 token = parse_expression(token->next, &stmt->case_to);
2424 return case_statement(token, stmt);
2427 static struct token *parse_default_statement(struct token *token, struct statement *stmt)
2429 return case_statement(token->next, stmt);
2432 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt)
2434 struct symbol *target = lookup_symbol(token->ident, NS_ITERATOR);
2435 stmt->type = STMT_GOTO;
2436 stmt->goto_label = target;
2437 if (!target)
2438 sparse_error(stmt->pos, "break/continue not in iterator scope");
2439 return expect(token->next, ';', "at end of statement");
2442 static struct token *parse_switch_statement(struct token *token, struct statement *stmt)
2444 stmt->type = STMT_SWITCH;
2445 start_switch(stmt);
2446 token = parens_expression(token->next, &stmt->switch_expression, "after 'switch'");
2447 token = statement(token, &stmt->switch_statement);
2448 end_switch(stmt);
2449 return token;
2452 static void warn_label_usage(struct position def, struct position use, struct ident *ident)
2454 const char *id = show_ident(ident);
2455 sparse_error(use, "label '%s' used outside statement expression", id);
2456 info(def, " label '%s' defined here", id);
2457 current_fn->bogus_linear = 1;
2460 void check_label_usage(struct symbol *label, struct position use_pos)
2462 struct statement *def = label->stmt;
2464 if (def) {
2465 if (!is_in_scope(def->label_scope, label_scope))
2466 warn_label_usage(def->pos, use_pos, label->ident);
2467 } else if (!label->label_scope) {
2468 label->label_scope = label_scope;
2469 label->label_pos = use_pos;
2473 static struct token *parse_goto_statement(struct token *token, struct statement *stmt)
2475 stmt->type = STMT_GOTO;
2476 token = token->next;
2477 if (match_op(token, '*')) {
2478 token = parse_expression(token->next, &stmt->goto_expression);
2479 add_statement(&function_computed_goto_list, stmt);
2480 } else if (token_type(token) == TOKEN_IDENT) {
2481 struct symbol *label = label_symbol(token, 1);
2482 stmt->goto_label = label;
2483 check_label_usage(label, stmt->pos);
2484 token = token->next;
2485 } else {
2486 sparse_error(token->pos, "Expected identifier or goto expression");
2488 return expect(token, ';', "at end of statement");
2491 static struct token *parse_context_statement(struct token *token, struct statement *stmt)
2493 stmt->type = STMT_CONTEXT;
2494 token = token->next;
2495 token = expect(token, '(', "after __context__ statement");
2496 token = assignment_expression(token, &stmt->expression);
2497 if (!stmt->expression)
2498 unexpected(token, "expression expected after '('");
2499 if (match_op(token, ',')) {
2500 token = token->next;
2501 stmt->context = stmt->expression;
2502 token = assignment_expression(token, &stmt->expression);
2503 if (!stmt->expression)
2504 unexpected(token, "expression expected after ','");
2506 token = expect(token, ')', "at end of __context__ statement");
2507 return expect(token, ';', "at end of statement");
2510 static struct token *parse_range_statement(struct token *token, struct statement *stmt)
2512 stmt->type = STMT_RANGE;
2513 token = token->next;
2514 token = expect(token, '(', "after __range__ statement");
2515 token = assignment_expression(token, &stmt->range_expression);
2516 token = expect(token, ',', "after range expression");
2517 token = assignment_expression(token, &stmt->range_low);
2518 token = expect(token, ',', "after low range");
2519 token = assignment_expression(token, &stmt->range_high);
2520 token = expect(token, ')', "after range statement");
2521 return expect(token, ';', "after range statement");
2524 static struct token *handle_label_attributes(struct token *token, struct symbol *label)
2526 struct decl_state ctx = { };
2528 token = handle_attributes(token, &ctx);
2529 label->label_modifiers = ctx.ctype.modifiers;
2530 return token;
2533 static struct token *statement(struct token *token, struct statement **tree)
2535 struct statement *stmt = alloc_statement(token->pos, STMT_NONE);
2537 *tree = stmt;
2538 if (token_type(token) == TOKEN_IDENT) {
2539 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2540 if (s && s->op->statement)
2541 return s->op->statement(token, stmt);
2543 if (match_op(token->next, ':')) {
2544 struct symbol *s = label_symbol(token, 0);
2545 token = handle_label_attributes(token->next->next, s);
2546 if (s->stmt) {
2547 sparse_error(stmt->pos, "label '%s' redefined", show_ident(s->ident));
2548 // skip the label to avoid multiple definitions
2549 return statement(token, tree);
2551 stmt->type = STMT_LABEL;
2552 stmt->label_identifier = s;
2553 stmt->label_scope = label_scope;
2554 if (s->label_scope) {
2555 if (!is_in_scope(label_scope, s->label_scope))
2556 warn_label_usage(stmt->pos, s->label_pos, s->ident);
2558 s->stmt = stmt;
2559 if (match_op(token, '}')) {
2560 warning(token->pos, "statement expected after label");
2561 stmt->label_statement = alloc_statement(token->pos, STMT_NONE);
2562 return token;
2564 return statement(token, &stmt->label_statement);
2568 if (match_op(token, '{')) {
2569 token = compound_statement(token->next, stmt);
2570 return expect(token, '}', "at end of compound statement");
2573 stmt->type = STMT_EXPRESSION;
2574 return expression_statement(token, &stmt->expression);
2577 /* gcc extension - __label__ ident-list; in the beginning of compound stmt */
2578 static struct token *label_statement(struct token *token)
2580 while (token_type(token) == TOKEN_IDENT) {
2581 struct symbol *sym = alloc_symbol(token->pos, SYM_LABEL);
2582 /* it's block-scope, but we want label namespace */
2583 bind_symbol_with_scope(sym, token->ident, NS_LABEL, block_scope);
2584 fn_local_symbol(sym);
2585 token = token->next;
2586 if (!match_op(token, ','))
2587 break;
2588 token = token->next;
2590 return expect(token, ';', "at end of label declaration");
2593 static struct token * statement_list(struct token *token, struct statement_list **list)
2595 int seen_statement = 0;
2596 while (token_type(token) == TOKEN_IDENT &&
2597 token->ident == &__label___ident)
2598 token = label_statement(token->next);
2599 for (;;) {
2600 struct statement * stmt;
2601 if (eof_token(token))
2602 break;
2603 if (match_op(token, '}'))
2604 break;
2605 if (match_ident(token, &_Static_assert_ident)) {
2606 token = parse_static_assert(token, NULL);
2607 continue;
2609 if (lookup_type(token)) {
2610 if (seen_statement) {
2611 warning(token->pos, "mixing declarations and code");
2612 seen_statement = 0;
2614 stmt = alloc_statement(token->pos, STMT_DECLARATION);
2615 token = external_declaration(token, &stmt->declaration, NULL);
2616 } else {
2617 seen_statement = Wdeclarationafterstatement;
2618 token = statement(token, &stmt);
2620 add_statement(list, stmt);
2622 return token;
2625 static struct token *identifier_list(struct token *token, struct symbol *fn)
2627 struct symbol_list **list = &fn->arguments;
2628 for (;;) {
2629 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2630 sym->ident = token->ident;
2631 token = token->next;
2632 sym->endpos = token->pos;
2633 sym->ctype.base_type = &incomplete_ctype;
2634 add_symbol(list, sym);
2635 if (!match_op(token, ',') ||
2636 token_type(token->next) != TOKEN_IDENT ||
2637 lookup_type(token->next))
2638 break;
2639 token = token->next;
2641 return token;
2644 static struct token *parameter_type_list(struct token *token, struct symbol *fn)
2646 struct symbol_list **list = &fn->arguments;
2648 for (;;) {
2649 struct symbol *sym;
2651 if (match_op(token, SPECIAL_ELLIPSIS)) {
2652 fn->variadic = 1;
2653 token = token->next;
2654 break;
2657 sym = alloc_symbol(token->pos, SYM_NODE);
2658 token = parameter_declaration(token, sym);
2659 if (sym->ctype.base_type == &void_ctype) {
2660 /* Special case: (void) */
2661 if (!*list && !sym->ident)
2662 break;
2663 warning(token->pos, "void parameter");
2665 add_symbol(list, sym);
2666 if (!match_op(token, ','))
2667 break;
2668 token = token->next;
2670 return token;
2673 struct token *compound_statement(struct token *token, struct statement *stmt)
2675 stmt->type = STMT_COMPOUND;
2676 start_block_scope(token->pos);
2677 token = statement_list(token, &stmt->stmts);
2678 end_block_scope();
2679 return token;
2682 static struct expression *identifier_expression(struct token *token)
2684 struct expression *expr = alloc_expression(token->pos, EXPR_IDENTIFIER);
2685 expr->expr_ident = token->ident;
2686 return expr;
2689 static struct expression *index_expression(struct expression *from, struct expression *to)
2691 int idx_from, idx_to;
2692 struct expression *expr = alloc_expression(from->pos, EXPR_INDEX);
2694 idx_from = const_expression_value(from);
2695 idx_to = idx_from;
2696 if (to) {
2697 idx_to = const_expression_value(to);
2698 if (idx_to < idx_from || idx_from < 0)
2699 warning(from->pos, "nonsense array initializer index range");
2701 expr->idx_from = idx_from;
2702 expr->idx_to = idx_to;
2703 return expr;
2706 static struct token *single_initializer(struct expression **ep, struct token *token)
2708 int expect_equal = 0;
2709 struct token *next = token->next;
2710 struct expression **tail = ep;
2711 int nested;
2713 *ep = NULL;
2715 if ((token_type(token) == TOKEN_IDENT) && match_op(next, ':')) {
2716 struct expression *expr = identifier_expression(token);
2717 if (Wold_initializer)
2718 warning(token->pos, "obsolete struct initializer, use C99 syntax");
2719 token = initializer(&expr->ident_expression, next->next);
2720 if (expr->ident_expression)
2721 *ep = expr;
2722 return token;
2725 for (tail = ep, nested = 0; ; nested++, next = token->next) {
2726 if (match_op(token, '.') && (token_type(next) == TOKEN_IDENT)) {
2727 struct expression *expr = identifier_expression(next);
2728 *tail = expr;
2729 tail = &expr->ident_expression;
2730 expect_equal = 1;
2731 token = next->next;
2732 } else if (match_op(token, '[')) {
2733 struct expression *from = NULL, *to = NULL, *expr;
2734 token = constant_expression(token->next, &from);
2735 if (!from) {
2736 sparse_error(token->pos, "Expected constant expression");
2737 break;
2739 if (match_op(token, SPECIAL_ELLIPSIS))
2740 token = constant_expression(token->next, &to);
2741 expr = index_expression(from, to);
2742 *tail = expr;
2743 tail = &expr->idx_expression;
2744 token = expect(token, ']', "at end of initializer index");
2745 if (nested)
2746 expect_equal = 1;
2747 } else {
2748 break;
2751 if (nested && !expect_equal) {
2752 if (!match_op(token, '='))
2753 warning(token->pos, "obsolete array initializer, use C99 syntax");
2754 else
2755 expect_equal = 1;
2757 if (expect_equal)
2758 token = expect(token, '=', "at end of initializer index");
2760 token = initializer(tail, token);
2761 if (!*tail)
2762 *ep = NULL;
2763 return token;
2766 static struct token *initializer_list(struct expression_list **list, struct token *token)
2768 struct expression *expr;
2770 for (;;) {
2771 token = single_initializer(&expr, token);
2772 if (!expr)
2773 break;
2774 add_expression(list, expr);
2775 if (!match_op(token, ','))
2776 break;
2777 token = token->next;
2779 return token;
2782 struct token *initializer(struct expression **tree, struct token *token)
2784 if (match_op(token, '{')) {
2785 struct expression *expr = alloc_expression(token->pos, EXPR_INITIALIZER);
2786 *tree = expr;
2787 if (!Wuniversal_initializer) {
2788 struct token *next = token->next;
2789 // '{ 0 }' is equivalent to '{ }' except for some
2790 // warnings, like using 0 to initialize a null-pointer.
2791 if (match_token_zero(next)) {
2792 if (match_op(next->next, '}'))
2793 expr->zero_init = 1;
2797 token = initializer_list(&expr->expr_list, token->next);
2798 return expect(token, '}', "at end of initializer");
2800 return assignment_expression(token, tree);
2803 static void declare_argument(struct symbol *sym, struct symbol *fn)
2805 if (!sym->ident) {
2806 sparse_error(sym->pos, "no identifier for function argument");
2807 return;
2809 if (sym->ctype.base_type == &incomplete_ctype) {
2810 sym->ctype.base_type = &int_ctype;
2812 if (Wimplicit_int) {
2813 sparse_error(sym->pos, "missing type declaration for parameter '%s'",
2814 show_ident(sym->ident));
2817 bind_symbol(sym, sym->ident, NS_SYMBOL);
2820 static int is_syscall(struct symbol *sym)
2822 char *macro;
2823 char *name;
2824 int is_syscall = 0;
2826 macro = get_macro_name(sym->pos);
2827 if (macro &&
2828 (strncmp("SYSCALL_DEFINE", macro, strlen("SYSCALL_DEFINE")) == 0 ||
2829 strncmp("COMPAT_SYSCALL_DEFINE", macro, strlen("COMPAT_SYSCALL_DEFINE")) == 0))
2830 is_syscall = 1;
2832 name = sym->ident->name;
2834 if (name && strncmp(name, "sys_", 4) ==0)
2835 is_syscall = 1;
2837 if (name && strncmp(name, "compat_sys_", 11) == 0)
2838 is_syscall = 1;
2840 return is_syscall;
2844 static struct token *parse_function_body(struct token *token, struct symbol *decl,
2845 struct symbol_list **list)
2847 struct symbol_list **old_symbol_list;
2848 struct symbol *base_type = decl->ctype.base_type;
2849 struct statement *stmt, **p;
2850 struct symbol *prev;
2851 struct symbol *arg;
2853 old_symbol_list = function_symbol_list;
2854 if (decl->ctype.modifiers & MOD_INLINE) {
2855 function_symbol_list = &decl->inline_symbol_list;
2856 p = &base_type->inline_stmt;
2857 } else {
2858 function_symbol_list = &decl->symbol_list;
2859 p = &base_type->stmt;
2861 function_computed_target_list = NULL;
2862 function_computed_goto_list = NULL;
2864 if ((decl->ctype.modifiers & (MOD_EXTERN|MOD_INLINE)) == MOD_EXTERN) {
2865 if (Wexternal_function_has_definition)
2866 warning(decl->pos, "function '%s' with external linkage has definition", show_ident(decl->ident));
2868 if (!(decl->ctype.modifiers & MOD_STATIC))
2869 decl->ctype.modifiers |= MOD_EXTERN;
2871 stmt = start_function(decl);
2872 *p = stmt;
2874 FOR_EACH_PTR (base_type->arguments, arg) {
2875 declare_argument(arg, base_type);
2876 } END_FOR_EACH_PTR(arg);
2878 token = statement_list(token->next, &stmt->stmts);
2879 end_function(decl);
2881 if (!(decl->ctype.modifiers & MOD_INLINE))
2882 add_symbol(list, decl);
2883 else if (is_syscall(decl)) {
2884 add_symbol(list, decl);
2886 printf("parse.c decl: %s\n", decl->ident->name);
2887 char *macro = get_macro_name(decl->pos);
2888 printf("decl macro: %s\n", macro);
2891 check_declaration(decl);
2892 decl->definition = decl;
2893 prev = decl->same_symbol;
2894 if (prev && prev->definition) {
2895 warning(decl->pos, "multiple definitions for function '%s'",
2896 show_ident(decl->ident));
2897 info(prev->definition->pos, " the previous one is here");
2898 } else {
2899 while (prev) {
2900 rebind_scope(prev, decl->scope);
2901 prev->definition = decl;
2902 prev = prev->same_symbol;
2905 function_symbol_list = old_symbol_list;
2906 if (function_computed_goto_list) {
2907 if (!function_computed_target_list)
2908 warning(decl->pos, "function '%s' has computed goto but no targets?", show_ident(decl->ident));
2909 else {
2910 FOR_EACH_PTR(function_computed_goto_list, stmt) {
2911 stmt->target_list = function_computed_target_list;
2912 } END_FOR_EACH_PTR(stmt);
2915 return expect(token, '}', "at end of function");
2918 static void promote_k_r_types(struct symbol *arg)
2920 struct symbol *base = arg->ctype.base_type;
2921 if (base && base->ctype.base_type == &int_type && base->rank < 0) {
2922 arg->ctype.base_type = &int_ctype;
2926 static void apply_k_r_types(struct symbol_list *argtypes, struct symbol *fn)
2928 struct symbol_list *real_args = fn->ctype.base_type->arguments;
2929 struct symbol *arg;
2931 FOR_EACH_PTR(real_args, arg) {
2932 struct symbol *type;
2934 /* This is quadratic in the number of arguments. We _really_ don't care */
2935 FOR_EACH_PTR(argtypes, type) {
2936 if (type->ident == arg->ident)
2937 goto match;
2938 } END_FOR_EACH_PTR(type);
2939 if (Wimplicit_int) {
2940 warning(arg->pos, "missing type declaration for parameter '%s'",
2941 show_ident(arg->ident));
2943 type = alloc_symbol(arg->pos, SYM_NODE);
2944 type->ident = arg->ident;
2945 type->ctype.base_type = &int_ctype;
2946 match:
2947 type->used = 1;
2948 /* "char" and "short" promote to "int" */
2949 promote_k_r_types(type);
2951 arg->ctype = type->ctype;
2952 } END_FOR_EACH_PTR(arg);
2954 FOR_EACH_PTR(argtypes, arg) {
2955 if (!arg->used)
2956 warning(arg->pos, "nonsensical parameter declaration '%s'", show_ident(arg->ident));
2957 } END_FOR_EACH_PTR(arg);
2961 static struct token *parse_k_r_arguments(struct token *token, struct symbol *decl,
2962 struct symbol_list **list)
2964 struct symbol_list *args = NULL;
2966 if (Wold_style_definition)
2967 warning(token->pos, "non-ANSI definition of function '%s'", show_ident(decl->ident));
2969 do {
2970 token = declaration_list(token, &args);
2971 if (!match_op(token, ';')) {
2972 sparse_error(token->pos, "expected ';' at end of parameter declaration");
2973 break;
2975 token = token->next;
2976 } while (lookup_type(token));
2978 apply_k_r_types(args, decl);
2980 if (!match_op(token, '{')) {
2981 sparse_error(token->pos, "expected function body");
2982 return token;
2984 return parse_function_body(token, decl, list);
2987 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list)
2989 struct symbol *anon = alloc_symbol(token->pos, SYM_NODE);
2990 struct symbol *fn = alloc_symbol(token->pos, SYM_FN);
2991 struct statement *stmt;
2993 anon->ctype.base_type = fn;
2994 stmt = alloc_statement(token->pos, STMT_NONE);
2995 fn->stmt = stmt;
2997 token = parse_asm_statement(token, stmt);
2999 // FIXME: add_symbol(list, anon);
3000 return token;
3003 struct token *external_declaration(struct token *token, struct symbol_list **list,
3004 validate_decl_t validate_decl)
3006 struct ident *ident = NULL;
3007 struct symbol *decl;
3008 struct decl_state ctx = { .ident = &ident };
3009 struct ctype saved;
3010 struct symbol *base_type;
3011 unsigned long mod;
3012 int is_typedef;
3014 if (match_ident(token, &_Pragma_ident))
3015 return parse_underscore_Pragma(token);
3017 /* Top-level inline asm or static assertion? */
3018 if (token_type(token) == TOKEN_IDENT) {
3019 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
3020 if (s && s->op->toplevel)
3021 return s->op->toplevel(token, list);
3024 /* Parse declaration-specifiers, if any */
3025 token = declaration_specifiers(token, &ctx);
3026 mod = decl_modifiers(&ctx);
3027 decl = alloc_symbol(token->pos, SYM_NODE);
3028 /* Just a type declaration? */
3029 if (match_op(token, ';')) {
3030 apply_modifiers(token->pos, &ctx);
3031 return token->next;
3034 saved = ctx.ctype;
3035 token = declarator(token, &ctx);
3036 token = handle_asm_name(token, &ctx);
3037 token = handle_attributes(token, &ctx);
3038 apply_modifiers(token->pos, &ctx);
3040 decl->ctype = ctx.ctype;
3041 decl->ctype.modifiers |= mod;
3042 decl->cleanup = ctx.cleanup;
3043 decl->endpos = token->pos;
3045 /* Just a type declaration? */
3046 if (!ident) {
3047 warning(token->pos, "missing identifier in declaration");
3048 return expect(token, ';', "at the end of type declaration");
3051 /* type define declaration? */
3052 is_typedef = ctx.storage_class == MOD_USERTYPE;
3054 /* Typedefs don't have meaningful storage */
3055 if (is_typedef)
3056 decl->ctype.modifiers |= MOD_USERTYPE;
3058 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
3060 base_type = decl->ctype.base_type;
3062 if (is_typedef) {
3063 if (base_type && !base_type->ident) {
3064 switch (base_type->type) {
3065 case SYM_STRUCT:
3066 case SYM_UNION:
3067 case SYM_ENUM:
3068 case SYM_RESTRICT:
3069 base_type->ident = ident;
3070 break;
3071 default:
3072 break;
3075 } else if (base_type && base_type->type == SYM_FN) {
3076 if (base_type->ctype.base_type == &autotype_ctype) {
3077 sparse_error(decl->pos, "'%s()' has __auto_type return type",
3078 show_ident(decl->ident));
3079 base_type->ctype.base_type = &int_ctype;
3081 if (base_type->ctype.base_type == &incomplete_ctype) {
3082 warning(decl->pos, "'%s()' has implicit return type",
3083 show_ident(decl->ident));
3084 base_type->ctype.base_type = &int_ctype;
3086 /* apply attributes placed after the declarator */
3087 decl->ctype.modifiers |= ctx.f_modifiers;
3089 /* K&R argument declaration? */
3090 if (lookup_type(token))
3091 return parse_k_r_arguments(token, decl, list);
3092 if (match_op(token, '{'))
3093 return parse_function_body(token, decl, list);
3095 if (!(decl->ctype.modifiers & MOD_STATIC))
3096 decl->ctype.modifiers |= MOD_EXTERN;
3097 } else if (base_type == &void_ctype && !(decl->ctype.modifiers & MOD_EXTERN)) {
3098 sparse_error(token->pos, "void declaration");
3100 if (base_type == &incomplete_ctype) {
3101 warning(decl->pos, "'%s' has implicit type", show_ident(decl->ident));
3102 decl->ctype.base_type = &int_ctype;;
3105 for (;;) {
3106 if (!is_typedef && match_op(token, '=')) {
3107 struct token *next = token->next;
3108 token = initializer(&decl->initializer, next);
3109 if (token == next)
3110 sparse_error(token->pos, "expression expected before '%s'", show_token(token));
3112 if (!is_typedef) {
3113 if (validate_decl)
3114 validate_decl(decl);
3116 if (decl->initializer && decl->ctype.modifiers & MOD_EXTERN) {
3117 warning(decl->pos, "symbol with external linkage has initializer");
3118 decl->ctype.modifiers &= ~MOD_EXTERN;
3121 if (!(decl->ctype.modifiers & (MOD_EXTERN | MOD_INLINE))) {
3122 add_symbol(list, decl);
3123 fn_local_symbol(decl);
3126 check_declaration(decl);
3127 if (decl->same_symbol) {
3128 decl->definition = decl->same_symbol->definition;
3129 decl->op = decl->same_symbol->op;
3130 if (is_typedef) {
3131 // TODO: handle -std=c89 --pedantic
3132 check_duplicates(decl);
3136 if (ctx.autotype) {
3137 const char *msg = NULL;
3138 if (decl->ctype.base_type != &autotype_ctype)
3139 msg = "on non-identifier";
3140 else if (match_op(token, ','))
3141 msg = "on declaration list";
3142 else if (!decl->initializer)
3143 msg = "without initializer";
3144 else if (decl->initializer->type == EXPR_SYMBOL &&
3145 decl->initializer->symbol == decl)
3146 msg = "on self-init var";
3147 if (msg) {
3148 sparse_error(decl->pos, "__auto_type %s", msg);
3149 decl->ctype.base_type = &bad_ctype;
3153 if (!match_op(token, ','))
3154 break;
3156 token = token->next;
3157 ident = NULL;
3158 decl = alloc_symbol(token->pos, SYM_NODE);
3159 ctx.ctype = saved;
3160 ctx.cleanup = NULL;
3161 token = handle_attributes(token, &ctx);
3162 token = declarator(token, &ctx);
3163 token = handle_asm_name(token, &ctx);
3164 token = handle_attributes(token, &ctx);
3165 apply_modifiers(token->pos, &ctx);
3166 decl->ctype = ctx.ctype;
3167 decl->ctype.modifiers |= mod;
3168 decl->cleanup = ctx.cleanup;
3169 decl->endpos = token->pos;
3170 if (!ident) {
3171 sparse_error(token->pos, "expected identifier name in type definition");
3172 return token;
3175 if (is_typedef)
3176 decl->ctype.modifiers |= MOD_USERTYPE;
3178 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
3180 /* Function declarations are automatically extern unless specifically static */
3181 base_type = decl->ctype.base_type;
3182 if (!is_typedef && base_type && base_type->type == SYM_FN) {
3183 if (!(decl->ctype.modifiers & MOD_STATIC))
3184 decl->ctype.modifiers |= MOD_EXTERN;
3187 return expect(token, ';', "at end of declaration");