unssa: update comment about the unneeded copies
[smatch.git] / parse.c
blob9b087f0a465a06f5232d0e23dfa6df59decac6ff
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, unsigned int keywords);
54 typedef struct token *declarator_t(struct token *, struct decl_state *);
55 static declarator_t
56 struct_specifier, union_specifier, enum_specifier,
57 attribute_specifier, typeof_specifier, parse_asm_declarator,
58 typedef_specifier, inline_specifier, auto_specifier,
59 register_specifier, static_specifier, extern_specifier,
60 thread_specifier, const_qualifier, volatile_qualifier;
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);
77 typedef struct token *attr_t(struct token *, struct symbol *,
78 struct decl_state *);
80 static attr_t
81 attribute_packed, attribute_aligned, attribute_modifier,
82 attribute_address_space, attribute_context,
83 attribute_designated_init,
84 attribute_transparent_union, ignore_attribute,
85 attribute_mode, attribute_force;
87 typedef struct symbol *to_mode_t(struct symbol *);
89 static to_mode_t
90 to_QI_mode, to_HI_mode, to_SI_mode, to_DI_mode, to_TI_mode, to_word_mode;
92 enum {
93 Set_T = 1,
94 Set_S = 2,
95 Set_Char = 4,
96 Set_Int = 8,
97 Set_Double = 16,
98 Set_Float = 32,
99 Set_Signed = 64,
100 Set_Unsigned = 128,
101 Set_Short = 256,
102 Set_Long = 512,
103 Set_Vlong = 1024,
104 Set_Int128 = 2048,
105 Set_Any = Set_T | Set_Short | Set_Long | Set_Signed | Set_Unsigned
108 enum {
109 CInt = 0, CSInt, CUInt, CReal, CChar, CSChar, CUChar,
112 enum {
113 SNone = 0, STypedef, SAuto, SRegister, SExtern, SStatic, SForced, SMax,
116 static struct symbol_op typedef_op = {
117 .type = KW_MODIFIER,
118 .declarator = typedef_specifier,
121 static struct symbol_op inline_op = {
122 .type = KW_MODIFIER,
123 .declarator = inline_specifier,
126 static struct symbol_op auto_op = {
127 .type = KW_MODIFIER,
128 .declarator = auto_specifier,
131 static struct symbol_op register_op = {
132 .type = KW_MODIFIER,
133 .declarator = register_specifier,
136 static struct symbol_op static_op = {
137 .type = KW_MODIFIER,
138 .declarator = static_specifier,
141 static struct symbol_op extern_op = {
142 .type = KW_MODIFIER,
143 .declarator = extern_specifier,
146 static struct symbol_op thread_op = {
147 .type = KW_MODIFIER,
148 .declarator = thread_specifier,
151 static struct symbol_op const_op = {
152 .type = KW_QUALIFIER,
153 .declarator = const_qualifier,
156 static struct symbol_op volatile_op = {
157 .type = KW_QUALIFIER,
158 .declarator = volatile_qualifier,
161 static struct symbol_op restrict_op = {
162 .type = KW_QUALIFIER,
165 static struct symbol_op typeof_op = {
166 .type = KW_SPECIFIER,
167 .declarator = typeof_specifier,
168 .test = Set_Any,
169 .set = Set_S|Set_T,
172 static struct symbol_op attribute_op = {
173 .type = KW_ATTRIBUTE,
174 .declarator = attribute_specifier,
177 static struct symbol_op struct_op = {
178 .type = KW_SPECIFIER,
179 .declarator = struct_specifier,
180 .test = Set_Any,
181 .set = Set_S|Set_T,
184 static struct symbol_op union_op = {
185 .type = KW_SPECIFIER,
186 .declarator = union_specifier,
187 .test = Set_Any,
188 .set = Set_S|Set_T,
191 static struct symbol_op enum_op = {
192 .type = KW_SPECIFIER,
193 .declarator = enum_specifier,
194 .test = Set_Any,
195 .set = Set_S|Set_T,
198 static struct symbol_op spec_op = {
199 .type = KW_SPECIFIER | KW_EXACT,
200 .test = Set_Any,
201 .set = Set_S|Set_T,
204 static struct symbol_op char_op = {
205 .type = KW_SPECIFIER,
206 .test = Set_T|Set_Long|Set_Short,
207 .set = Set_T|Set_Char,
208 .class = CChar,
211 static struct symbol_op int_op = {
212 .type = KW_SPECIFIER,
213 .test = Set_T,
214 .set = Set_T|Set_Int,
217 static struct symbol_op double_op = {
218 .type = KW_SPECIFIER,
219 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Vlong,
220 .set = Set_T|Set_Double,
221 .class = CReal,
224 static struct symbol_op float_op = {
225 .type = KW_SPECIFIER | KW_SHORT,
226 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Long,
227 .set = Set_T|Set_Float,
228 .class = CReal,
231 static struct symbol_op short_op = {
232 .type = KW_SPECIFIER | KW_SHORT,
233 .test = Set_S|Set_Char|Set_Float|Set_Double|Set_Long|Set_Short,
234 .set = Set_Short,
237 static struct symbol_op signed_op = {
238 .type = KW_SPECIFIER,
239 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
240 .set = Set_Signed,
241 .class = CSInt,
244 static struct symbol_op unsigned_op = {
245 .type = KW_SPECIFIER,
246 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
247 .set = Set_Unsigned,
248 .class = CUInt,
251 static struct symbol_op long_op = {
252 .type = KW_SPECIFIER | KW_LONG,
253 .test = Set_S|Set_Char|Set_Float|Set_Short|Set_Vlong,
254 .set = Set_Long,
257 static struct symbol_op int128_op = {
258 .type = KW_SPECIFIER | KW_LONG,
259 .test = Set_S|Set_T|Set_Char|Set_Short|Set_Int|Set_Float|Set_Double|Set_Long|Set_Vlong|Set_Int128,
260 .set = Set_T|Set_Int128,
263 static struct symbol_op if_op = {
264 .statement = parse_if_statement,
267 static struct symbol_op return_op = {
268 .statement = parse_return_statement,
271 static struct symbol_op loop_iter_op = {
272 .statement = parse_loop_iterator,
275 static struct symbol_op default_op = {
276 .statement = parse_default_statement,
279 static struct symbol_op case_op = {
280 .statement = parse_case_statement,
283 static struct symbol_op switch_op = {
284 .statement = parse_switch_statement,
287 static struct symbol_op for_op = {
288 .statement = parse_for_statement,
291 static struct symbol_op while_op = {
292 .statement = parse_while_statement,
295 static struct symbol_op do_op = {
296 .statement = parse_do_statement,
299 static struct symbol_op goto_op = {
300 .statement = parse_goto_statement,
303 static struct symbol_op __context___op = {
304 .statement = parse_context_statement,
307 static struct symbol_op range_op = {
308 .statement = parse_range_statement,
311 static struct symbol_op asm_op = {
312 .type = KW_ASM,
313 .declarator = parse_asm_declarator,
314 .statement = parse_asm_statement,
315 .toplevel = toplevel_asm_declaration,
318 static struct symbol_op packed_op = {
319 .attribute = attribute_packed,
322 static struct symbol_op aligned_op = {
323 .attribute = attribute_aligned,
326 static struct symbol_op attr_mod_op = {
327 .attribute = attribute_modifier,
330 static struct symbol_op attr_force_op = {
331 .attribute = attribute_force,
334 static struct symbol_op address_space_op = {
335 .attribute = attribute_address_space,
338 static struct symbol_op mode_op = {
339 .attribute = attribute_mode,
342 static struct symbol_op context_op = {
343 .attribute = attribute_context,
346 static struct symbol_op designated_init_op = {
347 .attribute = attribute_designated_init,
350 static struct symbol_op transparent_union_op = {
351 .attribute = attribute_transparent_union,
354 static struct symbol_op ignore_attr_op = {
355 .attribute = ignore_attribute,
358 static struct symbol_op mode_QI_op = {
359 .type = KW_MODE,
360 .to_mode = to_QI_mode
363 static struct symbol_op mode_HI_op = {
364 .type = KW_MODE,
365 .to_mode = to_HI_mode
368 static struct symbol_op mode_SI_op = {
369 .type = KW_MODE,
370 .to_mode = to_SI_mode
373 static struct symbol_op mode_DI_op = {
374 .type = KW_MODE,
375 .to_mode = to_DI_mode
378 static struct symbol_op mode_TI_op = {
379 .type = KW_MODE,
380 .to_mode = to_TI_mode
383 static struct symbol_op mode_word_op = {
384 .type = KW_MODE,
385 .to_mode = to_word_mode
388 static struct init_keyword {
389 const char *name;
390 enum namespace ns;
391 unsigned long modifiers;
392 struct symbol_op *op;
393 struct symbol *type;
394 } keyword_table[] = {
395 /* Type qualifiers */
396 { "const", NS_TYPEDEF, .op = &const_op },
397 { "__const", NS_TYPEDEF, .op = &const_op },
398 { "__const__", NS_TYPEDEF, .op = &const_op },
399 { "volatile", NS_TYPEDEF, .op = &volatile_op },
400 { "__volatile", NS_TYPEDEF, .op = &volatile_op },
401 { "__volatile__", NS_TYPEDEF, .op = &volatile_op },
403 /* Typedef.. */
404 { "typedef", NS_TYPEDEF, .op = &typedef_op },
406 /* Type specifiers */
407 { "void", NS_TYPEDEF, .type = &void_ctype, .op = &spec_op},
408 { "char", NS_TYPEDEF, .op = &char_op },
409 { "short", NS_TYPEDEF, .op = &short_op },
410 { "int", NS_TYPEDEF, .op = &int_op },
411 { "long", NS_TYPEDEF, .op = &long_op },
412 { "float", NS_TYPEDEF, .op = &float_op },
413 { "double", NS_TYPEDEF, .op = &double_op },
414 { "signed", NS_TYPEDEF, .op = &signed_op },
415 { "__signed", NS_TYPEDEF, .op = &signed_op },
416 { "__signed__", NS_TYPEDEF, .op = &signed_op },
417 { "unsigned", NS_TYPEDEF, .op = &unsigned_op },
418 { "__int128", NS_TYPEDEF, .op = &int128_op },
419 { "_Bool", NS_TYPEDEF, .type = &bool_ctype, .op = &spec_op },
421 /* Predeclared types */
422 { "__builtin_va_list", NS_TYPEDEF, .type = &ptr_ctype, .op = &spec_op },
423 { "__builtin_ms_va_list", NS_TYPEDEF, .type = &ptr_ctype, .op = &spec_op },
424 { "__int128_t", NS_TYPEDEF, .type = &lllong_ctype, .op = &spec_op },
425 { "__uint128_t",NS_TYPEDEF, .type = &ulllong_ctype, .op = &spec_op },
427 /* Extended types */
428 { "typeof", NS_TYPEDEF, .op = &typeof_op },
429 { "__typeof", NS_TYPEDEF, .op = &typeof_op },
430 { "__typeof__", NS_TYPEDEF, .op = &typeof_op },
432 { "__attribute", NS_TYPEDEF, .op = &attribute_op },
433 { "__attribute__", NS_TYPEDEF, .op = &attribute_op },
435 { "struct", NS_TYPEDEF, .op = &struct_op },
436 { "union", NS_TYPEDEF, .op = &union_op },
437 { "enum", NS_TYPEDEF, .op = &enum_op },
439 { "inline", NS_TYPEDEF, .op = &inline_op },
440 { "__inline", NS_TYPEDEF, .op = &inline_op },
441 { "__inline__", NS_TYPEDEF, .op = &inline_op },
443 /* Ignored for now.. */
444 { "restrict", NS_TYPEDEF, .op = &restrict_op},
445 { "__restrict", NS_TYPEDEF, .op = &restrict_op},
446 { "__restrict__", NS_TYPEDEF, .op = &restrict_op},
448 /* Storage class */
449 { "auto", NS_TYPEDEF, .op = &auto_op },
450 { "register", NS_TYPEDEF, .op = &register_op },
451 { "static", NS_TYPEDEF, .op = &static_op },
452 { "extern", NS_TYPEDEF, .op = &extern_op },
453 { "__thread", NS_TYPEDEF, .op = &thread_op },
455 /* Statement */
456 { "if", NS_KEYWORD, .op = &if_op },
457 { "return", NS_KEYWORD, .op = &return_op },
458 { "break", NS_KEYWORD, .op = &loop_iter_op },
459 { "continue", NS_KEYWORD, .op = &loop_iter_op },
460 { "default", NS_KEYWORD, .op = &default_op },
461 { "case", NS_KEYWORD, .op = &case_op },
462 { "switch", NS_KEYWORD, .op = &switch_op },
463 { "for", NS_KEYWORD, .op = &for_op },
464 { "while", NS_KEYWORD, .op = &while_op },
465 { "do", NS_KEYWORD, .op = &do_op },
466 { "goto", NS_KEYWORD, .op = &goto_op },
467 { "__context__",NS_KEYWORD, .op = &__context___op },
468 { "__range__", NS_KEYWORD, .op = &range_op },
469 { "asm", NS_KEYWORD, .op = &asm_op },
470 { "__asm", NS_KEYWORD, .op = &asm_op },
471 { "__asm__", NS_KEYWORD, .op = &asm_op },
473 /* Attribute */
474 { "packed", NS_KEYWORD, .op = &packed_op },
475 { "__packed__", NS_KEYWORD, .op = &packed_op },
476 { "aligned", NS_KEYWORD, .op = &aligned_op },
477 { "__aligned__",NS_KEYWORD, .op = &aligned_op },
478 { "nocast", NS_KEYWORD, MOD_NOCAST, .op = &attr_mod_op },
479 { "noderef", NS_KEYWORD, MOD_NODEREF, .op = &attr_mod_op },
480 { "safe", NS_KEYWORD, MOD_SAFE, .op = &attr_mod_op },
481 { "force", NS_KEYWORD, .op = &attr_force_op },
482 { "bitwise", NS_KEYWORD, MOD_BITWISE, .op = &attr_mod_op },
483 { "__bitwise__",NS_KEYWORD, MOD_BITWISE, .op = &attr_mod_op },
484 { "address_space",NS_KEYWORD, .op = &address_space_op },
485 { "mode", NS_KEYWORD, .op = &mode_op },
486 { "context", NS_KEYWORD, .op = &context_op },
487 { "designated_init", NS_KEYWORD, .op = &designated_init_op },
488 { "__transparent_union__", NS_KEYWORD, .op = &transparent_union_op },
489 { "noreturn", NS_KEYWORD, MOD_NORETURN, .op = &attr_mod_op },
490 { "__noreturn__", NS_KEYWORD, MOD_NORETURN, .op = &attr_mod_op },
491 { "pure", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
492 {"__pure__", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
493 {"const", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
494 {"__const", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
495 {"__const__", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
497 { "__mode__", NS_KEYWORD, .op = &mode_op },
498 { "QI", NS_KEYWORD, MOD_CHAR, .op = &mode_QI_op },
499 { "__QI__", NS_KEYWORD, MOD_CHAR, .op = &mode_QI_op },
500 { "HI", NS_KEYWORD, MOD_SHORT, .op = &mode_HI_op },
501 { "__HI__", NS_KEYWORD, MOD_SHORT, .op = &mode_HI_op },
502 { "SI", NS_KEYWORD, .op = &mode_SI_op },
503 { "__SI__", NS_KEYWORD, .op = &mode_SI_op },
504 { "DI", NS_KEYWORD, MOD_LONGLONG, .op = &mode_DI_op },
505 { "__DI__", NS_KEYWORD, MOD_LONGLONG, .op = &mode_DI_op },
506 { "TI", NS_KEYWORD, MOD_LONGLONGLONG, .op = &mode_TI_op },
507 { "__TI__", NS_KEYWORD, MOD_LONGLONGLONG, .op = &mode_TI_op },
508 { "word", NS_KEYWORD, MOD_LONG, .op = &mode_word_op },
509 { "__word__", NS_KEYWORD, MOD_LONG, .op = &mode_word_op },
512 static const char *ignored_attributes[] = {
513 "alias",
514 "__alias__",
515 "alloc_align",
516 "__alloc_align__",
517 "alloc_size",
518 "__alloc_size__",
519 "always_inline",
520 "__always_inline__",
521 "artificial",
522 "__artificial__",
523 "assume_aligned",
524 "__assume_aligned__",
525 "bounded",
526 "__bounded__",
527 "cdecl",
528 "__cdecl__",
529 "cold",
530 "__cold__",
531 "constructor",
532 "__constructor__",
533 "deprecated",
534 "__deprecated__",
535 "destructor",
536 "__destructor__",
537 "dllexport",
538 "__dllexport__",
539 "dllimport",
540 "__dllimport__",
541 "error",
542 "__error__",
543 "externally_visible",
544 "__externally_visible__",
545 "fastcall",
546 "__fastcall__",
547 "format",
548 "__format__",
549 "format_arg",
550 "__format_arg__",
551 "gnu_inline",
552 "__gnu_inline__",
553 "hot",
554 "__hot__",
555 "hotpatch",
556 "__hotpatch__",
557 "leaf",
558 "__leaf__",
559 "l1_text",
560 "__l1_text__",
561 "l1_data",
562 "__l1_data__",
563 "l2",
564 "__l2__",
565 "malloc",
566 "__malloc__",
567 "may_alias",
568 "__may_alias__",
569 "model",
570 "__model__",
571 "ms_abi",
572 "__ms_abi__",
573 "ms_hook_prologue",
574 "__ms_hook_prologue__",
575 "naked",
576 "__naked__",
577 "no_instrument_function",
578 "__no_instrument_function__",
579 "noclone",
580 "__noclone",
581 "__noclone__",
582 "noinline",
583 "__noinline__",
584 "nonnull",
585 "__nonnull",
586 "__nonnull__",
587 "nothrow",
588 "__nothrow",
589 "__nothrow__",
590 "regparm",
591 "__regparm__",
592 "section",
593 "__section__",
594 "sentinel",
595 "__sentinel__",
596 "signal",
597 "__signal__",
598 "stdcall",
599 "__stdcall__",
600 "syscall_linkage",
601 "__syscall_linkage__",
602 "sysv_abi",
603 "__sysv_abi__",
604 "unused",
605 "__unused__",
606 "used",
607 "__used__",
608 "vector_size",
609 "__vector_size__",
610 "visibility",
611 "__visibility__",
612 "warn_unused_result",
613 "__warn_unused_result__",
614 "warning",
615 "__warning__",
616 "weak",
617 "__weak__",
618 "no_sanitize_address",
619 "__no_sanitize_address__",
623 void init_parser(int stream)
625 int i;
626 for (i = 0; i < ARRAY_SIZE(keyword_table); i++) {
627 struct init_keyword *ptr = keyword_table + i;
628 struct symbol *sym = create_symbol(stream, ptr->name, SYM_KEYWORD, ptr->ns);
629 sym->ident->keyword = 1;
630 if (ptr->ns == NS_TYPEDEF)
631 sym->ident->reserved = 1;
632 sym->ctype.modifiers = ptr->modifiers;
633 sym->ctype.base_type = ptr->type;
634 sym->op = ptr->op;
637 for (i = 0; i < ARRAY_SIZE(ignored_attributes); i++) {
638 const char * name = ignored_attributes[i];
639 struct symbol *sym = create_symbol(stream, name, SYM_KEYWORD,
640 NS_KEYWORD);
641 sym->ident->keyword = 1;
642 sym->op = &ignore_attr_op;
647 // Add a symbol to the list of function-local symbols
648 static void fn_local_symbol(struct symbol *sym)
650 if (function_symbol_list)
651 add_symbol(function_symbol_list, sym);
654 static int SENTINEL_ATTR match_idents(struct token *token, ...)
656 va_list args;
657 struct ident * next;
659 if (token_type(token) != TOKEN_IDENT)
660 return 0;
662 va_start(args, token);
663 do {
664 next = va_arg(args, struct ident *);
665 } while (next && token->ident != next);
666 va_end(args);
668 return next && token->ident == next;
672 struct statement *alloc_statement(struct position pos, int type)
674 struct statement *stmt = __alloc_statement(0);
675 stmt->type = type;
676 stmt->pos = pos;
677 return stmt;
680 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list);
682 static void apply_modifiers(struct position pos, struct decl_state *ctx)
684 struct symbol *ctype;
685 if (!ctx->mode)
686 return;
687 ctype = ctx->mode->to_mode(ctx->ctype.base_type);
688 if (!ctype)
689 sparse_error(pos, "don't know how to apply mode to %s",
690 show_typename(ctx->ctype.base_type));
691 else
692 ctx->ctype.base_type = ctype;
696 static struct symbol * alloc_indirect_symbol(struct position pos, struct ctype *ctype, int type)
698 struct symbol *sym = alloc_symbol(pos, type);
700 sym->ctype.base_type = ctype->base_type;
701 sym->ctype.modifiers = ctype->modifiers;
703 ctype->base_type = sym;
704 ctype->modifiers = 0;
705 return sym;
709 * NOTE! NS_LABEL is not just a different namespace,
710 * it also ends up using function scope instead of the
711 * regular symbol scope.
713 struct symbol *label_symbol(struct token *token)
715 struct symbol *sym = lookup_symbol(token->ident, NS_LABEL);
716 if (!sym) {
717 sym = alloc_symbol(token->pos, SYM_LABEL);
718 bind_symbol(sym, token->ident, NS_LABEL);
719 fn_local_symbol(sym);
721 return sym;
724 static struct token *struct_union_enum_specifier(enum type type,
725 struct token *token, struct decl_state *ctx,
726 struct token *(*parse)(struct token *, struct symbol *))
728 struct symbol *sym;
729 struct position *repos;
731 token = handle_attributes(token, ctx, KW_ATTRIBUTE);
732 if (token_type(token) == TOKEN_IDENT) {
733 sym = lookup_symbol(token->ident, NS_STRUCT);
734 if (!sym ||
735 (is_outer_scope(sym->scope) &&
736 (match_op(token->next,';') || match_op(token->next,'{')))) {
737 // Either a new symbol, or else an out-of-scope
738 // symbol being redefined.
739 sym = alloc_symbol(token->pos, type);
740 bind_symbol(sym, token->ident, NS_STRUCT);
742 if (sym->type != type)
743 error_die(token->pos, "invalid tag applied to %s", show_typename (sym));
744 ctx->ctype.base_type = sym;
745 repos = &token->pos;
746 token = token->next;
747 if (match_op(token, '{')) {
748 // The following test is actually wrong for empty
749 // structs, but (1) they are not C99, (2) gcc does
750 // the same thing, and (3) it's easier.
751 if (sym->symbol_list)
752 error_die(token->pos, "redefinition of %s", show_typename (sym));
753 sym->pos = *repos;
754 token = parse(token->next, sym);
755 token = expect(token, '}', "at end of struct-union-enum-specifier");
757 // Mark the structure as needing re-examination
758 sym->examined = 0;
759 sym->endpos = token->pos;
761 return token;
764 // private struct/union/enum type
765 if (!match_op(token, '{')) {
766 sparse_error(token->pos, "expected declaration");
767 ctx->ctype.base_type = &bad_ctype;
768 return token;
771 sym = alloc_symbol(token->pos, type);
772 token = parse(token->next, sym);
773 ctx->ctype.base_type = sym;
774 token = expect(token, '}', "at end of specifier");
775 sym->endpos = token->pos;
777 return token;
780 static struct token *parse_struct_declaration(struct token *token, struct symbol *sym)
782 struct symbol *field, *last = NULL;
783 struct token *res;
784 res = struct_declaration_list(token, &sym->symbol_list);
785 FOR_EACH_PTR(sym->symbol_list, field) {
786 if (!field->ident) {
787 struct symbol *base = field->ctype.base_type;
788 if (base && base->type == SYM_BITFIELD)
789 continue;
791 if (last)
792 last->next_subobject = field;
793 last = field;
794 } END_FOR_EACH_PTR(field);
795 return res;
798 static struct token *parse_union_declaration(struct token *token, struct symbol *sym)
800 return struct_declaration_list(token, &sym->symbol_list);
803 static struct token *struct_specifier(struct token *token, struct decl_state *ctx)
805 return struct_union_enum_specifier(SYM_STRUCT, token, ctx, parse_struct_declaration);
808 static struct token *union_specifier(struct token *token, struct decl_state *ctx)
810 return struct_union_enum_specifier(SYM_UNION, token, ctx, parse_union_declaration);
814 typedef struct {
815 int x;
816 unsigned long long y;
817 } Num;
819 static void upper_boundary(Num *n, Num *v)
821 if (n->x > v->x)
822 return;
823 if (n->x < v->x) {
824 *n = *v;
825 return;
827 if (n->y < v->y)
828 n->y = v->y;
831 static void lower_boundary(Num *n, Num *v)
833 if (n->x < v->x)
834 return;
835 if (n->x > v->x) {
836 *n = *v;
837 return;
839 if (n->y > v->y)
840 n->y = v->y;
843 static int type_is_ok(struct symbol *type, Num *upper, Num *lower)
845 int shift = type->bit_size;
846 int is_unsigned = type->ctype.modifiers & MOD_UNSIGNED;
848 if (!is_unsigned)
849 shift--;
850 if (upper->x == 0 && upper->y >> shift)
851 return 0;
852 if (lower->x == 0 || (!is_unsigned && (~lower->y >> shift) == 0))
853 return 1;
854 return 0;
857 static struct symbol *bigger_enum_type(struct symbol *s1, struct symbol *s2)
859 if (s1->bit_size < s2->bit_size) {
860 s1 = s2;
861 } else if (s1->bit_size == s2->bit_size) {
862 if (s2->ctype.modifiers & MOD_UNSIGNED)
863 s1 = s2;
865 if (s1->bit_size < bits_in_int)
866 return &int_ctype;
867 return s1;
870 static void cast_enum_list(struct symbol_list *list, struct symbol *base_type)
872 struct symbol *sym;
874 FOR_EACH_PTR(list, sym) {
875 struct expression *expr = sym->initializer;
876 struct symbol *ctype;
877 if (expr->type != EXPR_VALUE)
878 continue;
879 ctype = expr->ctype;
880 if (ctype->bit_size == base_type->bit_size)
881 continue;
882 cast_value(expr, base_type, expr, ctype);
883 } END_FOR_EACH_PTR(sym);
886 static struct token *parse_enum_declaration(struct token *token, struct symbol *parent)
888 unsigned long long lastval = 0;
889 struct symbol *ctype = NULL, *base_type = NULL;
890 Num upper = {-1, 0}, lower = {1, 0};
892 parent->examined = 1;
893 parent->ctype.base_type = &int_ctype;
894 while (token_type(token) == TOKEN_IDENT) {
895 struct expression *expr = NULL;
896 struct token *next = token->next;
897 struct symbol *sym;
899 if (match_op(next, '=')) {
900 next = constant_expression(next->next, &expr);
901 lastval = get_expression_value(expr);
902 ctype = &void_ctype;
903 if (expr && expr->ctype)
904 ctype = expr->ctype;
905 } else if (!ctype) {
906 ctype = &int_ctype;
907 } else if (is_int_type(ctype)) {
908 lastval++;
909 } else {
910 error_die(token->pos, "can't increment the last enum member");
913 if (!expr) {
914 expr = alloc_expression(token->pos, EXPR_VALUE);
915 expr->value = lastval;
916 expr->ctype = ctype;
919 sym = alloc_symbol(token->pos, SYM_NODE);
920 bind_symbol(sym, token->ident, NS_SYMBOL);
921 sym->ctype.modifiers &= ~MOD_ADDRESSABLE;
922 sym->initializer = expr;
923 sym->enum_member = 1;
924 sym->ctype.base_type = parent;
925 add_ptr_list(&parent->symbol_list, sym);
927 if (base_type != &bad_ctype) {
928 if (ctype->type == SYM_NODE)
929 ctype = ctype->ctype.base_type;
930 if (ctype->type == SYM_ENUM) {
931 if (ctype == parent)
932 ctype = base_type;
933 else
934 ctype = ctype->ctype.base_type;
937 * base_type rules:
938 * - if all enums are of the same type, then
939 * the base_type is that type (two first
940 * cases)
941 * - if enums are of different types, they
942 * all have to be integer types, and the
943 * base type is at least "int_ctype".
944 * - otherwise the base_type is "bad_ctype".
946 if (!base_type) {
947 base_type = ctype;
948 } else if (ctype == base_type) {
949 /* nothing */
950 } else if (is_int_type(base_type) && is_int_type(ctype)) {
951 base_type = bigger_enum_type(base_type, ctype);
952 } else
953 base_type = &bad_ctype;
954 parent->ctype.base_type = base_type;
956 if (is_int_type(base_type)) {
957 Num v = {.y = lastval};
958 if (ctype->ctype.modifiers & MOD_UNSIGNED)
959 v.x = 0;
960 else if ((long long)lastval >= 0)
961 v.x = 0;
962 else
963 v.x = -1;
964 upper_boundary(&upper, &v);
965 lower_boundary(&lower, &v);
967 token = next;
969 sym->endpos = token->pos;
971 if (!match_op(token, ','))
972 break;
973 token = token->next;
975 if (!base_type) {
976 sparse_error(token->pos, "bad enum definition");
977 base_type = &bad_ctype;
979 else if (!is_int_type(base_type))
980 base_type = base_type;
981 else if (type_is_ok(base_type, &upper, &lower))
982 base_type = base_type;
983 else if (type_is_ok(&int_ctype, &upper, &lower))
984 base_type = &int_ctype;
985 else if (type_is_ok(&uint_ctype, &upper, &lower))
986 base_type = &uint_ctype;
987 else if (type_is_ok(&long_ctype, &upper, &lower))
988 base_type = &long_ctype;
989 else if (type_is_ok(&ulong_ctype, &upper, &lower))
990 base_type = &ulong_ctype;
991 else if (type_is_ok(&llong_ctype, &upper, &lower))
992 base_type = &llong_ctype;
993 else if (type_is_ok(&ullong_ctype, &upper, &lower))
994 base_type = &ullong_ctype;
995 else
996 base_type = &bad_ctype;
997 parent->ctype.base_type = base_type;
998 parent->ctype.modifiers |= (base_type->ctype.modifiers & MOD_UNSIGNED);
999 parent->examined = 0;
1001 cast_enum_list(parent->symbol_list, base_type);
1003 return token;
1006 static struct token *enum_specifier(struct token *token, struct decl_state *ctx)
1008 struct token *ret = struct_union_enum_specifier(SYM_ENUM, token, ctx, parse_enum_declaration);
1009 struct ctype *ctype = &ctx->ctype.base_type->ctype;
1011 if (!ctype->base_type)
1012 ctype->base_type = &incomplete_ctype;
1014 return ret;
1017 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype);
1019 static struct token *typeof_specifier(struct token *token, struct decl_state *ctx)
1021 struct symbol *sym;
1023 if (!match_op(token, '(')) {
1024 sparse_error(token->pos, "expected '(' after typeof");
1025 return token;
1027 if (lookup_type(token->next)) {
1028 token = typename(token->next, &sym, NULL);
1029 ctx->ctype.base_type = sym->ctype.base_type;
1030 apply_ctype(token->pos, &sym->ctype, &ctx->ctype);
1031 } else {
1032 struct symbol *typeof_sym = alloc_symbol(token->pos, SYM_TYPEOF);
1033 token = parse_expression(token->next, &typeof_sym->initializer);
1035 typeof_sym->endpos = token->pos;
1036 if (!typeof_sym->initializer) {
1037 sparse_error(token->pos, "expected expression after the '(' token");
1038 typeof_sym = &bad_ctype;
1040 ctx->ctype.base_type = typeof_sym;
1042 return expect(token, ')', "after typeof");
1045 static struct token *ignore_attribute(struct token *token, struct symbol *attr, struct decl_state *ctx)
1047 struct expression *expr = NULL;
1048 if (match_op(token, '('))
1049 token = parens_expression(token, &expr, "in attribute");
1050 return token;
1053 static struct token *attribute_packed(struct token *token, struct symbol *attr, struct decl_state *ctx)
1055 if (!ctx->ctype.alignment)
1056 ctx->ctype.alignment = 1;
1057 return token;
1060 static struct token *attribute_aligned(struct token *token, struct symbol *attr, struct decl_state *ctx)
1062 int alignment = max_alignment;
1063 struct expression *expr = NULL;
1065 if (match_op(token, '(')) {
1066 token = parens_expression(token, &expr, "in attribute");
1067 if (expr)
1068 alignment = const_expression_value(expr);
1070 if (alignment & (alignment-1)) {
1071 warning(token->pos, "I don't like non-power-of-2 alignments");
1072 return token;
1073 } else if (alignment > ctx->ctype.alignment)
1074 ctx->ctype.alignment = alignment;
1075 return token;
1078 static void apply_qualifier(struct position *pos, struct ctype *ctx, unsigned long qual)
1080 if (ctx->modifiers & qual)
1081 warning(*pos, "duplicate %s", modifier_string(qual));
1082 ctx->modifiers |= qual;
1085 static struct token *attribute_modifier(struct token *token, struct symbol *attr, struct decl_state *ctx)
1087 apply_qualifier(&token->pos, &ctx->ctype, attr->ctype.modifiers);
1088 return token;
1091 static struct token *attribute_address_space(struct token *token, struct symbol *attr, struct decl_state *ctx)
1093 struct expression *expr = NULL;
1094 int as;
1095 token = expect(token, '(', "after address_space attribute");
1096 token = conditional_expression(token, &expr);
1097 if (expr) {
1098 as = const_expression_value(expr);
1099 if (Waddress_space && as)
1100 ctx->ctype.as = as;
1102 token = expect(token, ')', "after address_space attribute");
1103 return token;
1106 static struct symbol *to_QI_mode(struct symbol *ctype)
1108 if (ctype->ctype.base_type != &int_type)
1109 return NULL;
1110 if (ctype == &char_ctype)
1111 return ctype;
1112 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uchar_ctype
1113 : &schar_ctype;
1116 static struct symbol *to_HI_mode(struct symbol *ctype)
1118 if (ctype->ctype.base_type != &int_type)
1119 return NULL;
1120 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ushort_ctype
1121 : &sshort_ctype;
1124 static struct symbol *to_SI_mode(struct symbol *ctype)
1126 if (ctype->ctype.base_type != &int_type)
1127 return NULL;
1128 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint_ctype
1129 : &sint_ctype;
1132 static struct symbol *to_DI_mode(struct symbol *ctype)
1134 if (ctype->ctype.base_type != &int_type)
1135 return NULL;
1136 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ullong_ctype
1137 : &sllong_ctype;
1140 static struct symbol *to_TI_mode(struct symbol *ctype)
1142 if (ctype->ctype.base_type != &int_type)
1143 return NULL;
1144 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulllong_ctype
1145 : &slllong_ctype;
1148 static struct symbol *to_word_mode(struct symbol *ctype)
1150 if (ctype->ctype.base_type != &int_type)
1151 return NULL;
1152 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulong_ctype
1153 : &slong_ctype;
1156 static struct token *attribute_mode(struct token *token, struct symbol *attr, struct decl_state *ctx)
1158 token = expect(token, '(', "after mode attribute");
1159 if (token_type(token) == TOKEN_IDENT) {
1160 struct symbol *mode = lookup_keyword(token->ident, NS_KEYWORD);
1161 if (mode && mode->op->type == KW_MODE)
1162 ctx->mode = mode->op;
1163 else
1164 sparse_error(token->pos, "unknown mode attribute %s\n", show_ident(token->ident));
1165 token = token->next;
1166 } else
1167 sparse_error(token->pos, "expect attribute mode symbol\n");
1168 token = expect(token, ')', "after mode attribute");
1169 return token;
1172 static struct token *attribute_context(struct token *token, struct symbol *attr, struct decl_state *ctx)
1174 struct context *context = alloc_context();
1175 struct expression *args[3];
1176 int argc = 0;
1178 token = expect(token, '(', "after context attribute");
1179 while (!match_op(token, ')')) {
1180 struct expression *expr = NULL;
1181 token = conditional_expression(token, &expr);
1182 if (!expr)
1183 break;
1184 if (argc < 3)
1185 args[argc++] = expr;
1186 if (!match_op(token, ','))
1187 break;
1188 token = token->next;
1191 switch(argc) {
1192 case 0:
1193 sparse_error(token->pos, "expected context input/output values");
1194 break;
1195 case 1:
1196 context->in = get_expression_value(args[0]);
1197 break;
1198 case 2:
1199 context->in = get_expression_value(args[0]);
1200 context->out = get_expression_value(args[1]);
1201 break;
1202 case 3:
1203 context->context = args[0];
1204 context->in = get_expression_value(args[1]);
1205 context->out = get_expression_value(args[2]);
1206 break;
1209 if (argc)
1210 add_ptr_list(&ctx->ctype.contexts, context);
1212 token = expect(token, ')', "after context attribute");
1213 return token;
1216 static struct token *attribute_designated_init(struct token *token, struct symbol *attr, struct decl_state *ctx)
1218 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_STRUCT)
1219 ctx->ctype.base_type->designated_init = 1;
1220 else
1221 warning(token->pos, "attribute designated_init applied to non-structure type");
1222 return token;
1225 static struct token *attribute_transparent_union(struct token *token, struct symbol *attr, struct decl_state *ctx)
1227 if (Wtransparent_union)
1228 warning(token->pos, "attribute __transparent_union__");
1230 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_UNION)
1231 ctx->ctype.base_type->transparent_union = 1;
1232 else
1233 warning(token->pos, "attribute __transparent_union__ applied to non-union type");
1234 return token;
1237 static struct token *recover_unknown_attribute(struct token *token)
1239 struct expression *expr = NULL;
1241 if (Wunknown_attribute)
1242 warning(token->pos, "attribute '%s': unknown attribute", show_ident(token->ident));
1243 token = token->next;
1244 if (match_op(token, '('))
1245 token = parens_expression(token, &expr, "in attribute");
1246 return token;
1249 static struct token *attribute_specifier(struct token *token, struct decl_state *ctx)
1251 token = expect(token, '(', "after attribute");
1252 token = expect(token, '(', "after attribute");
1254 for (;;) {
1255 struct ident *attribute_name;
1256 struct symbol *attr;
1258 if (eof_token(token))
1259 break;
1260 if (match_op(token, ';'))
1261 break;
1262 if (token_type(token) != TOKEN_IDENT)
1263 break;
1264 attribute_name = token->ident;
1265 attr = lookup_keyword(attribute_name, NS_KEYWORD);
1266 if (attr && attr->op->attribute)
1267 token = attr->op->attribute(token->next, attr, ctx);
1268 else
1269 token = recover_unknown_attribute(token);
1271 if (!match_op(token, ','))
1272 break;
1273 token = token->next;
1276 token = expect(token, ')', "after attribute");
1277 token = expect(token, ')', "after attribute");
1278 return token;
1281 static const char *storage_class[] =
1283 [STypedef] = "typedef",
1284 [SAuto] = "auto",
1285 [SExtern] = "extern",
1286 [SStatic] = "static",
1287 [SRegister] = "register",
1288 [SForced] = "[force]"
1291 static unsigned long storage_modifiers(struct decl_state *ctx)
1293 static unsigned long mod[SMax] =
1295 [SAuto] = MOD_AUTO,
1296 [SExtern] = MOD_EXTERN,
1297 [SStatic] = MOD_STATIC,
1298 [SRegister] = MOD_REGISTER
1300 return mod[ctx->storage_class] | (ctx->is_inline ? MOD_INLINE : 0)
1301 | (ctx->is_tls ? MOD_TLS : 0);
1304 static void set_storage_class(struct position *pos, struct decl_state *ctx, int class)
1306 /* __thread can be used alone, or with extern or static */
1307 if (ctx->is_tls && (class != SStatic && class != SExtern)) {
1308 sparse_error(*pos, "__thread can only be used alone, or with "
1309 "extern or static");
1310 return;
1313 if (!ctx->storage_class) {
1314 ctx->storage_class = class;
1315 return;
1317 if (ctx->storage_class == class)
1318 sparse_error(*pos, "duplicate %s", storage_class[class]);
1319 else
1320 sparse_error(*pos, "multiple storage classes");
1323 static struct token *typedef_specifier(struct token *next, struct decl_state *ctx)
1325 set_storage_class(&next->pos, ctx, STypedef);
1326 return next;
1329 static struct token *auto_specifier(struct token *next, struct decl_state *ctx)
1331 set_storage_class(&next->pos, ctx, SAuto);
1332 return next;
1335 static struct token *register_specifier(struct token *next, struct decl_state *ctx)
1337 set_storage_class(&next->pos, ctx, SRegister);
1338 return next;
1341 static struct token *static_specifier(struct token *next, struct decl_state *ctx)
1343 set_storage_class(&next->pos, ctx, SStatic);
1344 return next;
1347 static struct token *extern_specifier(struct token *next, struct decl_state *ctx)
1349 set_storage_class(&next->pos, ctx, SExtern);
1350 return next;
1353 static struct token *thread_specifier(struct token *next, struct decl_state *ctx)
1355 /* This GCC extension can be used alone, or with extern or static */
1356 if (!ctx->storage_class || ctx->storage_class == SStatic
1357 || ctx->storage_class == SExtern) {
1358 ctx->is_tls = 1;
1359 } else {
1360 sparse_error(next->pos, "__thread can only be used alone, or "
1361 "with extern or static");
1364 return next;
1367 static struct token *attribute_force(struct token *token, struct symbol *attr, struct decl_state *ctx)
1369 set_storage_class(&token->pos, ctx, SForced);
1370 return token;
1373 static struct token *inline_specifier(struct token *next, struct decl_state *ctx)
1375 ctx->is_inline = 1;
1376 return next;
1379 static struct token *const_qualifier(struct token *next, struct decl_state *ctx)
1381 apply_qualifier(&next->pos, &ctx->ctype, MOD_CONST);
1382 return next;
1385 static struct token *volatile_qualifier(struct token *next, struct decl_state *ctx)
1387 apply_qualifier(&next->pos, &ctx->ctype, MOD_VOLATILE);
1388 return next;
1391 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype)
1393 unsigned long mod = thistype->modifiers;
1395 if (mod)
1396 apply_qualifier(&pos, ctype, mod);
1398 /* Context */
1399 concat_ptr_list((struct ptr_list *)thistype->contexts,
1400 (struct ptr_list **)&ctype->contexts);
1402 /* Alignment */
1403 if (thistype->alignment > ctype->alignment)
1404 ctype->alignment = thistype->alignment;
1406 /* Address space */
1407 if (thistype->as)
1408 ctype->as = thistype->as;
1411 static void specifier_conflict(struct position pos, int what, struct ident *new)
1413 const char *old;
1414 if (what & (Set_S | Set_T))
1415 goto Catch_all;
1416 if (what & Set_Char)
1417 old = "char";
1418 else if (what & Set_Double)
1419 old = "double";
1420 else if (what & Set_Float)
1421 old = "float";
1422 else if (what & Set_Signed)
1423 old = "signed";
1424 else if (what & Set_Unsigned)
1425 old = "unsigned";
1426 else if (what & Set_Short)
1427 old = "short";
1428 else if (what & Set_Long)
1429 old = "long";
1430 else
1431 old = "long long";
1432 sparse_error(pos, "impossible combination of type specifiers: %s %s",
1433 old, show_ident(new));
1434 return;
1436 Catch_all:
1437 sparse_error(pos, "two or more data types in declaration specifiers");
1440 static struct symbol * const int_types[] =
1441 {&short_ctype, &int_ctype, &long_ctype, &llong_ctype, &lllong_ctype};
1442 static struct symbol * const signed_types[] =
1443 {&sshort_ctype, &sint_ctype, &slong_ctype, &sllong_ctype,
1444 &slllong_ctype};
1445 static struct symbol * const unsigned_types[] =
1446 {&ushort_ctype, &uint_ctype, &ulong_ctype, &ullong_ctype,
1447 &ulllong_ctype};
1448 static struct symbol * const real_types[] =
1449 {&float_ctype, &double_ctype, &ldouble_ctype};
1450 static struct symbol * const char_types[] =
1451 {&char_ctype, &schar_ctype, &uchar_ctype};
1452 static struct symbol * const * const types[] = {
1453 int_types + 1, signed_types + 1, unsigned_types + 1,
1454 real_types + 1, char_types, char_types + 1, char_types + 2
1457 struct symbol *ctype_integer(int size, int want_unsigned)
1459 return types[want_unsigned ? CUInt : CInt][size];
1462 static struct token *handle_qualifiers(struct token *t, struct decl_state *ctx)
1464 while (token_type(t) == TOKEN_IDENT) {
1465 struct symbol *s = lookup_symbol(t->ident, NS_TYPEDEF);
1466 if (!s)
1467 break;
1468 if (s->type != SYM_KEYWORD)
1469 break;
1470 if (!(s->op->type & (KW_ATTRIBUTE | KW_QUALIFIER)))
1471 break;
1472 t = t->next;
1473 if (s->op->declarator)
1474 t = s->op->declarator(t, ctx);
1476 return t;
1479 static struct token *declaration_specifiers(struct token *token, struct decl_state *ctx)
1481 int seen = 0;
1482 int class = CInt;
1483 int size = 0;
1485 while (token_type(token) == TOKEN_IDENT) {
1486 struct symbol *s = lookup_symbol(token->ident,
1487 NS_TYPEDEF | NS_SYMBOL);
1488 if (!s || !(s->namespace & NS_TYPEDEF))
1489 break;
1490 if (s->type != SYM_KEYWORD) {
1491 if (seen & Set_Any)
1492 break;
1493 seen |= Set_S | Set_T;
1494 ctx->ctype.base_type = s->ctype.base_type;
1495 apply_ctype(token->pos, &s->ctype, &ctx->ctype);
1496 token = token->next;
1497 continue;
1499 if (s->op->type & KW_SPECIFIER) {
1500 if (seen & s->op->test) {
1501 specifier_conflict(token->pos,
1502 seen & s->op->test,
1503 token->ident);
1504 break;
1506 seen |= s->op->set;
1507 class += s->op->class;
1508 if (s->op->set & Set_Int128)
1509 size = 2;
1510 if (s->op->type & KW_SHORT) {
1511 size = -1;
1512 } else if (s->op->type & KW_LONG && size++) {
1513 if (class == CReal) {
1514 specifier_conflict(token->pos,
1515 Set_Vlong,
1516 &double_ident);
1517 break;
1519 seen |= Set_Vlong;
1522 token = token->next;
1523 if (s->op->declarator)
1524 token = s->op->declarator(token, ctx);
1525 if (s->op->type & KW_EXACT) {
1526 ctx->ctype.base_type = s->ctype.base_type;
1527 ctx->ctype.modifiers |= s->ctype.modifiers;
1531 if (!(seen & Set_S)) { /* not set explicitly? */
1532 struct symbol *base = &incomplete_ctype;
1533 if (seen & Set_Any)
1534 base = types[class][size];
1535 ctx->ctype.base_type = base;
1538 if (ctx->ctype.modifiers & MOD_BITWISE) {
1539 struct symbol *type;
1540 ctx->ctype.modifiers &= ~MOD_BITWISE;
1541 if (!is_int_type(ctx->ctype.base_type)) {
1542 sparse_error(token->pos, "invalid modifier");
1543 return token;
1545 type = alloc_symbol(token->pos, SYM_BASETYPE);
1546 *type = *ctx->ctype.base_type;
1547 type->ctype.modifiers &= ~MOD_SPECIFIER;
1548 type->ctype.base_type = ctx->ctype.base_type;
1549 type->type = SYM_RESTRICT;
1550 ctx->ctype.base_type = type;
1551 create_fouled(type);
1553 return token;
1556 static struct token *abstract_array_static_declarator(struct token *token, int *has_static)
1558 while (token->ident == &static_ident) {
1559 if (*has_static)
1560 sparse_error(token->pos, "duplicate array static declarator");
1562 *has_static = 1;
1563 token = token->next;
1565 return token;
1569 static struct token *abstract_array_declarator(struct token *token, struct symbol *sym)
1571 struct expression *expr = NULL;
1572 int has_static = 0;
1574 token = abstract_array_static_declarator(token, &has_static);
1576 if (match_idents(token, &restrict_ident, &__restrict_ident, &__restrict___ident, NULL))
1577 token = abstract_array_static_declarator(token->next, &has_static);
1578 token = parse_expression(token, &expr);
1579 sym->array_size = expr;
1580 return token;
1583 static struct token *parameter_type_list(struct token *, struct symbol *);
1584 static struct token *identifier_list(struct token *, struct symbol *);
1585 static struct token *declarator(struct token *token, struct decl_state *ctx);
1587 static struct token *skip_attribute(struct token *token)
1589 token = token->next;
1590 if (match_op(token, '(')) {
1591 int depth = 1;
1592 token = token->next;
1593 while (depth && !eof_token(token)) {
1594 if (token_type(token) == TOKEN_SPECIAL) {
1595 if (token->special == '(')
1596 depth++;
1597 else if (token->special == ')')
1598 depth--;
1600 token = token->next;
1603 return token;
1606 static struct token *skip_attributes(struct token *token)
1608 struct symbol *keyword;
1609 for (;;) {
1610 if (token_type(token) != TOKEN_IDENT)
1611 break;
1612 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1613 if (!keyword || keyword->type != SYM_KEYWORD)
1614 break;
1615 if (!(keyword->op->type & KW_ATTRIBUTE))
1616 break;
1617 token = expect(token->next, '(', "after attribute");
1618 token = expect(token, '(', "after attribute");
1619 for (;;) {
1620 if (eof_token(token))
1621 break;
1622 if (match_op(token, ';'))
1623 break;
1624 if (token_type(token) != TOKEN_IDENT)
1625 break;
1626 token = skip_attribute(token);
1627 if (!match_op(token, ','))
1628 break;
1629 token = token->next;
1631 token = expect(token, ')', "after attribute");
1632 token = expect(token, ')', "after attribute");
1634 return token;
1637 static struct token *handle_attributes(struct token *token, struct decl_state *ctx, unsigned int keywords)
1639 struct symbol *keyword;
1640 for (;;) {
1641 if (token_type(token) != TOKEN_IDENT)
1642 break;
1643 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1644 if (!keyword || keyword->type != SYM_KEYWORD)
1645 break;
1646 if (!(keyword->op->type & keywords))
1647 break;
1648 token = keyword->op->declarator(token->next, ctx);
1649 keywords &= KW_ATTRIBUTE;
1651 return token;
1654 static int is_nested(struct token *token, struct token **p,
1655 int prefer_abstract)
1658 * This can be either a parameter list or a grouping.
1659 * For the direct (non-abstract) case, we know if must be
1660 * a parameter list if we already saw the identifier.
1661 * For the abstract case, we know if must be a parameter
1662 * list if it is empty or starts with a type.
1664 struct token *next = token->next;
1666 *p = next = skip_attributes(next);
1668 if (token_type(next) == TOKEN_IDENT) {
1669 if (lookup_type(next))
1670 return !prefer_abstract;
1671 return 1;
1674 if (match_op(next, ')') || match_op(next, SPECIAL_ELLIPSIS))
1675 return 0;
1677 return 1;
1680 enum kind {
1681 Empty, K_R, Proto, Bad_Func,
1684 static enum kind which_func(struct token *token,
1685 struct ident **n,
1686 int prefer_abstract)
1688 struct token *next = token->next;
1690 if (token_type(next) == TOKEN_IDENT) {
1691 if (lookup_type(next))
1692 return Proto;
1693 /* identifier list not in definition; complain */
1694 if (prefer_abstract)
1695 warning(token->pos,
1696 "identifier list not in definition");
1697 return K_R;
1700 if (token_type(next) != TOKEN_SPECIAL)
1701 return Bad_Func;
1703 if (next->special == ')') {
1704 /* don't complain about those */
1705 if (!n || match_op(next->next, ';'))
1706 return Empty;
1707 warning(next->pos,
1708 "non-ANSI function declaration of function '%s'",
1709 show_ident(*n));
1710 return Empty;
1713 if (next->special == SPECIAL_ELLIPSIS) {
1714 warning(next->pos,
1715 "variadic functions must have one named argument");
1716 return Proto;
1719 return Bad_Func;
1722 static struct token *direct_declarator(struct token *token, struct decl_state *ctx)
1724 struct ctype *ctype = &ctx->ctype;
1725 struct token *next;
1726 struct ident **p = ctx->ident;
1728 if (ctx->ident && token_type(token) == TOKEN_IDENT) {
1729 *ctx->ident = token->ident;
1730 token = token->next;
1731 } else if (match_op(token, '(') &&
1732 is_nested(token, &next, ctx->prefer_abstract)) {
1733 struct symbol *base_type = ctype->base_type;
1734 if (token->next != next)
1735 next = handle_attributes(token->next, ctx,
1736 KW_ATTRIBUTE);
1737 token = declarator(next, ctx);
1738 token = expect(token, ')', "in nested declarator");
1739 while (ctype->base_type != base_type)
1740 ctype = &ctype->base_type->ctype;
1741 p = NULL;
1744 if (match_op(token, '(')) {
1745 enum kind kind = which_func(token, p, ctx->prefer_abstract);
1746 struct symbol *fn;
1747 fn = alloc_indirect_symbol(token->pos, ctype, SYM_FN);
1748 token = token->next;
1749 if (kind == K_R)
1750 token = identifier_list(token, fn);
1751 else if (kind == Proto)
1752 token = parameter_type_list(token, fn);
1753 token = expect(token, ')', "in function declarator");
1754 fn->endpos = token->pos;
1755 return token;
1758 while (match_op(token, '[')) {
1759 struct symbol *array;
1760 array = alloc_indirect_symbol(token->pos, ctype, SYM_ARRAY);
1761 token = abstract_array_declarator(token->next, array);
1762 token = expect(token, ']', "in abstract_array_declarator");
1763 array->endpos = token->pos;
1764 ctype = &array->ctype;
1766 return token;
1769 static struct token *pointer(struct token *token, struct decl_state *ctx)
1771 while (match_op(token,'*')) {
1772 struct symbol *ptr = alloc_symbol(token->pos, SYM_PTR);
1773 ptr->ctype.modifiers = ctx->ctype.modifiers;
1774 ptr->ctype.base_type = ctx->ctype.base_type;
1775 ptr->ctype.as = ctx->ctype.as;
1776 ptr->ctype.contexts = ctx->ctype.contexts;
1777 ctx->ctype.modifiers = 0;
1778 ctx->ctype.base_type = ptr;
1779 ctx->ctype.as = 0;
1780 ctx->ctype.contexts = NULL;
1781 ctx->ctype.alignment = 0;
1783 token = handle_qualifiers(token->next, ctx);
1784 ctx->ctype.base_type->endpos = token->pos;
1786 return token;
1789 static struct token *declarator(struct token *token, struct decl_state *ctx)
1791 token = pointer(token, ctx);
1792 return direct_declarator(token, ctx);
1795 static struct token *handle_bitfield(struct token *token, struct decl_state *ctx)
1797 struct ctype *ctype = &ctx->ctype;
1798 struct expression *expr;
1799 struct symbol *bitfield;
1800 long long width;
1802 if (ctype->base_type != &int_type && !is_int_type(ctype->base_type)) {
1803 sparse_error(token->pos, "invalid bitfield specifier for type %s.",
1804 show_typename(ctype->base_type));
1805 // Parse this to recover gracefully.
1806 return conditional_expression(token->next, &expr);
1809 bitfield = alloc_indirect_symbol(token->pos, ctype, SYM_BITFIELD);
1810 token = conditional_expression(token->next, &expr);
1811 width = const_expression_value(expr);
1812 bitfield->bit_size = width;
1814 if (width < 0 || width > INT_MAX) {
1815 sparse_error(token->pos, "invalid bitfield width, %lld.", width);
1816 width = -1;
1817 } else if (*ctx->ident && width == 0) {
1818 sparse_error(token->pos, "invalid named zero-width bitfield `%s'",
1819 show_ident(*ctx->ident));
1820 width = -1;
1821 } else if (*ctx->ident) {
1822 struct symbol *base_type = bitfield->ctype.base_type;
1823 struct symbol *bitfield_type = base_type == &int_type ? bitfield : base_type;
1824 int is_signed = !(bitfield_type->ctype.modifiers & MOD_UNSIGNED);
1825 if (Wone_bit_signed_bitfield && width == 1 && is_signed) {
1826 // Valid values are either {-1;0} or {0}, depending on integer
1827 // representation. The latter makes for very efficient code...
1828 sparse_error(token->pos, "dubious one-bit signed bitfield");
1830 if (Wdefault_bitfield_sign &&
1831 bitfield_type->type != SYM_ENUM &&
1832 !(bitfield_type->ctype.modifiers & MOD_EXPLICITLY_SIGNED) &&
1833 is_signed) {
1834 // The sign of bitfields is unspecified by default.
1835 warning(token->pos, "dubious bitfield without explicit `signed' or `unsigned'");
1838 bitfield->bit_size = width;
1839 bitfield->endpos = token->pos;
1840 return token;
1843 static struct token *declaration_list(struct token *token, struct symbol_list **list)
1845 struct decl_state ctx = {.prefer_abstract = 0};
1846 struct ctype saved;
1847 unsigned long mod;
1849 token = declaration_specifiers(token, &ctx);
1850 mod = storage_modifiers(&ctx);
1851 saved = ctx.ctype;
1852 for (;;) {
1853 struct symbol *decl = alloc_symbol(token->pos, SYM_NODE);
1854 ctx.ident = &decl->ident;
1856 token = declarator(token, &ctx);
1857 if (match_op(token, ':'))
1858 token = handle_bitfield(token, &ctx);
1860 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
1861 apply_modifiers(token->pos, &ctx);
1863 decl->ctype = ctx.ctype;
1864 decl->ctype.modifiers |= mod;
1865 decl->endpos = token->pos;
1866 add_symbol(list, decl);
1867 if (!match_op(token, ','))
1868 break;
1869 token = token->next;
1870 ctx.ctype = saved;
1872 return token;
1875 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list)
1877 while (!match_op(token, '}')) {
1878 if (!match_op(token, ';'))
1879 token = declaration_list(token, list);
1880 if (!match_op(token, ';')) {
1881 sparse_error(token->pos, "expected ; at end of declaration");
1882 break;
1884 token = token->next;
1886 return token;
1889 static struct token *parameter_declaration(struct token *token, struct symbol *sym)
1891 struct decl_state ctx = {.prefer_abstract = 1};
1893 token = declaration_specifiers(token, &ctx);
1894 ctx.ident = &sym->ident;
1895 token = declarator(token, &ctx);
1896 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
1897 apply_modifiers(token->pos, &ctx);
1898 sym->ctype = ctx.ctype;
1899 sym->ctype.modifiers |= storage_modifiers(&ctx);
1900 sym->endpos = token->pos;
1901 sym->forced_arg = ctx.storage_class == SForced;
1902 return token;
1905 struct token *typename(struct token *token, struct symbol **p, int *forced)
1907 struct decl_state ctx = {.prefer_abstract = 1};
1908 int class;
1909 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
1910 *p = sym;
1911 token = declaration_specifiers(token, &ctx);
1912 token = declarator(token, &ctx);
1913 apply_modifiers(token->pos, &ctx);
1914 sym->ctype = ctx.ctype;
1915 sym->endpos = token->pos;
1916 class = ctx.storage_class;
1917 if (forced) {
1918 *forced = 0;
1919 if (class == SForced) {
1920 *forced = 1;
1921 class = 0;
1924 if (class)
1925 warning(sym->pos, "storage class in typename (%s %s)",
1926 storage_class[class], show_typename(sym));
1927 return token;
1930 static struct token *expression_statement(struct token *token, struct expression **tree)
1932 token = parse_expression(token, tree);
1933 return expect(token, ';', "at end of statement");
1936 static struct token *parse_asm_operands(struct token *token, struct statement *stmt,
1937 struct expression_list **inout)
1939 struct expression *expr;
1941 /* Allow empty operands */
1942 if (match_op(token->next, ':') || match_op(token->next, ')'))
1943 return token->next;
1944 do {
1945 struct ident *ident = NULL;
1946 if (match_op(token->next, '[') &&
1947 token_type(token->next->next) == TOKEN_IDENT &&
1948 match_op(token->next->next->next, ']')) {
1949 ident = token->next->next->ident;
1950 token = token->next->next->next;
1952 add_expression(inout, (struct expression *)ident); /* UGGLEE!!! */
1953 token = primary_expression(token->next, &expr);
1954 add_expression(inout, expr);
1955 token = parens_expression(token, &expr, "in asm parameter");
1956 add_expression(inout, expr);
1957 } while (match_op(token, ','));
1958 return token;
1961 static struct token *parse_asm_clobbers(struct token *token, struct statement *stmt,
1962 struct expression_list **clobbers)
1964 struct expression *expr;
1966 do {
1967 token = primary_expression(token->next, &expr);
1968 if (expr)
1969 add_expression(clobbers, expr);
1970 } while (match_op(token, ','));
1971 return token;
1974 static struct token *parse_asm_labels(struct token *token, struct statement *stmt,
1975 struct symbol_list **labels)
1977 struct symbol *label;
1979 do {
1980 token = token->next; /* skip ':' and ',' */
1981 if (token_type(token) != TOKEN_IDENT)
1982 return token;
1983 label = label_symbol(token);
1984 add_symbol(labels, label);
1985 token = token->next;
1986 } while (match_op(token, ','));
1987 return token;
1990 static struct token *parse_asm_statement(struct token *token, struct statement *stmt)
1992 int is_goto = 0;
1994 token = token->next;
1995 stmt->type = STMT_ASM;
1996 if (match_idents(token, &__volatile___ident, &__volatile_ident, &volatile_ident, NULL)) {
1997 token = token->next;
1999 if (token_type(token) == TOKEN_IDENT && token->ident == &goto_ident) {
2000 is_goto = 1;
2001 token = token->next;
2003 token = expect(token, '(', "after asm");
2004 token = parse_expression(token, &stmt->asm_string);
2005 if (match_op(token, ':'))
2006 token = parse_asm_operands(token, stmt, &stmt->asm_outputs);
2007 if (match_op(token, ':'))
2008 token = parse_asm_operands(token, stmt, &stmt->asm_inputs);
2009 if (match_op(token, ':'))
2010 token = parse_asm_clobbers(token, stmt, &stmt->asm_clobbers);
2011 if (is_goto && match_op(token, ':'))
2012 token = parse_asm_labels(token, stmt, &stmt->asm_labels);
2013 token = expect(token, ')', "after asm");
2014 return expect(token, ';', "at end of asm-statement");
2017 static struct token *parse_asm_declarator(struct token *token, struct decl_state *ctx)
2019 struct expression *expr;
2020 token = expect(token, '(', "after asm");
2021 token = parse_expression(token->next, &expr);
2022 token = expect(token, ')', "after asm");
2023 return token;
2026 /* Make a statement out of an expression */
2027 static struct statement *make_statement(struct expression *expr)
2029 struct statement *stmt;
2031 if (!expr)
2032 return NULL;
2033 stmt = alloc_statement(expr->pos, STMT_EXPRESSION);
2034 stmt->expression = expr;
2035 return stmt;
2039 * All iterators have two symbols associated with them:
2040 * the "continue" and "break" symbols, which are targets
2041 * for continue and break statements respectively.
2043 * They are in a special name-space, but they follow
2044 * all the normal visibility rules, so nested iterators
2045 * automatically work right.
2047 static void start_iterator(struct statement *stmt)
2049 struct symbol *cont, *brk;
2051 start_symbol_scope();
2052 cont = alloc_symbol(stmt->pos, SYM_NODE);
2053 bind_symbol(cont, &continue_ident, NS_ITERATOR);
2054 brk = alloc_symbol(stmt->pos, SYM_NODE);
2055 bind_symbol(brk, &break_ident, NS_ITERATOR);
2057 stmt->type = STMT_ITERATOR;
2058 stmt->iterator_break = brk;
2059 stmt->iterator_continue = cont;
2060 fn_local_symbol(brk);
2061 fn_local_symbol(cont);
2064 static void end_iterator(struct statement *stmt)
2066 end_symbol_scope();
2069 static struct statement *start_function(struct symbol *sym)
2071 struct symbol *ret;
2072 struct statement *stmt = alloc_statement(sym->pos, STMT_COMPOUND);
2074 start_function_scope();
2075 ret = alloc_symbol(sym->pos, SYM_NODE);
2076 ret->ctype = sym->ctype.base_type->ctype;
2077 ret->ctype.modifiers &= ~(MOD_STORAGE | MOD_CONST | MOD_VOLATILE | MOD_TLS | MOD_INLINE | MOD_ADDRESSABLE | MOD_NOCAST | MOD_NODEREF | MOD_ACCESSED | MOD_TOPLEVEL);
2078 ret->ctype.modifiers |= (MOD_AUTO | MOD_REGISTER);
2079 bind_symbol(ret, &return_ident, NS_ITERATOR);
2080 stmt->ret = ret;
2081 fn_local_symbol(ret);
2083 // Currently parsed symbol for __func__/__FUNCTION__/__PRETTY_FUNCTION__
2084 current_fn = sym;
2086 return stmt;
2089 static void end_function(struct symbol *sym)
2091 current_fn = NULL;
2092 end_function_scope();
2096 * A "switch()" statement, like an iterator, has a
2097 * the "break" symbol associated with it. It works
2098 * exactly like the iterator break - it's the target
2099 * for any break-statements in scope, and means that
2100 * "break" handling doesn't even need to know whether
2101 * it's breaking out of an iterator or a switch.
2103 * In addition, the "case" symbol is a marker for the
2104 * case/default statements to find the switch statement
2105 * that they are associated with.
2107 static void start_switch(struct statement *stmt)
2109 struct symbol *brk, *switch_case;
2111 start_symbol_scope();
2112 brk = alloc_symbol(stmt->pos, SYM_NODE);
2113 bind_symbol(brk, &break_ident, NS_ITERATOR);
2115 switch_case = alloc_symbol(stmt->pos, SYM_NODE);
2116 bind_symbol(switch_case, &case_ident, NS_ITERATOR);
2117 switch_case->stmt = stmt;
2119 stmt->type = STMT_SWITCH;
2120 stmt->switch_break = brk;
2121 stmt->switch_case = switch_case;
2123 fn_local_symbol(brk);
2124 fn_local_symbol(switch_case);
2127 static void end_switch(struct statement *stmt)
2129 if (!stmt->switch_case->symbol_list)
2130 warning(stmt->pos, "switch with no cases");
2131 end_symbol_scope();
2134 static void add_case_statement(struct statement *stmt)
2136 struct symbol *target = lookup_symbol(&case_ident, NS_ITERATOR);
2137 struct symbol *sym;
2139 if (!target) {
2140 sparse_error(stmt->pos, "not in switch scope");
2141 stmt->type = STMT_NONE;
2142 return;
2144 sym = alloc_symbol(stmt->pos, SYM_NODE);
2145 add_symbol(&target->symbol_list, sym);
2146 sym->stmt = stmt;
2147 stmt->case_label = sym;
2148 fn_local_symbol(sym);
2151 static struct token *parse_return_statement(struct token *token, struct statement *stmt)
2153 struct symbol *target = lookup_symbol(&return_ident, NS_ITERATOR);
2155 if (!target)
2156 error_die(token->pos, "internal error: return without a function target");
2157 stmt->type = STMT_RETURN;
2158 stmt->ret_target = target;
2159 return expression_statement(token->next, &stmt->ret_value);
2162 static struct token *parse_for_statement(struct token *token, struct statement *stmt)
2164 struct symbol_list *syms;
2165 struct expression *e1, *e2, *e3;
2166 struct statement *iterator;
2168 start_iterator(stmt);
2169 token = expect(token->next, '(', "after 'for'");
2171 syms = NULL;
2172 e1 = NULL;
2173 /* C99 variable declaration? */
2174 if (lookup_type(token)) {
2175 token = external_declaration(token, &syms);
2176 } else {
2177 token = parse_expression(token, &e1);
2178 token = expect(token, ';', "in 'for'");
2180 token = parse_expression(token, &e2);
2181 token = expect(token, ';', "in 'for'");
2182 token = parse_expression(token, &e3);
2183 token = expect(token, ')', "in 'for'");
2184 token = statement(token, &iterator);
2186 stmt->iterator_syms = syms;
2187 stmt->iterator_pre_statement = make_statement(e1);
2188 stmt->iterator_pre_condition = e2;
2189 stmt->iterator_post_statement = make_statement(e3);
2190 stmt->iterator_post_condition = NULL;
2191 stmt->iterator_statement = iterator;
2192 end_iterator(stmt);
2194 return token;
2197 static struct token *parse_while_statement(struct token *token, struct statement *stmt)
2199 struct expression *expr;
2200 struct statement *iterator;
2202 start_iterator(stmt);
2203 token = parens_expression(token->next, &expr, "after 'while'");
2204 token = statement(token, &iterator);
2206 stmt->iterator_pre_condition = expr;
2207 stmt->iterator_post_condition = NULL;
2208 stmt->iterator_statement = iterator;
2209 end_iterator(stmt);
2211 return token;
2214 static struct token *parse_do_statement(struct token *token, struct statement *stmt)
2216 struct expression *expr;
2217 struct statement *iterator;
2219 start_iterator(stmt);
2220 token = statement(token->next, &iterator);
2221 if (token_type(token) == TOKEN_IDENT && token->ident == &while_ident)
2222 token = token->next;
2223 else
2224 sparse_error(token->pos, "expected 'while' after 'do'");
2225 token = parens_expression(token, &expr, "after 'do-while'");
2227 stmt->iterator_post_condition = expr;
2228 stmt->iterator_statement = iterator;
2229 end_iterator(stmt);
2231 if (iterator && iterator->type != STMT_COMPOUND && Wdo_while)
2232 warning(iterator->pos, "do-while statement is not a compound statement");
2234 return expect(token, ';', "after statement");
2237 static struct token *parse_if_statement(struct token *token, struct statement *stmt)
2239 stmt->type = STMT_IF;
2240 token = parens_expression(token->next, &stmt->if_conditional, "after if");
2241 token = statement(token, &stmt->if_true);
2242 if (token_type(token) != TOKEN_IDENT)
2243 return token;
2244 if (token->ident != &else_ident)
2245 return token;
2246 return statement(token->next, &stmt->if_false);
2249 static inline struct token *case_statement(struct token *token, struct statement *stmt)
2251 stmt->type = STMT_CASE;
2252 token = expect(token, ':', "after default/case");
2253 add_case_statement(stmt);
2254 return statement(token, &stmt->case_statement);
2257 static struct token *parse_case_statement(struct token *token, struct statement *stmt)
2259 token = parse_expression(token->next, &stmt->case_expression);
2260 if (match_op(token, SPECIAL_ELLIPSIS))
2261 token = parse_expression(token->next, &stmt->case_to);
2262 return case_statement(token, stmt);
2265 static struct token *parse_default_statement(struct token *token, struct statement *stmt)
2267 return case_statement(token->next, stmt);
2270 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt)
2272 struct symbol *target = lookup_symbol(token->ident, NS_ITERATOR);
2273 stmt->type = STMT_GOTO;
2274 stmt->goto_label = target;
2275 if (!target)
2276 sparse_error(stmt->pos, "break/continue not in iterator scope");
2277 return expect(token->next, ';', "at end of statement");
2280 static struct token *parse_switch_statement(struct token *token, struct statement *stmt)
2282 stmt->type = STMT_SWITCH;
2283 start_switch(stmt);
2284 token = parens_expression(token->next, &stmt->switch_expression, "after 'switch'");
2285 token = statement(token, &stmt->switch_statement);
2286 end_switch(stmt);
2287 return token;
2290 static struct token *parse_goto_statement(struct token *token, struct statement *stmt)
2292 stmt->type = STMT_GOTO;
2293 token = token->next;
2294 if (match_op(token, '*')) {
2295 token = parse_expression(token->next, &stmt->goto_expression);
2296 add_statement(&function_computed_goto_list, stmt);
2297 } else if (token_type(token) == TOKEN_IDENT) {
2298 stmt->goto_label = label_symbol(token);
2299 token = token->next;
2300 } else {
2301 sparse_error(token->pos, "Expected identifier or goto expression");
2303 return expect(token, ';', "at end of statement");
2306 static struct token *parse_context_statement(struct token *token, struct statement *stmt)
2308 stmt->type = STMT_CONTEXT;
2309 token = parse_expression(token->next, &stmt->expression);
2310 if (stmt->expression->type == EXPR_PREOP
2311 && stmt->expression->op == '('
2312 && stmt->expression->unop->type == EXPR_COMMA) {
2313 struct expression *expr;
2314 expr = stmt->expression->unop;
2315 stmt->context = expr->left;
2316 stmt->expression = expr->right;
2318 return expect(token, ';', "at end of statement");
2321 static struct token *parse_range_statement(struct token *token, struct statement *stmt)
2323 stmt->type = STMT_RANGE;
2324 token = assignment_expression(token->next, &stmt->range_expression);
2325 token = expect(token, ',', "after range expression");
2326 token = assignment_expression(token, &stmt->range_low);
2327 token = expect(token, ',', "after low range");
2328 token = assignment_expression(token, &stmt->range_high);
2329 return expect(token, ';', "after range statement");
2332 static struct token *statement(struct token *token, struct statement **tree)
2334 struct statement *stmt = alloc_statement(token->pos, STMT_NONE);
2336 *tree = stmt;
2337 if (token_type(token) == TOKEN_IDENT) {
2338 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2339 if (s && s->op->statement)
2340 return s->op->statement(token, stmt);
2342 if (match_op(token->next, ':')) {
2343 struct symbol *s = label_symbol(token);
2344 stmt->type = STMT_LABEL;
2345 stmt->label_identifier = s;
2346 if (s->stmt)
2347 sparse_error(stmt->pos, "label '%s' redefined", show_ident(token->ident));
2348 s->stmt = stmt;
2349 token = skip_attributes(token->next->next);
2350 return statement(token, &stmt->label_statement);
2354 if (match_op(token, '{')) {
2355 stmt->type = STMT_COMPOUND;
2356 start_symbol_scope();
2357 token = compound_statement(token->next, stmt);
2358 end_symbol_scope();
2360 return expect(token, '}', "at end of compound statement");
2363 stmt->type = STMT_EXPRESSION;
2364 return expression_statement(token, &stmt->expression);
2367 /* gcc extension - __label__ ident-list; in the beginning of compound stmt */
2368 static struct token *label_statement(struct token *token)
2370 while (token_type(token) == TOKEN_IDENT) {
2371 struct symbol *sym = alloc_symbol(token->pos, SYM_LABEL);
2372 /* it's block-scope, but we want label namespace */
2373 bind_symbol(sym, token->ident, NS_SYMBOL);
2374 sym->namespace = NS_LABEL;
2375 fn_local_symbol(sym);
2376 token = token->next;
2377 if (!match_op(token, ','))
2378 break;
2379 token = token->next;
2381 return expect(token, ';', "at end of label declaration");
2384 static struct token * statement_list(struct token *token, struct statement_list **list)
2386 int seen_statement = 0;
2387 while (token_type(token) == TOKEN_IDENT &&
2388 token->ident == &__label___ident)
2389 token = label_statement(token->next);
2390 for (;;) {
2391 struct statement * stmt;
2392 if (eof_token(token))
2393 break;
2394 if (match_op(token, '}'))
2395 break;
2396 if (lookup_type(token)) {
2397 if (seen_statement) {
2398 warning(token->pos, "mixing declarations and code");
2399 seen_statement = 0;
2401 stmt = alloc_statement(token->pos, STMT_DECLARATION);
2402 token = external_declaration(token, &stmt->declaration);
2403 } else {
2404 seen_statement = Wdeclarationafterstatement;
2405 token = statement(token, &stmt);
2407 add_statement(list, stmt);
2409 return token;
2412 static struct token *identifier_list(struct token *token, struct symbol *fn)
2414 struct symbol_list **list = &fn->arguments;
2415 for (;;) {
2416 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2417 sym->ident = token->ident;
2418 token = token->next;
2419 sym->endpos = token->pos;
2420 sym->ctype.base_type = &incomplete_ctype;
2421 add_symbol(list, sym);
2422 if (!match_op(token, ',') ||
2423 token_type(token->next) != TOKEN_IDENT ||
2424 lookup_type(token->next))
2425 break;
2426 token = token->next;
2428 return token;
2431 static struct token *parameter_type_list(struct token *token, struct symbol *fn)
2433 struct symbol_list **list = &fn->arguments;
2435 for (;;) {
2436 struct symbol *sym;
2438 if (match_op(token, SPECIAL_ELLIPSIS)) {
2439 fn->variadic = 1;
2440 token = token->next;
2441 break;
2444 sym = alloc_symbol(token->pos, SYM_NODE);
2445 token = parameter_declaration(token, sym);
2446 if (sym->ctype.base_type == &void_ctype) {
2447 /* Special case: (void) */
2448 if (!*list && !sym->ident)
2449 break;
2450 warning(token->pos, "void parameter");
2452 add_symbol(list, sym);
2453 if (!match_op(token, ','))
2454 break;
2455 token = token->next;
2457 return token;
2460 struct token *compound_statement(struct token *token, struct statement *stmt)
2462 token = statement_list(token, &stmt->stmts);
2463 return token;
2466 static struct expression *identifier_expression(struct token *token)
2468 struct expression *expr = alloc_expression(token->pos, EXPR_IDENTIFIER);
2469 expr->expr_ident = token->ident;
2470 return expr;
2473 static struct expression *index_expression(struct expression *from, struct expression *to)
2475 int idx_from, idx_to;
2476 struct expression *expr = alloc_expression(from->pos, EXPR_INDEX);
2478 idx_from = const_expression_value(from);
2479 idx_to = idx_from;
2480 if (to) {
2481 idx_to = const_expression_value(to);
2482 if (idx_to < idx_from || idx_from < 0)
2483 warning(from->pos, "nonsense array initializer index range");
2485 expr->idx_from = idx_from;
2486 expr->idx_to = idx_to;
2487 return expr;
2490 static struct token *single_initializer(struct expression **ep, struct token *token)
2492 int expect_equal = 0;
2493 struct token *next = token->next;
2494 struct expression **tail = ep;
2495 int nested;
2497 *ep = NULL;
2499 if ((token_type(token) == TOKEN_IDENT) && match_op(next, ':')) {
2500 struct expression *expr = identifier_expression(token);
2501 if (Wold_initializer)
2502 warning(token->pos, "obsolete struct initializer, use C99 syntax");
2503 token = initializer(&expr->ident_expression, next->next);
2504 if (expr->ident_expression)
2505 *ep = expr;
2506 return token;
2509 for (tail = ep, nested = 0; ; nested++, next = token->next) {
2510 if (match_op(token, '.') && (token_type(next) == TOKEN_IDENT)) {
2511 struct expression *expr = identifier_expression(next);
2512 *tail = expr;
2513 tail = &expr->ident_expression;
2514 expect_equal = 1;
2515 token = next->next;
2516 } else if (match_op(token, '[')) {
2517 struct expression *from = NULL, *to = NULL, *expr;
2518 token = constant_expression(token->next, &from);
2519 if (!from) {
2520 sparse_error(token->pos, "Expected constant expression");
2521 break;
2523 if (match_op(token, SPECIAL_ELLIPSIS))
2524 token = constant_expression(token->next, &to);
2525 expr = index_expression(from, to);
2526 *tail = expr;
2527 tail = &expr->idx_expression;
2528 token = expect(token, ']', "at end of initializer index");
2529 if (nested)
2530 expect_equal = 1;
2531 } else {
2532 break;
2535 if (nested && !expect_equal) {
2536 if (!match_op(token, '='))
2537 warning(token->pos, "obsolete array initializer, use C99 syntax");
2538 else
2539 expect_equal = 1;
2541 if (expect_equal)
2542 token = expect(token, '=', "at end of initializer index");
2544 token = initializer(tail, token);
2545 if (!*tail)
2546 *ep = NULL;
2547 return token;
2550 static struct token *initializer_list(struct expression_list **list, struct token *token)
2552 struct expression *expr;
2554 for (;;) {
2555 token = single_initializer(&expr, token);
2556 if (!expr)
2557 break;
2558 add_expression(list, expr);
2559 if (!match_op(token, ','))
2560 break;
2561 token = token->next;
2563 return token;
2566 struct token *initializer(struct expression **tree, struct token *token)
2568 if (match_op(token, '{')) {
2569 struct expression *expr = alloc_expression(token->pos, EXPR_INITIALIZER);
2570 *tree = expr;
2571 token = initializer_list(&expr->expr_list, token->next);
2572 return expect(token, '}', "at end of initializer");
2574 return assignment_expression(token, tree);
2577 static void declare_argument(struct symbol *sym, struct symbol *fn)
2579 if (!sym->ident) {
2580 sparse_error(sym->pos, "no identifier for function argument");
2581 return;
2583 bind_symbol(sym, sym->ident, NS_SYMBOL);
2586 static struct token *parse_function_body(struct token *token, struct symbol *decl,
2587 struct symbol_list **list)
2589 struct symbol_list **old_symbol_list;
2590 struct symbol *base_type = decl->ctype.base_type;
2591 struct statement *stmt, **p;
2592 struct symbol *prev;
2593 struct symbol *arg;
2595 old_symbol_list = function_symbol_list;
2596 if (decl->ctype.modifiers & MOD_INLINE) {
2597 function_symbol_list = &decl->inline_symbol_list;
2598 p = &base_type->inline_stmt;
2599 } else {
2600 function_symbol_list = &decl->symbol_list;
2601 p = &base_type->stmt;
2603 function_computed_target_list = NULL;
2604 function_computed_goto_list = NULL;
2606 if (decl->ctype.modifiers & MOD_EXTERN) {
2607 if (!(decl->ctype.modifiers & MOD_INLINE))
2608 warning(decl->pos, "function '%s' with external linkage has definition", show_ident(decl->ident));
2610 if (!(decl->ctype.modifiers & MOD_STATIC))
2611 decl->ctype.modifiers |= MOD_EXTERN;
2613 stmt = start_function(decl);
2615 *p = stmt;
2616 FOR_EACH_PTR (base_type->arguments, arg) {
2617 declare_argument(arg, base_type);
2618 } END_FOR_EACH_PTR(arg);
2620 token = compound_statement(token->next, stmt);
2622 end_function(decl);
2623 if (!(decl->ctype.modifiers & MOD_INLINE))
2624 add_symbol(list, decl);
2625 check_declaration(decl);
2626 decl->definition = decl;
2627 prev = decl->same_symbol;
2628 if (prev && prev->definition) {
2629 warning(decl->pos, "multiple definitions for function '%s'",
2630 show_ident(decl->ident));
2631 info(prev->definition->pos, " the previous one is here");
2632 } else {
2633 while (prev) {
2634 rebind_scope(prev, decl->scope);
2635 prev->definition = decl;
2636 prev = prev->same_symbol;
2639 function_symbol_list = old_symbol_list;
2640 if (function_computed_goto_list) {
2641 if (!function_computed_target_list)
2642 warning(decl->pos, "function '%s' has computed goto but no targets?", show_ident(decl->ident));
2643 else {
2644 FOR_EACH_PTR(function_computed_goto_list, stmt) {
2645 stmt->target_list = function_computed_target_list;
2646 } END_FOR_EACH_PTR(stmt);
2649 return expect(token, '}', "at end of function");
2652 static void promote_k_r_types(struct symbol *arg)
2654 struct symbol *base = arg->ctype.base_type;
2655 if (base && base->ctype.base_type == &int_type && (base->ctype.modifiers & (MOD_CHAR | MOD_SHORT))) {
2656 arg->ctype.base_type = &int_ctype;
2660 static void apply_k_r_types(struct symbol_list *argtypes, struct symbol *fn)
2662 struct symbol_list *real_args = fn->ctype.base_type->arguments;
2663 struct symbol *arg;
2665 FOR_EACH_PTR(real_args, arg) {
2666 struct symbol *type;
2668 /* This is quadratic in the number of arguments. We _really_ don't care */
2669 FOR_EACH_PTR(argtypes, type) {
2670 if (type->ident == arg->ident)
2671 goto match;
2672 } END_FOR_EACH_PTR(type);
2673 sparse_error(arg->pos, "missing type declaration for parameter '%s'", show_ident(arg->ident));
2674 continue;
2675 match:
2676 type->used = 1;
2677 /* "char" and "short" promote to "int" */
2678 promote_k_r_types(type);
2680 arg->ctype = type->ctype;
2681 } END_FOR_EACH_PTR(arg);
2683 FOR_EACH_PTR(argtypes, arg) {
2684 if (!arg->used)
2685 warning(arg->pos, "nonsensical parameter declaration '%s'", show_ident(arg->ident));
2686 } END_FOR_EACH_PTR(arg);
2690 static struct token *parse_k_r_arguments(struct token *token, struct symbol *decl,
2691 struct symbol_list **list)
2693 struct symbol_list *args = NULL;
2695 warning(token->pos, "non-ANSI definition of function '%s'", show_ident(decl->ident));
2696 do {
2697 token = declaration_list(token, &args);
2698 if (!match_op(token, ';')) {
2699 sparse_error(token->pos, "expected ';' at end of parameter declaration");
2700 break;
2702 token = token->next;
2703 } while (lookup_type(token));
2705 apply_k_r_types(args, decl);
2707 if (!match_op(token, '{')) {
2708 sparse_error(token->pos, "expected function body");
2709 return token;
2711 return parse_function_body(token, decl, list);
2714 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list)
2716 struct symbol *anon = alloc_symbol(token->pos, SYM_NODE);
2717 struct symbol *fn = alloc_symbol(token->pos, SYM_FN);
2718 struct statement *stmt;
2720 anon->ctype.base_type = fn;
2721 stmt = alloc_statement(token->pos, STMT_NONE);
2722 fn->stmt = stmt;
2724 token = parse_asm_statement(token, stmt);
2726 add_symbol(list, anon);
2727 return token;
2730 struct token *external_declaration(struct token *token, struct symbol_list **list)
2732 struct ident *ident = NULL;
2733 struct symbol *decl;
2734 struct decl_state ctx = { .ident = &ident };
2735 struct ctype saved;
2736 struct symbol *base_type;
2737 unsigned long mod;
2738 int is_typedef;
2740 /* Top-level inline asm? */
2741 if (token_type(token) == TOKEN_IDENT) {
2742 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2743 if (s && s->op->toplevel)
2744 return s->op->toplevel(token, list);
2747 /* Parse declaration-specifiers, if any */
2748 token = declaration_specifiers(token, &ctx);
2749 mod = storage_modifiers(&ctx);
2750 decl = alloc_symbol(token->pos, SYM_NODE);
2751 /* Just a type declaration? */
2752 if (match_op(token, ';')) {
2753 apply_modifiers(token->pos, &ctx);
2754 return token->next;
2757 saved = ctx.ctype;
2758 token = declarator(token, &ctx);
2759 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
2760 apply_modifiers(token->pos, &ctx);
2762 decl->ctype = ctx.ctype;
2763 decl->ctype.modifiers |= mod;
2764 decl->endpos = token->pos;
2766 /* Just a type declaration? */
2767 if (!ident) {
2768 warning(token->pos, "missing identifier in declaration");
2769 return expect(token, ';', "at the end of type declaration");
2772 /* type define declaration? */
2773 is_typedef = ctx.storage_class == STypedef;
2775 /* Typedefs don't have meaningful storage */
2776 if (is_typedef)
2777 decl->ctype.modifiers |= MOD_USERTYPE;
2779 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
2781 base_type = decl->ctype.base_type;
2783 if (is_typedef) {
2784 if (base_type && !base_type->ident) {
2785 switch (base_type->type) {
2786 case SYM_STRUCT:
2787 case SYM_UNION:
2788 case SYM_ENUM:
2789 case SYM_RESTRICT:
2790 base_type->ident = ident;
2791 break;
2792 default:
2793 break;
2796 } else if (base_type && base_type->type == SYM_FN) {
2797 /* K&R argument declaration? */
2798 if (lookup_type(token))
2799 return parse_k_r_arguments(token, decl, list);
2800 if (match_op(token, '{'))
2801 return parse_function_body(token, decl, list);
2803 if (!(decl->ctype.modifiers & MOD_STATIC))
2804 decl->ctype.modifiers |= MOD_EXTERN;
2805 } else if (base_type == &void_ctype && !(decl->ctype.modifiers & MOD_EXTERN)) {
2806 sparse_error(token->pos, "void declaration");
2809 for (;;) {
2810 if (!is_typedef && match_op(token, '=')) {
2811 if (decl->ctype.modifiers & MOD_EXTERN) {
2812 warning(decl->pos, "symbol with external linkage has initializer");
2813 decl->ctype.modifiers &= ~MOD_EXTERN;
2815 token = initializer(&decl->initializer, token->next);
2817 if (!is_typedef) {
2818 if (!(decl->ctype.modifiers & (MOD_EXTERN | MOD_INLINE))) {
2819 add_symbol(list, decl);
2820 fn_local_symbol(decl);
2823 check_declaration(decl);
2824 if (decl->same_symbol)
2825 decl->definition = decl->same_symbol->definition;
2827 if (!match_op(token, ','))
2828 break;
2830 token = token->next;
2831 ident = NULL;
2832 decl = alloc_symbol(token->pos, SYM_NODE);
2833 ctx.ctype = saved;
2834 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
2835 token = declarator(token, &ctx);
2836 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
2837 apply_modifiers(token->pos, &ctx);
2838 decl->ctype = ctx.ctype;
2839 decl->ctype.modifiers |= mod;
2840 decl->endpos = token->pos;
2841 if (!ident) {
2842 sparse_error(token->pos, "expected identifier name in type definition");
2843 return token;
2846 if (is_typedef)
2847 decl->ctype.modifiers |= MOD_USERTYPE;
2849 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
2851 /* Function declarations are automatically extern unless specifically static */
2852 base_type = decl->ctype.base_type;
2853 if (!is_typedef && base_type && base_type->type == SYM_FN) {
2854 if (!(decl->ctype.modifiers & MOD_STATIC))
2855 decl->ctype.modifiers |= MOD_EXTERN;
2858 return expect(token, ';', "at end of declaration");