3 * Parser for C declarations.
4 * Copyright © 2011-2012, 2021, 2023 Nick Bowler
6 * This program is free software: you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation, either version 3 of the License, or
9 * (at your option) any later version.
11 * This program is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14 * GNU General Public License for more details.
16 * You should have received a copy of the GNU General Public License
17 * along with this program. If not, see <http://www.gnu.org/licenses/>.
21 %name-prefix "cdecl__yy"
22 %parse-param {void *scanner}
23 %parse-param {struct cdecl **out}
24 %lex-param {yyscan_t scanner}
36 #include "cdecl-internal.h"
39 #define FAIL(msg) do { \
40 yyerror(&yylloc, NULL, NULL, msg); \
44 #define ALLOC(ptr, size) do { \
45 (ptr) = malloc(size); \
47 cdecl__errmsg(CDECL__ENOMEM); \
52 #define ALLOC_STRUCT(ptr, type, ...) do { \
53 ALLOC(ptr, sizeof (type)); \
54 *(ptr) = (type) { __VA_ARGS__ }; \
58 * With the postprocessing performed by fix-yytname.awk, all the symbol
59 * name strings can be used directly in error messages and there is no
60 * need for any string processing.
62 #define yytnamerr(a, b) cdecl__strlcpy(a, b, (a) ? INT_MAX : 0)
70 void cdecl__free(struct cdecl *);
71 int cdecl__yyparse(void *scanner, struct cdecl **out);
72 const char *cdecl__token_name(unsigned token);
80 struct cdecl_declspec *declspec;
81 struct cdecl_declarator *declarator;
86 static void yyerror(YYLTYPE *, yyscan_t, struct cdecl **, const char *);
87 static void free_decl(struct cdecl *);
89 static void free_declspec(struct cdecl_declspec *x)
91 struct cdecl_declspec *p;
100 static void free_declarator(struct cdecl_declarator *x)
102 struct cdecl_declarator *p;
108 case CDECL_DECL_NULL:
110 case CDECL_DECL_IDENT:
113 case CDECL_DECL_POINTER:
114 free_declspec(x->u.pointer.qualifiers);
116 case CDECL_DECL_ARRAY:
117 free(x->u.array.vla);
119 case CDECL_DECL_FUNCTION:
120 free_decl(x->u.function.parameters);
131 static void free_decl(struct cdecl *x)
138 /* The specifiers may be shared by an entire chain. */
139 if (!p || p->specifiers != x->specifiers)
140 free_declspec(x->specifiers);
142 free_declarator(x->declarators);
148 void cdecl__free(struct cdecl *decl)
154 * Join two declaration specifier lists into a single list, with "a" being the
155 * head of the new list.
157 * The list "a" is assumed to be nonempty.
159 static void join_specs(struct cdecl_declspec *a, struct cdecl_declspec *b)
167 * Alter an abstract declarator (type name) to declare an identifier instead,
168 * used by the English parser rules to reduce "identifier as type" sequences.
170 static struct cdecl *insert_identifier(struct cdecl *decl, char *ident)
172 struct cdecl_declarator *d = decl->declarators;
177 d->type = CDECL_DECL_IDENT;
184 %destructor { free($$); } <strval>
185 %destructor { free_declspec($$); } <declspec>
186 %destructor { free_declarator($$); } <declarator>
187 %destructor { free_decl($$); } <decl>
193 %token <strval> T_IDENT "identifier"
194 %token <uintval> T_UINT "integer constant"
196 %token T_SEMICOLON ";"
197 %token T_ASTERISK "*"
200 %token T_LBRACKET "["
201 %token T_RBRACKET "]"
203 %token T_ELLIPSIS "..."
205 %token <spectype> T_TYPEDEF "typedef"
206 %token <spectype> T_EXTERN "extern"
207 %token <spectype> T_STATIC "static"
208 %token <spectype> T_AUTO "auto"
209 %token <spectype> T_REGISTER "register"
211 %token <spectype> T_INLINE "inline"
213 %token <spectype> T_RESTRICT "restrict"
214 %token <spectype> T_VOLATILE "volatile"
215 %token <spectype> T_CONST "const"
217 %token <spectype> T_VOID "void"
218 %token <spectype> T_CHAR "char"
219 %token <spectype> T_SHORT "short"
220 %token <spectype> T_INT "int"
221 %token <spectype> T_LONG "long"
222 %token <spectype> T_FLOAT "float"
223 %token <spectype> T_DOUBLE "double"
224 %token <spectype> T_SIGNED "signed"
225 %token <spectype> T_UNSIGNED "unsigned"
226 %token <spectype> T_BOOL "_Bool"
227 %token <spectype> T_COMPLEX "_Complex"
228 %token <spectype> T_IMAGINARY "_Imaginary"
230 %token <spectype> T_STRUCT "struct"
231 %token <spectype> T_UNION "union"
232 %token <spectype> T_ENUM "enum"
238 %token T_DECLARE "declare"
239 %token T_POINTER "pointer"
240 %token T_FUNCTION "function"
241 %token T_RETURNING "returning"
242 %token T_ARRAY "array"
246 %token T_VLA "variable-length"
248 %type <strval> vla_ident
249 %type <uintval> array_length
250 %type <boolval> varargs
251 %type <spectype> declspec_simple qualifier_simple
252 %type <spectype> typespec_simple typespec_tagged
253 %type <declspec> declspec_notype declspec_noid typespec_noid typespec
254 %type <declspec> qualifier qualifiers
255 %type <declspec> declspecs declspecs_noid
256 %type <declarator> direct_declarator declarator pointer array parens postfix
257 %type <declarator> direct_declarator_ish declarator_ish parameter_type_list
258 %type <decl> declaration declarators declarator_wrap
259 %type <decl> parameter
261 %type <strval> english_vla
262 %type <declspec> storage_func_specs post_specs
263 %type <declspec> type_qual_spec type_qual_specs typedef_name_qual
264 %type <declarator> english_declarator english_array english_function
265 %type <declarator> english_parameter_list null_decl
266 %type <decl> english english_declaration
267 %type <decl> english_parameter
270 * Harmless shift/reduce conflicts in english_parameter. See comments below
277 input: T_ENGLISH english {
285 declaration: declspecs declarators semi {
291 * We support parsing declarations using arbitrary identifiers as type
292 * specifiers (a la C typedef). To avoid confusion with identifiers that
293 * may also be used as declarators, note the following:
295 * (a) Every valid C declaration must have at least one type specifier, and
296 * (b) Valid declarations with typedef names have exactly one type specifier.
298 * So the rule applied when parsing specifiers is: an identifier is a type
299 * specifier only if we have not yet seen any type specifiers whatsoever
300 * (within one declaration specifier list).
302 * Treating identifiers as type specifiers by default can lead to strange and
303 * unexpected parses; libcdecl applies a simplification step to the resulting
304 * parse tree afterwards.
306 declspecs: declspec_notype declspecs {
309 } | typespec declspecs_noid {
314 declspecs_noid: { $$ = NULL; } | declspec_noid declspecs_noid {
319 qualifiers: { $$ = NULL; } | qualifiers qualifier {
324 declarators: declarator_wrap | declarator_wrap T_COMMA declarators {
329 declarator_wrap: declarator {
330 ALLOC_STRUCT($$, struct cdecl, .declarators = $1);
333 declspec_simple: T_AUTO
340 typespec_simple: T_VOID
353 typespec_tagged: T_STRUCT | T_UNION | T_ENUM | { $$ = CDECL_TYPE_IDENT; }
355 qualifier_simple: T_CONST
359 declspec_notype: qualifier | declspec_simple {
360 ALLOC_STRUCT($$, struct cdecl_declspec, .type = $1);
363 typespec_noid: typespec_simple {
364 ALLOC_STRUCT($$, struct cdecl_declspec, .type = $1);
367 qualifier: qualifier_simple {
368 ALLOC_STRUCT($$, struct cdecl_declspec, .type = $1);
371 typespec: typespec_noid | typespec_tagged T_IDENT {
372 ALLOC_STRUCT($$, struct cdecl_declspec,
377 declspec_noid: declspec_notype | typespec_noid
379 vla_ident: T_IDENT | T_ASTERISK {
380 ALLOC($$, sizeof "");
384 array: T_LBRACKET array_length T_RBRACKET {
385 ALLOC_STRUCT($$, struct cdecl_declarator,
386 .type = CDECL_DECL_ARRAY,
387 .u.array.length = $2);
388 } | T_LBRACKET vla_ident T_RBRACKET {
389 ALLOC_STRUCT($$, struct cdecl_declarator,
390 .type = CDECL_DECL_ARRAY,
394 parameter: declspecs declarator {
395 ALLOC_STRUCT($$, struct cdecl,
400 varargs: { $$ = false; } | T_COMMA T_ELLIPSIS { $$ = true; }
402 parameter_type_list: parameter varargs {
403 ALLOC_STRUCT($$, struct cdecl_declarator,
404 .type = CDECL_DECL_FUNCTION,
405 .u.function.parameters = $1,
406 .u.function.variadic = $2);
407 } | parameter T_COMMA parameter_type_list {
409 $1->next = $$->u.function.parameters;
410 $$->u.function.parameters = $1;
413 parens: T_LPAREN parameter_type_list T_RPAREN {
415 } | T_LPAREN declarator_ish T_RPAREN {
416 ALLOC_STRUCT($$, struct cdecl_declarator,
417 .type = CDECL_DECL_FUNCTION);
418 ALLOC_STRUCT($$->u.function.parameters, struct cdecl,
422 pointer: T_ASTERISK qualifiers direct_declarator {
423 ALLOC_STRUCT($$, struct cdecl_declarator,
424 .type = CDECL_DECL_POINTER,
425 .u.pointer.qualifiers = $2,
427 } | T_ASTERISK qualifiers pointer {
428 ALLOC_STRUCT($$, struct cdecl_declarator,
429 .type = CDECL_DECL_POINTER,
430 .u.pointer.qualifiers = $2,
434 declarator: direct_declarator | pointer
435 declarator_ish: direct_declarator_ish | pointer
436 postfix: array | parens
438 direct_declarator_ish: {
439 ALLOC_STRUCT($$, struct cdecl_declarator,
440 .type = CDECL_DECL_NULL);
441 } | direct_declarator_ish postfix {
447 ALLOC_STRUCT($$, struct cdecl_declarator,
448 .type = CDECL_DECL_NULL);
450 ALLOC_STRUCT($$, struct cdecl_declarator,
451 .type = CDECL_DECL_IDENT,
453 } | direct_declarator postfix {
458 english: T_DECLARE T_IDENT T_AS english_declaration {
459 $$ = insert_identifier($4, $2);
460 } | T_TYPE english_declaration {
464 storage_func_specs: { $$ = NULL; } | declspec_simple storage_func_specs {
465 ALLOC_STRUCT($$, struct cdecl_declspec,
470 type_qual_spec: typespec_noid | qualifier
472 type_qual_specs: { $$ = NULL; } | type_qual_spec type_qual_specs {
478 * The "qualifiers" nonterminal needs to be used here to avoid shift/reduce
479 * conflicts with pointer declarators. So we end up needing to stitch
480 * together three different specifiers lists.
482 post_specs: qualifiers typespec type_qual_specs {
488 english_declaration: storage_func_specs english_declarator post_specs {
490 ALLOC_STRUCT($$, struct cdecl,
495 english_declarator: {
496 ALLOC_STRUCT($$, struct cdecl_declarator,
497 .type = CDECL_DECL_NULL);
498 } | english_declarator qualifiers T_POINTER T_TO {
499 ALLOC_STRUCT($$, struct cdecl_declarator,
500 .type = CDECL_DECL_POINTER,
502 .u.pointer.qualifiers = $2);
503 } | english_declarator english_array {
506 } | english_declarator english_function {
511 english_function: T_FUNCTION T_RETURNING {
512 ALLOC_STRUCT($$, struct cdecl_declarator,
513 .type = CDECL_DECL_FUNCTION,
514 .u.function.parameters = NULL);
515 } | T_FUNCTION T_LPAREN english_parameter_list T_RPAREN T_RETURNING {
519 english_parameter_list: english_parameter varargs {
520 ALLOC_STRUCT($$, struct cdecl_declarator,
521 .type = CDECL_DECL_FUNCTION,
522 .u.function.parameters = $1,
523 .u.function.variadic = $2);
524 } | english_parameter T_COMMA english_parameter_list {
526 $1->next = $$->u.function.parameters;
527 $$->u.function.parameters = $1;
530 typedef_name_qual: T_IDENT qualifiers {
531 ALLOC_STRUCT($$, struct cdecl_declspec,
532 .type = CDECL_TYPE_IDENT,
538 ALLOC_STRUCT($$, struct cdecl_declarator,
539 .type = CDECL_DECL_NULL);
543 * There is a small shift/reduce conflict here. An unadorned identifier
544 * as the first thing in the parameter might be a typedef name deep in the
545 * first english_declaration (thus empty storage_func_specs and empty
546 * english_declarator need to be reduced) or it might be the identifier
547 * before the "as" (thus the identifier should be shifted).
549 * The typedef name conflict is the only issue, so treating it as a special
550 * case makes the shift harmless.
552 english_parameter: english_declaration | typedef_name_qual null_decl {
553 ALLOC_STRUCT($$, struct cdecl,
556 } | T_IDENT T_AS english_declaration {
557 $$ = insert_identifier($3, $1);
560 english_array: T_VLA T_ARRAY english_vla T_OF {
561 ALLOC_STRUCT($$, struct cdecl_declarator,
562 .type = CDECL_DECL_ARRAY,
564 } | T_ARRAY array_length T_OF {
565 ALLOC_STRUCT($$, struct cdecl_declarator,
566 .type = CDECL_DECL_ARRAY,
567 .u.array.length = $2);
570 array_length: { $$ = 0; }
571 array_length: T_UINT {
573 FAIL(_("array length must be positive"));
576 english_vla: T_IDENT | {
577 ALLOC($$, sizeof "");
584 * Expose the token string table to the rest of the library, in order to
585 * produce strings that match parser keywords.
587 * In order for this to work properly, the Bison output must be postprocessed
588 * by fix-yytname.awk to remove pointless quotation marks from the keyword
591 const char *cdecl__token_name(unsigned token)
593 return yytname[YYTRANSLATE(token)];
597 yyerror(YYLTYPE *loc, yyscan_t scanner, struct cdecl **out, const char *err)
599 if (strstr(err, yytname[YYTRANSLATE(T_LEX_ERROR)]))
602 cdecl__err(CDECL_ENOPARSE, "%s", err);