sparse, llvm: Fix SIGSEGV for extern symbols
[smatch.git] / parse.c
blobf8ade3e04970e7f4feab6aaac9ae4f9eb255794a
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 "leaf",
523 "__leaf__",
524 "l1_text",
525 "__l1_text__",
526 "l1_data",
527 "__l1_data__",
528 "l2",
529 "__l2__",
530 "may_alias",
531 "__may_alias__",
532 "malloc",
533 "__malloc__",
534 "may_alias",
535 "__may_alias__",
536 "model",
537 "__model__",
538 "ms_abi",
539 "__ms_abi__",
540 "ms_hook_prologue",
541 "__ms_hook_prologue__",
542 "naked",
543 "__naked__",
544 "no_instrument_function",
545 "__no_instrument_function__",
546 "noinline",
547 "__noinline__",
548 "nonnull",
549 "__nonnull",
550 "__nonnull__",
551 "nothrow",
552 "__nothrow",
553 "__nothrow__",
554 "regparm",
555 "__regparm__",
556 "section",
557 "__section__",
558 "sentinel",
559 "__sentinel__",
560 "signal",
561 "__signal__",
562 "stdcall",
563 "__stdcall__",
564 "syscall_linkage",
565 "__syscall_linkage__",
566 "sysv_abi",
567 "__sysv_abi__",
568 "unused",
569 "__unused__",
570 "used",
571 "__used__",
572 "vector_size",
573 "visibility",
574 "__visibility__",
575 "warn_unused_result",
576 "__warn_unused_result__",
577 "warning",
578 "__warning__",
579 "weak",
580 "__weak__",
584 void init_parser(int stream)
586 int i;
587 for (i = 0; i < ARRAY_SIZE(keyword_table); i++) {
588 struct init_keyword *ptr = keyword_table + i;
589 struct symbol *sym = create_symbol(stream, ptr->name, SYM_KEYWORD, ptr->ns);
590 sym->ident->keyword = 1;
591 if (ptr->ns == NS_TYPEDEF)
592 sym->ident->reserved = 1;
593 sym->ctype.modifiers = ptr->modifiers;
594 sym->ctype.base_type = ptr->type;
595 sym->op = ptr->op;
598 for (i = 0; i < ARRAY_SIZE(ignored_attributes); i++) {
599 const char * name = ignored_attributes[i];
600 struct symbol *sym = create_symbol(stream, name, SYM_KEYWORD,
601 NS_KEYWORD);
602 sym->ident->keyword = 1;
603 sym->op = &ignore_attr_op;
608 // Add a symbol to the list of function-local symbols
609 static void fn_local_symbol(struct symbol *sym)
611 if (function_symbol_list)
612 add_symbol(function_symbol_list, sym);
615 static int SENTINEL_ATTR match_idents(struct token *token, ...)
617 va_list args;
618 struct ident * next;
620 if (token_type(token) != TOKEN_IDENT)
621 return 0;
623 va_start(args, token);
624 do {
625 next = va_arg(args, struct ident *);
626 } while (next && token->ident != next);
627 va_end(args);
629 return next && token->ident == next;
633 struct statement *alloc_statement(struct position pos, int type)
635 struct statement *stmt = __alloc_statement(0);
636 stmt->type = type;
637 stmt->pos = pos;
638 return stmt;
641 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list);
643 static void apply_modifiers(struct position pos, struct decl_state *ctx)
645 struct symbol *ctype;
646 if (!ctx->mode)
647 return;
648 ctype = ctx->mode->to_mode(ctx->ctype.base_type);
649 if (!ctype)
650 sparse_error(pos, "don't know how to apply mode to %s",
651 show_typename(ctx->ctype.base_type));
652 else
653 ctx->ctype.base_type = ctype;
657 static struct symbol * alloc_indirect_symbol(struct position pos, struct ctype *ctype, int type)
659 struct symbol *sym = alloc_symbol(pos, type);
661 sym->ctype.base_type = ctype->base_type;
662 sym->ctype.modifiers = ctype->modifiers;
664 ctype->base_type = sym;
665 ctype->modifiers = 0;
666 return sym;
670 * NOTE! NS_LABEL is not just a different namespace,
671 * it also ends up using function scope instead of the
672 * regular symbol scope.
674 struct symbol *label_symbol(struct token *token)
676 struct symbol *sym = lookup_symbol(token->ident, NS_LABEL);
677 if (!sym) {
678 sym = alloc_symbol(token->pos, SYM_LABEL);
679 bind_symbol(sym, token->ident, NS_LABEL);
680 fn_local_symbol(sym);
682 return sym;
685 static struct token *struct_union_enum_specifier(enum type type,
686 struct token *token, struct decl_state *ctx,
687 struct token *(*parse)(struct token *, struct symbol *))
689 struct symbol *sym;
690 struct position *repos;
692 token = handle_attributes(token, ctx, KW_ATTRIBUTE);
693 if (token_type(token) == TOKEN_IDENT) {
694 sym = lookup_symbol(token->ident, NS_STRUCT);
695 if (!sym ||
696 (is_outer_scope(sym->scope) &&
697 (match_op(token->next,';') || match_op(token->next,'{')))) {
698 // Either a new symbol, or else an out-of-scope
699 // symbol being redefined.
700 sym = alloc_symbol(token->pos, type);
701 bind_symbol(sym, token->ident, NS_STRUCT);
703 if (sym->type != type)
704 error_die(token->pos, "invalid tag applied to %s", show_typename (sym));
705 ctx->ctype.base_type = sym;
706 repos = &token->pos;
707 token = token->next;
708 if (match_op(token, '{')) {
709 // The following test is actually wrong for empty
710 // structs, but (1) they are not C99, (2) gcc does
711 // the same thing, and (3) it's easier.
712 if (sym->symbol_list)
713 error_die(token->pos, "redefinition of %s", show_typename (sym));
714 sym->pos = *repos;
715 token = parse(token->next, sym);
716 token = expect(token, '}', "at end of struct-union-enum-specifier");
718 // Mark the structure as needing re-examination
719 sym->examined = 0;
720 sym->endpos = token->pos;
722 return token;
725 // private struct/union/enum type
726 if (!match_op(token, '{')) {
727 sparse_error(token->pos, "expected declaration");
728 ctx->ctype.base_type = &bad_ctype;
729 return token;
732 sym = alloc_symbol(token->pos, type);
733 token = parse(token->next, sym);
734 ctx->ctype.base_type = sym;
735 token = expect(token, '}', "at end of specifier");
736 sym->endpos = token->pos;
738 return token;
741 static struct token *parse_struct_declaration(struct token *token, struct symbol *sym)
743 struct symbol *field, *last = NULL;
744 struct token *res;
745 res = struct_declaration_list(token, &sym->symbol_list);
746 FOR_EACH_PTR(sym->symbol_list, field) {
747 if (!field->ident) {
748 struct symbol *base = field->ctype.base_type;
749 if (base && base->type == SYM_BITFIELD)
750 continue;
752 if (last)
753 last->next_subobject = field;
754 last = field;
755 } END_FOR_EACH_PTR(field);
756 return res;
759 static struct token *parse_union_declaration(struct token *token, struct symbol *sym)
761 return struct_declaration_list(token, &sym->symbol_list);
764 static struct token *struct_specifier(struct token *token, struct decl_state *ctx)
766 return struct_union_enum_specifier(SYM_STRUCT, token, ctx, parse_struct_declaration);
769 static struct token *union_specifier(struct token *token, struct decl_state *ctx)
771 return struct_union_enum_specifier(SYM_UNION, token, ctx, parse_union_declaration);
775 typedef struct {
776 int x;
777 unsigned long long y;
778 } Num;
780 static void upper_boundary(Num *n, Num *v)
782 if (n->x > v->x)
783 return;
784 if (n->x < v->x) {
785 *n = *v;
786 return;
788 if (n->y < v->y)
789 n->y = v->y;
792 static void lower_boundary(Num *n, Num *v)
794 if (n->x < v->x)
795 return;
796 if (n->x > v->x) {
797 *n = *v;
798 return;
800 if (n->y > v->y)
801 n->y = v->y;
804 static int type_is_ok(struct symbol *type, Num *upper, Num *lower)
806 int shift = type->bit_size;
807 int is_unsigned = type->ctype.modifiers & MOD_UNSIGNED;
809 if (!is_unsigned)
810 shift--;
811 if (upper->x == 0 && upper->y >> shift)
812 return 0;
813 if (lower->x == 0 || (!is_unsigned && (~lower->y >> shift) == 0))
814 return 1;
815 return 0;
818 static struct symbol *bigger_enum_type(struct symbol *s1, struct symbol *s2)
820 if (s1->bit_size < s2->bit_size) {
821 s1 = s2;
822 } else if (s1->bit_size == s2->bit_size) {
823 if (s2->ctype.modifiers & MOD_UNSIGNED)
824 s1 = s2;
826 if (s1->bit_size < bits_in_int)
827 return &int_ctype;
828 return s1;
831 static void cast_enum_list(struct symbol_list *list, struct symbol *base_type)
833 struct symbol *sym;
835 FOR_EACH_PTR(list, sym) {
836 struct expression *expr = sym->initializer;
837 struct symbol *ctype;
838 if (expr->type != EXPR_VALUE)
839 continue;
840 ctype = expr->ctype;
841 if (ctype->bit_size == base_type->bit_size)
842 continue;
843 cast_value(expr, base_type, expr, ctype);
844 } END_FOR_EACH_PTR(sym);
847 static struct token *parse_enum_declaration(struct token *token, struct symbol *parent)
849 unsigned long long lastval = 0;
850 struct symbol *ctype = NULL, *base_type = NULL;
851 Num upper = {-1, 0}, lower = {1, 0};
853 parent->examined = 1;
854 parent->ctype.base_type = &int_ctype;
855 while (token_type(token) == TOKEN_IDENT) {
856 struct expression *expr = NULL;
857 struct token *next = token->next;
858 struct symbol *sym;
860 if (match_op(next, '=')) {
861 next = constant_expression(next->next, &expr);
862 lastval = get_expression_value(expr);
863 ctype = &void_ctype;
864 if (expr && expr->ctype)
865 ctype = expr->ctype;
866 } else if (!ctype) {
867 ctype = &int_ctype;
868 } else if (is_int_type(ctype)) {
869 lastval++;
870 } else {
871 error_die(token->pos, "can't increment the last enum member");
874 if (!expr) {
875 expr = alloc_expression(token->pos, EXPR_VALUE);
876 expr->value = lastval;
877 expr->ctype = ctype;
880 sym = alloc_symbol(token->pos, SYM_NODE);
881 bind_symbol(sym, token->ident, NS_SYMBOL);
882 sym->ctype.modifiers &= ~MOD_ADDRESSABLE;
883 sym->initializer = expr;
884 sym->enum_member = 1;
885 sym->ctype.base_type = parent;
886 add_ptr_list(&parent->symbol_list, sym);
888 if (base_type != &bad_ctype) {
889 if (ctype->type == SYM_NODE)
890 ctype = ctype->ctype.base_type;
891 if (ctype->type == SYM_ENUM) {
892 if (ctype == parent)
893 ctype = base_type;
894 else
895 ctype = ctype->ctype.base_type;
898 * base_type rules:
899 * - if all enums are of the same type, then
900 * the base_type is that type (two first
901 * cases)
902 * - if enums are of different types, they
903 * all have to be integer types, and the
904 * base type is at least "int_ctype".
905 * - otherwise the base_type is "bad_ctype".
907 if (!base_type) {
908 base_type = ctype;
909 } else if (ctype == base_type) {
910 /* nothing */
911 } else if (is_int_type(base_type) && is_int_type(ctype)) {
912 base_type = bigger_enum_type(base_type, ctype);
913 } else
914 base_type = &bad_ctype;
915 parent->ctype.base_type = base_type;
917 if (is_int_type(base_type)) {
918 Num v = {.y = lastval};
919 if (ctype->ctype.modifiers & MOD_UNSIGNED)
920 v.x = 0;
921 else if ((long long)lastval >= 0)
922 v.x = 0;
923 else
924 v.x = -1;
925 upper_boundary(&upper, &v);
926 lower_boundary(&lower, &v);
928 token = next;
930 sym->endpos = token->pos;
932 if (!match_op(token, ','))
933 break;
934 token = token->next;
936 if (!base_type) {
937 sparse_error(token->pos, "bad enum definition");
938 base_type = &bad_ctype;
940 else if (!is_int_type(base_type))
941 base_type = base_type;
942 else if (type_is_ok(base_type, &upper, &lower))
943 base_type = base_type;
944 else if (type_is_ok(&int_ctype, &upper, &lower))
945 base_type = &int_ctype;
946 else if (type_is_ok(&uint_ctype, &upper, &lower))
947 base_type = &uint_ctype;
948 else if (type_is_ok(&long_ctype, &upper, &lower))
949 base_type = &long_ctype;
950 else if (type_is_ok(&ulong_ctype, &upper, &lower))
951 base_type = &ulong_ctype;
952 else if (type_is_ok(&llong_ctype, &upper, &lower))
953 base_type = &llong_ctype;
954 else if (type_is_ok(&ullong_ctype, &upper, &lower))
955 base_type = &ullong_ctype;
956 else
957 base_type = &bad_ctype;
958 parent->ctype.base_type = base_type;
959 parent->ctype.modifiers |= (base_type->ctype.modifiers & MOD_UNSIGNED);
960 parent->examined = 0;
962 cast_enum_list(parent->symbol_list, base_type);
964 return token;
967 static struct token *enum_specifier(struct token *token, struct decl_state *ctx)
969 struct token *ret = struct_union_enum_specifier(SYM_ENUM, token, ctx, parse_enum_declaration);
970 struct ctype *ctype = &ctx->ctype.base_type->ctype;
972 if (!ctype->base_type)
973 ctype->base_type = &incomplete_ctype;
975 return ret;
978 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype);
980 static struct token *typeof_specifier(struct token *token, struct decl_state *ctx)
982 struct symbol *sym;
984 if (!match_op(token, '(')) {
985 sparse_error(token->pos, "expected '(' after typeof");
986 return token;
988 if (lookup_type(token->next)) {
989 token = typename(token->next, &sym, NULL);
990 ctx->ctype.base_type = sym->ctype.base_type;
991 apply_ctype(token->pos, &sym->ctype, &ctx->ctype);
992 } else {
993 struct symbol *typeof_sym = alloc_symbol(token->pos, SYM_TYPEOF);
994 token = parse_expression(token->next, &typeof_sym->initializer);
996 typeof_sym->endpos = token->pos;
997 if (!typeof_sym->initializer) {
998 sparse_error(token->pos, "expected expression after the '(' token");
999 typeof_sym = &bad_ctype;
1001 ctx->ctype.base_type = typeof_sym;
1003 return expect(token, ')', "after typeof");
1006 static struct token *ignore_attribute(struct token *token, struct symbol *attr, struct decl_state *ctx)
1008 struct expression *expr = NULL;
1009 if (match_op(token, '('))
1010 token = parens_expression(token, &expr, "in attribute");
1011 return token;
1014 static struct token *attribute_packed(struct token *token, struct symbol *attr, struct decl_state *ctx)
1016 if (!ctx->ctype.alignment)
1017 ctx->ctype.alignment = 1;
1018 return token;
1021 static struct token *attribute_aligned(struct token *token, struct symbol *attr, struct decl_state *ctx)
1023 int alignment = max_alignment;
1024 struct expression *expr = NULL;
1026 if (match_op(token, '(')) {
1027 token = parens_expression(token, &expr, "in attribute");
1028 if (expr)
1029 alignment = const_expression_value(expr);
1031 if (alignment & (alignment-1)) {
1032 warning(token->pos, "I don't like non-power-of-2 alignments");
1033 return token;
1034 } else if (alignment > ctx->ctype.alignment)
1035 ctx->ctype.alignment = alignment;
1036 return token;
1039 static void apply_qualifier(struct position *pos, struct ctype *ctx, unsigned long qual)
1041 if (ctx->modifiers & qual)
1042 warning(*pos, "duplicate %s", modifier_string(qual));
1043 ctx->modifiers |= qual;
1046 static struct token *attribute_modifier(struct token *token, struct symbol *attr, struct decl_state *ctx)
1048 apply_qualifier(&token->pos, &ctx->ctype, attr->ctype.modifiers);
1049 return token;
1052 static struct token *attribute_address_space(struct token *token, struct symbol *attr, struct decl_state *ctx)
1054 struct expression *expr = NULL;
1055 int as;
1056 token = expect(token, '(', "after address_space attribute");
1057 token = conditional_expression(token, &expr);
1058 if (expr) {
1059 as = const_expression_value(expr);
1060 if (Waddress_space && as)
1061 ctx->ctype.as = as;
1063 token = expect(token, ')', "after address_space attribute");
1064 return token;
1067 static struct symbol *to_QI_mode(struct symbol *ctype)
1069 if (ctype->ctype.base_type != &int_type)
1070 return NULL;
1071 if (ctype == &char_ctype)
1072 return ctype;
1073 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uchar_ctype
1074 : &schar_ctype;
1077 static struct symbol *to_HI_mode(struct symbol *ctype)
1079 if (ctype->ctype.base_type != &int_type)
1080 return NULL;
1081 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ushort_ctype
1082 : &sshort_ctype;
1085 static struct symbol *to_SI_mode(struct symbol *ctype)
1087 if (ctype->ctype.base_type != &int_type)
1088 return NULL;
1089 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint_ctype
1090 : &sint_ctype;
1093 static struct symbol *to_DI_mode(struct symbol *ctype)
1095 if (ctype->ctype.base_type != &int_type)
1096 return NULL;
1097 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ullong_ctype
1098 : &sllong_ctype;
1101 static struct symbol *to_TI_mode(struct symbol *ctype)
1103 if (ctype->ctype.base_type != &int_type)
1104 return NULL;
1105 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulllong_ctype
1106 : &slllong_ctype;
1109 static struct symbol *to_word_mode(struct symbol *ctype)
1111 if (ctype->ctype.base_type != &int_type)
1112 return NULL;
1113 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulong_ctype
1114 : &slong_ctype;
1117 static struct token *attribute_mode(struct token *token, struct symbol *attr, struct decl_state *ctx)
1119 token = expect(token, '(', "after mode attribute");
1120 if (token_type(token) == TOKEN_IDENT) {
1121 struct symbol *mode = lookup_keyword(token->ident, NS_KEYWORD);
1122 if (mode && mode->op->type == KW_MODE)
1123 ctx->mode = mode->op;
1124 else
1125 sparse_error(token->pos, "unknown mode attribute %s\n", show_ident(token->ident));
1126 token = token->next;
1127 } else
1128 sparse_error(token->pos, "expect attribute mode symbol\n");
1129 token = expect(token, ')', "after mode attribute");
1130 return token;
1133 static struct token *attribute_context(struct token *token, struct symbol *attr, struct decl_state *ctx)
1135 struct context *context = alloc_context();
1136 struct expression *args[3];
1137 int argc = 0;
1139 token = expect(token, '(', "after context attribute");
1140 while (!match_op(token, ')')) {
1141 struct expression *expr = NULL;
1142 token = conditional_expression(token, &expr);
1143 if (!expr)
1144 break;
1145 if (argc < 3)
1146 args[argc++] = expr;
1147 if (!match_op(token, ','))
1148 break;
1149 token = token->next;
1152 switch(argc) {
1153 case 0:
1154 sparse_error(token->pos, "expected context input/output values");
1155 break;
1156 case 1:
1157 context->in = get_expression_value(args[0]);
1158 break;
1159 case 2:
1160 context->in = get_expression_value(args[0]);
1161 context->out = get_expression_value(args[1]);
1162 break;
1163 case 3:
1164 context->context = args[0];
1165 context->in = get_expression_value(args[1]);
1166 context->out = get_expression_value(args[2]);
1167 break;
1170 if (argc)
1171 add_ptr_list(&ctx->ctype.contexts, context);
1173 token = expect(token, ')', "after context attribute");
1174 return token;
1177 static struct token *attribute_designated_init(struct token *token, struct symbol *attr, struct decl_state *ctx)
1179 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_STRUCT)
1180 ctx->ctype.base_type->designated_init = 1;
1181 else
1182 warning(token->pos, "attribute designated_init applied to non-structure type");
1183 return token;
1186 static struct token *attribute_transparent_union(struct token *token, struct symbol *attr, struct decl_state *ctx)
1188 if (Wtransparent_union)
1189 warning(token->pos, "ignoring attribute __transparent_union__");
1190 return token;
1193 static struct token *recover_unknown_attribute(struct token *token)
1195 struct expression *expr = NULL;
1197 sparse_error(token->pos, "attribute '%s': unknown attribute", show_ident(token->ident));
1198 token = token->next;
1199 if (match_op(token, '('))
1200 token = parens_expression(token, &expr, "in attribute");
1201 return token;
1204 static struct token *attribute_specifier(struct token *token, struct decl_state *ctx)
1206 token = expect(token, '(', "after attribute");
1207 token = expect(token, '(', "after attribute");
1209 for (;;) {
1210 struct ident *attribute_name;
1211 struct symbol *attr;
1213 if (eof_token(token))
1214 break;
1215 if (match_op(token, ';'))
1216 break;
1217 if (token_type(token) != TOKEN_IDENT)
1218 break;
1219 attribute_name = token->ident;
1220 attr = lookup_keyword(attribute_name, NS_KEYWORD);
1221 if (attr && attr->op->attribute)
1222 token = attr->op->attribute(token->next, attr, ctx);
1223 else
1224 token = recover_unknown_attribute(token);
1226 if (!match_op(token, ','))
1227 break;
1228 token = token->next;
1231 token = expect(token, ')', "after attribute");
1232 token = expect(token, ')', "after attribute");
1233 return token;
1236 static const char *storage_class[] =
1238 [STypedef] = "typedef",
1239 [SAuto] = "auto",
1240 [SExtern] = "extern",
1241 [SStatic] = "static",
1242 [SRegister] = "register",
1243 [SForced] = "[force]"
1246 static unsigned long storage_modifiers(struct decl_state *ctx)
1248 static unsigned long mod[] =
1250 [SAuto] = MOD_AUTO,
1251 [SExtern] = MOD_EXTERN,
1252 [SStatic] = MOD_STATIC,
1253 [SRegister] = MOD_REGISTER
1255 return mod[ctx->storage_class] | (ctx->is_inline ? MOD_INLINE : 0)
1256 | (ctx->is_tls ? MOD_TLS : 0);
1259 static void set_storage_class(struct position *pos, struct decl_state *ctx, int class)
1261 /* __thread can be used alone, or with extern or static */
1262 if (ctx->is_tls && (class != SStatic && class != SExtern)) {
1263 sparse_error(*pos, "__thread can only be used alone, or with "
1264 "extern or static");
1265 return;
1268 if (!ctx->storage_class) {
1269 ctx->storage_class = class;
1270 return;
1272 if (ctx->storage_class == class)
1273 sparse_error(*pos, "duplicate %s", storage_class[class]);
1274 else
1275 sparse_error(*pos, "multiple storage classes");
1278 static struct token *typedef_specifier(struct token *next, struct decl_state *ctx)
1280 set_storage_class(&next->pos, ctx, STypedef);
1281 return next;
1284 static struct token *auto_specifier(struct token *next, struct decl_state *ctx)
1286 set_storage_class(&next->pos, ctx, SAuto);
1287 return next;
1290 static struct token *register_specifier(struct token *next, struct decl_state *ctx)
1292 set_storage_class(&next->pos, ctx, SRegister);
1293 return next;
1296 static struct token *static_specifier(struct token *next, struct decl_state *ctx)
1298 set_storage_class(&next->pos, ctx, SStatic);
1299 return next;
1302 static struct token *extern_specifier(struct token *next, struct decl_state *ctx)
1304 set_storage_class(&next->pos, ctx, SExtern);
1305 return next;
1308 static struct token *thread_specifier(struct token *next, struct decl_state *ctx)
1310 /* This GCC extension can be used alone, or with extern or static */
1311 if (!ctx->storage_class || ctx->storage_class == SStatic
1312 || ctx->storage_class == SExtern) {
1313 ctx->is_tls = 1;
1314 } else {
1315 sparse_error(next->pos, "__thread can only be used alone, or "
1316 "with extern or static");
1319 return next;
1322 static struct token *attribute_force(struct token *token, struct symbol *attr, struct decl_state *ctx)
1324 set_storage_class(&token->pos, ctx, SForced);
1325 return token;
1328 static struct token *inline_specifier(struct token *next, struct decl_state *ctx)
1330 ctx->is_inline = 1;
1331 return next;
1334 static struct token *const_qualifier(struct token *next, struct decl_state *ctx)
1336 apply_qualifier(&next->pos, &ctx->ctype, MOD_CONST);
1337 return next;
1340 static struct token *volatile_qualifier(struct token *next, struct decl_state *ctx)
1342 apply_qualifier(&next->pos, &ctx->ctype, MOD_VOLATILE);
1343 return next;
1346 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype)
1348 unsigned long mod = thistype->modifiers;
1350 if (mod)
1351 apply_qualifier(&pos, ctype, mod);
1353 /* Context */
1354 concat_ptr_list((struct ptr_list *)thistype->contexts,
1355 (struct ptr_list **)&ctype->contexts);
1357 /* Alignment */
1358 if (thistype->alignment > ctype->alignment)
1359 ctype->alignment = thistype->alignment;
1361 /* Address space */
1362 if (thistype->as)
1363 ctype->as = thistype->as;
1366 static void specifier_conflict(struct position pos, int what, struct ident *new)
1368 const char *old;
1369 if (what & (Set_S | Set_T))
1370 goto Catch_all;
1371 if (what & Set_Char)
1372 old = "char";
1373 else if (what & Set_Double)
1374 old = "double";
1375 else if (what & Set_Float)
1376 old = "float";
1377 else if (what & Set_Signed)
1378 old = "signed";
1379 else if (what & Set_Unsigned)
1380 old = "unsigned";
1381 else if (what & Set_Short)
1382 old = "short";
1383 else if (what & Set_Long)
1384 old = "long";
1385 else
1386 old = "long long";
1387 sparse_error(pos, "impossible combination of type specifiers: %s %s",
1388 old, show_ident(new));
1389 return;
1391 Catch_all:
1392 sparse_error(pos, "two or more data types in declaration specifiers");
1395 static struct symbol * const int_types[] =
1396 {&short_ctype, &int_ctype, &long_ctype, &llong_ctype};
1397 static struct symbol * const signed_types[] =
1398 {&sshort_ctype, &sint_ctype, &slong_ctype, &sllong_ctype,
1399 &slllong_ctype};
1400 static struct symbol * const unsigned_types[] =
1401 {&ushort_ctype, &uint_ctype, &ulong_ctype, &ullong_ctype,
1402 &ulllong_ctype};
1403 static struct symbol * const real_types[] =
1404 {&float_ctype, &double_ctype, &ldouble_ctype};
1405 static struct symbol * const char_types[] =
1406 {&char_ctype, &schar_ctype, &uchar_ctype};
1407 static struct symbol * const * const types[] = {
1408 int_types + 1, signed_types + 1, unsigned_types + 1,
1409 real_types + 1, char_types, char_types + 1, char_types + 2
1412 struct symbol *ctype_integer(int size, int want_unsigned)
1414 return types[want_unsigned ? CUInt : CInt][size];
1417 static struct token *handle_qualifiers(struct token *t, struct decl_state *ctx)
1419 while (token_type(t) == TOKEN_IDENT) {
1420 struct symbol *s = lookup_symbol(t->ident, NS_TYPEDEF);
1421 if (!s)
1422 break;
1423 if (s->type != SYM_KEYWORD)
1424 break;
1425 if (!(s->op->type & (KW_ATTRIBUTE | KW_QUALIFIER)))
1426 break;
1427 t = t->next;
1428 if (s->op->declarator)
1429 t = s->op->declarator(t, ctx);
1431 return t;
1434 static struct token *declaration_specifiers(struct token *token, struct decl_state *ctx)
1436 int seen = 0;
1437 int class = CInt;
1438 int size = 0;
1440 while (token_type(token) == TOKEN_IDENT) {
1441 struct symbol *s = lookup_symbol(token->ident,
1442 NS_TYPEDEF | NS_SYMBOL);
1443 if (!s || !(s->namespace & NS_TYPEDEF))
1444 break;
1445 if (s->type != SYM_KEYWORD) {
1446 if (seen & Set_Any)
1447 break;
1448 seen |= Set_S | Set_T;
1449 ctx->ctype.base_type = s->ctype.base_type;
1450 apply_ctype(token->pos, &s->ctype, &ctx->ctype);
1451 token = token->next;
1452 continue;
1454 if (s->op->type & KW_SPECIFIER) {
1455 if (seen & s->op->test) {
1456 specifier_conflict(token->pos,
1457 seen & s->op->test,
1458 token->ident);
1459 break;
1461 seen |= s->op->set;
1462 class += s->op->class;
1463 if (s->op->type & KW_SHORT) {
1464 size = -1;
1465 } else if (s->op->type & KW_LONG && size++) {
1466 if (class == CReal) {
1467 specifier_conflict(token->pos,
1468 Set_Vlong,
1469 &double_ident);
1470 break;
1472 seen |= Set_Vlong;
1475 token = token->next;
1476 if (s->op->declarator)
1477 token = s->op->declarator(token, ctx);
1478 if (s->op->type & KW_EXACT) {
1479 ctx->ctype.base_type = s->ctype.base_type;
1480 ctx->ctype.modifiers |= s->ctype.modifiers;
1484 if (!(seen & Set_S)) { /* not set explicitly? */
1485 struct symbol *base = &incomplete_ctype;
1486 if (seen & Set_Any)
1487 base = types[class][size];
1488 ctx->ctype.base_type = base;
1491 if (ctx->ctype.modifiers & MOD_BITWISE) {
1492 struct symbol *type;
1493 ctx->ctype.modifiers &= ~MOD_BITWISE;
1494 if (!is_int_type(ctx->ctype.base_type)) {
1495 sparse_error(token->pos, "invalid modifier");
1496 return token;
1498 type = alloc_symbol(token->pos, SYM_BASETYPE);
1499 *type = *ctx->ctype.base_type;
1500 type->ctype.modifiers &= ~MOD_SPECIFIER;
1501 type->ctype.base_type = ctx->ctype.base_type;
1502 type->type = SYM_RESTRICT;
1503 ctx->ctype.base_type = type;
1504 create_fouled(type);
1506 return token;
1509 static struct token *abstract_array_declarator(struct token *token, struct symbol *sym)
1511 struct expression *expr = NULL;
1513 if (match_idents(token, &restrict_ident, &__restrict_ident, NULL))
1514 token = token->next;
1515 token = parse_expression(token, &expr);
1516 sym->array_size = expr;
1517 return token;
1520 static struct token *parameter_type_list(struct token *, struct symbol *);
1521 static struct token *identifier_list(struct token *, struct symbol *);
1522 static struct token *declarator(struct token *token, struct decl_state *ctx);
1524 static struct token *skip_attribute(struct token *token)
1526 token = token->next;
1527 if (match_op(token, '(')) {
1528 int depth = 1;
1529 token = token->next;
1530 while (depth && !eof_token(token)) {
1531 if (token_type(token) == TOKEN_SPECIAL) {
1532 if (token->special == '(')
1533 depth++;
1534 else if (token->special == ')')
1535 depth--;
1537 token = token->next;
1540 return token;
1543 static struct token *skip_attributes(struct token *token)
1545 struct symbol *keyword;
1546 for (;;) {
1547 if (token_type(token) != TOKEN_IDENT)
1548 break;
1549 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1550 if (!keyword || keyword->type != SYM_KEYWORD)
1551 break;
1552 if (!(keyword->op->type & KW_ATTRIBUTE))
1553 break;
1554 token = expect(token->next, '(', "after attribute");
1555 token = expect(token, '(', "after attribute");
1556 for (;;) {
1557 if (eof_token(token))
1558 break;
1559 if (match_op(token, ';'))
1560 break;
1561 if (token_type(token) != TOKEN_IDENT)
1562 break;
1563 token = skip_attribute(token);
1564 if (!match_op(token, ','))
1565 break;
1566 token = token->next;
1568 token = expect(token, ')', "after attribute");
1569 token = expect(token, ')', "after attribute");
1571 return token;
1574 static struct token *handle_attributes(struct token *token, struct decl_state *ctx, unsigned int keywords)
1576 struct symbol *keyword;
1577 for (;;) {
1578 if (token_type(token) != TOKEN_IDENT)
1579 break;
1580 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1581 if (!keyword || keyword->type != SYM_KEYWORD)
1582 break;
1583 if (!(keyword->op->type & keywords))
1584 break;
1585 token = keyword->op->declarator(token->next, ctx);
1586 keywords &= KW_ATTRIBUTE;
1588 return token;
1591 static int is_nested(struct token *token, struct token **p,
1592 int prefer_abstract)
1595 * This can be either a parameter list or a grouping.
1596 * For the direct (non-abstract) case, we know if must be
1597 * a parameter list if we already saw the identifier.
1598 * For the abstract case, we know if must be a parameter
1599 * list if it is empty or starts with a type.
1601 struct token *next = token->next;
1603 *p = next = skip_attributes(next);
1605 if (token_type(next) == TOKEN_IDENT) {
1606 if (lookup_type(next))
1607 return !prefer_abstract;
1608 return 1;
1611 if (match_op(next, ')') || match_op(next, SPECIAL_ELLIPSIS))
1612 return 0;
1614 return 1;
1617 enum kind {
1618 Empty, K_R, Proto, Bad_Func,
1621 static enum kind which_func(struct token *token,
1622 struct ident **n,
1623 int prefer_abstract)
1625 struct token *next = token->next;
1627 if (token_type(next) == TOKEN_IDENT) {
1628 if (lookup_type(next))
1629 return Proto;
1630 /* identifier list not in definition; complain */
1631 if (prefer_abstract)
1632 warning(token->pos,
1633 "identifier list not in definition");
1634 return K_R;
1637 if (token_type(next) != TOKEN_SPECIAL)
1638 return Bad_Func;
1640 if (next->special == ')') {
1641 /* don't complain about those */
1642 if (!n || match_op(next->next, ';'))
1643 return Empty;
1644 warning(next->pos,
1645 "non-ANSI function declaration of function '%s'",
1646 show_ident(*n));
1647 return Empty;
1650 if (next->special == SPECIAL_ELLIPSIS) {
1651 warning(next->pos,
1652 "variadic functions must have one named argument");
1653 return Proto;
1656 return Bad_Func;
1659 static struct token *direct_declarator(struct token *token, struct decl_state *ctx)
1661 struct ctype *ctype = &ctx->ctype;
1662 struct token *next;
1663 struct ident **p = ctx->ident;
1665 if (ctx->ident && token_type(token) == TOKEN_IDENT) {
1666 *ctx->ident = token->ident;
1667 token = token->next;
1668 } else if (match_op(token, '(') &&
1669 is_nested(token, &next, ctx->prefer_abstract)) {
1670 struct symbol *base_type = ctype->base_type;
1671 if (token->next != next)
1672 next = handle_attributes(token->next, ctx,
1673 KW_ATTRIBUTE);
1674 token = declarator(next, ctx);
1675 token = expect(token, ')', "in nested declarator");
1676 while (ctype->base_type != base_type)
1677 ctype = &ctype->base_type->ctype;
1678 p = NULL;
1681 if (match_op(token, '(')) {
1682 enum kind kind = which_func(token, p, ctx->prefer_abstract);
1683 struct symbol *fn;
1684 fn = alloc_indirect_symbol(token->pos, ctype, SYM_FN);
1685 token = token->next;
1686 if (kind == K_R)
1687 token = identifier_list(token, fn);
1688 else if (kind == Proto)
1689 token = parameter_type_list(token, fn);
1690 token = expect(token, ')', "in function declarator");
1691 fn->endpos = token->pos;
1692 return token;
1695 while (match_op(token, '[')) {
1696 struct symbol *array;
1697 array = alloc_indirect_symbol(token->pos, ctype, SYM_ARRAY);
1698 token = abstract_array_declarator(token->next, array);
1699 token = expect(token, ']', "in abstract_array_declarator");
1700 array->endpos = token->pos;
1701 ctype = &array->ctype;
1703 return token;
1706 static struct token *pointer(struct token *token, struct decl_state *ctx)
1708 while (match_op(token,'*')) {
1709 struct symbol *ptr = alloc_symbol(token->pos, SYM_PTR);
1710 ptr->ctype.modifiers = ctx->ctype.modifiers;
1711 ptr->ctype.base_type = ctx->ctype.base_type;
1712 ptr->ctype.as = ctx->ctype.as;
1713 ptr->ctype.contexts = ctx->ctype.contexts;
1714 ctx->ctype.modifiers = 0;
1715 ctx->ctype.base_type = ptr;
1716 ctx->ctype.as = 0;
1717 ctx->ctype.contexts = NULL;
1718 ctx->ctype.alignment = 0;
1720 token = handle_qualifiers(token->next, ctx);
1721 ctx->ctype.base_type->endpos = token->pos;
1723 return token;
1726 static struct token *declarator(struct token *token, struct decl_state *ctx)
1728 token = pointer(token, ctx);
1729 return direct_declarator(token, ctx);
1732 static struct token *handle_bitfield(struct token *token, struct decl_state *ctx)
1734 struct ctype *ctype = &ctx->ctype;
1735 struct expression *expr;
1736 struct symbol *bitfield;
1737 long long width;
1739 if (ctype->base_type != &int_type && !is_int_type(ctype->base_type)) {
1740 sparse_error(token->pos, "invalid bitfield specifier for type %s.",
1741 show_typename(ctype->base_type));
1742 // Parse this to recover gracefully.
1743 return conditional_expression(token->next, &expr);
1746 bitfield = alloc_indirect_symbol(token->pos, ctype, SYM_BITFIELD);
1747 token = conditional_expression(token->next, &expr);
1748 width = const_expression_value(expr);
1749 bitfield->bit_size = width;
1751 if (width < 0 || width > INT_MAX) {
1752 sparse_error(token->pos, "invalid bitfield width, %lld.", width);
1753 width = -1;
1754 } else if (*ctx->ident && width == 0) {
1755 sparse_error(token->pos, "invalid named zero-width bitfield `%s'",
1756 show_ident(*ctx->ident));
1757 width = -1;
1758 } else if (*ctx->ident) {
1759 struct symbol *base_type = bitfield->ctype.base_type;
1760 struct symbol *bitfield_type = base_type == &int_type ? bitfield : base_type;
1761 int is_signed = !(bitfield_type->ctype.modifiers & MOD_UNSIGNED);
1762 if (Wone_bit_signed_bitfield && width == 1 && is_signed) {
1763 // Valid values are either {-1;0} or {0}, depending on integer
1764 // representation. The latter makes for very efficient code...
1765 sparse_error(token->pos, "dubious one-bit signed bitfield");
1767 if (Wdefault_bitfield_sign &&
1768 bitfield_type->type != SYM_ENUM &&
1769 !(bitfield_type->ctype.modifiers & MOD_EXPLICITLY_SIGNED) &&
1770 is_signed) {
1771 // The sign of bitfields is unspecified by default.
1772 warning(token->pos, "dubious bitfield without explicit `signed' or `unsigned'");
1775 bitfield->bit_size = width;
1776 bitfield->endpos = token->pos;
1777 return token;
1780 static struct token *declaration_list(struct token *token, struct symbol_list **list)
1782 struct decl_state ctx = {.prefer_abstract = 0};
1783 struct ctype saved;
1784 unsigned long mod;
1786 token = declaration_specifiers(token, &ctx);
1787 mod = storage_modifiers(&ctx);
1788 saved = ctx.ctype;
1789 for (;;) {
1790 struct symbol *decl = alloc_symbol(token->pos, SYM_NODE);
1791 ctx.ident = &decl->ident;
1793 token = declarator(token, &ctx);
1794 if (match_op(token, ':'))
1795 token = handle_bitfield(token, &ctx);
1797 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
1798 apply_modifiers(token->pos, &ctx);
1800 decl->ctype = ctx.ctype;
1801 decl->ctype.modifiers |= mod;
1802 decl->endpos = token->pos;
1803 add_symbol(list, decl);
1804 if (!match_op(token, ','))
1805 break;
1806 token = token->next;
1807 ctx.ctype = saved;
1809 return token;
1812 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list)
1814 while (!match_op(token, '}')) {
1815 if (!match_op(token, ';'))
1816 token = declaration_list(token, list);
1817 if (!match_op(token, ';')) {
1818 sparse_error(token->pos, "expected ; at end of declaration");
1819 break;
1821 token = token->next;
1823 return token;
1826 static struct token *parameter_declaration(struct token *token, struct symbol *sym)
1828 struct decl_state ctx = {.prefer_abstract = 1};
1830 token = declaration_specifiers(token, &ctx);
1831 ctx.ident = &sym->ident;
1832 token = declarator(token, &ctx);
1833 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
1834 apply_modifiers(token->pos, &ctx);
1835 sym->ctype = ctx.ctype;
1836 sym->ctype.modifiers |= storage_modifiers(&ctx);
1837 sym->endpos = token->pos;
1838 return token;
1841 struct token *typename(struct token *token, struct symbol **p, int *forced)
1843 struct decl_state ctx = {.prefer_abstract = 1};
1844 int class;
1845 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
1846 *p = sym;
1847 token = declaration_specifiers(token, &ctx);
1848 token = declarator(token, &ctx);
1849 apply_modifiers(token->pos, &ctx);
1850 sym->ctype = ctx.ctype;
1851 sym->endpos = token->pos;
1852 class = ctx.storage_class;
1853 if (forced) {
1854 *forced = 0;
1855 if (class == SForced) {
1856 *forced = 1;
1857 class = 0;
1860 if (class)
1861 warning(sym->pos, "storage class in typename (%s %s)",
1862 storage_class[class], show_typename(sym));
1863 return token;
1866 static struct token *expression_statement(struct token *token, struct expression **tree)
1868 token = parse_expression(token, tree);
1869 return expect(token, ';', "at end of statement");
1872 static struct token *parse_asm_operands(struct token *token, struct statement *stmt,
1873 struct expression_list **inout)
1875 struct expression *expr;
1877 /* Allow empty operands */
1878 if (match_op(token->next, ':') || match_op(token->next, ')'))
1879 return token->next;
1880 do {
1881 struct ident *ident = NULL;
1882 if (match_op(token->next, '[') &&
1883 token_type(token->next->next) == TOKEN_IDENT &&
1884 match_op(token->next->next->next, ']')) {
1885 ident = token->next->next->ident;
1886 token = token->next->next->next;
1888 add_expression(inout, (struct expression *)ident); /* UGGLEE!!! */
1889 token = primary_expression(token->next, &expr);
1890 add_expression(inout, expr);
1891 token = parens_expression(token, &expr, "in asm parameter");
1892 add_expression(inout, expr);
1893 } while (match_op(token, ','));
1894 return token;
1897 static struct token *parse_asm_clobbers(struct token *token, struct statement *stmt,
1898 struct expression_list **clobbers)
1900 struct expression *expr;
1902 do {
1903 token = primary_expression(token->next, &expr);
1904 if (expr)
1905 add_expression(clobbers, expr);
1906 } while (match_op(token, ','));
1907 return token;
1910 static struct token *parse_asm_labels(struct token *token, struct statement *stmt,
1911 struct symbol_list **labels)
1913 struct symbol *label;
1915 do {
1916 token = token->next; /* skip ':' and ',' */
1917 if (token_type(token) != TOKEN_IDENT)
1918 return token;
1919 label = label_symbol(token);
1920 add_symbol(labels, label);
1921 token = token->next;
1922 } while (match_op(token, ','));
1923 return token;
1926 static struct token *parse_asm_statement(struct token *token, struct statement *stmt)
1928 int is_goto = 0;
1930 token = token->next;
1931 stmt->type = STMT_ASM;
1932 if (match_idents(token, &__volatile___ident, &__volatile_ident, &volatile_ident, NULL)) {
1933 token = token->next;
1935 if (token_type(token) == TOKEN_IDENT && token->ident == &goto_ident) {
1936 is_goto = 1;
1937 token = token->next;
1939 token = expect(token, '(', "after asm");
1940 token = parse_expression(token, &stmt->asm_string);
1941 if (match_op(token, ':'))
1942 token = parse_asm_operands(token, stmt, &stmt->asm_outputs);
1943 if (match_op(token, ':'))
1944 token = parse_asm_operands(token, stmt, &stmt->asm_inputs);
1945 if (match_op(token, ':'))
1946 token = parse_asm_clobbers(token, stmt, &stmt->asm_clobbers);
1947 if (is_goto && match_op(token, ':'))
1948 token = parse_asm_labels(token, stmt, &stmt->asm_labels);
1949 token = expect(token, ')', "after asm");
1950 return expect(token, ';', "at end of asm-statement");
1953 static struct token *parse_asm_declarator(struct token *token, struct decl_state *ctx)
1955 struct expression *expr;
1956 token = expect(token, '(', "after asm");
1957 token = parse_expression(token->next, &expr);
1958 token = expect(token, ')', "after asm");
1959 return token;
1962 /* Make a statement out of an expression */
1963 static struct statement *make_statement(struct expression *expr)
1965 struct statement *stmt;
1967 if (!expr)
1968 return NULL;
1969 stmt = alloc_statement(expr->pos, STMT_EXPRESSION);
1970 stmt->expression = expr;
1971 return stmt;
1975 * All iterators have two symbols associated with them:
1976 * the "continue" and "break" symbols, which are targets
1977 * for continue and break statements respectively.
1979 * They are in a special name-space, but they follow
1980 * all the normal visibility rules, so nested iterators
1981 * automatically work right.
1983 static void start_iterator(struct statement *stmt)
1985 struct symbol *cont, *brk;
1987 start_symbol_scope();
1988 cont = alloc_symbol(stmt->pos, SYM_NODE);
1989 bind_symbol(cont, &continue_ident, NS_ITERATOR);
1990 brk = alloc_symbol(stmt->pos, SYM_NODE);
1991 bind_symbol(brk, &break_ident, NS_ITERATOR);
1993 stmt->type = STMT_ITERATOR;
1994 stmt->iterator_break = brk;
1995 stmt->iterator_continue = cont;
1996 fn_local_symbol(brk);
1997 fn_local_symbol(cont);
2000 static void end_iterator(struct statement *stmt)
2002 end_symbol_scope();
2005 static struct statement *start_function(struct symbol *sym)
2007 struct symbol *ret;
2008 struct statement *stmt = alloc_statement(sym->pos, STMT_COMPOUND);
2010 start_function_scope();
2011 ret = alloc_symbol(sym->pos, SYM_NODE);
2012 ret->ctype = sym->ctype.base_type->ctype;
2013 ret->ctype.modifiers &= ~(MOD_STORAGE | MOD_CONST | MOD_VOLATILE | MOD_TLS | MOD_INLINE | MOD_ADDRESSABLE | MOD_NOCAST | MOD_NODEREF | MOD_ACCESSED | MOD_TOPLEVEL);
2014 ret->ctype.modifiers |= (MOD_AUTO | MOD_REGISTER);
2015 bind_symbol(ret, &return_ident, NS_ITERATOR);
2016 stmt->ret = ret;
2017 fn_local_symbol(ret);
2019 // Currently parsed symbol for __func__/__FUNCTION__/__PRETTY_FUNCTION__
2020 current_fn = sym;
2022 return stmt;
2025 static void end_function(struct symbol *sym)
2027 current_fn = NULL;
2028 end_function_scope();
2032 * A "switch()" statement, like an iterator, has a
2033 * the "break" symbol associated with it. It works
2034 * exactly like the iterator break - it's the target
2035 * for any break-statements in scope, and means that
2036 * "break" handling doesn't even need to know whether
2037 * it's breaking out of an iterator or a switch.
2039 * In addition, the "case" symbol is a marker for the
2040 * case/default statements to find the switch statement
2041 * that they are associated with.
2043 static void start_switch(struct statement *stmt)
2045 struct symbol *brk, *switch_case;
2047 start_symbol_scope();
2048 brk = alloc_symbol(stmt->pos, SYM_NODE);
2049 bind_symbol(brk, &break_ident, NS_ITERATOR);
2051 switch_case = alloc_symbol(stmt->pos, SYM_NODE);
2052 bind_symbol(switch_case, &case_ident, NS_ITERATOR);
2053 switch_case->stmt = stmt;
2055 stmt->type = STMT_SWITCH;
2056 stmt->switch_break = brk;
2057 stmt->switch_case = switch_case;
2059 fn_local_symbol(brk);
2060 fn_local_symbol(switch_case);
2063 static void end_switch(struct statement *stmt)
2065 if (!stmt->switch_case->symbol_list)
2066 warning(stmt->pos, "switch with no cases");
2067 end_symbol_scope();
2070 static void add_case_statement(struct statement *stmt)
2072 struct symbol *target = lookup_symbol(&case_ident, NS_ITERATOR);
2073 struct symbol *sym;
2075 if (!target) {
2076 sparse_error(stmt->pos, "not in switch scope");
2077 stmt->type = STMT_NONE;
2078 return;
2080 sym = alloc_symbol(stmt->pos, SYM_NODE);
2081 add_symbol(&target->symbol_list, sym);
2082 sym->stmt = stmt;
2083 stmt->case_label = sym;
2084 fn_local_symbol(sym);
2087 static struct token *parse_return_statement(struct token *token, struct statement *stmt)
2089 struct symbol *target = lookup_symbol(&return_ident, NS_ITERATOR);
2091 if (!target)
2092 error_die(token->pos, "internal error: return without a function target");
2093 stmt->type = STMT_RETURN;
2094 stmt->ret_target = target;
2095 return expression_statement(token->next, &stmt->ret_value);
2098 static struct token *parse_for_statement(struct token *token, struct statement *stmt)
2100 struct symbol_list *syms;
2101 struct expression *e1, *e2, *e3;
2102 struct statement *iterator;
2104 start_iterator(stmt);
2105 token = expect(token->next, '(', "after 'for'");
2107 syms = NULL;
2108 e1 = NULL;
2109 /* C99 variable declaration? */
2110 if (lookup_type(token)) {
2111 token = external_declaration(token, &syms);
2112 } else {
2113 token = parse_expression(token, &e1);
2114 token = expect(token, ';', "in 'for'");
2116 token = parse_expression(token, &e2);
2117 token = expect(token, ';', "in 'for'");
2118 token = parse_expression(token, &e3);
2119 token = expect(token, ')', "in 'for'");
2120 token = statement(token, &iterator);
2122 stmt->iterator_syms = syms;
2123 stmt->iterator_pre_statement = make_statement(e1);
2124 stmt->iterator_pre_condition = e2;
2125 stmt->iterator_post_statement = make_statement(e3);
2126 stmt->iterator_post_condition = NULL;
2127 stmt->iterator_statement = iterator;
2128 end_iterator(stmt);
2130 return token;
2133 static struct token *parse_while_statement(struct token *token, struct statement *stmt)
2135 struct expression *expr;
2136 struct statement *iterator;
2138 start_iterator(stmt);
2139 token = parens_expression(token->next, &expr, "after 'while'");
2140 token = statement(token, &iterator);
2142 stmt->iterator_pre_condition = expr;
2143 stmt->iterator_post_condition = NULL;
2144 stmt->iterator_statement = iterator;
2145 end_iterator(stmt);
2147 return token;
2150 static struct token *parse_do_statement(struct token *token, struct statement *stmt)
2152 struct expression *expr;
2153 struct statement *iterator;
2155 start_iterator(stmt);
2156 token = statement(token->next, &iterator);
2157 if (token_type(token) == TOKEN_IDENT && token->ident == &while_ident)
2158 token = token->next;
2159 else
2160 sparse_error(token->pos, "expected 'while' after 'do'");
2161 token = parens_expression(token, &expr, "after 'do-while'");
2163 stmt->iterator_post_condition = expr;
2164 stmt->iterator_statement = iterator;
2165 end_iterator(stmt);
2167 if (iterator && iterator->type != STMT_COMPOUND && Wdo_while)
2168 warning(iterator->pos, "do-while statement is not a compound statement");
2170 return expect(token, ';', "after statement");
2173 static struct token *parse_if_statement(struct token *token, struct statement *stmt)
2175 stmt->type = STMT_IF;
2176 token = parens_expression(token->next, &stmt->if_conditional, "after if");
2177 token = statement(token, &stmt->if_true);
2178 if (token_type(token) != TOKEN_IDENT)
2179 return token;
2180 if (token->ident != &else_ident)
2181 return token;
2182 return statement(token->next, &stmt->if_false);
2185 static inline struct token *case_statement(struct token *token, struct statement *stmt)
2187 stmt->type = STMT_CASE;
2188 token = expect(token, ':', "after default/case");
2189 add_case_statement(stmt);
2190 return statement(token, &stmt->case_statement);
2193 static struct token *parse_case_statement(struct token *token, struct statement *stmt)
2195 token = parse_expression(token->next, &stmt->case_expression);
2196 if (match_op(token, SPECIAL_ELLIPSIS))
2197 token = parse_expression(token->next, &stmt->case_to);
2198 return case_statement(token, stmt);
2201 static struct token *parse_default_statement(struct token *token, struct statement *stmt)
2203 return case_statement(token->next, stmt);
2206 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt)
2208 struct symbol *target = lookup_symbol(token->ident, NS_ITERATOR);
2209 stmt->type = STMT_GOTO;
2210 stmt->goto_label = target;
2211 if (!target)
2212 sparse_error(stmt->pos, "break/continue not in iterator scope");
2213 return expect(token->next, ';', "at end of statement");
2216 static struct token *parse_switch_statement(struct token *token, struct statement *stmt)
2218 stmt->type = STMT_SWITCH;
2219 start_switch(stmt);
2220 token = parens_expression(token->next, &stmt->switch_expression, "after 'switch'");
2221 token = statement(token, &stmt->switch_statement);
2222 end_switch(stmt);
2223 return token;
2226 static struct token *parse_goto_statement(struct token *token, struct statement *stmt)
2228 stmt->type = STMT_GOTO;
2229 token = token->next;
2230 if (match_op(token, '*')) {
2231 token = parse_expression(token->next, &stmt->goto_expression);
2232 add_statement(&function_computed_goto_list, stmt);
2233 } else if (token_type(token) == TOKEN_IDENT) {
2234 stmt->goto_label = label_symbol(token);
2235 token = token->next;
2236 } else {
2237 sparse_error(token->pos, "Expected identifier or goto expression");
2239 return expect(token, ';', "at end of statement");
2242 static struct token *parse_context_statement(struct token *token, struct statement *stmt)
2244 stmt->type = STMT_CONTEXT;
2245 token = parse_expression(token->next, &stmt->expression);
2246 if (stmt->expression->type == EXPR_PREOP
2247 && stmt->expression->op == '('
2248 && stmt->expression->unop->type == EXPR_COMMA) {
2249 struct expression *expr;
2250 expr = stmt->expression->unop;
2251 stmt->context = expr->left;
2252 stmt->expression = expr->right;
2254 return expect(token, ';', "at end of statement");
2257 static struct token *parse_range_statement(struct token *token, struct statement *stmt)
2259 stmt->type = STMT_RANGE;
2260 token = assignment_expression(token->next, &stmt->range_expression);
2261 token = expect(token, ',', "after range expression");
2262 token = assignment_expression(token, &stmt->range_low);
2263 token = expect(token, ',', "after low range");
2264 token = assignment_expression(token, &stmt->range_high);
2265 return expect(token, ';', "after range statement");
2268 static struct token *statement(struct token *token, struct statement **tree)
2270 struct statement *stmt = alloc_statement(token->pos, STMT_NONE);
2272 *tree = stmt;
2273 if (token_type(token) == TOKEN_IDENT) {
2274 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2275 if (s && s->op->statement)
2276 return s->op->statement(token, stmt);
2278 if (match_op(token->next, ':')) {
2279 stmt->type = STMT_LABEL;
2280 stmt->label_identifier = label_symbol(token);
2281 token = skip_attributes(token->next->next);
2282 return statement(token, &stmt->label_statement);
2286 if (match_op(token, '{')) {
2287 stmt->type = STMT_COMPOUND;
2288 start_symbol_scope();
2289 token = compound_statement(token->next, stmt);
2290 end_symbol_scope();
2292 return expect(token, '}', "at end of compound statement");
2295 stmt->type = STMT_EXPRESSION;
2296 return expression_statement(token, &stmt->expression);
2299 /* gcc extension - __label__ ident-list; in the beginning of compound stmt */
2300 static struct token *label_statement(struct token *token)
2302 while (token_type(token) == TOKEN_IDENT) {
2303 struct symbol *sym = alloc_symbol(token->pos, SYM_LABEL);
2304 /* it's block-scope, but we want label namespace */
2305 bind_symbol(sym, token->ident, NS_SYMBOL);
2306 sym->namespace = NS_LABEL;
2307 fn_local_symbol(sym);
2308 token = token->next;
2309 if (!match_op(token, ','))
2310 break;
2311 token = token->next;
2313 return expect(token, ';', "at end of label declaration");
2316 static struct token * statement_list(struct token *token, struct statement_list **list)
2318 int seen_statement = 0;
2319 while (token_type(token) == TOKEN_IDENT &&
2320 token->ident == &__label___ident)
2321 token = label_statement(token->next);
2322 for (;;) {
2323 struct statement * stmt;
2324 if (eof_token(token))
2325 break;
2326 if (match_op(token, '}'))
2327 break;
2328 if (lookup_type(token)) {
2329 if (seen_statement) {
2330 warning(token->pos, "mixing declarations and code");
2331 seen_statement = 0;
2333 stmt = alloc_statement(token->pos, STMT_DECLARATION);
2334 token = external_declaration(token, &stmt->declaration);
2335 } else {
2336 seen_statement = Wdeclarationafterstatement;
2337 token = statement(token, &stmt);
2339 add_statement(list, stmt);
2341 return token;
2344 static struct token *identifier_list(struct token *token, struct symbol *fn)
2346 struct symbol_list **list = &fn->arguments;
2347 for (;;) {
2348 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2349 sym->ident = token->ident;
2350 token = token->next;
2351 sym->endpos = token->pos;
2352 sym->ctype.base_type = &incomplete_ctype;
2353 add_symbol(list, sym);
2354 if (!match_op(token, ',') ||
2355 token_type(token->next) != TOKEN_IDENT ||
2356 lookup_type(token->next))
2357 break;
2358 token = token->next;
2360 return token;
2363 static struct token *parameter_type_list(struct token *token, struct symbol *fn)
2365 struct symbol_list **list = &fn->arguments;
2367 for (;;) {
2368 struct symbol *sym;
2370 if (match_op(token, SPECIAL_ELLIPSIS)) {
2371 fn->variadic = 1;
2372 token = token->next;
2373 break;
2376 sym = alloc_symbol(token->pos, SYM_NODE);
2377 token = parameter_declaration(token, sym);
2378 if (sym->ctype.base_type == &void_ctype) {
2379 /* Special case: (void) */
2380 if (!*list && !sym->ident)
2381 break;
2382 warning(token->pos, "void parameter");
2384 add_symbol(list, sym);
2385 if (!match_op(token, ','))
2386 break;
2387 token = token->next;
2389 return token;
2392 struct token *compound_statement(struct token *token, struct statement *stmt)
2394 token = statement_list(token, &stmt->stmts);
2395 return token;
2398 static struct expression *identifier_expression(struct token *token)
2400 struct expression *expr = alloc_expression(token->pos, EXPR_IDENTIFIER);
2401 expr->expr_ident = token->ident;
2402 return expr;
2405 static struct expression *index_expression(struct expression *from, struct expression *to)
2407 int idx_from, idx_to;
2408 struct expression *expr = alloc_expression(from->pos, EXPR_INDEX);
2410 idx_from = const_expression_value(from);
2411 idx_to = idx_from;
2412 if (to) {
2413 idx_to = const_expression_value(to);
2414 if (idx_to < idx_from || idx_from < 0)
2415 warning(from->pos, "nonsense array initializer index range");
2417 expr->idx_from = idx_from;
2418 expr->idx_to = idx_to;
2419 return expr;
2422 static struct token *single_initializer(struct expression **ep, struct token *token)
2424 int expect_equal = 0;
2425 struct token *next = token->next;
2426 struct expression **tail = ep;
2427 int nested;
2429 *ep = NULL;
2431 if ((token_type(token) == TOKEN_IDENT) && match_op(next, ':')) {
2432 struct expression *expr = identifier_expression(token);
2433 if (Wold_initializer)
2434 warning(token->pos, "obsolete struct initializer, use C99 syntax");
2435 token = initializer(&expr->ident_expression, next->next);
2436 if (expr->ident_expression)
2437 *ep = expr;
2438 return token;
2441 for (tail = ep, nested = 0; ; nested++, next = token->next) {
2442 if (match_op(token, '.') && (token_type(next) == TOKEN_IDENT)) {
2443 struct expression *expr = identifier_expression(next);
2444 *tail = expr;
2445 tail = &expr->ident_expression;
2446 expect_equal = 1;
2447 token = next->next;
2448 } else if (match_op(token, '[')) {
2449 struct expression *from = NULL, *to = NULL, *expr;
2450 token = constant_expression(token->next, &from);
2451 if (!from) {
2452 sparse_error(token->pos, "Expected constant expression");
2453 break;
2455 if (match_op(token, SPECIAL_ELLIPSIS))
2456 token = constant_expression(token->next, &to);
2457 expr = index_expression(from, to);
2458 *tail = expr;
2459 tail = &expr->idx_expression;
2460 token = expect(token, ']', "at end of initializer index");
2461 if (nested)
2462 expect_equal = 1;
2463 } else {
2464 break;
2467 if (nested && !expect_equal) {
2468 if (!match_op(token, '='))
2469 warning(token->pos, "obsolete array initializer, use C99 syntax");
2470 else
2471 expect_equal = 1;
2473 if (expect_equal)
2474 token = expect(token, '=', "at end of initializer index");
2476 token = initializer(tail, token);
2477 if (!*tail)
2478 *ep = NULL;
2479 return token;
2482 static struct token *initializer_list(struct expression_list **list, struct token *token)
2484 struct expression *expr;
2486 for (;;) {
2487 token = single_initializer(&expr, token);
2488 if (!expr)
2489 break;
2490 add_expression(list, expr);
2491 if (!match_op(token, ','))
2492 break;
2493 token = token->next;
2495 return token;
2498 struct token *initializer(struct expression **tree, struct token *token)
2500 if (match_op(token, '{')) {
2501 struct expression *expr = alloc_expression(token->pos, EXPR_INITIALIZER);
2502 *tree = expr;
2503 token = initializer_list(&expr->expr_list, token->next);
2504 return expect(token, '}', "at end of initializer");
2506 return assignment_expression(token, tree);
2509 static void declare_argument(struct symbol *sym, struct symbol *fn)
2511 if (!sym->ident) {
2512 sparse_error(sym->pos, "no identifier for function argument");
2513 return;
2515 bind_symbol(sym, sym->ident, NS_SYMBOL);
2518 static struct token *parse_function_body(struct token *token, struct symbol *decl,
2519 struct symbol_list **list)
2521 struct symbol_list **old_symbol_list;
2522 struct symbol *base_type = decl->ctype.base_type;
2523 struct statement *stmt, **p;
2524 struct symbol *prev;
2525 struct symbol *arg;
2527 old_symbol_list = function_symbol_list;
2528 if (decl->ctype.modifiers & MOD_INLINE) {
2529 function_symbol_list = &decl->inline_symbol_list;
2530 p = &base_type->inline_stmt;
2531 } else {
2532 function_symbol_list = &decl->symbol_list;
2533 p = &base_type->stmt;
2535 function_computed_target_list = NULL;
2536 function_computed_goto_list = NULL;
2538 if (decl->ctype.modifiers & MOD_EXTERN) {
2539 if (!(decl->ctype.modifiers & MOD_INLINE))
2540 warning(decl->pos, "function '%s' with external linkage has definition", show_ident(decl->ident));
2542 if (!(decl->ctype.modifiers & MOD_STATIC))
2543 decl->ctype.modifiers |= MOD_EXTERN;
2545 stmt = start_function(decl);
2547 *p = stmt;
2548 FOR_EACH_PTR (base_type->arguments, arg) {
2549 declare_argument(arg, base_type);
2550 } END_FOR_EACH_PTR(arg);
2552 token = compound_statement(token->next, stmt);
2554 end_function(decl);
2555 if (!(decl->ctype.modifiers & MOD_INLINE))
2556 add_symbol(list, decl);
2557 check_declaration(decl);
2558 decl->definition = decl;
2559 prev = decl->same_symbol;
2560 if (prev && prev->definition) {
2561 warning(decl->pos, "multiple definitions for function '%s'",
2562 show_ident(decl->ident));
2563 info(prev->definition->pos, " the previous one is here");
2564 } else {
2565 while (prev) {
2566 prev->definition = decl;
2567 prev = prev->same_symbol;
2570 function_symbol_list = old_symbol_list;
2571 if (function_computed_goto_list) {
2572 if (!function_computed_target_list)
2573 warning(decl->pos, "function '%s' has computed goto but no targets?", show_ident(decl->ident));
2574 else {
2575 FOR_EACH_PTR(function_computed_goto_list, stmt) {
2576 stmt->target_list = function_computed_target_list;
2577 } END_FOR_EACH_PTR(stmt);
2580 return expect(token, '}', "at end of function");
2583 static void promote_k_r_types(struct symbol *arg)
2585 struct symbol *base = arg->ctype.base_type;
2586 if (base && base->ctype.base_type == &int_type && (base->ctype.modifiers & (MOD_CHAR | MOD_SHORT))) {
2587 arg->ctype.base_type = &int_ctype;
2591 static void apply_k_r_types(struct symbol_list *argtypes, struct symbol *fn)
2593 struct symbol_list *real_args = fn->ctype.base_type->arguments;
2594 struct symbol *arg;
2596 FOR_EACH_PTR(real_args, arg) {
2597 struct symbol *type;
2599 /* This is quadratic in the number of arguments. We _really_ don't care */
2600 FOR_EACH_PTR(argtypes, type) {
2601 if (type->ident == arg->ident)
2602 goto match;
2603 } END_FOR_EACH_PTR(type);
2604 sparse_error(arg->pos, "missing type declaration for parameter '%s'", show_ident(arg->ident));
2605 continue;
2606 match:
2607 type->used = 1;
2608 /* "char" and "short" promote to "int" */
2609 promote_k_r_types(type);
2611 arg->ctype = type->ctype;
2612 } END_FOR_EACH_PTR(arg);
2614 FOR_EACH_PTR(argtypes, arg) {
2615 if (!arg->used)
2616 warning(arg->pos, "nonsensical parameter declaration '%s'", show_ident(arg->ident));
2617 } END_FOR_EACH_PTR(arg);
2621 static struct token *parse_k_r_arguments(struct token *token, struct symbol *decl,
2622 struct symbol_list **list)
2624 struct symbol_list *args = NULL;
2626 warning(token->pos, "non-ANSI definition of function '%s'", show_ident(decl->ident));
2627 do {
2628 token = declaration_list(token, &args);
2629 if (!match_op(token, ';')) {
2630 sparse_error(token->pos, "expected ';' at end of parameter declaration");
2631 break;
2633 token = token->next;
2634 } while (lookup_type(token));
2636 apply_k_r_types(args, decl);
2638 if (!match_op(token, '{')) {
2639 sparse_error(token->pos, "expected function body");
2640 return token;
2642 return parse_function_body(token, decl, list);
2645 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list)
2647 struct symbol *anon = alloc_symbol(token->pos, SYM_NODE);
2648 struct symbol *fn = alloc_symbol(token->pos, SYM_FN);
2649 struct statement *stmt;
2651 anon->ctype.base_type = fn;
2652 stmt = alloc_statement(token->pos, STMT_NONE);
2653 fn->stmt = stmt;
2655 token = parse_asm_statement(token, stmt);
2657 add_symbol(list, anon);
2658 return token;
2661 struct token *external_declaration(struct token *token, struct symbol_list **list)
2663 struct ident *ident = NULL;
2664 struct symbol *decl;
2665 struct decl_state ctx = { .ident = &ident };
2666 struct ctype saved;
2667 struct symbol *base_type;
2668 unsigned long mod;
2669 int is_typedef;
2671 /* Top-level inline asm? */
2672 if (token_type(token) == TOKEN_IDENT) {
2673 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2674 if (s && s->op->toplevel)
2675 return s->op->toplevel(token, list);
2678 /* Parse declaration-specifiers, if any */
2679 token = declaration_specifiers(token, &ctx);
2680 mod = storage_modifiers(&ctx);
2681 decl = alloc_symbol(token->pos, SYM_NODE);
2682 /* Just a type declaration? */
2683 if (match_op(token, ';')) {
2684 apply_modifiers(token->pos, &ctx);
2685 return token->next;
2688 saved = ctx.ctype;
2689 token = declarator(token, &ctx);
2690 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
2691 apply_modifiers(token->pos, &ctx);
2693 decl->ctype = ctx.ctype;
2694 decl->ctype.modifiers |= mod;
2695 decl->endpos = token->pos;
2697 /* Just a type declaration? */
2698 if (!ident) {
2699 warning(token->pos, "missing identifier in declaration");
2700 return expect(token, ';', "at the end of type declaration");
2703 /* type define declaration? */
2704 is_typedef = ctx.storage_class == STypedef;
2706 /* Typedefs don't have meaningful storage */
2707 if (is_typedef)
2708 decl->ctype.modifiers |= MOD_USERTYPE;
2710 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
2712 base_type = decl->ctype.base_type;
2714 if (is_typedef) {
2715 if (base_type && !base_type->ident) {
2716 switch (base_type->type) {
2717 case SYM_STRUCT:
2718 case SYM_UNION:
2719 case SYM_ENUM:
2720 case SYM_RESTRICT:
2721 base_type->ident = ident;
2724 } else if (base_type && base_type->type == SYM_FN) {
2725 /* K&R argument declaration? */
2726 if (lookup_type(token))
2727 return parse_k_r_arguments(token, decl, list);
2728 if (match_op(token, '{'))
2729 return parse_function_body(token, decl, list);
2731 if (!(decl->ctype.modifiers & MOD_STATIC))
2732 decl->ctype.modifiers |= MOD_EXTERN;
2733 } else if (base_type == &void_ctype && !(decl->ctype.modifiers & MOD_EXTERN)) {
2734 sparse_error(token->pos, "void declaration");
2737 for (;;) {
2738 if (!is_typedef && match_op(token, '=')) {
2739 if (decl->ctype.modifiers & MOD_EXTERN) {
2740 warning(decl->pos, "symbol with external linkage has initializer");
2741 decl->ctype.modifiers &= ~MOD_EXTERN;
2743 token = initializer(&decl->initializer, token->next);
2745 if (!is_typedef) {
2746 if (!(decl->ctype.modifiers & (MOD_EXTERN | MOD_INLINE))) {
2747 add_symbol(list, decl);
2748 fn_local_symbol(decl);
2751 check_declaration(decl);
2752 if (decl->same_symbol)
2753 decl->definition = decl->same_symbol->definition;
2755 if (!match_op(token, ','))
2756 break;
2758 token = token->next;
2759 ident = NULL;
2760 decl = alloc_symbol(token->pos, SYM_NODE);
2761 ctx.ctype = saved;
2762 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
2763 token = declarator(token, &ctx);
2764 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
2765 apply_modifiers(token->pos, &ctx);
2766 decl->ctype = ctx.ctype;
2767 decl->ctype.modifiers |= mod;
2768 decl->endpos = token->pos;
2769 if (!ident) {
2770 sparse_error(token->pos, "expected identifier name in type definition");
2771 return token;
2774 if (is_typedef)
2775 decl->ctype.modifiers |= MOD_USERTYPE;
2777 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
2779 /* Function declarations are automatically extern unless specifically static */
2780 base_type = decl->ctype.base_type;
2781 if (!is_typedef && base_type && base_type->type == SYM_FN) {
2782 if (!(decl->ctype.modifiers & MOD_STATIC))
2783 decl->ctype.modifiers |= MOD_EXTERN;
2786 return expect(token, ';', "at end of declaration");