Merge branch 'buf_size' into devel
[smatch.git] / parse.c
bloba809c145e0cb8ab2bbfecd08ce9376c8ce85d59c
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 * Licensed under the Open Software License version 1.1
13 #include <stdarg.h>
14 #include <stdlib.h>
15 #include <stdio.h>
16 #include <string.h>
17 #include <ctype.h>
18 #include <unistd.h>
19 #include <fcntl.h>
20 #include <limits.h>
22 #include "lib.h"
23 #include "allocate.h"
24 #include "token.h"
25 #include "parse.h"
26 #include "symbol.h"
27 #include "scope.h"
28 #include "expression.h"
29 #include "target.h"
31 static struct symbol_list **function_symbol_list;
32 struct symbol_list *function_computed_target_list;
33 struct statement_list *function_computed_goto_list;
35 static struct token *statement(struct token *token, struct statement **tree);
36 static struct token *handle_attributes(struct token *token, struct decl_state *ctx, unsigned int keywords);
38 typedef struct token *declarator_t(struct token *, struct decl_state *);
39 static declarator_t
40 struct_specifier, union_specifier, enum_specifier,
41 attribute_specifier, typeof_specifier, parse_asm_declarator,
42 typedef_specifier, inline_specifier, auto_specifier,
43 register_specifier, static_specifier, extern_specifier,
44 thread_specifier, const_qualifier, volatile_qualifier;
46 static struct token *parse_if_statement(struct token *token, struct statement *stmt);
47 static struct token *parse_return_statement(struct token *token, struct statement *stmt);
48 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt);
49 static struct token *parse_default_statement(struct token *token, struct statement *stmt);
50 static struct token *parse_case_statement(struct token *token, struct statement *stmt);
51 static struct token *parse_switch_statement(struct token *token, struct statement *stmt);
52 static struct token *parse_for_statement(struct token *token, struct statement *stmt);
53 static struct token *parse_while_statement(struct token *token, struct statement *stmt);
54 static struct token *parse_do_statement(struct token *token, struct statement *stmt);
55 static struct token *parse_goto_statement(struct token *token, struct statement *stmt);
56 static struct token *parse_context_statement(struct token *token, struct statement *stmt);
57 static struct token *parse_range_statement(struct token *token, struct statement *stmt);
58 static struct token *parse_asm_statement(struct token *token, struct statement *stmt);
59 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list);
61 typedef struct token *attr_t(struct token *, struct symbol *,
62 struct decl_state *);
64 static attr_t
65 attribute_packed, attribute_aligned, attribute_modifier,
66 attribute_address_space, attribute_context,
67 attribute_designated_init,
68 attribute_transparent_union, ignore_attribute,
69 attribute_mode, attribute_force;
71 typedef struct symbol *to_mode_t(struct symbol *);
73 static to_mode_t
74 to_QI_mode, to_HI_mode, to_SI_mode, to_DI_mode, to_TI_mode, to_word_mode;
76 enum {
77 Set_T = 1,
78 Set_S = 2,
79 Set_Char = 4,
80 Set_Int = 8,
81 Set_Double = 16,
82 Set_Float = 32,
83 Set_Signed = 64,
84 Set_Unsigned = 128,
85 Set_Short = 256,
86 Set_Long = 512,
87 Set_Vlong = 1024,
88 Set_Any = Set_T | Set_Short | Set_Long | Set_Signed | Set_Unsigned
91 enum {
92 CInt = 0, CSInt, CUInt, CReal, CChar, CSChar, CUChar
95 enum {
96 SNone = 0, STypedef, SAuto, SRegister, SExtern, SStatic, SForced
99 static struct symbol_op typedef_op = {
100 .type = KW_MODIFIER,
101 .declarator = typedef_specifier,
104 static struct symbol_op inline_op = {
105 .type = KW_MODIFIER,
106 .declarator = inline_specifier,
109 static struct symbol_op auto_op = {
110 .type = KW_MODIFIER,
111 .declarator = auto_specifier,
114 static struct symbol_op register_op = {
115 .type = KW_MODIFIER,
116 .declarator = register_specifier,
119 static struct symbol_op static_op = {
120 .type = KW_MODIFIER,
121 .declarator = static_specifier,
124 static struct symbol_op extern_op = {
125 .type = KW_MODIFIER,
126 .declarator = extern_specifier,
129 static struct symbol_op thread_op = {
130 .type = KW_MODIFIER,
131 .declarator = thread_specifier,
134 static struct symbol_op const_op = {
135 .type = KW_QUALIFIER,
136 .declarator = const_qualifier,
139 static struct symbol_op volatile_op = {
140 .type = KW_QUALIFIER,
141 .declarator = volatile_qualifier,
144 static struct symbol_op restrict_op = {
145 .type = KW_QUALIFIER,
148 static struct symbol_op typeof_op = {
149 .type = KW_SPECIFIER,
150 .declarator = typeof_specifier,
151 .test = Set_Any,
152 .set = Set_S|Set_T,
155 static struct symbol_op attribute_op = {
156 .type = KW_ATTRIBUTE,
157 .declarator = attribute_specifier,
160 static struct symbol_op struct_op = {
161 .type = KW_SPECIFIER,
162 .declarator = struct_specifier,
163 .test = Set_Any,
164 .set = Set_S|Set_T,
167 static struct symbol_op union_op = {
168 .type = KW_SPECIFIER,
169 .declarator = union_specifier,
170 .test = Set_Any,
171 .set = Set_S|Set_T,
174 static struct symbol_op enum_op = {
175 .type = KW_SPECIFIER,
176 .declarator = enum_specifier,
177 .test = Set_Any,
178 .set = Set_S|Set_T,
181 static struct symbol_op spec_op = {
182 .type = KW_SPECIFIER | KW_EXACT,
183 .test = Set_Any,
184 .set = Set_S|Set_T,
187 static struct symbol_op char_op = {
188 .type = KW_SPECIFIER,
189 .test = Set_T|Set_Long|Set_Short,
190 .set = Set_T|Set_Char,
191 .class = CChar,
194 static struct symbol_op int_op = {
195 .type = KW_SPECIFIER,
196 .test = Set_T,
197 .set = Set_T|Set_Int,
200 static struct symbol_op double_op = {
201 .type = KW_SPECIFIER,
202 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Vlong,
203 .set = Set_T|Set_Double,
204 .class = CReal,
207 static struct symbol_op float_op = {
208 .type = KW_SPECIFIER | KW_SHORT,
209 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Long,
210 .set = Set_T|Set_Float,
211 .class = CReal,
214 static struct symbol_op short_op = {
215 .type = KW_SPECIFIER | KW_SHORT,
216 .test = Set_S|Set_Char|Set_Float|Set_Double|Set_Long|Set_Short,
217 .set = Set_Short,
220 static struct symbol_op signed_op = {
221 .type = KW_SPECIFIER,
222 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
223 .set = Set_Signed,
224 .class = CSInt,
227 static struct symbol_op unsigned_op = {
228 .type = KW_SPECIFIER,
229 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
230 .set = Set_Unsigned,
231 .class = CUInt,
234 static struct symbol_op long_op = {
235 .type = KW_SPECIFIER | KW_LONG,
236 .test = Set_S|Set_Char|Set_Float|Set_Short|Set_Vlong,
237 .set = Set_Long,
240 static struct symbol_op if_op = {
241 .statement = parse_if_statement,
244 static struct symbol_op return_op = {
245 .statement = parse_return_statement,
248 static struct symbol_op loop_iter_op = {
249 .statement = parse_loop_iterator,
252 static struct symbol_op default_op = {
253 .statement = parse_default_statement,
256 static struct symbol_op case_op = {
257 .statement = parse_case_statement,
260 static struct symbol_op switch_op = {
261 .statement = parse_switch_statement,
264 static struct symbol_op for_op = {
265 .statement = parse_for_statement,
268 static struct symbol_op while_op = {
269 .statement = parse_while_statement,
272 static struct symbol_op do_op = {
273 .statement = parse_do_statement,
276 static struct symbol_op goto_op = {
277 .statement = parse_goto_statement,
280 static struct symbol_op __context___op = {
281 .statement = parse_context_statement,
284 static struct symbol_op range_op = {
285 .statement = parse_range_statement,
288 static struct symbol_op asm_op = {
289 .type = KW_ASM,
290 .declarator = parse_asm_declarator,
291 .statement = parse_asm_statement,
292 .toplevel = toplevel_asm_declaration,
295 static struct symbol_op packed_op = {
296 .attribute = attribute_packed,
299 static struct symbol_op aligned_op = {
300 .attribute = attribute_aligned,
303 static struct symbol_op attr_mod_op = {
304 .attribute = attribute_modifier,
307 static struct symbol_op attr_force_op = {
308 .attribute = attribute_force,
311 static struct symbol_op address_space_op = {
312 .attribute = attribute_address_space,
315 static struct symbol_op mode_op = {
316 .attribute = attribute_mode,
319 static struct symbol_op context_op = {
320 .attribute = attribute_context,
323 static struct symbol_op designated_init_op = {
324 .attribute = attribute_designated_init,
327 static struct symbol_op transparent_union_op = {
328 .attribute = attribute_transparent_union,
331 static struct symbol_op ignore_attr_op = {
332 .attribute = ignore_attribute,
335 static struct symbol_op mode_QI_op = {
336 .type = KW_MODE,
337 .to_mode = to_QI_mode
340 static struct symbol_op mode_HI_op = {
341 .type = KW_MODE,
342 .to_mode = to_HI_mode
345 static struct symbol_op mode_SI_op = {
346 .type = KW_MODE,
347 .to_mode = to_SI_mode
350 static struct symbol_op mode_DI_op = {
351 .type = KW_MODE,
352 .to_mode = to_DI_mode
355 static struct symbol_op mode_TI_op = {
356 .type = KW_MODE,
357 .to_mode = to_TI_mode
360 static struct symbol_op mode_word_op = {
361 .type = KW_MODE,
362 .to_mode = to_word_mode
365 static struct init_keyword {
366 const char *name;
367 enum namespace ns;
368 unsigned long modifiers;
369 struct symbol_op *op;
370 struct symbol *type;
371 } keyword_table[] = {
372 /* Type qualifiers */
373 { "const", NS_TYPEDEF, .op = &const_op },
374 { "__const", NS_TYPEDEF, .op = &const_op },
375 { "__const__", NS_TYPEDEF, .op = &const_op },
376 { "volatile", NS_TYPEDEF, .op = &volatile_op },
377 { "__volatile", NS_TYPEDEF, .op = &volatile_op },
378 { "__volatile__", NS_TYPEDEF, .op = &volatile_op },
380 /* Typedef.. */
381 { "typedef", NS_TYPEDEF, .op = &typedef_op },
383 /* Type specifiers */
384 { "void", NS_TYPEDEF, .type = &void_ctype, .op = &spec_op},
385 { "char", NS_TYPEDEF, .op = &char_op },
386 { "short", NS_TYPEDEF, .op = &short_op },
387 { "int", NS_TYPEDEF, .op = &int_op },
388 { "long", NS_TYPEDEF, .op = &long_op },
389 { "float", NS_TYPEDEF, .op = &float_op },
390 { "double", NS_TYPEDEF, .op = &double_op },
391 { "signed", NS_TYPEDEF, .op = &signed_op },
392 { "__signed", NS_TYPEDEF, .op = &signed_op },
393 { "__signed__", NS_TYPEDEF, .op = &signed_op },
394 { "unsigned", NS_TYPEDEF, .op = &unsigned_op },
395 { "_Bool", NS_TYPEDEF, .type = &bool_ctype, .op = &spec_op },
397 /* Predeclared types */
398 { "__builtin_va_list", NS_TYPEDEF, .type = &ptr_ctype, .op = &spec_op },
399 { "__builtin_ms_va_list", NS_TYPEDEF, .type = &ptr_ctype, .op = &spec_op },
400 { "__int128_t", NS_TYPEDEF, .type = &lllong_ctype, .op = &spec_op },
401 { "__uint128_t",NS_TYPEDEF, .type = &ulllong_ctype, .op = &spec_op },
403 /* Extended types */
404 { "typeof", NS_TYPEDEF, .op = &typeof_op },
405 { "__typeof", NS_TYPEDEF, .op = &typeof_op },
406 { "__typeof__", NS_TYPEDEF, .op = &typeof_op },
408 { "__attribute", NS_TYPEDEF, .op = &attribute_op },
409 { "__attribute__", NS_TYPEDEF, .op = &attribute_op },
411 { "struct", NS_TYPEDEF, .op = &struct_op },
412 { "union", NS_TYPEDEF, .op = &union_op },
413 { "enum", NS_TYPEDEF, .op = &enum_op },
415 { "inline", NS_TYPEDEF, .op = &inline_op },
416 { "__inline", NS_TYPEDEF, .op = &inline_op },
417 { "__inline__", NS_TYPEDEF, .op = &inline_op },
419 /* Ignored for now.. */
420 { "restrict", NS_TYPEDEF, .op = &restrict_op},
421 { "__restrict", NS_TYPEDEF, .op = &restrict_op},
423 /* Storage class */
424 { "auto", NS_TYPEDEF, .op = &auto_op },
425 { "register", NS_TYPEDEF, .op = &register_op },
426 { "static", NS_TYPEDEF, .op = &static_op },
427 { "extern", NS_TYPEDEF, .op = &extern_op },
428 { "__thread", NS_TYPEDEF, .op = &thread_op },
430 /* Statement */
431 { "if", NS_KEYWORD, .op = &if_op },
432 { "return", NS_KEYWORD, .op = &return_op },
433 { "break", NS_KEYWORD, .op = &loop_iter_op },
434 { "continue", NS_KEYWORD, .op = &loop_iter_op },
435 { "default", NS_KEYWORD, .op = &default_op },
436 { "case", NS_KEYWORD, .op = &case_op },
437 { "switch", NS_KEYWORD, .op = &switch_op },
438 { "for", NS_KEYWORD, .op = &for_op },
439 { "while", NS_KEYWORD, .op = &while_op },
440 { "do", NS_KEYWORD, .op = &do_op },
441 { "goto", NS_KEYWORD, .op = &goto_op },
442 { "__context__",NS_KEYWORD, .op = &__context___op },
443 { "__range__", NS_KEYWORD, .op = &range_op },
444 { "asm", NS_KEYWORD, .op = &asm_op },
445 { "__asm", NS_KEYWORD, .op = &asm_op },
446 { "__asm__", NS_KEYWORD, .op = &asm_op },
448 /* Attribute */
449 { "packed", NS_KEYWORD, .op = &packed_op },
450 { "__packed__", NS_KEYWORD, .op = &packed_op },
451 { "aligned", NS_KEYWORD, .op = &aligned_op },
452 { "__aligned__",NS_KEYWORD, .op = &aligned_op },
453 { "nocast", NS_KEYWORD, MOD_NOCAST, .op = &attr_mod_op },
454 { "noderef", NS_KEYWORD, MOD_NODEREF, .op = &attr_mod_op },
455 { "safe", NS_KEYWORD, MOD_SAFE, .op = &attr_mod_op },
456 { "force", NS_KEYWORD, .op = &attr_force_op },
457 { "bitwise", NS_KEYWORD, MOD_BITWISE, .op = &attr_mod_op },
458 { "__bitwise__",NS_KEYWORD, MOD_BITWISE, .op = &attr_mod_op },
459 { "address_space",NS_KEYWORD, .op = &address_space_op },
460 { "mode", NS_KEYWORD, .op = &mode_op },
461 { "context", NS_KEYWORD, .op = &context_op },
462 { "designated_init", NS_KEYWORD, .op = &designated_init_op },
463 { "__transparent_union__", NS_KEYWORD, .op = &transparent_union_op },
464 { "noreturn", NS_KEYWORD, MOD_NORETURN, .op = &attr_mod_op },
465 { "__noreturn__", NS_KEYWORD, MOD_NORETURN, .op = &attr_mod_op },
466 { "pure", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
467 {"__pure__", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
468 {"const", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
469 {"__const", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
470 {"__const__", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
472 { "__mode__", NS_KEYWORD, .op = &mode_op },
473 { "QI", NS_KEYWORD, MOD_CHAR, .op = &mode_QI_op },
474 { "__QI__", NS_KEYWORD, MOD_CHAR, .op = &mode_QI_op },
475 { "HI", NS_KEYWORD, MOD_SHORT, .op = &mode_HI_op },
476 { "__HI__", NS_KEYWORD, MOD_SHORT, .op = &mode_HI_op },
477 { "SI", NS_KEYWORD, .op = &mode_SI_op },
478 { "__SI__", NS_KEYWORD, .op = &mode_SI_op },
479 { "DI", NS_KEYWORD, MOD_LONGLONG, .op = &mode_DI_op },
480 { "__DI__", NS_KEYWORD, MOD_LONGLONG, .op = &mode_DI_op },
481 { "TI", NS_KEYWORD, MOD_LONGLONGLONG, .op = &mode_TI_op },
482 { "__TI__", NS_KEYWORD, MOD_LONGLONGLONG, .op = &mode_TI_op },
483 { "word", NS_KEYWORD, MOD_LONG, .op = &mode_word_op },
484 { "__word__", NS_KEYWORD, MOD_LONG, .op = &mode_word_op },
487 const char *ignored_attributes[] = {
488 "alias",
489 "__alias__",
490 "alloc_size",
491 "__alloc_size__",
492 "always_inline",
493 "__always_inline__",
494 "artificial",
495 "__artificial__",
496 "bounded",
497 "__bounded__",
498 "cdecl",
499 "__cdecl__",
500 "cold",
501 "__cold__",
502 "constructor",
503 "__constructor__",
504 "deprecated",
505 "__deprecated__",
506 "destructor",
507 "__destructor__",
508 "dllexport",
509 "__dllexport__",
510 "dllimport",
511 "__dllimport__",
512 "externally_visible",
513 "__externally_visible__",
514 "fastcall",
515 "__fastcall__",
516 "format",
517 "__format__",
518 "format_arg",
519 "__format_arg__",
520 "hot",
521 "__hot__",
522 "l1_text",
523 "__l1_text__",
524 "l1_data",
525 "__l1_data__",
526 "l2",
527 "__l2__",
528 "may_alias",
529 "__may_alias__",
530 "malloc",
531 "__malloc__",
532 "may_alias",
533 "__may_alias__",
534 "model",
535 "__model__",
536 "ms_abi",
537 "__ms_abi__",
538 "ms_hook_prologue",
539 "__ms_hook_prologue__",
540 "naked",
541 "__naked__",
542 "no_instrument_function",
543 "__no_instrument_function__",
544 "noinline",
545 "__noinline__",
546 "nonnull",
547 "__nonnull",
548 "__nonnull__",
549 "nothrow",
550 "__nothrow",
551 "__nothrow__",
552 "regparm",
553 "__regparm__",
554 "section",
555 "__section__",
556 "sentinel",
557 "__sentinel__",
558 "signal",
559 "__signal__",
560 "stdcall",
561 "__stdcall__",
562 "syscall_linkage",
563 "__syscall_linkage__",
564 "sysv_abi",
565 "__sysv_abi__",
566 "unused",
567 "__unused__",
568 "used",
569 "__used__",
570 "vector_size",
571 "visibility",
572 "__visibility__",
573 "warn_unused_result",
574 "__warn_unused_result__",
575 "warning",
576 "__warning__",
577 "weak",
578 "__weak__",
582 void init_parser(int stream)
584 int i;
585 for (i = 0; i < ARRAY_SIZE(keyword_table); i++) {
586 struct init_keyword *ptr = keyword_table + i;
587 struct symbol *sym = create_symbol(stream, ptr->name, SYM_KEYWORD, ptr->ns);
588 sym->ident->keyword = 1;
589 if (ptr->ns == NS_TYPEDEF)
590 sym->ident->reserved = 1;
591 sym->ctype.modifiers = ptr->modifiers;
592 sym->ctype.base_type = ptr->type;
593 sym->op = ptr->op;
596 for (i = 0; i < ARRAY_SIZE(ignored_attributes); i++) {
597 const char * name = ignored_attributes[i];
598 struct symbol *sym = create_symbol(stream, name, SYM_KEYWORD,
599 NS_KEYWORD);
600 sym->ident->keyword = 1;
601 sym->op = &ignore_attr_op;
606 // Add a symbol to the list of function-local symbols
607 static void fn_local_symbol(struct symbol *sym)
609 if (function_symbol_list)
610 add_symbol(function_symbol_list, sym);
613 static int SENTINEL_ATTR match_idents(struct token *token, ...)
615 va_list args;
616 struct ident * next;
618 if (token_type(token) != TOKEN_IDENT)
619 return 0;
621 va_start(args, token);
622 do {
623 next = va_arg(args, struct ident *);
624 } while (next && token->ident != next);
625 va_end(args);
627 return next && token->ident == next;
631 struct statement *alloc_statement(struct position pos, int type)
633 struct statement *stmt = __alloc_statement(0);
634 stmt->type = type;
635 stmt->pos = pos;
636 return stmt;
639 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list);
641 static void apply_modifiers(struct position pos, struct decl_state *ctx)
643 struct symbol *ctype;
644 if (!ctx->mode)
645 return;
646 ctype = ctx->mode->to_mode(ctx->ctype.base_type);
647 if (!ctype)
648 sparse_error(pos, "don't know how to apply mode to %s",
649 show_typename(ctx->ctype.base_type));
650 else
651 ctx->ctype.base_type = ctype;
655 static struct symbol * alloc_indirect_symbol(struct position pos, struct ctype *ctype, int type)
657 struct symbol *sym = alloc_symbol(pos, type);
659 sym->ctype.base_type = ctype->base_type;
660 sym->ctype.modifiers = ctype->modifiers;
662 ctype->base_type = sym;
663 ctype->modifiers = 0;
664 return sym;
668 * NOTE! NS_LABEL is not just a different namespace,
669 * it also ends up using function scope instead of the
670 * regular symbol scope.
672 struct symbol *label_symbol(struct token *token)
674 struct symbol *sym = lookup_symbol(token->ident, NS_LABEL);
675 if (!sym) {
676 sym = alloc_symbol(token->pos, SYM_LABEL);
677 bind_symbol(sym, token->ident, NS_LABEL);
678 fn_local_symbol(sym);
680 return sym;
683 static struct token *struct_union_enum_specifier(enum type type,
684 struct token *token, struct decl_state *ctx,
685 struct token *(*parse)(struct token *, struct symbol *))
687 struct symbol *sym;
688 struct position *repos;
690 token = handle_attributes(token, ctx, KW_ATTRIBUTE);
691 if (token_type(token) == TOKEN_IDENT) {
692 sym = lookup_symbol(token->ident, NS_STRUCT);
693 if (!sym ||
694 (is_outer_scope(sym->scope) &&
695 (match_op(token->next,';') || match_op(token->next,'{')))) {
696 // Either a new symbol, or else an out-of-scope
697 // symbol being redefined.
698 sym = alloc_symbol(token->pos, type);
699 bind_symbol(sym, token->ident, NS_STRUCT);
701 if (sym->type != type)
702 error_die(token->pos, "invalid tag applied to %s", show_typename (sym));
703 ctx->ctype.base_type = sym;
704 repos = &token->pos;
705 token = token->next;
706 if (match_op(token, '{')) {
707 // The following test is actually wrong for empty
708 // structs, but (1) they are not C99, (2) gcc does
709 // the same thing, and (3) it's easier.
710 if (sym->symbol_list)
711 error_die(token->pos, "redefinition of %s", show_typename (sym));
712 sym->pos = *repos;
713 token = parse(token->next, sym);
714 token = expect(token, '}', "at end of struct-union-enum-specifier");
716 // Mark the structure as needing re-examination
717 sym->examined = 0;
718 sym->endpos = token->pos;
720 return token;
723 // private struct/union/enum type
724 if (!match_op(token, '{')) {
725 sparse_error(token->pos, "expected declaration");
726 ctx->ctype.base_type = &bad_ctype;
727 return token;
730 sym = alloc_symbol(token->pos, type);
731 token = parse(token->next, sym);
732 ctx->ctype.base_type = sym;
733 token = expect(token, '}', "at end of specifier");
734 sym->endpos = token->pos;
736 return token;
739 static struct token *parse_struct_declaration(struct token *token, struct symbol *sym)
741 struct symbol *field, *last = NULL;
742 struct token *res;
743 res = struct_declaration_list(token, &sym->symbol_list);
744 FOR_EACH_PTR(sym->symbol_list, field) {
745 if (!field->ident) {
746 struct symbol *base = field->ctype.base_type;
747 if (base && base->type == SYM_BITFIELD)
748 continue;
750 if (last)
751 last->next_subobject = field;
752 last = field;
753 } END_FOR_EACH_PTR(field);
754 return res;
757 static struct token *parse_union_declaration(struct token *token, struct symbol *sym)
759 return struct_declaration_list(token, &sym->symbol_list);
762 static struct token *struct_specifier(struct token *token, struct decl_state *ctx)
764 return struct_union_enum_specifier(SYM_STRUCT, token, ctx, parse_struct_declaration);
767 static struct token *union_specifier(struct token *token, struct decl_state *ctx)
769 return struct_union_enum_specifier(SYM_UNION, token, ctx, parse_union_declaration);
773 typedef struct {
774 int x;
775 unsigned long long y;
776 } Num;
778 static void upper_boundary(Num *n, Num *v)
780 if (n->x > v->x)
781 return;
782 if (n->x < v->x) {
783 *n = *v;
784 return;
786 if (n->y < v->y)
787 n->y = v->y;
790 static void lower_boundary(Num *n, Num *v)
792 if (n->x < v->x)
793 return;
794 if (n->x > v->x) {
795 *n = *v;
796 return;
798 if (n->y > v->y)
799 n->y = v->y;
802 static int type_is_ok(struct symbol *type, Num *upper, Num *lower)
804 int shift = type->bit_size;
805 int is_unsigned = type->ctype.modifiers & MOD_UNSIGNED;
807 if (!is_unsigned)
808 shift--;
809 if (upper->x == 0 && upper->y >> shift)
810 return 0;
811 if (lower->x == 0 || (!is_unsigned && (~lower->y >> shift) == 0))
812 return 1;
813 return 0;
816 static struct symbol *bigger_enum_type(struct symbol *s1, struct symbol *s2)
818 if (s1->bit_size < s2->bit_size) {
819 s1 = s2;
820 } else if (s1->bit_size == s2->bit_size) {
821 if (s2->ctype.modifiers & MOD_UNSIGNED)
822 s1 = s2;
824 if (s1->bit_size < bits_in_int)
825 return &int_ctype;
826 return s1;
829 static void cast_enum_list(struct symbol_list *list, struct symbol *base_type)
831 struct symbol *sym;
833 FOR_EACH_PTR(list, sym) {
834 struct expression *expr = sym->initializer;
835 struct symbol *ctype;
836 if (expr->type != EXPR_VALUE)
837 continue;
838 ctype = expr->ctype;
839 if (ctype->bit_size == base_type->bit_size)
840 continue;
841 cast_value(expr, base_type, expr, ctype);
842 } END_FOR_EACH_PTR(sym);
845 static struct token *parse_enum_declaration(struct token *token, struct symbol *parent)
847 unsigned long long lastval = 0;
848 struct symbol *ctype = NULL, *base_type = NULL;
849 Num upper = {-1, 0}, lower = {1, 0};
851 parent->examined = 1;
852 parent->ctype.base_type = &int_ctype;
853 while (token_type(token) == TOKEN_IDENT) {
854 struct expression *expr = NULL;
855 struct token *next = token->next;
856 struct symbol *sym;
858 if (match_op(next, '=')) {
859 next = constant_expression(next->next, &expr);
860 lastval = get_expression_value(expr);
861 ctype = &void_ctype;
862 if (expr && expr->ctype)
863 ctype = expr->ctype;
864 } else if (!ctype) {
865 ctype = &int_ctype;
866 } else if (is_int_type(ctype)) {
867 lastval++;
868 } else {
869 error_die(token->pos, "can't increment the last enum member");
872 if (!expr) {
873 expr = alloc_expression(token->pos, EXPR_VALUE);
874 expr->value = lastval;
875 expr->ctype = ctype;
878 sym = alloc_symbol(token->pos, SYM_NODE);
879 bind_symbol(sym, token->ident, NS_SYMBOL);
880 sym->ctype.modifiers &= ~MOD_ADDRESSABLE;
881 sym->initializer = expr;
882 sym->enum_member = 1;
883 sym->ctype.base_type = parent;
884 add_ptr_list(&parent->symbol_list, sym);
886 if (base_type != &bad_ctype) {
887 if (ctype->type == SYM_NODE)
888 ctype = ctype->ctype.base_type;
889 if (ctype->type == SYM_ENUM) {
890 if (ctype == parent)
891 ctype = base_type;
892 else
893 ctype = ctype->ctype.base_type;
896 * base_type rules:
897 * - if all enums are of the same type, then
898 * the base_type is that type (two first
899 * cases)
900 * - if enums are of different types, they
901 * all have to be integer types, and the
902 * base type is at least "int_ctype".
903 * - otherwise the base_type is "bad_ctype".
905 if (!base_type) {
906 base_type = ctype;
907 } else if (ctype == base_type) {
908 /* nothing */
909 } else if (is_int_type(base_type) && is_int_type(ctype)) {
910 base_type = bigger_enum_type(base_type, ctype);
911 } else
912 base_type = &bad_ctype;
913 parent->ctype.base_type = base_type;
915 if (is_int_type(base_type)) {
916 Num v = {.y = lastval};
917 if (ctype->ctype.modifiers & MOD_UNSIGNED)
918 v.x = 0;
919 else if ((long long)lastval >= 0)
920 v.x = 0;
921 else
922 v.x = -1;
923 upper_boundary(&upper, &v);
924 lower_boundary(&lower, &v);
926 token = next;
928 sym->endpos = token->pos;
930 if (!match_op(token, ','))
931 break;
932 token = token->next;
934 if (!base_type) {
935 sparse_error(token->pos, "bad enum definition");
936 base_type = &bad_ctype;
938 else if (!is_int_type(base_type))
939 base_type = base_type;
940 else if (type_is_ok(base_type, &upper, &lower))
941 base_type = base_type;
942 else if (type_is_ok(&int_ctype, &upper, &lower))
943 base_type = &int_ctype;
944 else if (type_is_ok(&uint_ctype, &upper, &lower))
945 base_type = &uint_ctype;
946 else if (type_is_ok(&long_ctype, &upper, &lower))
947 base_type = &long_ctype;
948 else if (type_is_ok(&ulong_ctype, &upper, &lower))
949 base_type = &ulong_ctype;
950 else if (type_is_ok(&llong_ctype, &upper, &lower))
951 base_type = &llong_ctype;
952 else if (type_is_ok(&ullong_ctype, &upper, &lower))
953 base_type = &ullong_ctype;
954 else
955 base_type = &bad_ctype;
956 parent->ctype.base_type = base_type;
957 parent->ctype.modifiers |= (base_type->ctype.modifiers & MOD_UNSIGNED);
958 parent->examined = 0;
960 cast_enum_list(parent->symbol_list, base_type);
962 return token;
965 static struct token *enum_specifier(struct token *token, struct decl_state *ctx)
967 struct token *ret = struct_union_enum_specifier(SYM_ENUM, token, ctx, parse_enum_declaration);
968 struct ctype *ctype = &ctx->ctype.base_type->ctype;
970 if (!ctype->base_type)
971 ctype->base_type = &incomplete_ctype;
973 return ret;
976 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype);
978 static struct token *typeof_specifier(struct token *token, struct decl_state *ctx)
980 struct symbol *sym;
982 if (!match_op(token, '(')) {
983 sparse_error(token->pos, "expected '(' after typeof");
984 return token;
986 if (lookup_type(token->next)) {
987 token = typename(token->next, &sym, NULL);
988 ctx->ctype.base_type = sym->ctype.base_type;
989 apply_ctype(token->pos, &sym->ctype, &ctx->ctype);
990 } else {
991 struct symbol *typeof_sym = alloc_symbol(token->pos, SYM_TYPEOF);
992 token = parse_expression(token->next, &typeof_sym->initializer);
994 typeof_sym->endpos = token->pos;
995 if (!typeof_sym->initializer) {
996 sparse_error(token->pos, "expected expression after the '(' token");
997 typeof_sym = &bad_ctype;
999 ctx->ctype.base_type = typeof_sym;
1001 return expect(token, ')', "after typeof");
1004 static struct token *ignore_attribute(struct token *token, struct symbol *attr, struct decl_state *ctx)
1006 struct expression *expr = NULL;
1007 if (match_op(token, '('))
1008 token = parens_expression(token, &expr, "in attribute");
1009 return token;
1012 static struct token *attribute_packed(struct token *token, struct symbol *attr, struct decl_state *ctx)
1014 if (!ctx->ctype.alignment)
1015 ctx->ctype.alignment = 1;
1016 return token;
1019 static struct token *attribute_aligned(struct token *token, struct symbol *attr, struct decl_state *ctx)
1021 int alignment = max_alignment;
1022 struct expression *expr = NULL;
1024 if (match_op(token, '(')) {
1025 token = parens_expression(token, &expr, "in attribute");
1026 if (expr)
1027 alignment = const_expression_value(expr);
1029 if (alignment & (alignment-1)) {
1030 warning(token->pos, "I don't like non-power-of-2 alignments");
1031 return token;
1032 } else if (alignment > ctx->ctype.alignment)
1033 ctx->ctype.alignment = alignment;
1034 return token;
1037 static void apply_qualifier(struct position *pos, struct ctype *ctx, unsigned long qual)
1039 if (ctx->modifiers & qual)
1040 warning(*pos, "duplicate %s", modifier_string(qual));
1041 ctx->modifiers |= qual;
1044 static struct token *attribute_modifier(struct token *token, struct symbol *attr, struct decl_state *ctx)
1046 apply_qualifier(&token->pos, &ctx->ctype, attr->ctype.modifiers);
1047 return token;
1050 static struct token *attribute_address_space(struct token *token, struct symbol *attr, struct decl_state *ctx)
1052 struct expression *expr = NULL;
1053 int as;
1054 token = expect(token, '(', "after address_space attribute");
1055 token = conditional_expression(token, &expr);
1056 if (expr) {
1057 as = const_expression_value(expr);
1058 if (Waddress_space && as)
1059 ctx->ctype.as = as;
1061 token = expect(token, ')', "after address_space attribute");
1062 return token;
1065 static struct symbol *to_QI_mode(struct symbol *ctype)
1067 if (ctype->ctype.base_type != &int_type)
1068 return NULL;
1069 if (ctype == &char_ctype)
1070 return ctype;
1071 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uchar_ctype
1072 : &schar_ctype;
1075 static struct symbol *to_HI_mode(struct symbol *ctype)
1077 if (ctype->ctype.base_type != &int_type)
1078 return NULL;
1079 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ushort_ctype
1080 : &sshort_ctype;
1083 static struct symbol *to_SI_mode(struct symbol *ctype)
1085 if (ctype->ctype.base_type != &int_type)
1086 return NULL;
1087 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint_ctype
1088 : &sint_ctype;
1091 static struct symbol *to_DI_mode(struct symbol *ctype)
1093 if (ctype->ctype.base_type != &int_type)
1094 return NULL;
1095 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ullong_ctype
1096 : &sllong_ctype;
1099 static struct symbol *to_TI_mode(struct symbol *ctype)
1101 if (ctype->ctype.base_type != &int_type)
1102 return NULL;
1103 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulllong_ctype
1104 : &slllong_ctype;
1107 static struct symbol *to_word_mode(struct symbol *ctype)
1109 if (ctype->ctype.base_type != &int_type)
1110 return NULL;
1111 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulong_ctype
1112 : &slong_ctype;
1115 static struct token *attribute_mode(struct token *token, struct symbol *attr, struct decl_state *ctx)
1117 token = expect(token, '(', "after mode attribute");
1118 if (token_type(token) == TOKEN_IDENT) {
1119 struct symbol *mode = lookup_keyword(token->ident, NS_KEYWORD);
1120 if (mode && mode->op->type == KW_MODE)
1121 ctx->mode = mode->op;
1122 else
1123 sparse_error(token->pos, "unknown mode attribute %s\n", show_ident(token->ident));
1124 token = token->next;
1125 } else
1126 sparse_error(token->pos, "expect attribute mode symbol\n");
1127 token = expect(token, ')', "after mode attribute");
1128 return token;
1131 static struct token *attribute_context(struct token *token, struct symbol *attr, struct decl_state *ctx)
1133 struct context *context = alloc_context();
1134 struct expression *args[3];
1135 int argc = 0;
1137 token = expect(token, '(', "after context attribute");
1138 while (!match_op(token, ')')) {
1139 struct expression *expr = NULL;
1140 token = conditional_expression(token, &expr);
1141 if (!expr)
1142 break;
1143 if (argc < 3)
1144 args[argc++] = expr;
1145 if (!match_op(token, ','))
1146 break;
1147 token = token->next;
1150 switch(argc) {
1151 case 0:
1152 sparse_error(token->pos, "expected context input/output values");
1153 break;
1154 case 1:
1155 context->in = get_expression_value(args[0]);
1156 break;
1157 case 2:
1158 context->in = get_expression_value(args[0]);
1159 context->out = get_expression_value(args[1]);
1160 break;
1161 case 3:
1162 context->context = args[0];
1163 context->in = get_expression_value(args[1]);
1164 context->out = get_expression_value(args[2]);
1165 break;
1168 if (argc)
1169 add_ptr_list(&ctx->ctype.contexts, context);
1171 token = expect(token, ')', "after context attribute");
1172 return token;
1175 static struct token *attribute_designated_init(struct token *token, struct symbol *attr, struct decl_state *ctx)
1177 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_STRUCT)
1178 ctx->ctype.base_type->designated_init = 1;
1179 else
1180 warning(token->pos, "attribute designated_init applied to non-structure type");
1181 return token;
1184 static struct token *attribute_transparent_union(struct token *token, struct symbol *attr, struct decl_state *ctx)
1186 if (Wtransparent_union)
1187 warning(token->pos, "ignoring attribute __transparent_union__");
1188 return token;
1191 static struct token *recover_unknown_attribute(struct token *token)
1193 struct expression *expr = NULL;
1195 sparse_error(token->pos, "attribute '%s': unknown attribute", show_ident(token->ident));
1196 token = token->next;
1197 if (match_op(token, '('))
1198 token = parens_expression(token, &expr, "in attribute");
1199 return token;
1202 static struct token *attribute_specifier(struct token *token, struct decl_state *ctx)
1204 token = expect(token, '(', "after attribute");
1205 token = expect(token, '(', "after attribute");
1207 for (;;) {
1208 struct ident *attribute_name;
1209 struct symbol *attr;
1211 if (eof_token(token))
1212 break;
1213 if (match_op(token, ';'))
1214 break;
1215 if (token_type(token) != TOKEN_IDENT)
1216 break;
1217 attribute_name = token->ident;
1218 attr = lookup_keyword(attribute_name, NS_KEYWORD);
1219 if (attr && attr->op->attribute)
1220 token = attr->op->attribute(token->next, attr, ctx);
1221 else
1222 token = recover_unknown_attribute(token);
1224 if (!match_op(token, ','))
1225 break;
1226 token = token->next;
1229 token = expect(token, ')', "after attribute");
1230 token = expect(token, ')', "after attribute");
1231 return token;
1234 static const char *storage_class[] =
1236 [STypedef] = "typedef",
1237 [SAuto] = "auto",
1238 [SExtern] = "extern",
1239 [SStatic] = "static",
1240 [SRegister] = "register",
1241 [SForced] = "[force]"
1244 static unsigned long storage_modifiers(struct decl_state *ctx)
1246 static unsigned long mod[] =
1248 [SAuto] = MOD_AUTO,
1249 [SExtern] = MOD_EXTERN,
1250 [SStatic] = MOD_STATIC,
1251 [SRegister] = MOD_REGISTER
1253 return mod[ctx->storage_class] | (ctx->is_inline ? MOD_INLINE : 0)
1254 | (ctx->is_tls ? MOD_TLS : 0);
1257 static void set_storage_class(struct position *pos, struct decl_state *ctx, int class)
1259 /* __thread can be used alone, or with extern or static */
1260 if (ctx->is_tls && (class != SStatic && class != SExtern)) {
1261 sparse_error(*pos, "__thread can only be used alone, or with "
1262 "extern or static");
1263 return;
1266 if (!ctx->storage_class) {
1267 ctx->storage_class = class;
1268 return;
1270 if (ctx->storage_class == class)
1271 sparse_error(*pos, "duplicate %s", storage_class[class]);
1272 else
1273 sparse_error(*pos, "multiple storage classes");
1276 static struct token *typedef_specifier(struct token *next, struct decl_state *ctx)
1278 set_storage_class(&next->pos, ctx, STypedef);
1279 return next;
1282 static struct token *auto_specifier(struct token *next, struct decl_state *ctx)
1284 set_storage_class(&next->pos, ctx, SAuto);
1285 return next;
1288 static struct token *register_specifier(struct token *next, struct decl_state *ctx)
1290 set_storage_class(&next->pos, ctx, SRegister);
1291 return next;
1294 static struct token *static_specifier(struct token *next, struct decl_state *ctx)
1296 set_storage_class(&next->pos, ctx, SStatic);
1297 return next;
1300 static struct token *extern_specifier(struct token *next, struct decl_state *ctx)
1302 set_storage_class(&next->pos, ctx, SExtern);
1303 return next;
1306 static struct token *thread_specifier(struct token *next, struct decl_state *ctx)
1308 /* This GCC extension can be used alone, or with extern or static */
1309 if (!ctx->storage_class || ctx->storage_class == SStatic
1310 || ctx->storage_class == SExtern) {
1311 ctx->is_tls = 1;
1312 } else {
1313 sparse_error(next->pos, "__thread can only be used alone, or "
1314 "with extern or static");
1317 return next;
1320 static struct token *attribute_force(struct token *token, struct symbol *attr, struct decl_state *ctx)
1322 set_storage_class(&token->pos, ctx, SForced);
1323 return token;
1326 static struct token *inline_specifier(struct token *next, struct decl_state *ctx)
1328 ctx->is_inline = 1;
1329 return next;
1332 static struct token *const_qualifier(struct token *next, struct decl_state *ctx)
1334 apply_qualifier(&next->pos, &ctx->ctype, MOD_CONST);
1335 return next;
1338 static struct token *volatile_qualifier(struct token *next, struct decl_state *ctx)
1340 apply_qualifier(&next->pos, &ctx->ctype, MOD_VOLATILE);
1341 return next;
1344 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype)
1346 unsigned long mod = thistype->modifiers;
1348 if (mod)
1349 apply_qualifier(&pos, ctype, mod);
1351 /* Context */
1352 concat_ptr_list((struct ptr_list *)thistype->contexts,
1353 (struct ptr_list **)&ctype->contexts);
1355 /* Alignment */
1356 if (thistype->alignment > ctype->alignment)
1357 ctype->alignment = thistype->alignment;
1359 /* Address space */
1360 if (thistype->as)
1361 ctype->as = thistype->as;
1364 static void specifier_conflict(struct position pos, int what, struct ident *new)
1366 const char *old;
1367 if (what & (Set_S | Set_T))
1368 goto Catch_all;
1369 if (what & Set_Char)
1370 old = "char";
1371 else if (what & Set_Double)
1372 old = "double";
1373 else if (what & Set_Float)
1374 old = "float";
1375 else if (what & Set_Signed)
1376 old = "signed";
1377 else if (what & Set_Unsigned)
1378 old = "unsigned";
1379 else if (what & Set_Short)
1380 old = "short";
1381 else if (what & Set_Long)
1382 old = "long";
1383 else
1384 old = "long long";
1385 sparse_error(pos, "impossible combination of type specifiers: %s %s",
1386 old, show_ident(new));
1387 return;
1389 Catch_all:
1390 sparse_error(pos, "two or more data types in declaration specifiers");
1393 static struct symbol * const int_types[] =
1394 {&short_ctype, &int_ctype, &long_ctype, &llong_ctype};
1395 static struct symbol * const signed_types[] =
1396 {&sshort_ctype, &sint_ctype, &slong_ctype, &sllong_ctype,
1397 &slllong_ctype};
1398 static struct symbol * const unsigned_types[] =
1399 {&ushort_ctype, &uint_ctype, &ulong_ctype, &ullong_ctype,
1400 &ulllong_ctype};
1401 static struct symbol * const real_types[] =
1402 {&float_ctype, &double_ctype, &ldouble_ctype};
1403 static struct symbol * const char_types[] =
1404 {&char_ctype, &schar_ctype, &uchar_ctype};
1405 static struct symbol * const * const types[] = {
1406 int_types + 1, signed_types + 1, unsigned_types + 1,
1407 real_types + 1, char_types, char_types + 1, char_types + 2
1410 struct symbol *ctype_integer(int size, int want_unsigned)
1412 return types[want_unsigned ? CUInt : CInt][size];
1415 static struct token *handle_qualifiers(struct token *t, struct decl_state *ctx)
1417 while (token_type(t) == TOKEN_IDENT) {
1418 struct symbol *s = lookup_symbol(t->ident, NS_TYPEDEF);
1419 if (!s)
1420 break;
1421 if (s->type != SYM_KEYWORD)
1422 break;
1423 if (!(s->op->type & (KW_ATTRIBUTE | KW_QUALIFIER)))
1424 break;
1425 t = t->next;
1426 if (s->op->declarator)
1427 t = s->op->declarator(t, ctx);
1429 return t;
1432 static struct token *declaration_specifiers(struct token *token, struct decl_state *ctx)
1434 int seen = 0;
1435 int class = CInt;
1436 int size = 0;
1438 while (token_type(token) == TOKEN_IDENT) {
1439 struct symbol *s = lookup_symbol(token->ident,
1440 NS_TYPEDEF | NS_SYMBOL);
1441 if (!s || !(s->namespace & NS_TYPEDEF))
1442 break;
1443 if (s->type != SYM_KEYWORD) {
1444 if (seen & Set_Any)
1445 break;
1446 seen |= Set_S | Set_T;
1447 ctx->ctype.base_type = s->ctype.base_type;
1448 apply_ctype(token->pos, &s->ctype, &ctx->ctype);
1449 token = token->next;
1450 continue;
1452 if (s->op->type & KW_SPECIFIER) {
1453 if (seen & s->op->test) {
1454 specifier_conflict(token->pos,
1455 seen & s->op->test,
1456 token->ident);
1457 break;
1459 seen |= s->op->set;
1460 class += s->op->class;
1461 if (s->op->type & KW_SHORT) {
1462 size = -1;
1463 } else if (s->op->type & KW_LONG && size++) {
1464 if (class == CReal) {
1465 specifier_conflict(token->pos,
1466 Set_Vlong,
1467 &double_ident);
1468 break;
1470 seen |= Set_Vlong;
1473 token = token->next;
1474 if (s->op->declarator)
1475 token = s->op->declarator(token, ctx);
1476 if (s->op->type & KW_EXACT) {
1477 ctx->ctype.base_type = s->ctype.base_type;
1478 ctx->ctype.modifiers |= s->ctype.modifiers;
1482 if (!(seen & Set_S)) { /* not set explicitly? */
1483 struct symbol *base = &incomplete_ctype;
1484 if (seen & Set_Any)
1485 base = types[class][size];
1486 ctx->ctype.base_type = base;
1489 if (ctx->ctype.modifiers & MOD_BITWISE) {
1490 struct symbol *type;
1491 ctx->ctype.modifiers &= ~MOD_BITWISE;
1492 if (!is_int_type(ctx->ctype.base_type)) {
1493 sparse_error(token->pos, "invalid modifier");
1494 return token;
1496 type = alloc_symbol(token->pos, SYM_BASETYPE);
1497 *type = *ctx->ctype.base_type;
1498 type->ctype.modifiers &= ~MOD_SPECIFIER;
1499 type->ctype.base_type = ctx->ctype.base_type;
1500 type->type = SYM_RESTRICT;
1501 ctx->ctype.base_type = type;
1502 create_fouled(type);
1504 return token;
1507 static struct token *abstract_array_declarator(struct token *token, struct symbol *sym)
1509 struct expression *expr = NULL;
1511 if (match_idents(token, &restrict_ident, &__restrict_ident, NULL))
1512 token = token->next;
1513 token = parse_expression(token, &expr);
1514 sym->array_size = expr;
1515 return token;
1518 static struct token *parameter_type_list(struct token *, struct symbol *);
1519 static struct token *identifier_list(struct token *, struct symbol *);
1520 static struct token *declarator(struct token *token, struct decl_state *ctx);
1522 static struct token *skip_attribute(struct token *token)
1524 token = token->next;
1525 if (match_op(token, '(')) {
1526 int depth = 1;
1527 token = token->next;
1528 while (depth && !eof_token(token)) {
1529 if (token_type(token) == TOKEN_SPECIAL) {
1530 if (token->special == '(')
1531 depth++;
1532 else if (token->special == ')')
1533 depth--;
1535 token = token->next;
1538 return token;
1541 static struct token *skip_attributes(struct token *token)
1543 struct symbol *keyword;
1544 for (;;) {
1545 if (token_type(token) != TOKEN_IDENT)
1546 break;
1547 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1548 if (!keyword || keyword->type != SYM_KEYWORD)
1549 break;
1550 if (!(keyword->op->type & KW_ATTRIBUTE))
1551 break;
1552 token = expect(token->next, '(', "after attribute");
1553 token = expect(token, '(', "after attribute");
1554 for (;;) {
1555 if (eof_token(token))
1556 break;
1557 if (match_op(token, ';'))
1558 break;
1559 if (token_type(token) != TOKEN_IDENT)
1560 break;
1561 token = skip_attribute(token);
1562 if (!match_op(token, ','))
1563 break;
1564 token = token->next;
1566 token = expect(token, ')', "after attribute");
1567 token = expect(token, ')', "after attribute");
1569 return token;
1572 static struct token *handle_attributes(struct token *token, struct decl_state *ctx, unsigned int keywords)
1574 struct symbol *keyword;
1575 for (;;) {
1576 if (token_type(token) != TOKEN_IDENT)
1577 break;
1578 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1579 if (!keyword || keyword->type != SYM_KEYWORD)
1580 break;
1581 if (!(keyword->op->type & keywords))
1582 break;
1583 token = keyword->op->declarator(token->next, ctx);
1584 keywords &= KW_ATTRIBUTE;
1586 return token;
1589 static int is_nested(struct token *token, struct token **p,
1590 int prefer_abstract)
1593 * This can be either a parameter list or a grouping.
1594 * For the direct (non-abstract) case, we know if must be
1595 * a parameter list if we already saw the identifier.
1596 * For the abstract case, we know if must be a parameter
1597 * list if it is empty or starts with a type.
1599 struct token *next = token->next;
1601 *p = next = skip_attributes(next);
1603 if (token_type(next) == TOKEN_IDENT) {
1604 if (lookup_type(next))
1605 return !prefer_abstract;
1606 return 1;
1609 if (match_op(next, ')') || match_op(next, SPECIAL_ELLIPSIS))
1610 return 0;
1612 return 1;
1615 enum kind {
1616 Empty, K_R, Proto, Bad_Func,
1619 static enum kind which_func(struct token *token,
1620 struct ident **n,
1621 int prefer_abstract)
1623 struct token *next = token->next;
1625 if (token_type(next) == TOKEN_IDENT) {
1626 if (lookup_type(next))
1627 return Proto;
1628 /* identifier list not in definition; complain */
1629 if (prefer_abstract)
1630 warning(token->pos,
1631 "identifier list not in definition");
1632 return K_R;
1635 if (token_type(next) != TOKEN_SPECIAL)
1636 return Bad_Func;
1638 if (next->special == ')') {
1639 /* don't complain about those */
1640 if (!n || match_op(next->next, ';'))
1641 return Empty;
1642 warning(next->pos,
1643 "non-ANSI function declaration of function '%s'",
1644 show_ident(*n));
1645 return Empty;
1648 if (next->special == SPECIAL_ELLIPSIS) {
1649 warning(next->pos,
1650 "variadic functions must have one named argument");
1651 return Proto;
1654 return Bad_Func;
1657 static struct token *direct_declarator(struct token *token, struct decl_state *ctx)
1659 struct ctype *ctype = &ctx->ctype;
1660 struct token *next;
1661 struct ident **p = ctx->ident;
1663 if (ctx->ident && token_type(token) == TOKEN_IDENT) {
1664 *ctx->ident = token->ident;
1665 token = token->next;
1666 } else if (match_op(token, '(') &&
1667 is_nested(token, &next, ctx->prefer_abstract)) {
1668 struct symbol *base_type = ctype->base_type;
1669 if (token->next != next)
1670 next = handle_attributes(token->next, ctx,
1671 KW_ATTRIBUTE);
1672 token = declarator(next, ctx);
1673 token = expect(token, ')', "in nested declarator");
1674 while (ctype->base_type != base_type)
1675 ctype = &ctype->base_type->ctype;
1676 p = NULL;
1679 if (match_op(token, '(')) {
1680 enum kind kind = which_func(token, p, ctx->prefer_abstract);
1681 struct symbol *fn;
1682 fn = alloc_indirect_symbol(token->pos, ctype, SYM_FN);
1683 token = token->next;
1684 if (kind == K_R)
1685 token = identifier_list(token, fn);
1686 else if (kind == Proto)
1687 token = parameter_type_list(token, fn);
1688 token = expect(token, ')', "in function declarator");
1689 fn->endpos = token->pos;
1690 return token;
1693 while (match_op(token, '[')) {
1694 struct symbol *array;
1695 array = alloc_indirect_symbol(token->pos, ctype, SYM_ARRAY);
1696 token = abstract_array_declarator(token->next, array);
1697 token = expect(token, ']', "in abstract_array_declarator");
1698 array->endpos = token->pos;
1699 ctype = &array->ctype;
1701 return token;
1704 static struct token *pointer(struct token *token, struct decl_state *ctx)
1706 while (match_op(token,'*')) {
1707 struct symbol *ptr = alloc_symbol(token->pos, SYM_PTR);
1708 ptr->ctype.modifiers = ctx->ctype.modifiers;
1709 ptr->ctype.base_type = ctx->ctype.base_type;
1710 ptr->ctype.as = ctx->ctype.as;
1711 ptr->ctype.contexts = ctx->ctype.contexts;
1712 ctx->ctype.modifiers = 0;
1713 ctx->ctype.base_type = ptr;
1714 ctx->ctype.as = 0;
1715 ctx->ctype.contexts = NULL;
1716 ctx->ctype.alignment = 0;
1718 token = handle_qualifiers(token->next, ctx);
1719 ctx->ctype.base_type->endpos = token->pos;
1721 return token;
1724 static struct token *declarator(struct token *token, struct decl_state *ctx)
1726 token = pointer(token, ctx);
1727 return direct_declarator(token, ctx);
1730 static struct token *handle_bitfield(struct token *token, struct decl_state *ctx)
1732 struct ctype *ctype = &ctx->ctype;
1733 struct expression *expr;
1734 struct symbol *bitfield;
1735 long long width;
1737 if (ctype->base_type != &int_type && !is_int_type(ctype->base_type)) {
1738 sparse_error(token->pos, "invalid bitfield specifier for type %s.",
1739 show_typename(ctype->base_type));
1740 // Parse this to recover gracefully.
1741 return conditional_expression(token->next, &expr);
1744 bitfield = alloc_indirect_symbol(token->pos, ctype, SYM_BITFIELD);
1745 token = conditional_expression(token->next, &expr);
1746 width = const_expression_value(expr);
1747 bitfield->bit_size = width;
1749 if (width < 0 || width > INT_MAX) {
1750 sparse_error(token->pos, "invalid bitfield width, %lld.", width);
1751 width = -1;
1752 } else if (*ctx->ident && width == 0) {
1753 sparse_error(token->pos, "invalid named zero-width bitfield `%s'",
1754 show_ident(*ctx->ident));
1755 width = -1;
1756 } else if (*ctx->ident) {
1757 struct symbol *base_type = bitfield->ctype.base_type;
1758 struct symbol *bitfield_type = base_type == &int_type ? bitfield : base_type;
1759 int is_signed = !(bitfield_type->ctype.modifiers & MOD_UNSIGNED);
1760 if (Wone_bit_signed_bitfield && width == 1 && is_signed) {
1761 // Valid values are either {-1;0} or {0}, depending on integer
1762 // representation. The latter makes for very efficient code...
1763 sparse_error(token->pos, "dubious one-bit signed bitfield");
1765 if (Wdefault_bitfield_sign &&
1766 bitfield_type->type != SYM_ENUM &&
1767 !(bitfield_type->ctype.modifiers & MOD_EXPLICITLY_SIGNED) &&
1768 is_signed) {
1769 // The sign of bitfields is unspecified by default.
1770 warning(token->pos, "dubious bitfield without explicit `signed' or `unsigned'");
1773 bitfield->bit_size = width;
1774 bitfield->endpos = token->pos;
1775 return token;
1778 static struct token *declaration_list(struct token *token, struct symbol_list **list)
1780 struct decl_state ctx = {.prefer_abstract = 0};
1781 struct ctype saved;
1782 unsigned long mod;
1784 token = declaration_specifiers(token, &ctx);
1785 mod = storage_modifiers(&ctx);
1786 saved = ctx.ctype;
1787 for (;;) {
1788 struct symbol *decl = alloc_symbol(token->pos, SYM_NODE);
1789 ctx.ident = &decl->ident;
1791 token = declarator(token, &ctx);
1792 if (match_op(token, ':'))
1793 token = handle_bitfield(token, &ctx);
1795 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
1796 apply_modifiers(token->pos, &ctx);
1798 decl->ctype = ctx.ctype;
1799 decl->ctype.modifiers |= mod;
1800 decl->endpos = token->pos;
1801 add_symbol(list, decl);
1802 if (!match_op(token, ','))
1803 break;
1804 token = token->next;
1805 ctx.ctype = saved;
1807 return token;
1810 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list)
1812 while (!match_op(token, '}')) {
1813 if (!match_op(token, ';'))
1814 token = declaration_list(token, list);
1815 if (!match_op(token, ';')) {
1816 sparse_error(token->pos, "expected ; at end of declaration");
1817 break;
1819 token = token->next;
1821 return token;
1824 static struct token *parameter_declaration(struct token *token, struct symbol *sym)
1826 struct decl_state ctx = {.prefer_abstract = 1};
1828 token = declaration_specifiers(token, &ctx);
1829 ctx.ident = &sym->ident;
1830 token = declarator(token, &ctx);
1831 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
1832 apply_modifiers(token->pos, &ctx);
1833 sym->ctype = ctx.ctype;
1834 sym->ctype.modifiers |= storage_modifiers(&ctx);
1835 sym->endpos = token->pos;
1836 return token;
1839 struct token *typename(struct token *token, struct symbol **p, int *forced)
1841 struct decl_state ctx = {.prefer_abstract = 1};
1842 int class;
1843 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
1844 *p = sym;
1845 token = declaration_specifiers(token, &ctx);
1846 token = declarator(token, &ctx);
1847 apply_modifiers(token->pos, &ctx);
1848 sym->ctype = ctx.ctype;
1849 sym->endpos = token->pos;
1850 class = ctx.storage_class;
1851 if (forced) {
1852 *forced = 0;
1853 if (class == SForced) {
1854 *forced = 1;
1855 class = 0;
1858 if (class)
1859 warning(sym->pos, "storage class in typename (%s %s)",
1860 storage_class[class], show_typename(sym));
1861 return token;
1864 static struct token *expression_statement(struct token *token, struct expression **tree)
1866 token = parse_expression(token, tree);
1867 return expect(token, ';', "at end of statement");
1870 static struct token *parse_asm_operands(struct token *token, struct statement *stmt,
1871 struct expression_list **inout)
1873 struct expression *expr;
1875 /* Allow empty operands */
1876 if (match_op(token->next, ':') || match_op(token->next, ')'))
1877 return token->next;
1878 do {
1879 struct ident *ident = NULL;
1880 if (match_op(token->next, '[') &&
1881 token_type(token->next->next) == TOKEN_IDENT &&
1882 match_op(token->next->next->next, ']')) {
1883 ident = token->next->next->ident;
1884 token = token->next->next->next;
1886 add_expression(inout, (struct expression *)ident); /* UGGLEE!!! */
1887 token = primary_expression(token->next, &expr);
1888 add_expression(inout, expr);
1889 token = parens_expression(token, &expr, "in asm parameter");
1890 add_expression(inout, expr);
1891 } while (match_op(token, ','));
1892 return token;
1895 static struct token *parse_asm_clobbers(struct token *token, struct statement *stmt,
1896 struct expression_list **clobbers)
1898 struct expression *expr;
1900 do {
1901 token = primary_expression(token->next, &expr);
1902 if (expr)
1903 add_expression(clobbers, expr);
1904 } while (match_op(token, ','));
1905 return token;
1908 static struct token *parse_asm_labels(struct token *token, struct statement *stmt,
1909 struct symbol_list **labels)
1911 struct symbol *label;
1913 do {
1914 token = token->next; /* skip ':' and ',' */
1915 if (token_type(token) != TOKEN_IDENT)
1916 return token;
1917 label = label_symbol(token);
1918 add_symbol(labels, label);
1919 token = token->next;
1920 } while (match_op(token, ','));
1921 return token;
1924 static struct token *parse_asm_statement(struct token *token, struct statement *stmt)
1926 int is_goto = 0;
1928 token = token->next;
1929 stmt->type = STMT_ASM;
1930 if (match_idents(token, &__volatile___ident, &__volatile_ident, &volatile_ident, NULL)) {
1931 token = token->next;
1933 if (token_type(token) == TOKEN_IDENT && token->ident == &goto_ident) {
1934 is_goto = 1;
1935 token = token->next;
1937 token = expect(token, '(', "after asm");
1938 token = parse_expression(token, &stmt->asm_string);
1939 if (match_op(token, ':'))
1940 token = parse_asm_operands(token, stmt, &stmt->asm_outputs);
1941 if (match_op(token, ':'))
1942 token = parse_asm_operands(token, stmt, &stmt->asm_inputs);
1943 if (match_op(token, ':'))
1944 token = parse_asm_clobbers(token, stmt, &stmt->asm_clobbers);
1945 if (is_goto && match_op(token, ':'))
1946 token = parse_asm_labels(token, stmt, &stmt->asm_labels);
1947 token = expect(token, ')', "after asm");
1948 return expect(token, ';', "at end of asm-statement");
1951 static struct token *parse_asm_declarator(struct token *token, struct decl_state *ctx)
1953 struct expression *expr;
1954 token = expect(token, '(', "after asm");
1955 token = parse_expression(token->next, &expr);
1956 token = expect(token, ')', "after asm");
1957 return token;
1960 /* Make a statement out of an expression */
1961 static struct statement *make_statement(struct expression *expr)
1963 struct statement *stmt;
1965 if (!expr)
1966 return NULL;
1967 stmt = alloc_statement(expr->pos, STMT_EXPRESSION);
1968 stmt->expression = expr;
1969 return stmt;
1973 * All iterators have two symbols associated with them:
1974 * the "continue" and "break" symbols, which are targets
1975 * for continue and break statements respectively.
1977 * They are in a special name-space, but they follow
1978 * all the normal visibility rules, so nested iterators
1979 * automatically work right.
1981 static void start_iterator(struct statement *stmt)
1983 struct symbol *cont, *brk;
1985 start_symbol_scope(stmt->pos);
1986 cont = alloc_symbol(stmt->pos, SYM_NODE);
1987 bind_symbol(cont, &continue_ident, NS_ITERATOR);
1988 brk = alloc_symbol(stmt->pos, SYM_NODE);
1989 bind_symbol(brk, &break_ident, NS_ITERATOR);
1991 stmt->type = STMT_ITERATOR;
1992 stmt->iterator_break = brk;
1993 stmt->iterator_continue = cont;
1994 fn_local_symbol(brk);
1995 fn_local_symbol(cont);
1998 static void end_iterator(struct statement *stmt)
2000 end_symbol_scope();
2003 static struct statement *start_function(struct symbol *sym)
2005 struct symbol *ret;
2006 struct statement *stmt = alloc_statement(sym->pos, STMT_COMPOUND);
2008 start_function_scope(stmt->pos);
2009 ret = alloc_symbol(sym->pos, SYM_NODE);
2010 ret->ctype = sym->ctype.base_type->ctype;
2011 ret->ctype.modifiers &= ~(MOD_STORAGE | MOD_CONST | MOD_VOLATILE | MOD_TLS | MOD_INLINE | MOD_ADDRESSABLE | MOD_NOCAST | MOD_NODEREF | MOD_ACCESSED | MOD_TOPLEVEL);
2012 ret->ctype.modifiers |= (MOD_AUTO | MOD_REGISTER);
2013 bind_symbol(ret, &return_ident, NS_ITERATOR);
2014 stmt->ret = ret;
2015 fn_local_symbol(ret);
2017 // Currently parsed symbol for __func__/__FUNCTION__/__PRETTY_FUNCTION__
2018 current_fn = sym;
2020 return stmt;
2023 static void end_function(struct symbol *sym)
2025 current_fn = NULL;
2026 end_function_scope();
2030 * A "switch()" statement, like an iterator, has a
2031 * the "break" symbol associated with it. It works
2032 * exactly like the iterator break - it's the target
2033 * for any break-statements in scope, and means that
2034 * "break" handling doesn't even need to know whether
2035 * it's breaking out of an iterator or a switch.
2037 * In addition, the "case" symbol is a marker for the
2038 * case/default statements to find the switch statement
2039 * that they are associated with.
2041 static void start_switch(struct statement *stmt)
2043 struct symbol *brk, *switch_case;
2045 start_symbol_scope(stmt->pos);
2046 brk = alloc_symbol(stmt->pos, SYM_NODE);
2047 bind_symbol(brk, &break_ident, NS_ITERATOR);
2049 switch_case = alloc_symbol(stmt->pos, SYM_NODE);
2050 bind_symbol(switch_case, &case_ident, NS_ITERATOR);
2051 switch_case->stmt = stmt;
2053 stmt->type = STMT_SWITCH;
2054 stmt->switch_break = brk;
2055 stmt->switch_case = switch_case;
2057 fn_local_symbol(brk);
2058 fn_local_symbol(switch_case);
2061 static void end_switch(struct statement *stmt)
2063 if (!stmt->switch_case->symbol_list)
2064 warning(stmt->pos, "switch with no cases");
2065 end_symbol_scope();
2068 static void add_case_statement(struct statement *stmt)
2070 struct symbol *target = lookup_symbol(&case_ident, NS_ITERATOR);
2071 struct symbol *sym;
2073 if (!target) {
2074 sparse_error(stmt->pos, "not in switch scope");
2075 stmt->type = STMT_NONE;
2076 return;
2078 sym = alloc_symbol(stmt->pos, SYM_NODE);
2079 add_symbol(&target->symbol_list, sym);
2080 sym->stmt = stmt;
2081 stmt->case_label = sym;
2082 fn_local_symbol(sym);
2085 static struct token *parse_return_statement(struct token *token, struct statement *stmt)
2087 struct symbol *target = lookup_symbol(&return_ident, NS_ITERATOR);
2089 if (!target)
2090 error_die(token->pos, "internal error: return without a function target");
2091 stmt->type = STMT_RETURN;
2092 stmt->ret_target = target;
2093 return expression_statement(token->next, &stmt->ret_value);
2096 static struct token *parse_for_statement(struct token *token, struct statement *stmt)
2098 struct symbol_list *syms;
2099 struct expression *e1, *e2, *e3;
2100 struct statement *iterator;
2102 start_iterator(stmt);
2103 token = expect(token->next, '(', "after 'for'");
2105 syms = NULL;
2106 e1 = NULL;
2107 /* C99 variable declaration? */
2108 if (lookup_type(token)) {
2109 token = external_declaration(token, &syms);
2110 } else {
2111 token = parse_expression(token, &e1);
2112 token = expect(token, ';', "in 'for'");
2114 token = parse_expression(token, &e2);
2115 token = expect(token, ';', "in 'for'");
2116 token = parse_expression(token, &e3);
2117 token = expect(token, ')', "in 'for'");
2118 token = statement(token, &iterator);
2120 stmt->iterator_syms = syms;
2121 stmt->iterator_pre_statement = make_statement(e1);
2122 stmt->iterator_pre_condition = e2;
2123 stmt->iterator_post_statement = make_statement(e3);
2124 stmt->iterator_post_condition = NULL;
2125 stmt->iterator_statement = iterator;
2126 end_iterator(stmt);
2128 return token;
2131 static struct token *parse_while_statement(struct token *token, struct statement *stmt)
2133 struct expression *expr;
2134 struct statement *iterator;
2136 start_iterator(stmt);
2137 token = parens_expression(token->next, &expr, "after 'while'");
2138 token = statement(token, &iterator);
2140 stmt->iterator_pre_condition = expr;
2141 stmt->iterator_post_condition = NULL;
2142 stmt->iterator_statement = iterator;
2143 end_iterator(stmt);
2145 return token;
2148 static struct token *parse_do_statement(struct token *token, struct statement *stmt)
2150 struct expression *expr;
2151 struct statement *iterator;
2153 start_iterator(stmt);
2154 token = statement(token->next, &iterator);
2155 if (token_type(token) == TOKEN_IDENT && token->ident == &while_ident)
2156 token = token->next;
2157 else
2158 sparse_error(token->pos, "expected 'while' after 'do'");
2159 token = parens_expression(token, &expr, "after 'do-while'");
2161 stmt->iterator_post_condition = expr;
2162 stmt->iterator_statement = iterator;
2163 end_iterator(stmt);
2165 if (iterator && iterator->type != STMT_COMPOUND && Wdo_while)
2166 warning(iterator->pos, "do-while statement is not a compound statement");
2168 return expect(token, ';', "after statement");
2171 static struct token *parse_if_statement(struct token *token, struct statement *stmt)
2173 stmt->type = STMT_IF;
2174 token = parens_expression(token->next, &stmt->if_conditional, "after if");
2175 token = statement(token, &stmt->if_true);
2176 if (token_type(token) != TOKEN_IDENT)
2177 return token;
2178 if (token->ident != &else_ident)
2179 return token;
2180 return statement(token->next, &stmt->if_false);
2183 static inline struct token *case_statement(struct token *token, struct statement *stmt)
2185 stmt->type = STMT_CASE;
2186 token = expect(token, ':', "after default/case");
2187 add_case_statement(stmt);
2188 return statement(token, &stmt->case_statement);
2191 static struct token *parse_case_statement(struct token *token, struct statement *stmt)
2193 token = parse_expression(token->next, &stmt->case_expression);
2194 if (match_op(token, SPECIAL_ELLIPSIS))
2195 token = parse_expression(token->next, &stmt->case_to);
2196 return case_statement(token, stmt);
2199 static struct token *parse_default_statement(struct token *token, struct statement *stmt)
2201 return case_statement(token->next, stmt);
2204 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt)
2206 struct symbol *target = lookup_symbol(token->ident, NS_ITERATOR);
2207 stmt->type = STMT_GOTO;
2208 stmt->goto_label = target;
2209 if (!target)
2210 sparse_error(stmt->pos, "break/continue not in iterator scope");
2211 return expect(token->next, ';', "at end of statement");
2214 static struct token *parse_switch_statement(struct token *token, struct statement *stmt)
2216 stmt->type = STMT_SWITCH;
2217 start_switch(stmt);
2218 token = parens_expression(token->next, &stmt->switch_expression, "after 'switch'");
2219 token = statement(token, &stmt->switch_statement);
2220 end_switch(stmt);
2221 return token;
2224 static struct token *parse_goto_statement(struct token *token, struct statement *stmt)
2226 stmt->type = STMT_GOTO;
2227 token = token->next;
2228 if (match_op(token, '*')) {
2229 token = parse_expression(token->next, &stmt->goto_expression);
2230 add_statement(&function_computed_goto_list, stmt);
2231 } else if (token_type(token) == TOKEN_IDENT) {
2232 stmt->goto_label = label_symbol(token);
2233 token = token->next;
2234 } else {
2235 sparse_error(token->pos, "Expected identifier or goto expression");
2237 return expect(token, ';', "at end of statement");
2240 static struct token *parse_context_statement(struct token *token, struct statement *stmt)
2242 stmt->type = STMT_CONTEXT;
2243 token = parse_expression(token->next, &stmt->expression);
2244 if (stmt->expression->type == EXPR_PREOP
2245 && stmt->expression->op == '('
2246 && stmt->expression->unop->type == EXPR_COMMA) {
2247 struct expression *expr;
2248 expr = stmt->expression->unop;
2249 stmt->context = expr->left;
2250 stmt->expression = expr->right;
2252 return expect(token, ';', "at end of statement");
2255 static struct token *parse_range_statement(struct token *token, struct statement *stmt)
2257 stmt->type = STMT_RANGE;
2258 token = assignment_expression(token->next, &stmt->range_expression);
2259 token = expect(token, ',', "after range expression");
2260 token = assignment_expression(token, &stmt->range_low);
2261 token = expect(token, ',', "after low range");
2262 token = assignment_expression(token, &stmt->range_high);
2263 return expect(token, ';', "after range statement");
2266 static struct token *statement(struct token *token, struct statement **tree)
2268 struct statement *stmt = alloc_statement(token->pos, STMT_NONE);
2270 *tree = stmt;
2271 if (token_type(token) == TOKEN_IDENT) {
2272 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2273 if (s && s->op->statement)
2274 return s->op->statement(token, stmt);
2276 if (match_op(token->next, ':')) {
2277 stmt->type = STMT_LABEL;
2278 stmt->label_identifier = label_symbol(token);
2279 token = skip_attributes(token->next->next);
2280 return statement(token, &stmt->label_statement);
2284 if (match_op(token, '{')) {
2285 stmt->type = STMT_COMPOUND;
2286 start_symbol_scope(stmt->pos);
2287 token = compound_statement(token->next, stmt);
2288 end_symbol_scope();
2290 return expect(token, '}', "at end of compound statement");
2293 stmt->type = STMT_EXPRESSION;
2294 return expression_statement(token, &stmt->expression);
2297 /* gcc extension - __label__ ident-list; in the beginning of compound stmt */
2298 static struct token *label_statement(struct token *token)
2300 while (token_type(token) == TOKEN_IDENT) {
2301 struct symbol *sym = alloc_symbol(token->pos, SYM_LABEL);
2302 /* it's block-scope, but we want label namespace */
2303 bind_symbol(sym, token->ident, NS_SYMBOL);
2304 sym->namespace = NS_LABEL;
2305 fn_local_symbol(sym);
2306 token = token->next;
2307 if (!match_op(token, ','))
2308 break;
2309 token = token->next;
2311 return expect(token, ';', "at end of label declaration");
2314 static struct token * statement_list(struct token *token, struct statement_list **list)
2316 int seen_statement = 0;
2317 while (token_type(token) == TOKEN_IDENT &&
2318 token->ident == &__label___ident)
2319 token = label_statement(token->next);
2320 for (;;) {
2321 struct statement * stmt;
2322 if (eof_token(token))
2323 break;
2324 if (match_op(token, '}'))
2325 break;
2326 if (lookup_type(token)) {
2327 if (seen_statement) {
2328 warning(token->pos, "mixing declarations and code");
2329 seen_statement = 0;
2331 stmt = alloc_statement(token->pos, STMT_DECLARATION);
2332 token = external_declaration(token, &stmt->declaration);
2333 } else {
2334 seen_statement = Wdeclarationafterstatement;
2335 token = statement(token, &stmt);
2337 add_statement(list, stmt);
2339 return token;
2342 static struct token *identifier_list(struct token *token, struct symbol *fn)
2344 struct symbol_list **list = &fn->arguments;
2345 for (;;) {
2346 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2347 sym->ident = token->ident;
2348 token = token->next;
2349 sym->endpos = token->pos;
2350 sym->ctype.base_type = &incomplete_ctype;
2351 add_symbol(list, sym);
2352 if (!match_op(token, ',') ||
2353 token_type(token->next) != TOKEN_IDENT ||
2354 lookup_type(token->next))
2355 break;
2356 token = token->next;
2358 return token;
2361 static struct token *parameter_type_list(struct token *token, struct symbol *fn)
2363 struct symbol_list **list = &fn->arguments;
2365 for (;;) {
2366 struct symbol *sym;
2368 if (match_op(token, SPECIAL_ELLIPSIS)) {
2369 fn->variadic = 1;
2370 token = token->next;
2371 break;
2374 sym = alloc_symbol(token->pos, SYM_NODE);
2375 token = parameter_declaration(token, sym);
2376 if (sym->ctype.base_type == &void_ctype) {
2377 /* Special case: (void) */
2378 if (!*list && !sym->ident)
2379 break;
2380 warning(token->pos, "void parameter");
2382 add_symbol(list, sym);
2383 if (!match_op(token, ','))
2384 break;
2385 token = token->next;
2387 return token;
2390 struct token *compound_statement(struct token *token, struct statement *stmt)
2392 token = statement_list(token, &stmt->stmts);
2393 return token;
2396 static struct expression *identifier_expression(struct token *token)
2398 struct expression *expr = alloc_expression(token->pos, EXPR_IDENTIFIER);
2399 expr->expr_ident = token->ident;
2400 return expr;
2403 static struct expression *index_expression(struct expression *from, struct expression *to)
2405 int idx_from, idx_to;
2406 struct expression *expr = alloc_expression(from->pos, EXPR_INDEX);
2408 idx_from = const_expression_value(from);
2409 idx_to = idx_from;
2410 if (to) {
2411 idx_to = const_expression_value(to);
2412 if (idx_to < idx_from || idx_from < 0)
2413 warning(from->pos, "nonsense array initializer index range");
2415 expr->idx_from = idx_from;
2416 expr->idx_to = idx_to;
2417 return expr;
2420 static struct token *single_initializer(struct expression **ep, struct token *token)
2422 int expect_equal = 0;
2423 struct token *next = token->next;
2424 struct expression **tail = ep;
2425 int nested;
2427 *ep = NULL;
2429 if ((token_type(token) == TOKEN_IDENT) && match_op(next, ':')) {
2430 struct expression *expr = identifier_expression(token);
2431 if (Wold_initializer)
2432 warning(token->pos, "obsolete struct initializer, use C99 syntax");
2433 token = initializer(&expr->ident_expression, next->next);
2434 if (expr->ident_expression)
2435 *ep = expr;
2436 return token;
2439 for (tail = ep, nested = 0; ; nested++, next = token->next) {
2440 if (match_op(token, '.') && (token_type(next) == TOKEN_IDENT)) {
2441 struct expression *expr = identifier_expression(next);
2442 *tail = expr;
2443 tail = &expr->ident_expression;
2444 expect_equal = 1;
2445 token = next->next;
2446 } else if (match_op(token, '[')) {
2447 struct expression *from = NULL, *to = NULL, *expr;
2448 token = constant_expression(token->next, &from);
2449 if (!from) {
2450 sparse_error(token->pos, "Expected constant expression");
2451 break;
2453 if (match_op(token, SPECIAL_ELLIPSIS))
2454 token = constant_expression(token->next, &to);
2455 expr = index_expression(from, to);
2456 *tail = expr;
2457 tail = &expr->idx_expression;
2458 token = expect(token, ']', "at end of initializer index");
2459 if (nested)
2460 expect_equal = 1;
2461 } else {
2462 break;
2465 if (nested && !expect_equal) {
2466 if (!match_op(token, '='))
2467 warning(token->pos, "obsolete array initializer, use C99 syntax");
2468 else
2469 expect_equal = 1;
2471 if (expect_equal)
2472 token = expect(token, '=', "at end of initializer index");
2474 token = initializer(tail, token);
2475 if (!*tail)
2476 *ep = NULL;
2477 return token;
2480 static struct token *initializer_list(struct expression_list **list, struct token *token)
2482 struct expression *expr;
2484 for (;;) {
2485 token = single_initializer(&expr, token);
2486 if (!expr)
2487 break;
2488 add_expression(list, expr);
2489 if (!match_op(token, ','))
2490 break;
2491 token = token->next;
2493 return token;
2496 struct token *initializer(struct expression **tree, struct token *token)
2498 if (match_op(token, '{')) {
2499 struct expression *expr = alloc_expression(token->pos, EXPR_INITIALIZER);
2500 *tree = expr;
2501 token = initializer_list(&expr->expr_list, token->next);
2502 return expect(token, '}', "at end of initializer");
2504 return assignment_expression(token, tree);
2507 static void declare_argument(struct symbol *sym, struct symbol *fn)
2509 if (!sym->ident) {
2510 sparse_error(sym->pos, "no identifier for function argument");
2511 return;
2513 bind_symbol(sym, sym->ident, NS_SYMBOL);
2516 static struct token *parse_function_body(struct token *token, struct symbol *decl,
2517 struct symbol_list **list)
2519 struct symbol_list **old_symbol_list;
2520 struct symbol *base_type = decl->ctype.base_type;
2521 struct statement *stmt, **p;
2522 struct symbol *prev;
2523 struct symbol *arg;
2525 old_symbol_list = function_symbol_list;
2526 if (decl->ctype.modifiers & MOD_INLINE) {
2527 function_symbol_list = &decl->inline_symbol_list;
2528 p = &base_type->inline_stmt;
2529 } else {
2530 function_symbol_list = &decl->symbol_list;
2531 p = &base_type->stmt;
2533 function_computed_target_list = NULL;
2534 function_computed_goto_list = NULL;
2536 if (decl->ctype.modifiers & MOD_EXTERN) {
2537 if (!(decl->ctype.modifiers & MOD_INLINE))
2538 warning(decl->pos, "function '%s' with external linkage has definition", show_ident(decl->ident));
2540 if (!(decl->ctype.modifiers & MOD_STATIC))
2541 decl->ctype.modifiers |= MOD_EXTERN;
2543 stmt = start_function(decl);
2545 *p = stmt;
2546 FOR_EACH_PTR (base_type->arguments, arg) {
2547 declare_argument(arg, base_type);
2548 } END_FOR_EACH_PTR(arg);
2550 token = compound_statement(token->next, stmt);
2552 end_function(decl);
2553 if (!(decl->ctype.modifiers & MOD_INLINE))
2554 add_symbol(list, decl);
2555 check_declaration(decl);
2556 decl->definition = decl;
2557 prev = decl->same_symbol;
2558 if (prev && prev->definition) {
2559 warning(decl->pos, "multiple definitions for function '%s'",
2560 show_ident(decl->ident));
2561 info(prev->definition->pos, " the previous one is here");
2562 } else {
2563 while (prev) {
2564 prev->definition = decl;
2565 prev = prev->same_symbol;
2568 function_symbol_list = old_symbol_list;
2569 if (function_computed_goto_list) {
2570 if (!function_computed_target_list)
2571 warning(decl->pos, "function '%s' has computed goto but no targets?", show_ident(decl->ident));
2572 else {
2573 FOR_EACH_PTR(function_computed_goto_list, stmt) {
2574 stmt->target_list = function_computed_target_list;
2575 } END_FOR_EACH_PTR(stmt);
2578 return expect(token, '}', "at end of function");
2581 static void promote_k_r_types(struct symbol *arg)
2583 struct symbol *base = arg->ctype.base_type;
2584 if (base && base->ctype.base_type == &int_type && (base->ctype.modifiers & (MOD_CHAR | MOD_SHORT))) {
2585 arg->ctype.base_type = &int_ctype;
2589 static void apply_k_r_types(struct symbol_list *argtypes, struct symbol *fn)
2591 struct symbol_list *real_args = fn->ctype.base_type->arguments;
2592 struct symbol *arg;
2594 FOR_EACH_PTR(real_args, arg) {
2595 struct symbol *type;
2597 /* This is quadratic in the number of arguments. We _really_ don't care */
2598 FOR_EACH_PTR(argtypes, type) {
2599 if (type->ident == arg->ident)
2600 goto match;
2601 } END_FOR_EACH_PTR(type);
2602 sparse_error(arg->pos, "missing type declaration for parameter '%s'", show_ident(arg->ident));
2603 continue;
2604 match:
2605 type->used = 1;
2606 /* "char" and "short" promote to "int" */
2607 promote_k_r_types(type);
2609 arg->ctype = type->ctype;
2610 } END_FOR_EACH_PTR(arg);
2612 FOR_EACH_PTR(argtypes, arg) {
2613 if (!arg->used)
2614 warning(arg->pos, "nonsensical parameter declaration '%s'", show_ident(arg->ident));
2615 } END_FOR_EACH_PTR(arg);
2619 static struct token *parse_k_r_arguments(struct token *token, struct symbol *decl,
2620 struct symbol_list **list)
2622 struct symbol_list *args = NULL;
2624 warning(token->pos, "non-ANSI definition of function '%s'", show_ident(decl->ident));
2625 do {
2626 token = declaration_list(token, &args);
2627 if (!match_op(token, ';')) {
2628 sparse_error(token->pos, "expected ';' at end of parameter declaration");
2629 break;
2631 token = token->next;
2632 } while (lookup_type(token));
2634 apply_k_r_types(args, decl);
2636 if (!match_op(token, '{')) {
2637 sparse_error(token->pos, "expected function body");
2638 return token;
2640 return parse_function_body(token, decl, list);
2643 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list)
2645 struct symbol *anon = alloc_symbol(token->pos, SYM_NODE);
2646 struct symbol *fn = alloc_symbol(token->pos, SYM_FN);
2647 struct statement *stmt;
2649 anon->ctype.base_type = fn;
2650 stmt = alloc_statement(token->pos, STMT_NONE);
2651 fn->stmt = stmt;
2653 token = parse_asm_statement(token, stmt);
2655 add_symbol(list, anon);
2656 return token;
2659 struct token *external_declaration(struct token *token, struct symbol_list **list)
2661 struct ident *ident = NULL;
2662 struct symbol *decl;
2663 struct decl_state ctx = { .ident = &ident };
2664 struct ctype saved;
2665 struct symbol *base_type;
2666 unsigned long mod;
2667 int is_typedef;
2669 /* Top-level inline asm? */
2670 if (token_type(token) == TOKEN_IDENT) {
2671 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2672 if (s && s->op->toplevel)
2673 return s->op->toplevel(token, list);
2676 /* Parse declaration-specifiers, if any */
2677 token = declaration_specifiers(token, &ctx);
2678 mod = storage_modifiers(&ctx);
2679 decl = alloc_symbol(token->pos, SYM_NODE);
2680 /* Just a type declaration? */
2681 if (match_op(token, ';')) {
2682 apply_modifiers(token->pos, &ctx);
2683 return token->next;
2686 saved = ctx.ctype;
2687 token = declarator(token, &ctx);
2688 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
2689 apply_modifiers(token->pos, &ctx);
2691 decl->ctype = ctx.ctype;
2692 decl->ctype.modifiers |= mod;
2693 decl->endpos = token->pos;
2695 /* Just a type declaration? */
2696 if (!ident) {
2697 warning(token->pos, "missing identifier in declaration");
2698 return expect(token, ';', "at the end of type declaration");
2701 /* type define declaration? */
2702 is_typedef = ctx.storage_class == STypedef;
2704 /* Typedefs don't have meaningful storage */
2705 if (is_typedef)
2706 decl->ctype.modifiers |= MOD_USERTYPE;
2708 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
2710 base_type = decl->ctype.base_type;
2712 if (is_typedef) {
2713 if (base_type && !base_type->ident) {
2714 switch (base_type->type) {
2715 case SYM_STRUCT:
2716 case SYM_UNION:
2717 case SYM_ENUM:
2718 case SYM_RESTRICT:
2719 base_type->ident = ident;
2722 } else if (base_type && base_type->type == SYM_FN) {
2723 /* K&R argument declaration? */
2724 if (lookup_type(token))
2725 return parse_k_r_arguments(token, decl, list);
2726 if (match_op(token, '{'))
2727 return parse_function_body(token, decl, list);
2729 if (!(decl->ctype.modifiers & MOD_STATIC))
2730 decl->ctype.modifiers |= MOD_EXTERN;
2731 } else if (base_type == &void_ctype && !(decl->ctype.modifiers & MOD_EXTERN)) {
2732 sparse_error(token->pos, "void declaration");
2735 for (;;) {
2736 if (!is_typedef && match_op(token, '=')) {
2737 if (decl->ctype.modifiers & MOD_EXTERN) {
2738 warning(decl->pos, "symbol with external linkage has initializer");
2739 decl->ctype.modifiers &= ~MOD_EXTERN;
2741 token = initializer(&decl->initializer, token->next);
2743 if (!is_typedef) {
2744 if (!(decl->ctype.modifiers & (MOD_EXTERN | MOD_INLINE))) {
2745 add_symbol(list, decl);
2746 fn_local_symbol(decl);
2749 check_declaration(decl);
2750 if (decl->same_symbol)
2751 decl->definition = decl->same_symbol->definition;
2753 if (!match_op(token, ','))
2754 break;
2756 token = token->next;
2757 ident = NULL;
2758 decl = alloc_symbol(token->pos, SYM_NODE);
2759 ctx.ctype = saved;
2760 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
2761 token = declarator(token, &ctx);
2762 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
2763 apply_modifiers(token->pos, &ctx);
2764 decl->ctype = ctx.ctype;
2765 decl->ctype.modifiers |= mod;
2766 decl->endpos = token->pos;
2767 if (!ident) {
2768 sparse_error(token->pos, "expected identifier name in type definition");
2769 return token;
2772 if (is_typedef)
2773 decl->ctype.modifiers |= MOD_USERTYPE;
2775 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
2777 /* Function declarations are automatically extern unless specifically static */
2778 base_type = decl->ctype.base_type;
2779 if (!is_typedef && base_type && base_type->type == SYM_FN) {
2780 if (!(decl->ctype.modifiers & MOD_STATIC))
2781 decl->ctype.modifiers |= MOD_EXTERN;
2784 return expect(token, ';', "at end of declaration");