1 /*
2 * Stupid C parser, version 1e-6.
3 *
4 * Let's see how hard this is to do.
5 *
6 * Copyright (C) 2003 Transmeta Corp.
7 * 2003-2004 Linus Torvalds
8 * Copyright (C) 2004 Christopher Li
9 *
10 * Permission is hereby granted, free of charge, to any person obtaining a copy
11 * of this software and associated documentation files (the "Software"), to deal
12 * in the Software without restriction, including without limitation the rights
13 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
14 * copies of the Software, and to permit persons to whom the Software is
15 * furnished to do so, subject to the following conditions:
16 *
17 * The above copyright notice and this permission notice shall be included in
18 * all copies or substantial portions of the Software.
19 *
20 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
21 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
23 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
24 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
25 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
26 * THE SOFTWARE.
27 */
28
29 #include <stdarg.h>
30 #include <stdlib.h>
31 #include <stdio.h>
32 #include <string.h>
33 #include <ctype.h>
34 #include <unistd.h>
35 #include <fcntl.h>
36 #include <limits.h>
37
38 #include "lib.h"
39 #include "allocate.h"
40 #include "token.h"
41 #include "parse.h"
42 #include "symbol.h"
43 #include "scope.h"
44 #include "expression.h"
45 #include "target.h"
46
47 static struct symbol_list **function_symbol_list;
48 struct symbol_list *function_computed_target_list;
49 struct statement_list *function_computed_goto_list;
50
51 static struct token *statement(struct token *token, struct statement **tree);
52 static struct token *handle_attributes(struct token *token, struct decl_state *ctx, unsigned int keywords);
53
54 typedef struct token *declarator_t(struct token *, struct decl_state *);
55 static declarator_t
56 struct_specifier, union_specifier, enum_specifier,
57 attribute_specifier, typeof_specifier, parse_asm_declarator,
58 typedef_specifier, inline_specifier, auto_specifier,
59 register_specifier, static_specifier, extern_specifier,
60 thread_specifier, const_qualifier, volatile_qualifier;
61 static declarator_t restrict_qualifier;
62 static declarator_t atomic_qualifier;
63
64 static struct token *parse_if_statement(struct token *token, struct statement *stmt);
65 static struct token *parse_return_statement(struct token *token, struct statement *stmt);
66 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt);
67 static struct token *parse_default_statement(struct token *token, struct statement *stmt);
68 static struct token *parse_case_statement(struct token *token, struct statement *stmt);
69 static struct token *parse_switch_statement(struct token *token, struct statement *stmt);
70 static struct token *parse_for_statement(struct token *token, struct statement *stmt);
71 static struct token *parse_while_statement(struct token *token, struct statement *stmt);
72 static struct token *parse_do_statement(struct token *token, struct statement *stmt);
73 static struct token *parse_goto_statement(struct token *token, struct statement *stmt);
74 static struct token *parse_context_statement(struct token *token, struct statement *stmt);
75 static struct token *parse_range_statement(struct token *token, struct statement *stmt);
76 static struct token *parse_asm_statement(struct token *token, struct statement *stmt);
77 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list);
78 static struct token *parse_static_assert(struct token *token, struct symbol_list **unused);
79
80 typedef struct token *attr_t(struct token *, struct symbol *,
81 struct decl_state *);
82
83 static attr_t
84 attribute_packed, attribute_aligned, attribute_modifier,
85 attribute_ext_visible,
86 attribute_bitwise,
87 attribute_address_space, attribute_context,
88 attribute_designated_init,
89 attribute_transparent_union, ignore_attribute,
90 attribute_mode, attribute_force;
91
92 typedef struct symbol *to_mode_t(struct symbol *);
93
94 static to_mode_t
95 to_QI_mode, to_HI_mode, to_SI_mode, to_DI_mode, to_TI_mode;
96 static to_mode_t to_pointer_mode, to_word_mode;
97
98 enum {
99 Set_T = 1,
100 Set_S = 2,
101 Set_Char = 4,
102 Set_Int = 8,
103 Set_Double = 16,
104 Set_Float = 32,
105 Set_Signed = 64,
106 Set_Unsigned = 128,
107 Set_Short = 256,
108 Set_Long = 512,
109 Set_Vlong = 1024,
110 Set_Int128 = 2048,
111 Set_Any = Set_T | Set_Short | Set_Long | Set_Signed | Set_Unsigned
112 };
113
114 enum {
115 CInt = 0, CSInt, CUInt, CReal, CChar, CSChar, CUChar,
116 };
117
118 enum {
119 SNone = 0, STypedef, SAuto, SRegister, SExtern, SStatic, SForced, SMax,
120 };
121
asm_modifier(struct token * token,unsigned long * mods,unsigned long mod)122 static void asm_modifier(struct token *token, unsigned long *mods, unsigned long mod)
123 {
124 if (*mods & mod)
125 warning(token->pos, "duplicated asm modifier");
126 *mods |= mod;
127 }
128
asm_modifier_volatile(struct token * token,unsigned long * mods)129 static void asm_modifier_volatile(struct token *token, unsigned long *mods)
130 {
131 asm_modifier(token, mods, MOD_VOLATILE);
132 }
133
asm_modifier_inline(struct token * token,unsigned long * mods)134 static void asm_modifier_inline(struct token *token, unsigned long *mods)
135 {
136 asm_modifier(token, mods, MOD_INLINE);
137 }
138
139 static struct symbol_op typedef_op = {
140 .type = KW_MODIFIER,
141 .declarator = typedef_specifier,
142 };
143
144 static struct symbol_op inline_op = {
145 .type = KW_MODIFIER,
146 .declarator = inline_specifier,
147 .asm_modifier = asm_modifier_inline,
148 };
149
150 static declarator_t noreturn_specifier;
151 static struct symbol_op noreturn_op = {
152 .type = KW_MODIFIER,
153 .declarator = noreturn_specifier,
154 };
155
156 static declarator_t alignas_specifier;
157 static struct symbol_op alignas_op = {
158 .type = KW_MODIFIER,
159 .declarator = alignas_specifier,
160 };
161
162 static struct symbol_op auto_op = {
163 .type = KW_MODIFIER,
164 .declarator = auto_specifier,
165 };
166
167 static struct symbol_op register_op = {
168 .type = KW_MODIFIER,
169 .declarator = register_specifier,
170 };
171
172 static struct symbol_op static_op = {
173 .type = KW_MODIFIER,
174 .declarator = static_specifier,
175 };
176
177 static struct symbol_op extern_op = {
178 .type = KW_MODIFIER,
179 .declarator = extern_specifier,
180 };
181
182 static struct symbol_op thread_op = {
183 .type = KW_MODIFIER,
184 .declarator = thread_specifier,
185 };
186
187 static struct symbol_op const_op = {
188 .type = KW_QUALIFIER,
189 .declarator = const_qualifier,
190 };
191
192 static struct symbol_op volatile_op = {
193 .type = KW_QUALIFIER,
194 .declarator = volatile_qualifier,
195 .asm_modifier = asm_modifier_volatile,
196 };
197
198 static struct symbol_op restrict_op = {
199 .type = KW_QUALIFIER,
200 .declarator = restrict_qualifier,
201 };
202
203 static struct symbol_op atomic_op = {
204 .type = KW_QUALIFIER,
205 .declarator = atomic_qualifier,
206 };
207
208 static struct symbol_op typeof_op = {
209 .type = KW_SPECIFIER,
210 .declarator = typeof_specifier,
211 .test = Set_Any,
212 .set = Set_S|Set_T,
213 };
214
215 static struct symbol_op attribute_op = {
216 .type = KW_ATTRIBUTE,
217 .declarator = attribute_specifier,
218 };
219
220 static struct symbol_op struct_op = {
221 .type = KW_SPECIFIER,
222 .declarator = struct_specifier,
223 .test = Set_Any,
224 .set = Set_S|Set_T,
225 };
226
227 static struct symbol_op union_op = {
228 .type = KW_SPECIFIER,
229 .declarator = union_specifier,
230 .test = Set_Any,
231 .set = Set_S|Set_T,
232 };
233
234 static struct symbol_op enum_op = {
235 .type = KW_SPECIFIER,
236 .declarator = enum_specifier,
237 .test = Set_Any,
238 .set = Set_S|Set_T,
239 };
240
241 static struct symbol_op spec_op = {
242 .type = KW_SPECIFIER | KW_EXACT,
243 .test = Set_Any,
244 .set = Set_S|Set_T,
245 };
246
247 static struct symbol_op char_op = {
248 .type = KW_SPECIFIER,
249 .test = Set_T|Set_Long|Set_Short,
250 .set = Set_T|Set_Char,
251 .class = CChar,
252 };
253
254 static struct symbol_op int_op = {
255 .type = KW_SPECIFIER,
256 .test = Set_T,
257 .set = Set_T|Set_Int,
258 };
259
260 static struct symbol_op double_op = {
261 .type = KW_SPECIFIER,
262 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Vlong,
263 .set = Set_T|Set_Double,
264 .class = CReal,
265 };
266
267 static struct symbol_op float_op = {
268 .type = KW_SPECIFIER | KW_SHORT,
269 .test = Set_T|Set_Signed|Set_Unsigned|Set_Short|Set_Long,
270 .set = Set_T|Set_Float,
271 .class = CReal,
272 };
273
274 static struct symbol_op short_op = {
275 .type = KW_SPECIFIER | KW_SHORT,
276 .test = Set_S|Set_Char|Set_Float|Set_Double|Set_Long|Set_Short,
277 .set = Set_Short,
278 };
279
280 static struct symbol_op signed_op = {
281 .type = KW_SPECIFIER,
282 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
283 .set = Set_Signed,
284 .class = CSInt,
285 };
286
287 static struct symbol_op unsigned_op = {
288 .type = KW_SPECIFIER,
289 .test = Set_S|Set_Float|Set_Double|Set_Signed|Set_Unsigned,
290 .set = Set_Unsigned,
291 .class = CUInt,
292 };
293
294 static struct symbol_op long_op = {
295 .type = KW_SPECIFIER | KW_LONG,
296 .test = Set_S|Set_Char|Set_Float|Set_Short|Set_Vlong,
297 .set = Set_Long,
298 };
299
300 static struct symbol_op int128_op = {
301 .type = KW_SPECIFIER | KW_LONG,
302 .test = Set_S|Set_T|Set_Char|Set_Short|Set_Int|Set_Float|Set_Double|Set_Long|Set_Vlong|Set_Int128,
303 .set = Set_T|Set_Int128,
304 };
305
306 static struct symbol_op if_op = {
307 .statement = parse_if_statement,
308 };
309
310 static struct symbol_op return_op = {
311 .statement = parse_return_statement,
312 };
313
314 static struct symbol_op loop_iter_op = {
315 .statement = parse_loop_iterator,
316 };
317
318 static struct symbol_op default_op = {
319 .statement = parse_default_statement,
320 };
321
322 static struct symbol_op case_op = {
323 .statement = parse_case_statement,
324 };
325
326 static struct symbol_op switch_op = {
327 .statement = parse_switch_statement,
328 };
329
330 static struct symbol_op for_op = {
331 .statement = parse_for_statement,
332 };
333
334 static struct symbol_op while_op = {
335 .statement = parse_while_statement,
336 };
337
338 static struct symbol_op do_op = {
339 .statement = parse_do_statement,
340 };
341
342 static struct symbol_op goto_op = {
343 .statement = parse_goto_statement,
344 };
345
346 static struct symbol_op __context___op = {
347 .statement = parse_context_statement,
348 };
349
350 static struct symbol_op range_op = {
351 .statement = parse_range_statement,
352 };
353
354 static struct symbol_op asm_op = {
355 .type = KW_ASM,
356 .declarator = parse_asm_declarator,
357 .statement = parse_asm_statement,
358 .toplevel = toplevel_asm_declaration,
359 };
360
361 static struct symbol_op static_assert_op = {
362 .toplevel = parse_static_assert,
363 };
364
365 static struct symbol_op packed_op = {
366 .attribute = attribute_packed,
367 };
368
369 static struct symbol_op aligned_op = {
370 .attribute = attribute_aligned,
371 };
372
373 static struct symbol_op attr_mod_op = {
374 .attribute = attribute_modifier,
375 };
376
377 static struct symbol_op ext_visible_op = {
378 .attribute = attribute_ext_visible,
379 };
380
381 static struct symbol_op attr_bitwise_op = {
382 .attribute = attribute_bitwise,
383 };
384
385 static struct symbol_op attr_force_op = {
386 .attribute = attribute_force,
387 };
388
389 static struct symbol_op address_space_op = {
390 .attribute = attribute_address_space,
391 };
392
393 static struct symbol_op mode_op = {
394 .attribute = attribute_mode,
395 };
396
397 static struct symbol_op context_op = {
398 .attribute = attribute_context,
399 };
400
401 static struct symbol_op designated_init_op = {
402 .attribute = attribute_designated_init,
403 };
404
405 static struct symbol_op transparent_union_op = {
406 .attribute = attribute_transparent_union,
407 };
408
409 static struct symbol_op ignore_attr_op = {
410 .attribute = ignore_attribute,
411 };
412
413 static struct symbol_op mode_QI_op = {
414 .type = KW_MODE,
415 .to_mode = to_QI_mode
416 };
417
418 static struct symbol_op mode_HI_op = {
419 .type = KW_MODE,
420 .to_mode = to_HI_mode
421 };
422
423 static struct symbol_op mode_SI_op = {
424 .type = KW_MODE,
425 .to_mode = to_SI_mode
426 };
427
428 static struct symbol_op mode_DI_op = {
429 .type = KW_MODE,
430 .to_mode = to_DI_mode
431 };
432
433 static struct symbol_op mode_TI_op = {
434 .type = KW_MODE,
435 .to_mode = to_TI_mode
436 };
437
438 static struct symbol_op mode_pointer_op = {
439 .type = KW_MODE,
440 .to_mode = to_pointer_mode
441 };
442
443 static struct symbol_op mode_word_op = {
444 .type = KW_MODE,
445 .to_mode = to_word_mode
446 };
447
448 /* Using NS_TYPEDEF will also make the keyword a reserved one */
449 static struct init_keyword {
450 const char *name;
451 enum namespace ns;
452 unsigned long modifiers;
453 struct symbol_op *op;
454 struct symbol *type;
455 } keyword_table[] = {
456 /* Type qualifiers */
457 { "const", NS_TYPEDEF, .op = &const_op },
458 { "__const", NS_TYPEDEF, .op = &const_op },
459 { "__const__", NS_TYPEDEF, .op = &const_op },
460 { "volatile", NS_TYPEDEF, .op = &volatile_op },
461 { "__volatile", NS_TYPEDEF, .op = &volatile_op },
462 { "__volatile__", NS_TYPEDEF, .op = &volatile_op },
463 { "restrict", NS_TYPEDEF, .op = &restrict_op},
464 { "__restrict", NS_TYPEDEF, .op = &restrict_op},
465 { "__restrict__", NS_TYPEDEF, .op = &restrict_op},
466 { "_Atomic", NS_TYPEDEF, .op = &atomic_op},
467
468 /* Typedef.. */
469 { "typedef", NS_TYPEDEF, .op = &typedef_op },
470
471 /* Type specifiers */
472 { "void", NS_TYPEDEF, .type = &void_ctype, .op = &spec_op},
473 { "char", NS_TYPEDEF, .op = &char_op },
474 { "short", NS_TYPEDEF, .op = &short_op },
475 { "int", NS_TYPEDEF, .op = &int_op },
476 { "long", NS_TYPEDEF, .op = &long_op },
477 { "float", NS_TYPEDEF, .op = &float_op },
478 { "double", NS_TYPEDEF, .op = &double_op },
479 { "signed", NS_TYPEDEF, .op = &signed_op },
480 { "__signed", NS_TYPEDEF, .op = &signed_op },
481 { "__signed__", NS_TYPEDEF, .op = &signed_op },
482 { "unsigned", NS_TYPEDEF, .op = &unsigned_op },
483 { "__int128", NS_TYPEDEF, .op = &int128_op },
484 { "_Bool", NS_TYPEDEF, .type = &bool_ctype, .op = &spec_op },
485
486 /* Predeclared types */
487 { "__builtin_va_list", NS_TYPEDEF, .type = &ptr_ctype, .op = &spec_op },
488 { "__builtin_ms_va_list", NS_TYPEDEF, .type = &ptr_ctype, .op = &spec_op },
489 { "__int128_t", NS_TYPEDEF, .type = &lllong_ctype, .op = &spec_op },
490 { "__uint128_t",NS_TYPEDEF, .type = &ulllong_ctype, .op = &spec_op },
491 { "_Float32", NS_TYPEDEF, .type = &float32_ctype, .op = &spec_op },
492 { "_Float32x", NS_TYPEDEF, .type = &float32x_ctype, .op = &spec_op },
493 { "_Float64", NS_TYPEDEF, .type = &float64_ctype, .op = &spec_op },
494 { "_Float64x", NS_TYPEDEF, .type = &float64x_ctype, .op = &spec_op },
495 { "_Float128", NS_TYPEDEF, .type = &float128_ctype, .op = &spec_op },
496
497 /* Extended types */
498 { "typeof", NS_TYPEDEF, .op = &typeof_op },
499 { "__typeof", NS_TYPEDEF, .op = &typeof_op },
500 { "__typeof__", NS_TYPEDEF, .op = &typeof_op },
501
502 { "__attribute", NS_TYPEDEF, .op = &attribute_op },
503 { "__attribute__", NS_TYPEDEF, .op = &attribute_op },
504
505 { "struct", NS_TYPEDEF, .op = &struct_op },
506 { "union", NS_TYPEDEF, .op = &union_op },
507 { "enum", NS_TYPEDEF, .op = &enum_op },
508
509 { "inline", NS_TYPEDEF, .op = &inline_op },
510 { "__inline", NS_TYPEDEF, .op = &inline_op },
511 { "__inline__", NS_TYPEDEF, .op = &inline_op },
512
513 { "_Noreturn", NS_TYPEDEF, .op = &noreturn_op },
514
515 { "_Alignas", NS_TYPEDEF, .op = &alignas_op },
516
517 /* Static assertion */
518 { "_Static_assert", NS_KEYWORD, .op = &static_assert_op },
519
520 /* Storage class */
521 { "auto", NS_TYPEDEF, .op = &auto_op },
522 { "register", NS_TYPEDEF, .op = ®ister_op },
523 { "static", NS_TYPEDEF, .op = &static_op },
524 { "extern", NS_TYPEDEF, .op = &extern_op },
525 { "__thread", NS_TYPEDEF, .op = &thread_op },
526 { "_Thread_local", NS_TYPEDEF, .op = &thread_op },
527
528 /* Statement */
529 { "if", NS_KEYWORD, .op = &if_op },
530 { "return", NS_KEYWORD, .op = &return_op },
531 { "break", NS_KEYWORD, .op = &loop_iter_op },
532 { "continue", NS_KEYWORD, .op = &loop_iter_op },
533 { "default", NS_KEYWORD, .op = &default_op },
534 { "case", NS_KEYWORD, .op = &case_op },
535 { "switch", NS_KEYWORD, .op = &switch_op },
536 { "for", NS_KEYWORD, .op = &for_op },
537 { "while", NS_KEYWORD, .op = &while_op },
538 { "do", NS_KEYWORD, .op = &do_op },
539 { "goto", NS_KEYWORD, .op = &goto_op },
540 { "__context__",NS_KEYWORD, .op = &__context___op },
541 { "__range__", NS_KEYWORD, .op = &range_op },
542 { "asm", NS_KEYWORD, .op = &asm_op },
543 { "__asm", NS_KEYWORD, .op = &asm_op },
544 { "__asm__", NS_KEYWORD, .op = &asm_op },
545
546 /* Attribute */
547 { "packed", NS_KEYWORD, .op = &packed_op },
548 { "__packed__", NS_KEYWORD, .op = &packed_op },
549 { "aligned", NS_KEYWORD, .op = &aligned_op },
550 { "__aligned__",NS_KEYWORD, .op = &aligned_op },
551 { "nocast", NS_KEYWORD, MOD_NOCAST, .op = &attr_mod_op },
552 { "noderef", NS_KEYWORD, MOD_NODEREF, .op = &attr_mod_op },
553 { "safe", NS_KEYWORD, MOD_SAFE, .op = &attr_mod_op },
554 { "force", NS_KEYWORD, .op = &attr_force_op },
555 { "bitwise", NS_KEYWORD, MOD_BITWISE, .op = &attr_bitwise_op },
556 { "__bitwise__",NS_KEYWORD, MOD_BITWISE, .op = &attr_bitwise_op },
557 { "address_space",NS_KEYWORD, .op = &address_space_op },
558 { "context", NS_KEYWORD, .op = &context_op },
559 { "designated_init", NS_KEYWORD, .op = &designated_init_op },
560 { "__designated_init__", NS_KEYWORD, .op = &designated_init_op },
561 { "transparent_union", NS_KEYWORD, .op = &transparent_union_op },
562 { "__transparent_union__", NS_KEYWORD, .op = &transparent_union_op },
563 { "noreturn", NS_KEYWORD, MOD_NORETURN, .op = &attr_mod_op },
564 { "__noreturn__", NS_KEYWORD, MOD_NORETURN, .op = &attr_mod_op },
565 { "pure", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
566 {"__pure__", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
567 {"const", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
568 {"__const", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
569 {"__const__", NS_KEYWORD, MOD_PURE, .op = &attr_mod_op },
570 {"externally_visible", NS_KEYWORD, .op = &ext_visible_op },
571 {"__externally_visible__", NS_KEYWORD, .op = &ext_visible_op },
572
573 { "mode", NS_KEYWORD, .op = &mode_op },
574 { "__mode__", NS_KEYWORD, .op = &mode_op },
575 { "QI", NS_KEYWORD, .op = &mode_QI_op },
576 { "__QI__", NS_KEYWORD, .op = &mode_QI_op },
577 { "HI", NS_KEYWORD, .op = &mode_HI_op },
578 { "__HI__", NS_KEYWORD, .op = &mode_HI_op },
579 { "SI", NS_KEYWORD, .op = &mode_SI_op },
580 { "__SI__", NS_KEYWORD, .op = &mode_SI_op },
581 { "DI", NS_KEYWORD, .op = &mode_DI_op },
582 { "__DI__", NS_KEYWORD, .op = &mode_DI_op },
583 { "TI", NS_KEYWORD, .op = &mode_TI_op },
584 { "__TI__", NS_KEYWORD, .op = &mode_TI_op },
585 { "byte", NS_KEYWORD, .op = &mode_QI_op },
586 { "__byte__", NS_KEYWORD, .op = &mode_QI_op },
587 { "pointer", NS_KEYWORD, .op = &mode_pointer_op },
588 { "__pointer__",NS_KEYWORD, .op = &mode_pointer_op },
589 { "word", NS_KEYWORD, .op = &mode_word_op },
590 { "__word__", NS_KEYWORD, .op = &mode_word_op },
591 };
592
593
594 static const char *ignored_attributes[] = {
595
596 #define GCC_ATTR(x) \
597 STRINGIFY(x), \
598 STRINGIFY(__##x##__),
599
600 #include "gcc-attr-list.h"
601
602 #undef GCC_ATTR
603
604 "bounded",
605 "__bounded__",
606 "__noclone",
607 "__nonnull",
608 "__nothrow",
609 };
610
611
init_parser(int stream)612 void init_parser(int stream)
613 {
614 int i;
615 for (i = 0; i < ARRAY_SIZE(keyword_table); i++) {
616 struct init_keyword *ptr = keyword_table + i;
617 struct symbol *sym = create_symbol(stream, ptr->name, SYM_KEYWORD, ptr->ns);
618 sym->ident->keyword = 1;
619 if (ptr->ns == NS_TYPEDEF)
620 sym->ident->reserved = 1;
621 sym->ctype.modifiers = ptr->modifiers;
622 sym->ctype.base_type = ptr->type;
623 sym->op = ptr->op;
624 }
625
626 for (i = 0; i < ARRAY_SIZE(ignored_attributes); i++) {
627 const char * name = ignored_attributes[i];
628 struct symbol *sym = create_symbol(stream, name, SYM_KEYWORD,
629 NS_KEYWORD);
630 if (!sym->op) {
631 sym->ident->keyword = 1;
632 sym->op = &ignore_attr_op;
633 }
634 }
635 }
636
637
638 // Add a symbol to the list of function-local symbols
fn_local_symbol(struct symbol * sym)639 static void fn_local_symbol(struct symbol *sym)
640 {
641 if (function_symbol_list)
642 add_symbol(function_symbol_list, sym);
643 }
644
match_idents(struct token * token,...)645 static int SENTINEL_ATTR match_idents(struct token *token, ...)
646 {
647 va_list args;
648 struct ident * next;
649
650 if (token_type(token) != TOKEN_IDENT)
651 return 0;
652
653 va_start(args, token);
654 do {
655 next = va_arg(args, struct ident *);
656 } while (next && token->ident != next);
657 va_end(args);
658
659 return next && token->ident == next;
660 }
661
662
alloc_statement(struct position pos,int type)663 struct statement *alloc_statement(struct position pos, int type)
664 {
665 struct statement *stmt = __alloc_statement(0);
666 stmt->type = type;
667 stmt->pos = pos;
668 return stmt;
669 }
670
671 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list);
672
673 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype);
674
apply_modifiers(struct position pos,struct decl_state * ctx)675 static void apply_modifiers(struct position pos, struct decl_state *ctx)
676 {
677 struct symbol *ctype;
678 if (!ctx->mode)
679 return;
680 ctype = ctx->mode->to_mode(ctx->ctype.base_type);
681 if (!ctype)
682 sparse_error(pos, "don't know how to apply mode to %s",
683 show_typename(ctx->ctype.base_type));
684 else
685 ctx->ctype.base_type = ctype;
686
687 }
688
alloc_indirect_symbol(struct position pos,struct ctype * ctype,int type)689 static struct symbol * alloc_indirect_symbol(struct position pos, struct ctype *ctype, int type)
690 {
691 struct symbol *sym = alloc_symbol(pos, type);
692
693 sym->ctype.base_type = ctype->base_type;
694 sym->ctype.modifiers = ctype->modifiers;
695
696 ctype->base_type = sym;
697 ctype->modifiers = 0;
698 return sym;
699 }
700
701 /*
702 * NOTE! NS_LABEL is not just a different namespace,
703 * it also ends up using function scope instead of the
704 * regular symbol scope.
705 */
label_symbol(struct token * token)706 struct symbol *label_symbol(struct token *token)
707 {
708 struct symbol *sym = lookup_symbol(token->ident, NS_LABEL);
709 if (!sym) {
710 sym = alloc_symbol(token->pos, SYM_LABEL);
711 bind_symbol(sym, token->ident, NS_LABEL);
712 fn_local_symbol(sym);
713 }
714 return sym;
715 }
716
struct_union_enum_specifier(enum type type,struct token * token,struct decl_state * ctx,struct token * (* parse)(struct token *,struct symbol *))717 static struct token *struct_union_enum_specifier(enum type type,
718 struct token *token, struct decl_state *ctx,
719 struct token *(*parse)(struct token *, struct symbol *))
720 {
721 struct symbol *sym;
722 struct position *repos;
723
724 token = handle_attributes(token, ctx, KW_ATTRIBUTE);
725 if (token_type(token) == TOKEN_IDENT) {
726 sym = lookup_symbol(token->ident, NS_STRUCT);
727 if (!sym ||
728 (is_outer_scope(sym->scope) &&
729 (match_op(token->next,';') || match_op(token->next,'{')))) {
730 // Either a new symbol, or else an out-of-scope
731 // symbol being redefined.
732 sym = alloc_symbol(token->pos, type);
733 bind_symbol(sym, token->ident, NS_STRUCT);
734 }
735 if (sym->type != type)
736 error_die(token->pos, "invalid tag applied to %s", show_typename (sym));
737 ctx->ctype.base_type = sym;
738 repos = &token->pos;
739 token = token->next;
740 if (match_op(token, '{')) {
741 struct decl_state attr = { .ctype.base_type = sym, };
742
743 // The following test is actually wrong for empty
744 // structs, but (1) they are not C99, (2) gcc does
745 // the same thing, and (3) it's easier.
746 if (sym->symbol_list)
747 error_die(token->pos, "redefinition of %s", show_typename (sym));
748 sym->pos = *repos;
749 token = parse(token->next, sym);
750 token = expect(token, '}', "at end of struct-union-enum-specifier");
751
752 token = handle_attributes(token, &attr, KW_ATTRIBUTE);
753 apply_ctype(token->pos, &attr.ctype, &sym->ctype);
754
755 // Mark the structure as needing re-examination
756 sym->examined = 0;
757 sym->endpos = token->pos;
758 }
759 return token;
760 }
761
762 // private struct/union/enum type
763 if (!match_op(token, '{')) {
764 sparse_error(token->pos, "expected declaration");
765 ctx->ctype.base_type = &bad_ctype;
766 return token;
767 }
768
769 sym = alloc_symbol(token->pos, type);
770 token = parse(token->next, sym);
771 ctx->ctype.base_type = sym;
772 token = expect(token, '}', "at end of specifier");
773 sym->endpos = token->pos;
774
775 return token;
776 }
777
parse_struct_declaration(struct token * token,struct symbol * sym)778 static struct token *parse_struct_declaration(struct token *token, struct symbol *sym)
779 {
780 struct symbol *field, *last = NULL;
781 struct token *res;
782 res = struct_declaration_list(token, &sym->symbol_list);
783 FOR_EACH_PTR(sym->symbol_list, field) {
784 if (!field->ident) {
785 struct symbol *base = field->ctype.base_type;
786 if (base && base->type == SYM_BITFIELD)
787 continue;
788 }
789 if (last)
790 last->next_subobject = field;
791 last = field;
792 } END_FOR_EACH_PTR(field);
793 return res;
794 }
795
parse_union_declaration(struct token * token,struct symbol * sym)796 static struct token *parse_union_declaration(struct token *token, struct symbol *sym)
797 {
798 return struct_declaration_list(token, &sym->symbol_list);
799 }
800
struct_specifier(struct token * token,struct decl_state * ctx)801 static struct token *struct_specifier(struct token *token, struct decl_state *ctx)
802 {
803 return struct_union_enum_specifier(SYM_STRUCT, token, ctx, parse_struct_declaration);
804 }
805
union_specifier(struct token * token,struct decl_state * ctx)806 static struct token *union_specifier(struct token *token, struct decl_state *ctx)
807 {
808 return struct_union_enum_specifier(SYM_UNION, token, ctx, parse_union_declaration);
809 }
810
811 ///
812 // safe right shift
813 //
814 // This allow to use a shift amount as big (or bigger)
815 // than the width of the value to be shifted, in which case
816 // the result is, of course, 0.
rshift(unsigned long long val,unsigned int n)817 static unsigned long long rshift(unsigned long long val, unsigned int n)
818 {
819 if (n >= (sizeof(val) * 8))
820 return 0;
821 return val >> n;
822 }
823
824 struct range {
825 long long neg;
826 unsigned long long pos;
827 };
828
update_range(struct range * range,unsigned long long uval,struct symbol * vtype)829 static void update_range(struct range *range, unsigned long long uval, struct symbol *vtype)
830 {
831 long long sval = uval;
832
833 if (is_signed_type(vtype) && (sval < 0)) {
834 if (sval < range->neg)
835 range->neg = sval;
836 } else {
837 if (uval > range->pos)
838 range->pos = uval;
839 }
840 }
841
type_is_ok(struct symbol * type,struct range range)842 static int type_is_ok(struct symbol *type, struct range range)
843 {
844 int shift = type->bit_size;
845 int is_unsigned = type->ctype.modifiers & MOD_UNSIGNED;
846
847 if (!is_unsigned)
848 shift--;
849 if (rshift(range.pos, shift))
850 return 0;
851 if (range.neg == 0)
852 return 1;
853 if (is_unsigned)
854 return 0;
855 if (rshift(~range.neg, shift))
856 return 0;
857 return 1;
858 }
859
type_range(struct symbol * type)860 static struct range type_range(struct symbol *type)
861 {
862 struct range range;
863 unsigned int size = type->bit_size;
864 unsigned long long max;
865 long long min;
866
867 if (is_signed_type(type)) {
868 min = sign_bit(size);
869 max = min - 1;
870 } else {
871 min = 0;
872 max = bits_mask(size);
873 }
874
875 range.pos = max;
876 range.neg = min;
877 return range;
878 }
879
val_in_range(struct range * range,long long sval,struct symbol * vtype)880 static int val_in_range(struct range *range, long long sval, struct symbol *vtype)
881 {
882 unsigned long long uval = sval;
883
884 if (is_signed_type(vtype) && (sval < 0))
885 return range->neg <= sval;
886 else
887 return uval <= range->pos;
888 }
889
cast_enum_list(struct symbol_list * list,struct symbol * base_type)890 static void cast_enum_list(struct symbol_list *list, struct symbol *base_type)
891 {
892 struct range irange = type_range(&int_ctype);
893 struct symbol *sym;
894
895 FOR_EACH_PTR(list, sym) {
896 struct expression *expr = sym->initializer;
897 struct symbol *ctype;
898 long long val;
899 if (expr->type != EXPR_VALUE)
900 continue;
901 ctype = expr->ctype;
902 val = get_expression_value(expr);
903 if (is_int_type(ctype) && val_in_range(&irange, val, ctype)) {
904 expr->ctype = &int_ctype;
905 continue;
906 }
907 if (ctype->bit_size == base_type->bit_size) {
908 expr->ctype = base_type;
909 continue;
910 }
911 cast_value(expr, base_type, expr, ctype);
912 expr->ctype = base_type;
913 } END_FOR_EACH_PTR(sym);
914 }
915
parse_enum_declaration(struct token * token,struct symbol * parent)916 static struct token *parse_enum_declaration(struct token *token, struct symbol *parent)
917 {
918 unsigned long long lastval = 0;
919 struct symbol *ctype = NULL, *base_type = NULL;
920 struct range range = { };
921 int mix_bitwise = 0;
922
923 parent->examined = 1;
924 parent->ctype.base_type = &int_ctype;
925 while (token_type(token) == TOKEN_IDENT) {
926 struct expression *expr = NULL;
927 struct token *next = token->next;
928 struct symbol *sym;
929
930 if (match_op(next, '=')) {
931 next = constant_expression(next->next, &expr);
932 lastval = get_expression_value(expr);
933 ctype = &void_ctype;
934 if (expr && expr->ctype)
935 ctype = expr->ctype;
936 } else if (!ctype) {
937 ctype = &int_ctype;
938 } else if (is_int_type(ctype)) {
939 lastval++;
940 } else {
941 error_die(token->pos, "can't increment the last enum member");
942 }
943
944 if (!expr) {
945 expr = alloc_expression(token->pos, EXPR_VALUE);
946 expr->value = lastval;
947 expr->ctype = ctype;
948 }
949
950 sym = alloc_symbol(token->pos, SYM_NODE);
951 bind_symbol(sym, token->ident, NS_SYMBOL);
952 sym->ctype.modifiers &= ~MOD_ADDRESSABLE;
953 sym->initializer = expr;
954 sym->enum_member = 1;
955 sym->ctype.base_type = parent;
956 add_ptr_list(&parent->symbol_list, sym);
957
958 if (base_type != &bad_ctype) {
959 if (ctype->type == SYM_NODE)
960 ctype = ctype->ctype.base_type;
961 if (ctype->type == SYM_ENUM) {
962 if (ctype == parent)
963 ctype = base_type;
964 else
965 ctype = ctype->ctype.base_type;
966 }
967 /*
968 * base_type rules:
969 * - if all enums are of the same type, then
970 * the base_type is that type (two first
971 * cases)
972 * - if enums are of different types, they
973 * all have to be integer types, and the
974 * base type is at least "int_ctype".
975 * - otherwise the base_type is "bad_ctype".
976 */
977 if (!base_type || ctype == &bad_ctype) {
978 base_type = ctype;
979 } else if (ctype == base_type) {
980 /* nothing */
981 } else if (is_int_type(base_type) && is_int_type(ctype)) {
982 base_type = &int_ctype;
983 } else if (is_restricted_type(base_type) != is_restricted_type(ctype)) {
984 if (!mix_bitwise++) {
985 warning(expr->pos, "mixed bitwiseness");
986 }
987 } else if (is_restricted_type(base_type) && base_type != ctype) {
988 sparse_error(expr->pos, "incompatible restricted type");
989 info(expr->pos, " expected: %s", show_typename(base_type));
990 info(expr->pos, " got: %s", show_typename(ctype));
991 base_type = &bad_ctype;
992 } else if (base_type != &bad_ctype) {
993 sparse_error(token->pos, "bad enum definition");
994 base_type = &bad_ctype;
995 }
996 parent->ctype.base_type = base_type;
997 }
998 if (is_int_type(base_type)) {
999 update_range(&range, lastval, ctype);
1000 }
1001 token = next;
1002
1003 sym->endpos = token->pos;
1004
1005 if (!match_op(token, ','))
1006 break;
1007 token = token->next;
1008 }
1009 if (!base_type) {
1010 sparse_error(token->pos, "empty enum definition");
1011 base_type = &bad_ctype;
1012 }
1013 else if (!is_int_type(base_type))
1014 ;
1015 else if (type_is_ok(&uint_ctype, range))
1016 base_type = &uint_ctype;
1017 else if (type_is_ok(&int_ctype, range))
1018 base_type = &int_ctype;
1019 else if (type_is_ok(&ulong_ctype, range))
1020 base_type = &ulong_ctype;
1021 else if (type_is_ok(&long_ctype, range))
1022 base_type = &long_ctype;
1023 else if (type_is_ok(&ullong_ctype, range))
1024 base_type = &ullong_ctype;
1025 else if (type_is_ok(&llong_ctype, range))
1026 base_type = &llong_ctype;
1027 else
1028 base_type = &bad_ctype;
1029 parent->ctype.base_type = base_type;
1030 parent->ctype.modifiers |= (base_type->ctype.modifiers & MOD_UNSIGNED);
1031 parent->examined = 0;
1032
1033 if (mix_bitwise)
1034 return token;
1035 cast_enum_list(parent->symbol_list, base_type);
1036
1037 return token;
1038 }
1039
enum_specifier(struct token * token,struct decl_state * ctx)1040 static struct token *enum_specifier(struct token *token, struct decl_state *ctx)
1041 {
1042 struct token *ret = struct_union_enum_specifier(SYM_ENUM, token, ctx, parse_enum_declaration);
1043 struct ctype *ctype = &ctx->ctype.base_type->ctype;
1044
1045 if (!ctype->base_type)
1046 ctype->base_type = &incomplete_ctype;
1047
1048 return ret;
1049 }
1050
typeof_specifier(struct token * token,struct decl_state * ctx)1051 static struct token *typeof_specifier(struct token *token, struct decl_state *ctx)
1052 {
1053 struct symbol *sym;
1054
1055 if (!match_op(token, '(')) {
1056 sparse_error(token->pos, "expected '(' after typeof");
1057 return token;
1058 }
1059 if (lookup_type(token->next)) {
1060 token = typename(token->next, &sym, NULL);
1061 ctx->ctype.base_type = sym->ctype.base_type;
1062 apply_ctype(token->pos, &sym->ctype, &ctx->ctype);
1063 } else {
1064 struct symbol *typeof_sym = alloc_symbol(token->pos, SYM_TYPEOF);
1065 token = parse_expression(token->next, &typeof_sym->initializer);
1066
1067 typeof_sym->endpos = token->pos;
1068 if (!typeof_sym->initializer) {
1069 sparse_error(token->pos, "expected expression after the '(' token");
1070 typeof_sym = &bad_ctype;
1071 }
1072 ctx->ctype.base_type = typeof_sym;
1073 }
1074 return expect(token, ')', "after typeof");
1075 }
1076
ignore_attribute(struct token * token,struct symbol * attr,struct decl_state * ctx)1077 static struct token *ignore_attribute(struct token *token, struct symbol *attr, struct decl_state *ctx)
1078 {
1079 struct expression *expr = NULL;
1080 if (match_op(token, '('))
1081 token = parens_expression(token, &expr, "in attribute");
1082 return token;
1083 }
1084
attribute_packed(struct token * token,struct symbol * attr,struct decl_state * ctx)1085 static struct token *attribute_packed(struct token *token, struct symbol *attr, struct decl_state *ctx)
1086 {
1087 if (!ctx->ctype.alignment)
1088 ctx->ctype.alignment = 1;
1089 return token;
1090 }
1091
attribute_aligned(struct token * token,struct symbol * attr,struct decl_state * ctx)1092 static struct token *attribute_aligned(struct token *token, struct symbol *attr, struct decl_state *ctx)
1093 {
1094 int alignment = max_alignment;
1095 struct expression *expr = NULL;
1096
1097 if (match_op(token, '(')) {
1098 token = parens_expression(token, &expr, "in attribute");
1099 if (expr)
1100 alignment = const_expression_value(expr);
1101 }
1102 if (alignment & (alignment-1)) {
1103 warning(token->pos, "I don't like non-power-of-2 alignments");
1104 return token;
1105 } else if (alignment > ctx->ctype.alignment)
1106 ctx->ctype.alignment = alignment;
1107 return token;
1108 }
1109
apply_qualifier(struct position * pos,struct ctype * ctx,unsigned long qual)1110 static void apply_qualifier(struct position *pos, struct ctype *ctx, unsigned long qual)
1111 {
1112 if (ctx->modifiers & qual)
1113 warning(*pos, "duplicate %s", modifier_string(qual));
1114 ctx->modifiers |= qual;
1115 }
1116
attribute_modifier(struct token * token,struct symbol * attr,struct decl_state * ctx)1117 static struct token *attribute_modifier(struct token *token, struct symbol *attr, struct decl_state *ctx)
1118 {
1119 apply_qualifier(&token->pos, &ctx->ctype, attr->ctype.modifiers);
1120 return token;
1121 }
1122
attribute_ext_visible(struct token * token,struct symbol * attr,struct decl_state * ctx)1123 static struct token *attribute_ext_visible(struct token *token, struct symbol *attr, struct decl_state *ctx)
1124 {
1125 ctx->is_ext_visible = 1;
1126 return token;
1127 }
1128
attribute_bitwise(struct token * token,struct symbol * attr,struct decl_state * ctx)1129 static struct token *attribute_bitwise(struct token *token, struct symbol *attr, struct decl_state *ctx)
1130 {
1131 if (Wbitwise)
1132 attribute_modifier(token, attr, ctx);
1133 return token;
1134 }
1135
numerical_address_space(int asn)1136 static struct ident *numerical_address_space(int asn)
1137 {
1138 char buff[32];
1139
1140 if (!asn)
1141 return NULL;
1142 sprintf(buff, "<asn:%d>", asn);
1143 return built_in_ident(buff);
1144 }
1145
attribute_address_space(struct token * token,struct symbol * attr,struct decl_state * ctx)1146 static struct token *attribute_address_space(struct token *token, struct symbol *attr, struct decl_state *ctx)
1147 {
1148 struct expression *expr = NULL;
1149 struct ident *as = NULL;
1150 struct token *next;
1151
1152 token = expect(token, '(', "after address_space attribute");
1153 switch (token_type(token)) {
1154 case TOKEN_NUMBER:
1155 next = primary_expression(token, &expr);
1156 if (expr->type != EXPR_VALUE)
1157 goto invalid;
1158 as = numerical_address_space(expr->value);
1159 break;
1160 case TOKEN_IDENT:
1161 next = token->next;
1162 as = token->ident;
1163 break;
1164 default:
1165 next = token->next;
1166 invalid:
1167 as = NULL;
1168 warning(token->pos, "invalid address space name");
1169 }
1170
1171 if (Waddress_space && as) {
1172 if (ctx->ctype.as)
1173 sparse_error(token->pos,
1174 "multiple address space given: %s & %s",
1175 show_as(ctx->ctype.as), show_as(as));
1176 ctx->ctype.as = as;
1177 }
1178 token = expect(next, ')', "after address_space attribute");
1179 return token;
1180 }
1181
to_QI_mode(struct symbol * ctype)1182 static struct symbol *to_QI_mode(struct symbol *ctype)
1183 {
1184 if (ctype->ctype.base_type != &int_type)
1185 return NULL;
1186 if (ctype == &char_ctype)
1187 return ctype;
1188 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uchar_ctype
1189 : &schar_ctype;
1190 }
1191
to_HI_mode(struct symbol * ctype)1192 static struct symbol *to_HI_mode(struct symbol *ctype)
1193 {
1194 if (ctype->ctype.base_type != &int_type)
1195 return NULL;
1196 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ushort_ctype
1197 : &sshort_ctype;
1198 }
1199
to_SI_mode(struct symbol * ctype)1200 static struct symbol *to_SI_mode(struct symbol *ctype)
1201 {
1202 if (ctype->ctype.base_type != &int_type)
1203 return NULL;
1204 return ctype->ctype.modifiers & MOD_UNSIGNED ? &uint_ctype
1205 : &sint_ctype;
1206 }
1207
to_DI_mode(struct symbol * ctype)1208 static struct symbol *to_DI_mode(struct symbol *ctype)
1209 {
1210 if (ctype->ctype.base_type != &int_type)
1211 return NULL;
1212 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ullong_ctype
1213 : &sllong_ctype;
1214 }
1215
to_TI_mode(struct symbol * ctype)1216 static struct symbol *to_TI_mode(struct symbol *ctype)
1217 {
1218 if (ctype->ctype.base_type != &int_type)
1219 return NULL;
1220 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulllong_ctype
1221 : &slllong_ctype;
1222 }
1223
to_pointer_mode(struct symbol * ctype)1224 static struct symbol *to_pointer_mode(struct symbol *ctype)
1225 {
1226 if (ctype->ctype.base_type != &int_type)
1227 return NULL;
1228 return ctype->ctype.modifiers & MOD_UNSIGNED ? uintptr_ctype
1229 : intptr_ctype;
1230 }
1231
to_word_mode(struct symbol * ctype)1232 static struct symbol *to_word_mode(struct symbol *ctype)
1233 {
1234 if (ctype->ctype.base_type != &int_type)
1235 return NULL;
1236 return ctype->ctype.modifiers & MOD_UNSIGNED ? &ulong_ctype
1237 : &slong_ctype;
1238 }
1239
attribute_mode(struct token * token,struct symbol * attr,struct decl_state * ctx)1240 static struct token *attribute_mode(struct token *token, struct symbol *attr, struct decl_state *ctx)
1241 {
1242 token = expect(token, '(', "after mode attribute");
1243 if (token_type(token) == TOKEN_IDENT) {
1244 struct symbol *mode = lookup_keyword(token->ident, NS_KEYWORD);
1245 if (mode && mode->op->type == KW_MODE)
1246 ctx->mode = mode->op;
1247 else
1248 sparse_error(token->pos, "unknown mode attribute %s", show_ident(token->ident));
1249 token = token->next;
1250 } else
1251 sparse_error(token->pos, "expect attribute mode symbol\n");
1252 token = expect(token, ')', "after mode attribute");
1253 return token;
1254 }
1255
attribute_context(struct token * token,struct symbol * attr,struct decl_state * ctx)1256 static struct token *attribute_context(struct token *token, struct symbol *attr, struct decl_state *ctx)
1257 {
1258 struct context *context = alloc_context();
1259 struct expression *args[3];
1260 int idx = 0;
1261
1262 token = expect(token, '(', "after context attribute");
1263 token = conditional_expression(token, &args[0]);
1264 token = expect(token, ',', "after context 1st argument");
1265 token = conditional_expression(token, &args[1]);
1266 if (match_op(token, ',')) {
1267 token = token->next;
1268 token = conditional_expression(token, &args[2]);
1269 token = expect(token, ')', "after context 3rd argument");
1270 context->context = args[0];
1271 idx++;
1272 } else {
1273 token = expect(token, ')', "after context 2nd argument");
1274 }
1275 context->in = get_expression_value(args[idx++]);
1276 context->out = get_expression_value(args[idx++]);
1277 add_ptr_list(&ctx->ctype.contexts, context);
1278 return token;
1279 }
1280
attribute_designated_init(struct token * token,struct symbol * attr,struct decl_state * ctx)1281 static struct token *attribute_designated_init(struct token *token, struct symbol *attr, struct decl_state *ctx)
1282 {
1283 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_STRUCT)
1284 ctx->ctype.base_type->designated_init = 1;
1285 else
1286 warning(token->pos, "attribute designated_init applied to non-structure type");
1287 return token;
1288 }
1289
attribute_transparent_union(struct token * token,struct symbol * attr,struct decl_state * ctx)1290 static struct token *attribute_transparent_union(struct token *token, struct symbol *attr, struct decl_state *ctx)
1291 {
1292 if (Wtransparent_union)
1293 warning(token->pos, "attribute __transparent_union__");
1294
1295 if (ctx->ctype.base_type && ctx->ctype.base_type->type == SYM_UNION)
1296 ctx->ctype.base_type->transparent_union = 1;
1297 else
1298 warning(token->pos, "attribute __transparent_union__ applied to non-union type");
1299 return token;
1300 }
1301
recover_unknown_attribute(struct token * token)1302 static struct token *recover_unknown_attribute(struct token *token)
1303 {
1304 struct expression *expr = NULL;
1305
1306 if (Wunknown_attribute)
1307 warning(token->pos, "unknown attribute '%s'", show_ident(token->ident));
1308 token = token->next;
1309 if (match_op(token, '('))
1310 token = parens_expression(token, &expr, "in attribute");
1311 return token;
1312 }
1313
attribute_specifier(struct token * token,struct decl_state * ctx)1314 static struct token *attribute_specifier(struct token *token, struct decl_state *ctx)
1315 {
1316 token = expect(token, '(', "after attribute");
1317 token = expect(token, '(', "after attribute");
1318
1319 for (;;) {
1320 struct ident *attribute_name;
1321 struct symbol *attr;
1322
1323 if (eof_token(token))
1324 break;
1325 if (match_op(token, ';'))
1326 break;
1327 if (token_type(token) != TOKEN_IDENT)
1328 break;
1329 attribute_name = token->ident;
1330 attr = lookup_keyword(attribute_name, NS_KEYWORD);
1331 if (attr && attr->op->attribute)
1332 token = attr->op->attribute(token->next, attr, ctx);
1333 else
1334 token = recover_unknown_attribute(token);
1335
1336 if (!match_op(token, ','))
1337 break;
1338 token = token->next;
1339 }
1340
1341 token = expect(token, ')', "after attribute");
1342 token = expect(token, ')', "after attribute");
1343 return token;
1344 }
1345
1346 static const char *storage_class[] =
1347 {
1348 [STypedef] = "typedef",
1349 [SAuto] = "auto",
1350 [SExtern] = "extern",
1351 [SStatic] = "static",
1352 [SRegister] = "register",
1353 [SForced] = "[force]"
1354 };
1355
storage_modifiers(struct decl_state * ctx)1356 static unsigned long storage_modifiers(struct decl_state *ctx)
1357 {
1358 static unsigned long mod[SMax] =
1359 {
1360 [SAuto] = MOD_AUTO,
1361 [SExtern] = MOD_EXTERN,
1362 [SStatic] = MOD_STATIC,
1363 [SRegister] = MOD_REGISTER
1364 };
1365 return mod[ctx->storage_class] | (ctx->is_inline ? MOD_INLINE : 0)
1366 | (ctx->is_tls ? MOD_TLS : 0)
1367 | (ctx->is_ext_visible ? MOD_EXT_VISIBLE : 0);
1368 }
1369
set_storage_class(struct position * pos,struct decl_state * ctx,int class)1370 static void set_storage_class(struct position *pos, struct decl_state *ctx, int class)
1371 {
1372 /* __thread can be used alone, or with extern or static */
1373 if (ctx->is_tls && (class != SStatic && class != SExtern)) {
1374 sparse_error(*pos, "__thread can only be used alone, or with "
1375 "extern or static");
1376 return;
1377 }
1378
1379 if (!ctx->storage_class) {
1380 ctx->storage_class = class;
1381 return;
1382 }
1383 if (ctx->storage_class == class)
1384 sparse_error(*pos, "duplicate %s", storage_class[class]);
1385 else
1386 sparse_error(*pos, "multiple storage classes");
1387 }
1388
typedef_specifier(struct token * next,struct decl_state * ctx)1389 static struct token *typedef_specifier(struct token *next, struct decl_state *ctx)
1390 {
1391 set_storage_class(&next->pos, ctx, STypedef);
1392 return next;
1393 }
1394
auto_specifier(struct token * next,struct decl_state * ctx)1395 static struct token *auto_specifier(struct token *next, struct decl_state *ctx)
1396 {
1397 set_storage_class(&next->pos, ctx, SAuto);
1398 return next;
1399 }
1400
register_specifier(struct token * next,struct decl_state * ctx)1401 static struct token *register_specifier(struct token *next, struct decl_state *ctx)
1402 {
1403 set_storage_class(&next->pos, ctx, SRegister);
1404 return next;
1405 }
1406
static_specifier(struct token * next,struct decl_state * ctx)1407 static struct token *static_specifier(struct token *next, struct decl_state *ctx)
1408 {
1409 set_storage_class(&next->pos, ctx, SStatic);
1410 return next;
1411 }
1412
extern_specifier(struct token * next,struct decl_state * ctx)1413 static struct token *extern_specifier(struct token *next, struct decl_state *ctx)
1414 {
1415 set_storage_class(&next->pos, ctx, SExtern);
1416 return next;
1417 }
1418
thread_specifier(struct token * next,struct decl_state * ctx)1419 static struct token *thread_specifier(struct token *next, struct decl_state *ctx)
1420 {
1421 /* This GCC extension can be used alone, or with extern or static */
1422 if (!ctx->storage_class || ctx->storage_class == SStatic
1423 || ctx->storage_class == SExtern) {
1424 ctx->is_tls = 1;
1425 } else {
1426 sparse_error(next->pos, "__thread can only be used alone, or "
1427 "with extern or static");
1428 }
1429
1430 return next;
1431 }
1432
attribute_force(struct token * token,struct symbol * attr,struct decl_state * ctx)1433 static struct token *attribute_force(struct token *token, struct symbol *attr, struct decl_state *ctx)
1434 {
1435 set_storage_class(&token->pos, ctx, SForced);
1436 return token;
1437 }
1438
inline_specifier(struct token * next,struct decl_state * ctx)1439 static struct token *inline_specifier(struct token *next, struct decl_state *ctx)
1440 {
1441 ctx->is_inline = 1;
1442 return next;
1443 }
1444
noreturn_specifier(struct token * next,struct decl_state * ctx)1445 static struct token *noreturn_specifier(struct token *next, struct decl_state *ctx)
1446 {
1447 apply_qualifier(&next->pos, &ctx->ctype, MOD_NORETURN);
1448 return next;
1449 }
1450
alignas_specifier(struct token * token,struct decl_state * ctx)1451 static struct token *alignas_specifier(struct token *token, struct decl_state *ctx)
1452 {
1453 int alignment = 0;
1454
1455 if (!match_op(token, '(')) {
1456 sparse_error(token->pos, "expected '(' after _Alignas");
1457 return token;
1458 }
1459 if (lookup_type(token->next)) {
1460 struct symbol *sym = NULL;
1461 token = typename(token->next, &sym, NULL);
1462 sym = examine_symbol_type(sym);
1463 alignment = sym->ctype.alignment;
1464 token = expect(token, ')', "after _Alignas(...");
1465 } else {
1466 struct expression *expr = NULL;
1467 token = parens_expression(token, &expr, "after _Alignas");
1468 if (!expr)
1469 return token;
1470 alignment = const_expression_value(expr);
1471 }
1472
1473 if (alignment < 0) {
1474 warning(token->pos, "non-positive alignment");
1475 return token;
1476 }
1477 if (alignment & (alignment-1)) {
1478 warning(token->pos, "non-power-of-2 alignment");
1479 return token;
1480 }
1481 if (alignment > ctx->ctype.alignment)
1482 ctx->ctype.alignment = alignment;
1483 return token;
1484 }
1485
const_qualifier(struct token * next,struct decl_state * ctx)1486 static struct token *const_qualifier(struct token *next, struct decl_state *ctx)
1487 {
1488 apply_qualifier(&next->pos, &ctx->ctype, MOD_CONST);
1489 return next;
1490 }
1491
volatile_qualifier(struct token * next,struct decl_state * ctx)1492 static struct token *volatile_qualifier(struct token *next, struct decl_state *ctx)
1493 {
1494 apply_qualifier(&next->pos, &ctx->ctype, MOD_VOLATILE);
1495 return next;
1496 }
1497
restrict_qualifier(struct token * next,struct decl_state * ctx)1498 static struct token *restrict_qualifier(struct token *next, struct decl_state *ctx)
1499 {
1500 apply_qualifier(&next->pos, &ctx->ctype, MOD_RESTRICT);
1501 return next;
1502 }
1503
atomic_qualifier(struct token * next,struct decl_state * ctx)1504 static struct token *atomic_qualifier(struct token *next, struct decl_state *ctx)
1505 {
1506 apply_qualifier(&next->pos, &ctx->ctype, MOD_ATOMIC);
1507 return next;
1508 }
1509
apply_ctype(struct position pos,struct ctype * thistype,struct ctype * ctype)1510 static void apply_ctype(struct position pos, struct ctype *thistype, struct ctype *ctype)
1511 {
1512 unsigned long mod = thistype->modifiers;
1513
1514 if (mod)
1515 apply_qualifier(&pos, ctype, mod);
1516
1517 /* Context */
1518 concat_ptr_list((struct ptr_list *)thistype->contexts,
1519 (struct ptr_list **)&ctype->contexts);
1520
1521 /* Alignment */
1522 if (thistype->alignment > ctype->alignment)
1523 ctype->alignment = thistype->alignment;
1524
1525 /* Address space */
1526 if (thistype->as)
1527 ctype->as = thistype->as;
1528 }
1529
specifier_conflict(struct position pos,int what,struct ident * new)1530 static void specifier_conflict(struct position pos, int what, struct ident *new)
1531 {
1532 const char *old;
1533 if (what & (Set_S | Set_T))
1534 goto Catch_all;
1535 if (what & Set_Char)
1536 old = "char";
1537 else if (what & Set_Double)
1538 old = "double";
1539 else if (what & Set_Float)
1540 old = "float";
1541 else if (what & Set_Signed)
1542 old = "signed";
1543 else if (what & Set_Unsigned)
1544 old = "unsigned";
1545 else if (what & Set_Short)
1546 old = "short";
1547 else if (what & Set_Long)
1548 old = "long";
1549 else
1550 old = "long long";
1551 sparse_error(pos, "impossible combination of type specifiers: %s %s",
1552 old, show_ident(new));
1553 return;
1554
1555 Catch_all:
1556 sparse_error(pos, "two or more data types in declaration specifiers");
1557 }
1558
1559 static struct symbol * const int_types[] =
1560 {&short_ctype, &int_ctype, &long_ctype, &llong_ctype, &lllong_ctype};
1561 static struct symbol * const signed_types[] =
1562 {&sshort_ctype, &sint_ctype, &slong_ctype, &sllong_ctype,
1563 &slllong_ctype};
1564 static struct symbol * const unsigned_types[] =
1565 {&ushort_ctype, &uint_ctype, &ulong_ctype, &ullong_ctype,
1566 &ulllong_ctype};
1567 static struct symbol * const real_types[] =
1568 {&float_ctype, &double_ctype, &ldouble_ctype};
1569 static struct symbol * const char_types[] =
1570 {&char_ctype, &schar_ctype, &uchar_ctype};
1571 static struct symbol * const * const types[] = {
1572 int_types + 1, signed_types + 1, unsigned_types + 1,
1573 real_types + 1, char_types, char_types + 1, char_types + 2
1574 };
1575
ctype_integer(int size,int want_unsigned)1576 struct symbol *ctype_integer(int size, int want_unsigned)
1577 {
1578 return types[want_unsigned ? CUInt : CInt][size];
1579 }
1580
handle_qualifiers(struct token * t,struct decl_state * ctx)1581 static struct token *handle_qualifiers(struct token *t, struct decl_state *ctx)
1582 {
1583 while (token_type(t) == TOKEN_IDENT) {
1584 struct symbol *s = lookup_symbol(t->ident, NS_TYPEDEF);
1585 if (!s)
1586 break;
1587 if (s->type != SYM_KEYWORD)
1588 break;
1589 if (!(s->op->type & (KW_ATTRIBUTE | KW_QUALIFIER)))
1590 break;
1591 t = t->next;
1592 if (s->op->declarator)
1593 t = s->op->declarator(t, ctx);
1594 }
1595 return t;
1596 }
1597
declaration_specifiers(struct token * token,struct decl_state * ctx)1598 static struct token *declaration_specifiers(struct token *token, struct decl_state *ctx)
1599 {
1600 int seen = 0;
1601 int class = CInt;
1602 int size = 0;
1603
1604 while (token_type(token) == TOKEN_IDENT) {
1605 struct symbol *s = lookup_symbol(token->ident,
1606 NS_TYPEDEF | NS_SYMBOL);
1607 if (!s || !(s->namespace & NS_TYPEDEF))
1608 break;
1609 if (s->type != SYM_KEYWORD) {
1610 if (seen & Set_Any)
1611 break;
1612 seen |= Set_S | Set_T;
1613 ctx->ctype.base_type = s->ctype.base_type;
1614 apply_ctype(token->pos, &s->ctype, &ctx->ctype);
1615 token = token->next;
1616 continue;
1617 }
1618 if (s->op->type & KW_SPECIFIER) {
1619 if (seen & s->op->test) {
1620 specifier_conflict(token->pos,
1621 seen & s->op->test,
1622 token->ident);
1623 break;
1624 }
1625 seen |= s->op->set;
1626 class += s->op->class;
1627 if (s->op->set & Set_Int128)
1628 size = 2;
1629 if (s->op->type & KW_SHORT) {
1630 size = -1;
1631 } else if (s->op->type & KW_LONG && size++) {
1632 if (class == CReal) {
1633 specifier_conflict(token->pos,
1634 Set_Vlong,
1635 &double_ident);
1636 break;
1637 }
1638 seen |= Set_Vlong;
1639 }
1640 }
1641 token = token->next;
1642 if (s->op->declarator)
1643 token = s->op->declarator(token, ctx);
1644 if (s->op->type & KW_EXACT) {
1645 ctx->ctype.base_type = s->ctype.base_type;
1646 ctx->ctype.modifiers |= s->ctype.modifiers;
1647 }
1648 }
1649
1650 if (!(seen & Set_S)) { /* not set explicitly? */
1651 struct symbol *base = &incomplete_ctype;
1652 if (seen & Set_Any)
1653 base = types[class][size];
1654 ctx->ctype.base_type = base;
1655 }
1656
1657 if (ctx->ctype.modifiers & MOD_BITWISE) {
1658 struct symbol *type;
1659 ctx->ctype.modifiers &= ~MOD_BITWISE;
1660 if (!is_int_type(ctx->ctype.base_type)) {
1661 sparse_error(token->pos, "invalid modifier");
1662 return token;
1663 }
1664 type = alloc_symbol(token->pos, SYM_BASETYPE);
1665 *type = *ctx->ctype.base_type;
1666 type->ctype.modifiers &= ~MOD_SPECIFIER;
1667 type->ctype.base_type = ctx->ctype.base_type;
1668 type->type = SYM_RESTRICT;
1669 ctx->ctype.base_type = type;
1670 create_fouled(type);
1671 }
1672 return token;
1673 }
1674
abstract_array_static_declarator(struct token * token,int * has_static)1675 static struct token *abstract_array_static_declarator(struct token *token, int *has_static)
1676 {
1677 while (token->ident == &static_ident) {
1678 if (*has_static)
1679 sparse_error(token->pos, "duplicate array static declarator");
1680
1681 *has_static = 1;
1682 token = token->next;
1683 }
1684 return token;
1685
1686 }
1687
abstract_array_declarator(struct token * token,struct symbol * sym)1688 static struct token *abstract_array_declarator(struct token *token, struct symbol *sym)
1689 {
1690 struct expression *expr = NULL;
1691 int has_static = 0;
1692
1693 token = abstract_array_static_declarator(token, &has_static);
1694
1695 if (match_idents(token, &restrict_ident, &__restrict_ident, &__restrict___ident, NULL))
1696 token = abstract_array_static_declarator(token->next, &has_static);
1697 token = parse_expression(token, &expr);
1698 sym->array_size = expr;
1699 return token;
1700 }
1701
1702 static struct token *parameter_type_list(struct token *, struct symbol *);
1703 static struct token *identifier_list(struct token *, struct symbol *);
1704 static struct token *declarator(struct token *token, struct decl_state *ctx);
1705
skip_attribute(struct token * token)1706 static struct token *skip_attribute(struct token *token)
1707 {
1708 token = token->next;
1709 if (match_op(token, '(')) {
1710 int depth = 1;
1711 token = token->next;
1712 while (depth && !eof_token(token)) {
1713 if (token_type(token) == TOKEN_SPECIAL) {
1714 if (token->special == '(')
1715 depth++;
1716 else if (token->special == ')')
1717 depth--;
1718 }
1719 token = token->next;
1720 }
1721 }
1722 return token;
1723 }
1724
skip_attributes(struct token * token)1725 static struct token *skip_attributes(struct token *token)
1726 {
1727 struct symbol *keyword;
1728 for (;;) {
1729 if (token_type(token) != TOKEN_IDENT)
1730 break;
1731 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1732 if (!keyword || keyword->type != SYM_KEYWORD)
1733 break;
1734 if (!(keyword->op->type & KW_ATTRIBUTE))
1735 break;
1736 token = expect(token->next, '(', "after attribute");
1737 token = expect(token, '(', "after attribute");
1738 for (;;) {
1739 if (eof_token(token))
1740 break;
1741 if (match_op(token, ';'))
1742 break;
1743 if (token_type(token) != TOKEN_IDENT)
1744 break;
1745 token = skip_attribute(token);
1746 if (!match_op(token, ','))
1747 break;
1748 token = token->next;
1749 }
1750 token = expect(token, ')', "after attribute");
1751 token = expect(token, ')', "after attribute");
1752 }
1753 return token;
1754 }
1755
handle_attributes(struct token * token,struct decl_state * ctx,unsigned int keywords)1756 static struct token *handle_attributes(struct token *token, struct decl_state *ctx, unsigned int keywords)
1757 {
1758 struct symbol *keyword;
1759 for (;;) {
1760 if (token_type(token) != TOKEN_IDENT)
1761 break;
1762 keyword = lookup_keyword(token->ident, NS_KEYWORD | NS_TYPEDEF);
1763 if (!keyword || keyword->type != SYM_KEYWORD)
1764 break;
1765 if (!(keyword->op->type & keywords))
1766 break;
1767 token = keyword->op->declarator(token->next, ctx);
1768 keywords &= KW_ATTRIBUTE;
1769 }
1770 return token;
1771 }
1772
is_nested(struct token * token,struct token ** p,int prefer_abstract)1773 static int is_nested(struct token *token, struct token **p,
1774 int prefer_abstract)
1775 {
1776 /*
1777 * This can be either a parameter list or a grouping.
1778 * For the direct (non-abstract) case, we know if must be
1779 * a parameter list if we already saw the identifier.
1780 * For the abstract case, we know if must be a parameter
1781 * list if it is empty or starts with a type.
1782 */
1783 struct token *next = token->next;
1784
1785 *p = next = skip_attributes(next);
1786
1787 if (token_type(next) == TOKEN_IDENT) {
1788 if (lookup_type(next))
1789 return !prefer_abstract;
1790 return 1;
1791 }
1792
1793 if (match_op(next, ')') || match_op(next, SPECIAL_ELLIPSIS))
1794 return 0;
1795
1796 return 1;
1797 }
1798
1799 enum kind {
1800 Empty, K_R, Proto, Bad_Func,
1801 };
1802
which_func(struct token * token,struct ident ** n,int prefer_abstract)1803 static enum kind which_func(struct token *token,
1804 struct ident **n,
1805 int prefer_abstract)
1806 {
1807 struct token *next = token->next;
1808
1809 if (token_type(next) == TOKEN_IDENT) {
1810 if (lookup_type(next))
1811 return Proto;
1812 /* identifier list not in definition; complain */
1813 if (prefer_abstract)
1814 warning(token->pos,
1815 "identifier list not in definition");
1816 return K_R;
1817 }
1818
1819 if (token_type(next) != TOKEN_SPECIAL)
1820 return Bad_Func;
1821
1822 if (next->special == ')') {
1823 /* don't complain about those */
1824 if (!n || match_op(next->next, ';') || match_op(next->next, ','))
1825 return Empty;
1826 if (Wstrict_prototypes)
1827 warning(next->pos,
1828 "non-ANSI function declaration of function '%s'",
1829 show_ident(*n));
1830 return Empty;
1831 }
1832
1833 if (next->special == SPECIAL_ELLIPSIS) {
1834 warning(next->pos,
1835 "variadic functions must have one named argument");
1836 return Proto;
1837 }
1838
1839 return Bad_Func;
1840 }
1841
direct_declarator(struct token * token,struct decl_state * ctx)1842 static struct token *direct_declarator(struct token *token, struct decl_state *ctx)
1843 {
1844 struct ctype *ctype = &ctx->ctype;
1845 struct token *next;
1846 struct ident **p = ctx->ident;
1847
1848 if (ctx->ident && token_type(token) == TOKEN_IDENT) {
1849 *ctx->ident = token->ident;
1850 token = token->next;
1851 } else if (match_op(token, '(') &&
1852 is_nested(token, &next, ctx->prefer_abstract)) {
1853 struct symbol *base_type = ctype->base_type;
1854 if (token->next != next)
1855 next = handle_attributes(token->next, ctx,
1856 KW_ATTRIBUTE);
1857 token = declarator(next, ctx);
1858 token = expect(token, ')', "in nested declarator");
1859 while (ctype->base_type != base_type)
1860 ctype = &ctype->base_type->ctype;
1861 p = NULL;
1862 }
1863
1864 if (match_op(token, '(')) {
1865 enum kind kind = which_func(token, p, ctx->prefer_abstract);
1866 struct symbol *fn;
1867 fn = alloc_indirect_symbol(token->pos, ctype, SYM_FN);
1868 token = token->next;
1869 if (kind == K_R)
1870 token = identifier_list(token, fn);
1871 else if (kind == Proto)
1872 token = parameter_type_list(token, fn);
1873 token = expect(token, ')', "in function declarator");
1874 fn->endpos = token->pos;
1875 return token;
1876 }
1877
1878 while (match_op(token, '[')) {
1879 struct symbol *array;
1880 array = alloc_indirect_symbol(token->pos, ctype, SYM_ARRAY);
1881 token = abstract_array_declarator(token->next, array);
1882 token = expect(token, ']', "in abstract_array_declarator");
1883 array->endpos = token->pos;
1884 ctype = &array->ctype;
1885 }
1886 return token;
1887 }
1888
pointer(struct token * token,struct decl_state * ctx)1889 static struct token *pointer(struct token *token, struct decl_state *ctx)
1890 {
1891 while (match_op(token,'*')) {
1892 struct symbol *ptr = alloc_symbol(token->pos, SYM_PTR);
1893 ptr->ctype.modifiers = ctx->ctype.modifiers;
1894 ptr->ctype.base_type = ctx->ctype.base_type;
1895 ptr->ctype.as = ctx->ctype.as;
1896 ptr->ctype.contexts = ctx->ctype.contexts;
1897 ctx->ctype.modifiers = 0;
1898 ctx->ctype.base_type = ptr;
1899 ctx->ctype.as = NULL;
1900 ctx->ctype.contexts = NULL;
1901 ctx->ctype.alignment = 0;
1902
1903 token = handle_qualifiers(token->next, ctx);
1904 ctx->ctype.base_type->endpos = token->pos;
1905 }
1906 return token;
1907 }
1908
declarator(struct token * token,struct decl_state * ctx)1909 static struct token *declarator(struct token *token, struct decl_state *ctx)
1910 {
1911 token = pointer(token, ctx);
1912 return direct_declarator(token, ctx);
1913 }
1914
handle_bitfield(struct token * token,struct decl_state * ctx)1915 static struct token *handle_bitfield(struct token *token, struct decl_state *ctx)
1916 {
1917 struct ctype *ctype = &ctx->ctype;
1918 struct expression *expr;
1919 struct symbol *bitfield;
1920 long long width;
1921
1922 if (ctype->base_type != &int_type && !is_int_type(ctype->base_type)) {
1923 sparse_error(token->pos, "invalid bitfield specifier for type %s.",
1924 show_typename(ctype->base_type));
1925 // Parse this to recover gracefully.
1926 return conditional_expression(token->next, &expr);
1927 }
1928
1929 bitfield = alloc_indirect_symbol(token->pos, ctype, SYM_BITFIELD);
1930 token = conditional_expression(token->next, &expr);
1931 width = const_expression_value(expr);
1932 bitfield->bit_size = width;
1933
1934 if (width < 0 || width > INT_MAX) {
1935 sparse_error(token->pos, "invalid bitfield width, %lld.", width);
1936 width = -1;
1937 } else if (*ctx->ident && width == 0) {
1938 sparse_error(token->pos, "invalid named zero-width bitfield `%s'",
1939 show_ident(*ctx->ident));
1940 width = -1;
1941 } else if (*ctx->ident) {
1942 struct symbol *base_type = bitfield->ctype.base_type;
1943 struct symbol *bitfield_type = base_type == &int_type ? bitfield : base_type;
1944 int is_signed = !(bitfield_type->ctype.modifiers & MOD_UNSIGNED);
1945 if (Wone_bit_signed_bitfield && width == 1 && is_signed) {
1946 // Valid values are either {-1;0} or {0}, depending on integer
1947 // representation. The latter makes for very efficient code...
1948 sparse_error(token->pos, "dubious one-bit signed bitfield");
1949 }
1950 if (Wdefault_bitfield_sign &&
1951 bitfield_type->type != SYM_ENUM &&
1952 !(bitfield_type->ctype.modifiers & MOD_EXPLICITLY_SIGNED) &&
1953 is_signed) {
1954 // The sign of bitfields is unspecified by default.
1955 warning(token->pos, "dubious bitfield without explicit `signed' or `unsigned'");
1956 }
1957 }
1958 bitfield->bit_size = width;
1959 bitfield->endpos = token->pos;
1960 return token;
1961 }
1962
declaration_list(struct token * token,struct symbol_list ** list)1963 static struct token *declaration_list(struct token *token, struct symbol_list **list)
1964 {
1965 struct decl_state ctx = {.prefer_abstract = 0};
1966 struct ctype saved;
1967 unsigned long mod;
1968
1969 token = declaration_specifiers(token, &ctx);
1970 mod = storage_modifiers(&ctx);
1971 saved = ctx.ctype;
1972 for (;;) {
1973 struct symbol *decl = alloc_symbol(token->pos, SYM_NODE);
1974 ctx.ident = &decl->ident;
1975
1976 token = declarator(token, &ctx);
1977 if (match_op(token, ':'))
1978 token = handle_bitfield(token, &ctx);
1979
1980 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
1981 apply_modifiers(token->pos, &ctx);
1982
1983 decl->ctype = ctx.ctype;
1984 decl->ctype.modifiers |= mod;
1985 decl->endpos = token->pos;
1986 add_symbol(list, decl);
1987 if (!match_op(token, ','))
1988 break;
1989 token = token->next;
1990 ctx.ctype = saved;
1991 }
1992 return token;
1993 }
1994
struct_declaration_list(struct token * token,struct symbol_list ** list)1995 static struct token *struct_declaration_list(struct token *token, struct symbol_list **list)
1996 {
1997 while (!match_op(token, '}')) {
1998 if (match_ident(token, &_Static_assert_ident)) {
1999 token = parse_static_assert(token, NULL);
2000 continue;
2001 }
2002 if (!match_op(token, ';'))
2003 token = declaration_list(token, list);
2004 if (!match_op(token, ';')) {
2005 sparse_error(token->pos, "expected ; at end of declaration");
2006 break;
2007 }
2008 token = token->next;
2009 }
2010 return token;
2011 }
2012
parameter_declaration(struct token * token,struct symbol * sym)2013 static struct token *parameter_declaration(struct token *token, struct symbol *sym)
2014 {
2015 struct decl_state ctx = {.prefer_abstract = 1};
2016
2017 token = declaration_specifiers(token, &ctx);
2018 ctx.ident = &sym->ident;
2019 token = declarator(token, &ctx);
2020 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
2021 apply_modifiers(token->pos, &ctx);
2022 sym->ctype = ctx.ctype;
2023 sym->ctype.modifiers |= storage_modifiers(&ctx);
2024 sym->endpos = token->pos;
2025 sym->forced_arg = ctx.storage_class == SForced;
2026 return token;
2027 }
2028
typename(struct token * token,struct symbol ** p,int * forced)2029 struct token *typename(struct token *token, struct symbol **p, int *forced)
2030 {
2031 struct decl_state ctx = {.prefer_abstract = 1};
2032 int class;
2033 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2034 *p = sym;
2035 token = declaration_specifiers(token, &ctx);
2036 token = declarator(token, &ctx);
2037 apply_modifiers(token->pos, &ctx);
2038 sym->ctype = ctx.ctype;
2039 sym->endpos = token->pos;
2040 class = ctx.storage_class;
2041 if (forced) {
2042 *forced = 0;
2043 if (class == SForced) {
2044 *forced = 1;
2045 class = 0;
2046 }
2047 }
2048 if (class)
2049 warning(sym->pos, "storage class in typename (%s %s)",
2050 storage_class[class], show_typename(sym));
2051 return token;
2052 }
2053
parse_underscore_Pragma(struct token * token)2054 static struct token *parse_underscore_Pragma(struct token *token)
2055 {
2056 struct token *next;
2057
2058 next = token->next;
2059 if (!match_op(next, '('))
2060 return next;
2061 next = next->next;
2062 if (next->pos.type != TOKEN_STRING)
2063 return next;
2064 next = next->next;
2065 if (!match_op(next, ')'))
2066 return next;
2067 return next->next;
2068 }
2069
expression_statement(struct token * token,struct expression ** tree)2070 static struct token *expression_statement(struct token *token, struct expression **tree)
2071 {
2072 if (match_ident(token, &_Pragma_ident))
2073 return parse_underscore_Pragma(token);
2074
2075 token = parse_expression(token, tree);
2076 return expect(token, ';', "at end of statement");
2077 }
2078
parse_asm_operands(struct token * token,struct statement * stmt,struct expression_list ** inout)2079 static struct token *parse_asm_operands(struct token *token, struct statement *stmt,
2080 struct expression_list **inout)
2081 {
2082 /* Allow empty operands */
2083 if (match_op(token->next, ':') || match_op(token->next, ')'))
2084 return token->next;
2085 do {
2086 struct expression *op = alloc_expression(token->pos, EXPR_ASM_OPERAND);
2087 if (match_op(token->next, '[') &&
2088 token_type(token->next->next) == TOKEN_IDENT &&
2089 match_op(token->next->next->next, ']')) {
2090 op->name = token->next->next->ident;
2091 token = token->next->next->next;
2092 }
2093 token = primary_expression(token->next, &op->constraint);
2094 token = parens_expression(token, &op->expr, "in asm parameter");
2095 add_expression(inout, op);
2096 } while (match_op(token, ','));
2097 return token;
2098 }
2099
parse_asm_clobbers(struct token * token,struct statement * stmt,struct expression_list ** clobbers)2100 static struct token *parse_asm_clobbers(struct token *token, struct statement *stmt,
2101 struct expression_list **clobbers)
2102 {
2103 struct expression *expr;
2104
2105 do {
2106 token = primary_expression(token->next, &expr);
2107 if (expr)
2108 add_expression(clobbers, expr);
2109 } while (match_op(token, ','));
2110 return token;
2111 }
2112
parse_asm_labels(struct token * token,struct statement * stmt,struct symbol_list ** labels)2113 static struct token *parse_asm_labels(struct token *token, struct statement *stmt,
2114 struct symbol_list **labels)
2115 {
2116 struct symbol *label;
2117
2118 do {
2119 token = token->next; /* skip ':' and ',' */
2120 if (token_type(token) != TOKEN_IDENT)
2121 return token;
2122 label = label_symbol(token);
2123 add_symbol(labels, label);
2124 token = token->next;
2125 } while (match_op(token, ','));
2126 return token;
2127 }
2128
parse_asm_statement(struct token * token,struct statement * stmt)2129 static struct token *parse_asm_statement(struct token *token, struct statement *stmt)
2130 {
2131 unsigned long mods = 0;
2132
2133 token = token->next;
2134 stmt->type = STMT_ASM;
2135 while (token_type(token) == TOKEN_IDENT) {
2136 struct symbol *s = lookup_keyword(token->ident, NS_TYPEDEF);
2137 if (s && s->op && s->op->asm_modifier)
2138 s->op->asm_modifier(token, &mods);
2139 else if (token->ident == &goto_ident)
2140 asm_modifier(token, &mods, MOD_ASM_GOTO);
2141 token = token->next;
2142 }
2143 token = expect(token, '(', "after asm");
2144 token = parse_expression(token, &stmt->asm_string);
2145 if (match_op(token, ':'))
2146 token = parse_asm_operands(token, stmt, &stmt->asm_outputs);
2147 if (match_op(token, ':'))
2148 token = parse_asm_operands(token, stmt, &stmt->asm_inputs);
2149 if (match_op(token, ':'))
2150 token = parse_asm_clobbers(token, stmt, &stmt->asm_clobbers);
2151 if (match_op(token, ':') && (mods & MOD_ASM_GOTO))
2152 token = parse_asm_labels(token, stmt, &stmt->asm_labels);
2153 token = expect(token, ')', "after asm");
2154 return expect(token, ';', "at end of asm-statement");
2155 }
2156
parse_asm_declarator(struct token * token,struct decl_state * ctx)2157 static struct token *parse_asm_declarator(struct token *token, struct decl_state *ctx)
2158 {
2159 struct expression *expr;
2160 token = expect(token, '(', "after asm");
2161 token = parse_expression(token->next, &expr);
2162 token = expect(token, ')', "after asm");
2163 return token;
2164 }
2165
parse_static_assert(struct token * token,struct symbol_list ** unused)2166 static struct token *parse_static_assert(struct token *token, struct symbol_list **unused)
2167 {
2168 struct expression *cond = NULL, *message = NULL;
2169
2170 token = expect(token->next, '(', "after _Static_assert");
2171 token = constant_expression(token, &cond);
2172 if (!cond)
2173 sparse_error(token->pos, "Expected constant expression");
2174 token = expect(token, ',', "after conditional expression in _Static_assert");
2175 token = parse_expression(token, &message);
2176 if (!message || message->type != EXPR_STRING) {
2177 struct position pos;
2178
2179 pos = message ? message->pos : token->pos;
2180 sparse_error(pos, "bad or missing string literal");
2181 cond = NULL;
2182 }
2183 token = expect(token, ')', "after diagnostic message in _Static_assert");
2184
2185 token = expect(token, ';', "after _Static_assert()");
2186
2187 if (cond && !const_expression_value(cond) && cond->type == EXPR_VALUE)
2188 sparse_error(cond->pos, "static assertion failed: %s",
2189 show_string(message->string));
2190 return token;
2191 }
2192
2193 /* Make a statement out of an expression */
make_statement(struct expression * expr)2194 static struct statement *make_statement(struct expression *expr)
2195 {
2196 struct statement *stmt;
2197
2198 if (!expr)
2199 return NULL;
2200 stmt = alloc_statement(expr->pos, STMT_EXPRESSION);
2201 stmt->expression = expr;
2202 return stmt;
2203 }
2204
2205 /*
2206 * All iterators have two symbols associated with them:
2207 * the "continue" and "break" symbols, which are targets
2208 * for continue and break statements respectively.
2209 *
2210 * They are in a special name-space, but they follow
2211 * all the normal visibility rules, so nested iterators
2212 * automatically work right.
2213 */
start_iterator(struct statement * stmt)2214 static void start_iterator(struct statement *stmt)
2215 {
2216 struct symbol *cont, *brk;
2217
2218 start_symbol_scope(stmt->pos);
2219 cont = alloc_symbol(stmt->pos, SYM_NODE);
2220 bind_symbol(cont, &continue_ident, NS_ITERATOR);
2221 brk = alloc_symbol(stmt->pos, SYM_NODE);
2222 bind_symbol(brk, &break_ident, NS_ITERATOR);
2223
2224 stmt->type = STMT_ITERATOR;
2225 stmt->iterator_break = brk;
2226 stmt->iterator_continue = cont;
2227 fn_local_symbol(brk);
2228 fn_local_symbol(cont);
2229 }
2230
end_iterator(struct statement * stmt)2231 static void end_iterator(struct statement *stmt)
2232 {
2233 end_symbol_scope();
2234 }
2235
start_function(struct symbol * sym)2236 static struct statement *start_function(struct symbol *sym)
2237 {
2238 struct symbol *ret;
2239 struct statement *stmt = alloc_statement(sym->pos, STMT_COMPOUND);
2240
2241 start_function_scope(sym->pos);
2242 ret = alloc_symbol(sym->pos, SYM_NODE);
2243 ret->ctype = sym->ctype.base_type->ctype;
2244 ret->ctype.modifiers &= ~(MOD_STORAGE | MOD_QUALIFIER | MOD_TLS | MOD_ACCESS | MOD_NOCAST | MOD_NODEREF);
2245 ret->ctype.modifiers |= (MOD_AUTO | MOD_REGISTER);
2246 bind_symbol(ret, &return_ident, NS_ITERATOR);
2247 stmt->ret = ret;
2248 fn_local_symbol(ret);
2249
2250 // Currently parsed symbol for __func__/__FUNCTION__/__PRETTY_FUNCTION__
2251 current_fn = sym;
2252
2253 return stmt;
2254 }
2255
end_function(struct symbol * sym)2256 static void end_function(struct symbol *sym)
2257 {
2258 current_fn = NULL;
2259 end_function_scope();
2260 }
2261
2262 /*
2263 * A "switch()" statement, like an iterator, has a
2264 * the "break" symbol associated with it. It works
2265 * exactly like the iterator break - it's the target
2266 * for any break-statements in scope, and means that
2267 * "break" handling doesn't even need to know whether
2268 * it's breaking out of an iterator or a switch.
2269 *
2270 * In addition, the "case" symbol is a marker for the
2271 * case/default statements to find the switch statement
2272 * that they are associated with.
2273 */
start_switch(struct statement * stmt)2274 static void start_switch(struct statement *stmt)
2275 {
2276 struct symbol *brk, *switch_case;
2277
2278 start_symbol_scope(stmt->pos);
2279 brk = alloc_symbol(stmt->pos, SYM_NODE);
2280 bind_symbol(brk, &break_ident, NS_ITERATOR);
2281
2282 switch_case = alloc_symbol(stmt->pos, SYM_NODE);
2283 bind_symbol(switch_case, &case_ident, NS_ITERATOR);
2284 switch_case->stmt = stmt;
2285
2286 stmt->type = STMT_SWITCH;
2287 stmt->switch_break = brk;
2288 stmt->switch_case = switch_case;
2289
2290 fn_local_symbol(brk);
2291 fn_local_symbol(switch_case);
2292 }
2293
end_switch(struct statement * stmt)2294 static void end_switch(struct statement *stmt)
2295 {
2296 if (!stmt->switch_case->symbol_list)
2297 warning(stmt->pos, "switch with no cases");
2298 end_symbol_scope();
2299 }
2300
add_case_statement(struct statement * stmt)2301 static void add_case_statement(struct statement *stmt)
2302 {
2303 struct symbol *target = lookup_symbol(&case_ident, NS_ITERATOR);
2304 struct symbol *sym;
2305
2306 if (!target) {
2307 sparse_error(stmt->pos, "not in switch scope");
2308 stmt->type = STMT_NONE;
2309 return;
2310 }
2311 sym = alloc_symbol(stmt->pos, SYM_NODE);
2312 add_symbol(&target->symbol_list, sym);
2313 sym->stmt = stmt;
2314 stmt->case_label = sym;
2315 fn_local_symbol(sym);
2316 }
2317
parse_return_statement(struct token * token,struct statement * stmt)2318 static struct token *parse_return_statement(struct token *token, struct statement *stmt)
2319 {
2320 struct symbol *target = lookup_symbol(&return_ident, NS_ITERATOR);
2321
2322 if (!target)
2323 error_die(token->pos, "internal error: return without a function target");
2324 stmt->type = STMT_RETURN;
2325 stmt->ret_target = target;
2326 return expression_statement(token->next, &stmt->ret_value);
2327 }
2328
validate_for_loop_decl(struct symbol * sym)2329 static void validate_for_loop_decl(struct symbol *sym)
2330 {
2331 unsigned long storage = sym->ctype.modifiers & MOD_STORAGE;
2332
2333 if (storage & ~(MOD_AUTO | MOD_REGISTER)) {
2334 const char *name = show_ident(sym->ident);
2335 sparse_error(sym->pos, "non-local var '%s' in for-loop initializer", name);
2336 sym->ctype.modifiers &= ~MOD_STORAGE;
2337 }
2338 }
2339
parse_for_statement(struct token * token,struct statement * stmt)2340 static struct token *parse_for_statement(struct token *token, struct statement *stmt)
2341 {
2342 struct symbol_list *syms;
2343 struct expression *e1, *e2, *e3;
2344 struct statement *iterator;
2345
2346 start_iterator(stmt);
2347 token = expect(token->next, '(', "after 'for'");
2348
2349 syms = NULL;
2350 e1 = NULL;
2351 /* C99 variable declaration? */
2352 if (lookup_type(token)) {
2353 token = external_declaration(token, &syms, validate_for_loop_decl);
2354 } else {
2355 token = parse_expression(token, &e1);
2356 token = expect(token, ';', "in 'for'");
2357 }
2358 token = parse_expression(token, &e2);
2359 token = expect(token, ';', "in 'for'");
2360 token = parse_expression(token, &e3);
2361 token = expect(token, ')', "in 'for'");
2362 token = statement(token, &iterator);
2363
2364 stmt->iterator_syms = syms;
2365 stmt->iterator_pre_statement = make_statement(e1);
2366 stmt->iterator_pre_condition = e2;
2367 stmt->iterator_post_statement = make_statement(e3);
2368 stmt->iterator_post_condition = NULL;
2369 stmt->iterator_statement = iterator;
2370 end_iterator(stmt);
2371
2372 return token;
2373 }
2374
parse_while_statement(struct token * token,struct statement * stmt)2375 static struct token *parse_while_statement(struct token *token, struct statement *stmt)
2376 {
2377 struct expression *expr;
2378 struct statement *iterator;
2379
2380 start_iterator(stmt);
2381 token = parens_expression(token->next, &expr, "after 'while'");
2382 token = statement(token, &iterator);
2383
2384 stmt->iterator_pre_condition = expr;
2385 stmt->iterator_post_condition = NULL;
2386 stmt->iterator_statement = iterator;
2387 end_iterator(stmt);
2388
2389 return token;
2390 }
2391
parse_do_statement(struct token * token,struct statement * stmt)2392 static struct token *parse_do_statement(struct token *token, struct statement *stmt)
2393 {
2394 struct expression *expr;
2395 struct statement *iterator;
2396
2397 start_iterator(stmt);
2398 token = statement(token->next, &iterator);
2399 if (token_type(token) == TOKEN_IDENT && token->ident == &while_ident)
2400 token = token->next;
2401 else
2402 sparse_error(token->pos, "expected 'while' after 'do'");
2403 token = parens_expression(token, &expr, "after 'do-while'");
2404
2405 stmt->iterator_post_condition = expr;
2406 stmt->iterator_statement = iterator;
2407 end_iterator(stmt);
2408
2409 if (iterator && iterator->type != STMT_COMPOUND && Wdo_while)
2410 warning(iterator->pos, "do-while statement is not a compound statement");
2411
2412 return expect(token, ';', "after statement");
2413 }
2414
parse_if_statement(struct token * token,struct statement * stmt)2415 static struct token *parse_if_statement(struct token *token, struct statement *stmt)
2416 {
2417 stmt->type = STMT_IF;
2418 token = parens_expression(token->next, &stmt->if_conditional, "after if");
2419 token = statement(token, &stmt->if_true);
2420 if (token_type(token) != TOKEN_IDENT)
2421 return token;
2422 if (token->ident != &else_ident)
2423 return token;
2424 return statement(token->next, &stmt->if_false);
2425 }
2426
case_statement(struct token * token,struct statement * stmt)2427 static inline struct token *case_statement(struct token *token, struct statement *stmt)
2428 {
2429 stmt->type = STMT_CASE;
2430 token = expect(token, ':', "after default/case");
2431 add_case_statement(stmt);
2432 return statement(token, &stmt->case_statement);
2433 }
2434
parse_case_statement(struct token * token,struct statement * stmt)2435 static struct token *parse_case_statement(struct token *token, struct statement *stmt)
2436 {
2437 token = parse_expression(token->next, &stmt->case_expression);
2438 if (match_op(token, SPECIAL_ELLIPSIS))
2439 token = parse_expression(token->next, &stmt->case_to);
2440 return case_statement(token, stmt);
2441 }
2442
parse_default_statement(struct token * token,struct statement * stmt)2443 static struct token *parse_default_statement(struct token *token, struct statement *stmt)
2444 {
2445 return case_statement(token->next, stmt);
2446 }
2447
parse_loop_iterator(struct token * token,struct statement * stmt)2448 static struct token *parse_loop_iterator(struct token *token, struct statement *stmt)
2449 {
2450 struct symbol *target = lookup_symbol(token->ident, NS_ITERATOR);
2451 stmt->type = STMT_GOTO;
2452 stmt->goto_label = target;
2453 if (!target)
2454 sparse_error(stmt->pos, "break/continue not in iterator scope");
2455 return expect(token->next, ';', "at end of statement");
2456 }
2457
parse_switch_statement(struct token * token,struct statement * stmt)2458 static struct token *parse_switch_statement(struct token *token, struct statement *stmt)
2459 {
2460 stmt->type = STMT_SWITCH;
2461 start_switch(stmt);
2462 token = parens_expression(token->next, &stmt->switch_expression, "after 'switch'");
2463 token = statement(token, &stmt->switch_statement);
2464 end_switch(stmt);
2465 return token;
2466 }
2467
parse_goto_statement(struct token * token,struct statement * stmt)2468 static struct token *parse_goto_statement(struct token *token, struct statement *stmt)
2469 {
2470 stmt->type = STMT_GOTO;
2471 token = token->next;
2472 if (match_op(token, '*')) {
2473 token = parse_expression(token->next, &stmt->goto_expression);
2474 add_statement(&function_computed_goto_list, stmt);
2475 } else if (token_type(token) == TOKEN_IDENT) {
2476 stmt->goto_label = label_symbol(token);
2477 token = token->next;
2478 } else {
2479 sparse_error(token->pos, "Expected identifier or goto expression");
2480 }
2481 return expect(token, ';', "at end of statement");
2482 }
2483
parse_context_statement(struct token * token,struct statement * stmt)2484 static struct token *parse_context_statement(struct token *token, struct statement *stmt)
2485 {
2486 stmt->type = STMT_CONTEXT;
2487 token = token->next;
2488 token = expect(token, '(', "after __context__ statement");
2489 token = assignment_expression(token, &stmt->expression);
2490 if (!stmt->expression)
2491 unexpected(token, "expression expected after '('");
2492 if (match_op(token, ',')) {
2493 token = token->next;
2494 stmt->context = stmt->expression;
2495 token = assignment_expression(token, &stmt->expression);
2496 if (!stmt->expression)
2497 unexpected(token, "expression expected after ','");
2498 }
2499 token = expect(token, ')', "at end of __context__ statement");
2500 return expect(token, ';', "at end of statement");
2501 }
2502
parse_range_statement(struct token * token,struct statement * stmt)2503 static struct token *parse_range_statement(struct token *token, struct statement *stmt)
2504 {
2505 stmt->type = STMT_RANGE;
2506 token = token->next;
2507 token = expect(token, '(', "after __range__ statement");
2508 token = assignment_expression(token, &stmt->range_expression);
2509 token = expect(token, ',', "after range expression");
2510 token = assignment_expression(token, &stmt->range_low);
2511 token = expect(token, ',', "after low range");
2512 token = assignment_expression(token, &stmt->range_high);
2513 token = expect(token, ')', "after range statement");
2514 return expect(token, ';', "after range statement");
2515 }
2516
statement(struct token * token,struct statement ** tree)2517 static struct token *statement(struct token *token, struct statement **tree)
2518 {
2519 struct statement *stmt = alloc_statement(token->pos, STMT_NONE);
2520
2521 *tree = stmt;
2522 if (token_type(token) == TOKEN_IDENT) {
2523 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2524 if (s && s->op->statement)
2525 return s->op->statement(token, stmt);
2526
2527 if (match_op(token->next, ':')) {
2528 struct symbol *s = label_symbol(token);
2529 token = skip_attributes(token->next->next);
2530 if (s->stmt) {
2531 sparse_error(stmt->pos, "label '%s' redefined", show_ident(s->ident));
2532 // skip the label to avoid multiple definitions
2533 return statement(token, tree);
2534 }
2535 stmt->type = STMT_LABEL;
2536 stmt->label_identifier = s;
2537 s->stmt = stmt;
2538 return statement(token, &stmt->label_statement);
2539 }
2540 }
2541
2542 if (match_op(token, '{')) {
2543 stmt->type = STMT_COMPOUND;
2544 start_symbol_scope(stmt->pos);
2545 token = compound_statement(token->next, stmt);
2546 end_symbol_scope();
2547
2548 return expect(token, '}', "at end of compound statement");
2549 }
2550
2551 stmt->type = STMT_EXPRESSION;
2552 return expression_statement(token, &stmt->expression);
2553 }
2554
2555 /* gcc extension - __label__ ident-list; in the beginning of compound stmt */
label_statement(struct token * token)2556 static struct token *label_statement(struct token *token)
2557 {
2558 while (token_type(token) == TOKEN_IDENT) {
2559 struct symbol *sym = alloc_symbol(token->pos, SYM_LABEL);
2560 /* it's block-scope, but we want label namespace */
2561 bind_symbol(sym, token->ident, NS_SYMBOL);
2562 sym->namespace = NS_LABEL;
2563 fn_local_symbol(sym);
2564 token = token->next;
2565 if (!match_op(token, ','))
2566 break;
2567 token = token->next;
2568 }
2569 return expect(token, ';', "at end of label declaration");
2570 }
2571
statement_list(struct token * token,struct statement_list ** list)2572 static struct token * statement_list(struct token *token, struct statement_list **list)
2573 {
2574 int seen_statement = 0;
2575 while (token_type(token) == TOKEN_IDENT &&
2576 token->ident == &__label___ident)
2577 token = label_statement(token->next);
2578 for (;;) {
2579 struct statement * stmt;
2580 if (eof_token(token))
2581 break;
2582 if (match_op(token, '}'))
2583 break;
2584 if (match_ident(token, &_Static_assert_ident)) {
2585 token = parse_static_assert(token, NULL);
2586 continue;
2587 }
2588 if (lookup_type(token)) {
2589 if (seen_statement) {
2590 warning(token->pos, "mixing declarations and code");
2591 seen_statement = 0;
2592 }
2593 stmt = alloc_statement(token->pos, STMT_DECLARATION);
2594 token = external_declaration(token, &stmt->declaration, NULL);
2595 } else {
2596 seen_statement = Wdeclarationafterstatement;
2597 token = statement(token, &stmt);
2598 }
2599 add_statement(list, stmt);
2600 }
2601 return token;
2602 }
2603
identifier_list(struct token * token,struct symbol * fn)2604 static struct token *identifier_list(struct token *token, struct symbol *fn)
2605 {
2606 struct symbol_list **list = &fn->arguments;
2607 for (;;) {
2608 struct symbol *sym = alloc_symbol(token->pos, SYM_NODE);
2609 sym->ident = token->ident;
2610 token = token->next;
2611 sym->endpos = token->pos;
2612 sym->ctype.base_type = &incomplete_ctype;
2613 add_symbol(list, sym);
2614 if (!match_op(token, ',') ||
2615 token_type(token->next) != TOKEN_IDENT ||
2616 lookup_type(token->next))
2617 break;
2618 token = token->next;
2619 }
2620 return token;
2621 }
2622
parameter_type_list(struct token * token,struct symbol * fn)2623 static struct token *parameter_type_list(struct token *token, struct symbol *fn)
2624 {
2625 struct symbol_list **list = &fn->arguments;
2626
2627 for (;;) {
2628 struct symbol *sym;
2629
2630 if (match_op(token, SPECIAL_ELLIPSIS)) {
2631 fn->variadic = 1;
2632 token = token->next;
2633 break;
2634 }
2635
2636 sym = alloc_symbol(token->pos, SYM_NODE);
2637 token = parameter_declaration(token, sym);
2638 if (sym->ctype.base_type == &void_ctype) {
2639 /* Special case: (void) */
2640 if (!*list && !sym->ident)
2641 break;
2642 warning(token->pos, "void parameter");
2643 }
2644 add_symbol(list, sym);
2645 if (!match_op(token, ','))
2646 break;
2647 token = token->next;
2648 }
2649 return token;
2650 }
2651
compound_statement(struct token * token,struct statement * stmt)2652 struct token *compound_statement(struct token *token, struct statement *stmt)
2653 {
2654 token = statement_list(token, &stmt->stmts);
2655 return token;
2656 }
2657
identifier_expression(struct token * token)2658 static struct expression *identifier_expression(struct token *token)
2659 {
2660 struct expression *expr = alloc_expression(token->pos, EXPR_IDENTIFIER);
2661 expr->expr_ident = token->ident;
2662 return expr;
2663 }
2664
index_expression(struct expression * from,struct expression * to)2665 static struct expression *index_expression(struct expression *from, struct expression *to)
2666 {
2667 int idx_from, idx_to;
2668 struct expression *expr = alloc_expression(from->pos, EXPR_INDEX);
2669
2670 idx_from = const_expression_value(from);
2671 idx_to = idx_from;
2672 if (to) {
2673 idx_to = const_expression_value(to);
2674 if (idx_to < idx_from || idx_from < 0)
2675 warning(from->pos, "nonsense array initializer index range");
2676 }
2677 expr->idx_from = idx_from;
2678 expr->idx_to = idx_to;
2679 return expr;
2680 }
2681
single_initializer(struct expression ** ep,struct token * token)2682 static struct token *single_initializer(struct expression **ep, struct token *token)
2683 {
2684 int expect_equal = 0;
2685 struct token *next = token->next;
2686 struct expression **tail = ep;
2687 int nested;
2688
2689 *ep = NULL;
2690
2691 if ((token_type(token) == TOKEN_IDENT) && match_op(next, ':')) {
2692 struct expression *expr = identifier_expression(token);
2693 if (Wold_initializer)
2694 warning(token->pos, "obsolete struct initializer, use C99 syntax");
2695 token = initializer(&expr->ident_expression, next->next);
2696 if (expr->ident_expression)
2697 *ep = expr;
2698 return token;
2699 }
2700
2701 for (tail = ep, nested = 0; ; nested++, next = token->next) {
2702 if (match_op(token, '.') && (token_type(next) == TOKEN_IDENT)) {
2703 struct expression *expr = identifier_expression(next);
2704 *tail = expr;
2705 tail = &expr->ident_expression;
2706 expect_equal = 1;
2707 token = next->next;
2708 } else if (match_op(token, '[')) {
2709 struct expression *from = NULL, *to = NULL, *expr;
2710 token = constant_expression(token->next, &from);
2711 if (!from) {
2712 sparse_error(token->pos, "Expected constant expression");
2713 break;
2714 }
2715 if (match_op(token, SPECIAL_ELLIPSIS))
2716 token = constant_expression(token->next, &to);
2717 expr = index_expression(from, to);
2718 *tail = expr;
2719 tail = &expr->idx_expression;
2720 token = expect(token, ']', "at end of initializer index");
2721 if (nested)
2722 expect_equal = 1;
2723 } else {
2724 break;
2725 }
2726 }
2727 if (nested && !expect_equal) {
2728 if (!match_op(token, '='))
2729 warning(token->pos, "obsolete array initializer, use C99 syntax");
2730 else
2731 expect_equal = 1;
2732 }
2733 if (expect_equal)
2734 token = expect(token, '=', "at end of initializer index");
2735
2736 token = initializer(tail, token);
2737 if (!*tail)
2738 *ep = NULL;
2739 return token;
2740 }
2741
initializer_list(struct expression_list ** list,struct token * token)2742 static struct token *initializer_list(struct expression_list **list, struct token *token)
2743 {
2744 struct expression *expr;
2745
2746 for (;;) {
2747 token = single_initializer(&expr, token);
2748 if (!expr)
2749 break;
2750 add_expression(list, expr);
2751 if (!match_op(token, ','))
2752 break;
2753 token = token->next;
2754 }
2755 return token;
2756 }
2757
initializer(struct expression ** tree,struct token * token)2758 struct token *initializer(struct expression **tree, struct token *token)
2759 {
2760 if (match_op(token, '{')) {
2761 struct expression *expr = alloc_expression(token->pos, EXPR_INITIALIZER);
2762 *tree = expr;
2763 token = initializer_list(&expr->expr_list, token->next);
2764 return expect(token, '}', "at end of initializer");
2765 }
2766 return assignment_expression(token, tree);
2767 }
2768
declare_argument(struct symbol * sym,struct symbol * fn)2769 static void declare_argument(struct symbol *sym, struct symbol *fn)
2770 {
2771 if (!sym->ident) {
2772 sparse_error(sym->pos, "no identifier for function argument");
2773 return;
2774 }
2775 bind_symbol(sym, sym->ident, NS_SYMBOL);
2776 }
2777
is_syscall(struct symbol * sym)2778 static int is_syscall(struct symbol *sym)
2779 {
2780 char *macro;
2781 char *name;
2782 int is_syscall = 0;
2783
2784 macro = get_macro_name(sym->pos);
2785 if (macro &&
2786 (strncmp("SYSCALL_DEFINE", macro, strlen("SYSCALL_DEFINE")) == 0 ||
2787 strncmp("COMPAT_SYSCALL_DEFINE", macro, strlen("COMPAT_SYSCALL_DEFINE")) == 0))
2788 is_syscall = 1;
2789
2790 name = sym->ident->name;
2791
2792 if (name && strncmp(name, "sys_", 4) ==0)
2793 is_syscall = 1;
2794
2795 if (name && strncmp(name, "compat_sys_", 11) == 0)
2796 is_syscall = 1;
2797
2798 return is_syscall;
2799 }
2800
2801
parse_function_body(struct token * token,struct symbol * decl,struct symbol_list ** list)2802 static struct token *parse_function_body(struct token *token, struct symbol *decl,
2803 struct symbol_list **list)
2804 {
2805 struct symbol_list **old_symbol_list;
2806 struct symbol *base_type = decl->ctype.base_type;
2807 struct statement *stmt, **p;
2808 struct symbol *prev;
2809 struct symbol *arg;
2810
2811 old_symbol_list = function_symbol_list;
2812 if (decl->ctype.modifiers & MOD_INLINE) {
2813 function_symbol_list = &decl->inline_symbol_list;
2814 p = &base_type->inline_stmt;
2815 } else {
2816 function_symbol_list = &decl->symbol_list;
2817 p = &base_type->stmt;
2818 }
2819 function_computed_target_list = NULL;
2820 function_computed_goto_list = NULL;
2821
2822 if ((decl->ctype.modifiers & (MOD_EXTERN|MOD_INLINE)) == MOD_EXTERN) {
2823 if (Wexternal_function_has_definition)
2824 warning(decl->pos, "function '%s' with external linkage has definition", show_ident(decl->ident));
2825 }
2826 if (!(decl->ctype.modifiers & MOD_STATIC))
2827 decl->ctype.modifiers |= MOD_EXTERN;
2828
2829 stmt = start_function(decl);
2830
2831 *p = stmt;
2832 FOR_EACH_PTR (base_type->arguments, arg) {
2833 declare_argument(arg, base_type);
2834 } END_FOR_EACH_PTR(arg);
2835
2836 token = compound_statement(token->next, stmt);
2837
2838 end_function(decl);
2839 if (!(decl->ctype.modifiers & MOD_INLINE))
2840 add_symbol(list, decl);
2841 else if (is_syscall(decl)) {
2842 add_symbol(list, decl);
2843 /*
2844 printf("parse.c decl: %s\n", decl->ident->name);
2845 char *macro = get_macro_name(decl->pos);
2846 printf("decl macro: %s\n", macro);
2847 */
2848 }
2849 check_declaration(decl);
2850 decl->definition = decl;
2851 prev = decl->same_symbol;
2852 if (prev && prev->definition) {
2853 warning(decl->pos, "multiple definitions for function '%s'",
2854 show_ident(decl->ident));
2855 info(prev->definition->pos, " the previous one is here");
2856 } else {
2857 while (prev) {
2858 rebind_scope(prev, decl->scope);
2859 prev->definition = decl;
2860 prev = prev->same_symbol;
2861 }
2862 }
2863 function_symbol_list = old_symbol_list;
2864 if (function_computed_goto_list) {
2865 if (!function_computed_target_list)
2866 warning(decl->pos, "function '%s' has computed goto but no targets?", show_ident(decl->ident));
2867 else {
2868 FOR_EACH_PTR(function_computed_goto_list, stmt) {
2869 stmt->target_list = function_computed_target_list;
2870 } END_FOR_EACH_PTR(stmt);
2871 }
2872 }
2873 return expect(token, '}', "at end of function");
2874 }
2875
promote_k_r_types(struct symbol * arg)2876 static void promote_k_r_types(struct symbol *arg)
2877 {
2878 struct symbol *base = arg->ctype.base_type;
2879 if (base && base->ctype.base_type == &int_type && (base->ctype.modifiers & (MOD_CHAR | MOD_SHORT))) {
2880 arg->ctype.base_type = &int_ctype;
2881 }
2882 }
2883
apply_k_r_types(struct symbol_list * argtypes,struct symbol * fn)2884 static void apply_k_r_types(struct symbol_list *argtypes, struct symbol *fn)
2885 {
2886 struct symbol_list *real_args = fn->ctype.base_type->arguments;
2887 struct symbol *arg;
2888
2889 FOR_EACH_PTR(real_args, arg) {
2890 struct symbol *type;
2891
2892 /* This is quadratic in the number of arguments. We _really_ don't care */
2893 FOR_EACH_PTR(argtypes, type) {
2894 if (type->ident == arg->ident)
2895 goto match;
2896 } END_FOR_EACH_PTR(type);
2897 if (Wimplicit_int) {
2898 sparse_error(arg->pos, "missing type declaration for parameter '%s'",
2899 show_ident(arg->ident));
2900 }
2901 type = alloc_symbol(arg->pos, SYM_NODE);
2902 type->ident = arg->ident;
2903 type->ctype.base_type = &int_ctype;
2904 match:
2905 type->used = 1;
2906 /* "char" and "short" promote to "int" */
2907 promote_k_r_types(type);
2908
2909 arg->ctype = type->ctype;
2910 } END_FOR_EACH_PTR(arg);
2911
2912 FOR_EACH_PTR(argtypes, arg) {
2913 if (!arg->used)
2914 warning(arg->pos, "nonsensical parameter declaration '%s'", show_ident(arg->ident));
2915 } END_FOR_EACH_PTR(arg);
2916
2917 }
2918
parse_k_r_arguments(struct token * token,struct symbol * decl,struct symbol_list ** list)2919 static struct token *parse_k_r_arguments(struct token *token, struct symbol *decl,
2920 struct symbol_list **list)
2921 {
2922 struct symbol_list *args = NULL;
2923
2924 if (Wold_style_definition)
2925 warning(token->pos, "non-ANSI definition of function '%s'", show_ident(decl->ident));
2926
2927 do {
2928 token = declaration_list(token, &args);
2929 if (!match_op(token, ';')) {
2930 sparse_error(token->pos, "expected ';' at end of parameter declaration");
2931 break;
2932 }
2933 token = token->next;
2934 } while (lookup_type(token));
2935
2936 apply_k_r_types(args, decl);
2937
2938 if (!match_op(token, '{')) {
2939 sparse_error(token->pos, "expected function body");
2940 return token;
2941 }
2942 return parse_function_body(token, decl, list);
2943 }
2944
toplevel_asm_declaration(struct token * token,struct symbol_list ** list)2945 static struct token *toplevel_asm_declaration(struct token *token, struct symbol_list **list)
2946 {
2947 struct symbol *anon = alloc_symbol(token->pos, SYM_NODE);
2948 struct symbol *fn = alloc_symbol(token->pos, SYM_FN);
2949 struct statement *stmt;
2950
2951 anon->ctype.base_type = fn;
2952 stmt = alloc_statement(token->pos, STMT_NONE);
2953 fn->stmt = stmt;
2954
2955 token = parse_asm_statement(token, stmt);
2956
2957 add_symbol(list, anon);
2958 return token;
2959 }
2960
external_declaration(struct token * token,struct symbol_list ** list,validate_decl_t validate_decl)2961 struct token *external_declaration(struct token *token, struct symbol_list **list,
2962 validate_decl_t validate_decl)
2963 {
2964 struct ident *ident = NULL;
2965 struct symbol *decl;
2966 struct decl_state ctx = { .ident = &ident };
2967 struct ctype saved;
2968 struct symbol *base_type;
2969 unsigned long mod;
2970 int is_typedef;
2971
2972 if (match_ident(token, &_Pragma_ident))
2973 return parse_underscore_Pragma(token);
2974
2975 /* Top-level inline asm or static assertion? */
2976 if (token_type(token) == TOKEN_IDENT) {
2977 struct symbol *s = lookup_keyword(token->ident, NS_KEYWORD);
2978 if (s && s->op->toplevel)
2979 return s->op->toplevel(token, list);
2980 }
2981
2982 /* Parse declaration-specifiers, if any */
2983 token = declaration_specifiers(token, &ctx);
2984 mod = storage_modifiers(&ctx);
2985 mod |= ctx.ctype.modifiers & MOD_NORETURN;
2986 decl = alloc_symbol(token->pos, SYM_NODE);
2987 /* Just a type declaration? */
2988 if (match_op(token, ';')) {
2989 apply_modifiers(token->pos, &ctx);
2990 return token->next;
2991 }
2992
2993 saved = ctx.ctype;
2994 token = declarator(token, &ctx);
2995 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
2996 apply_modifiers(token->pos, &ctx);
2997
2998 decl->ctype = ctx.ctype;
2999 decl->ctype.modifiers |= mod;
3000 decl->endpos = token->pos;
3001
3002 /* Just a type declaration? */
3003 if (!ident) {
3004 warning(token->pos, "missing identifier in declaration");
3005 return expect(token, ';', "at the end of type declaration");
3006 }
3007
3008 /* type define declaration? */
3009 is_typedef = ctx.storage_class == STypedef;
3010
3011 /* Typedefs don't have meaningful storage */
3012 if (is_typedef)
3013 decl->ctype.modifiers |= MOD_USERTYPE;
3014
3015 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
3016
3017 base_type = decl->ctype.base_type;
3018
3019 if (is_typedef) {
3020 if (base_type && !base_type->ident) {
3021 switch (base_type->type) {
3022 case SYM_STRUCT:
3023 case SYM_UNION:
3024 case SYM_ENUM:
3025 case SYM_RESTRICT:
3026 base_type->ident = ident;
3027 break;
3028 default:
3029 break;
3030 }
3031 }
3032 } else if (base_type && base_type->type == SYM_FN) {
3033 if (base_type->ctype.base_type == &incomplete_ctype) {
3034 warning(decl->pos, "'%s()' has implicit return type",
3035 show_ident(decl->ident));
3036 base_type->ctype.base_type = &int_ctype;
3037 }
3038 /* K&R argument declaration? */
3039 if (lookup_type(token))
3040 return parse_k_r_arguments(token, decl, list);
3041 if (match_op(token, '{'))
3042 return parse_function_body(token, decl, list);
3043
3044 if (!(decl->ctype.modifiers & MOD_STATIC))
3045 decl->ctype.modifiers |= MOD_EXTERN;
3046 } else if (base_type == &void_ctype && !(decl->ctype.modifiers & MOD_EXTERN)) {
3047 sparse_error(token->pos, "void declaration");
3048 }
3049 if (base_type == &incomplete_ctype) {
3050 warning(decl->pos, "'%s' has implicit type", show_ident(decl->ident));
3051 decl->ctype.base_type = &int_ctype;;
3052 }
3053
3054 for (;;) {
3055 if (!is_typedef && match_op(token, '=')) {
3056 token = initializer(&decl->initializer, token->next);
3057 }
3058 if (!is_typedef) {
3059 if (validate_decl)
3060 validate_decl(decl);
3061
3062 if (decl->initializer && decl->ctype.modifiers & MOD_EXTERN) {
3063 warning(decl->pos, "symbol with external linkage has initializer");
3064 decl->ctype.modifiers &= ~MOD_EXTERN;
3065 }
3066
3067 if (!(decl->ctype.modifiers & (MOD_EXTERN | MOD_INLINE))) {
3068 add_symbol(list, decl);
3069 fn_local_symbol(decl);
3070 }
3071 }
3072 check_declaration(decl);
3073 if (decl->same_symbol) {
3074 decl->definition = decl->same_symbol->definition;
3075 decl->op = decl->same_symbol->op;
3076 if (is_typedef) {
3077 // TODO: handle -std=c89 --pedantic
3078 check_duplicates(decl);
3079 }
3080 }
3081
3082 if (!match_op(token, ','))
3083 break;
3084
3085 token = token->next;
3086 ident = NULL;
3087 decl = alloc_symbol(token->pos, SYM_NODE);
3088 ctx.ctype = saved;
3089 token = handle_attributes(token, &ctx, KW_ATTRIBUTE);
3090 token = declarator(token, &ctx);
3091 token = handle_attributes(token, &ctx, KW_ATTRIBUTE | KW_ASM);
3092 apply_modifiers(token->pos, &ctx);
3093 decl->ctype = ctx.ctype;
3094 decl->ctype.modifiers |= mod;
3095 decl->endpos = token->pos;
3096 if (!ident) {
3097 sparse_error(token->pos, "expected identifier name in type definition");
3098 return token;
3099 }
3100
3101 if (is_typedef)
3102 decl->ctype.modifiers |= MOD_USERTYPE;
3103
3104 bind_symbol(decl, ident, is_typedef ? NS_TYPEDEF: NS_SYMBOL);
3105
3106 /* Function declarations are automatically extern unless specifically static */
3107 base_type = decl->ctype.base_type;
3108 if (!is_typedef && base_type && base_type->type == SYM_FN) {
3109 if (!(decl->ctype.modifiers & MOD_STATIC))
3110 decl->ctype.modifiers |= MOD_EXTERN;
3111 }
3112 }
3113 return expect(token, ';', "at end of declaration");
3114 }
3115