lex.c revision 1.62 1 /* $NetBSD: lex.c,v 1.62 2021/08/01 08:03:43 rillig Exp $ */
2
3 /*
4 * Copyright (c) 1996 Christopher G. Demetriou. All Rights Reserved.
5 * Copyright (c) 1994, 1995 Jochen Pohl
6 * All Rights Reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by Jochen Pohl for
19 * The NetBSD Project.
20 * 4. The name of the author may not be used to endorse or promote products
21 * derived from this software without specific prior written permission.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
24 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
25 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
26 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
27 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
28 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
32 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
33 */
34
35 #if HAVE_NBTOOL_CONFIG_H
36 #include "nbtool_config.h"
37 #endif
38
39 #include <sys/cdefs.h>
40 #if defined(__RCSID) && !defined(lint)
41 __RCSID("$NetBSD: lex.c,v 1.62 2021/08/01 08:03:43 rillig Exp $");
42 #endif
43
44 #include <ctype.h>
45 #include <errno.h>
46 #include <float.h>
47 #include <limits.h>
48 #include <math.h>
49 #include <stdlib.h>
50 #include <string.h>
51
52 #include "lint1.h"
53 #include "cgram.h"
54
55 #define CHAR_MASK ((int)(~(~0U << CHAR_SIZE)))
56
57
58 /* Current position (it's also updated when an included file is parsed) */
59 pos_t curr_pos = { "", 1, 0 };
60
61 /*
62 * Current position in C source (not updated when an included file is
63 * parsed).
64 */
65 pos_t csrc_pos = { "", 1, 0 };
66
67 /* Are we parsing a gcc attribute? */
68 bool attron;
69
70 bool in_system_header = false;
71
72 static sbuf_t *allocsb(void);
73 static void freesb(sbuf_t *);
74 static int inpc(void);
75 static int hash(const char *);
76 static sym_t * search(sbuf_t *);
77 static int keyw(sym_t *);
78 static int get_escaped_char(int);
79
80 void
81 lex_next_line(void)
82 {
83 curr_pos.p_line++;
84 curr_pos.p_uniq = 0;
85 debug_step("parsing %s:%d", curr_pos.p_file, curr_pos.p_line);
86 if (curr_pos.p_file == csrc_pos.p_file) {
87 csrc_pos.p_line++;
88 csrc_pos.p_uniq = 0;
89 }
90 }
91
92 void
93 lex_unknown_character(int c)
94 {
95
96 /* unknown character \%o */
97 error(250, c);
98 }
99
100 #define kwdef(name, token, scl, tspec, tqual, c89, c99, gcc, attr, deco) \
101 { \
102 name, token, scl, tspec, tqual, \
103 (c89) > 0, (c99) > 0, (gcc) > 0, (attr) > 0, deco, \
104 }
105 #define kwdef_token(name, token, c89, c99, gcc, attr, deco) \
106 kwdef(name, token, 0, 0, 0, c89, c99, gcc, attr, deco)
107 #define kwdef_sclass(name, sclass, c89, c99, gcc, attr, deco) \
108 kwdef(name, T_SCLASS, sclass, 0, 0, c89, c99, gcc, attr, deco)
109 #define kwdef_type(name, tspec, c89, c99, gcc, attr, deco) \
110 kwdef(name, T_TYPE, 0, tspec, 0, c89, c99, gcc, attr, deco)
111 #define kwdef_tqual(name, tqual, c89, c99, gcc, attr, deco) \
112 kwdef(name, T_QUAL, 0, 0, tqual, c89, c99, gcc, attr, deco)
113 #define kwdef_keyword(name, token) \
114 kwdef(name, token, 0, 0, 0, 0, 0, 0, 0, 1)
115 #define kwdef_gcc_attr(name, token) \
116 kwdef(name, token, 0, 0, 0, 0, 0, 1, 1, 5)
117
118 /*
119 * Keywords.
120 * During initialization they are written to the symbol table.
121 */
122 static struct kwtab {
123 const char *kw_name; /* keyword */
124 int kw_token; /* token returned by yylex() */
125 scl_t kw_scl; /* storage class if kw_token T_SCLASS */
126 tspec_t kw_tspec; /* type spec. if kw_token
127 * T_TYPE or T_STRUCT_OR_UNION */
128 tqual_t kw_tqual; /* type qual. if kw_token T_QUAL */
129 bool kw_c89 : 1; /* C89 keyword */
130 bool kw_c99 : 1; /* C99 keyword */
131 bool kw_gcc : 1; /* GCC keyword */
132 bool kw_attr : 1; /* GCC attribute, keyword */
133 u_int kw_deco : 3; /* 1 = name, 2 = __name, 4 = __name__ */
134 } kwtab[] = {
135 kwdef_gcc_attr( "alias", T_AT_ALIAS),
136 kwdef_keyword( "_Alignas", T_ALIGNAS),
137 kwdef_keyword( "_Alignof", T_ALIGNOF),
138 kwdef_gcc_attr( "aligned", T_AT_ALIGNED),
139 kwdef_token( "__alignof__", T_ALIGNOF, 0,0,0,0,1),
140 kwdef_gcc_attr( "alloc_size", T_AT_ALLOC_SIZE),
141 kwdef_gcc_attr( "always_inline",T_AT_ALWAYS_INLINE),
142 kwdef_token( "asm", T_ASM, 0,0,1,0,7),
143 kwdef_token( "attribute", T_ATTRIBUTE, 0,0,1,0,6),
144 kwdef_sclass( "auto", AUTO, 0,0,0,0,1),
145 kwdef_type( "_Bool", BOOL, 0,1,0,0,1),
146 kwdef_gcc_attr( "bounded", T_AT_BOUNDED),
147 kwdef_keyword( "break", T_BREAK),
148 kwdef_gcc_attr( "buffer", T_AT_BUFFER),
149 kwdef_token( "__builtin_offsetof", T_BUILTIN_OFFSETOF, 0,0,1,0,1),
150 kwdef_keyword( "case", T_CASE),
151 kwdef_type( "char", CHAR, 0,0,0,0,1),
152 kwdef_gcc_attr( "cold", T_AT_COLD),
153 kwdef_gcc_attr( "common", T_AT_COMMON),
154 kwdef_type( "_Complex", COMPLEX, 0,1,0,0,1),
155 kwdef_tqual( "const", CONST, 1,0,0,0,7),
156 kwdef_gcc_attr( "constructor", T_AT_CONSTRUCTOR),
157 kwdef_keyword( "continue", T_CONTINUE),
158 kwdef_keyword( "default", T_DEFAULT),
159 kwdef_gcc_attr( "deprecated", T_AT_DEPRECATED),
160 kwdef_gcc_attr( "destructor", T_AT_DESTRUCTOR),
161 kwdef_keyword( "do", T_DO),
162 kwdef_type( "double", DOUBLE, 0,0,0,0,1),
163 kwdef_keyword( "else", T_ELSE),
164 kwdef_keyword( "enum", T_ENUM),
165 kwdef_token( "__extension__",T_EXTENSION, 0,0,1,0,1),
166 kwdef_sclass( "extern", EXTERN, 0,0,0,0,1),
167 kwdef_gcc_attr( "fallthrough", T_AT_FALLTHROUGH),
168 kwdef_type( "float", FLOAT, 0,0,0,0,1),
169 kwdef_keyword( "for", T_FOR),
170 kwdef_gcc_attr( "format", T_AT_FORMAT),
171 kwdef_gcc_attr( "format_arg", T_AT_FORMAT_ARG),
172 kwdef_token( "_Generic", T_GENERIC, 0,1,0,0,1),
173 kwdef_gcc_attr( "gnu_inline", T_AT_GNU_INLINE),
174 kwdef_gcc_attr( "gnu_printf", T_AT_FORMAT_GNU_PRINTF),
175 kwdef_keyword( "goto", T_GOTO),
176 kwdef_gcc_attr( "hot", T_AT_HOT),
177 kwdef_keyword( "if", T_IF),
178 kwdef_token( "__imag__", T_IMAG, 0,0,1,0,1),
179 kwdef_sclass( "inline", INLINE, 0,1,0,0,7),
180 kwdef_type( "int", INT, 0,0,0,0,1),
181 #ifdef INT128_SIZE
182 kwdef_type( "__int128_t", INT128, 0,1,0,0,1),
183 #endif
184 kwdef_type( "long", LONG, 0,0,0,0,1),
185 kwdef_gcc_attr( "malloc", T_AT_MALLOC),
186 kwdef_gcc_attr( "may_alias", T_AT_MAY_ALIAS),
187 kwdef_gcc_attr( "minbytes", T_AT_MINBYTES),
188 kwdef_gcc_attr( "mode", T_AT_MODE),
189 kwdef_gcc_attr("no_instrument_function",
190 T_AT_NO_INSTRUMENT_FUNCTION),
191 kwdef_gcc_attr( "noinline", T_AT_NOINLINE),
192 kwdef_gcc_attr( "nonnull", T_AT_NONNULL),
193 kwdef_gcc_attr( "nonstring", T_AT_NONSTRING),
194 kwdef_token( "_Noreturn", T_NORETURN, 0,1,0,0,1),
195 kwdef_gcc_attr( "noreturn", T_AT_NORETURN),
196 kwdef_gcc_attr( "nothrow", T_AT_NOTHROW),
197 kwdef_gcc_attr( "optimize", T_AT_OPTIMIZE),
198 kwdef_gcc_attr( "packed", T_AT_PACKED),
199 kwdef_token( "__packed", T_PACKED, 0,0,0,0,1),
200 kwdef_gcc_attr( "pcs", T_AT_PCS),
201 kwdef_gcc_attr( "printf", T_AT_FORMAT_PRINTF),
202 kwdef_gcc_attr( "pure", T_AT_PURE),
203 kwdef_token( "__real__", T_REAL, 0,0,1,0,1),
204 kwdef_sclass( "register", REG, 0,0,0,0,1),
205 kwdef_tqual( "restrict", RESTRICT, 0,1,0,0,5),
206 kwdef_keyword( "return", T_RETURN),
207 kwdef_gcc_attr( "returns_twice",T_AT_RETURNS_TWICE),
208 kwdef_gcc_attr( "scanf", T_AT_FORMAT_SCANF),
209 kwdef_token( "section", T_AT_SECTION, 0,0,1,1,7),
210 kwdef_gcc_attr( "sentinel", T_AT_SENTINEL),
211 kwdef_type( "short", SHORT, 0,0,0,0,1),
212 kwdef_type( "signed", SIGNED, 1,0,0,0,3),
213 kwdef_keyword( "sizeof", T_SIZEOF),
214 kwdef_sclass( "static", STATIC, 0,0,0,0,1),
215 kwdef_gcc_attr( "strfmon", T_AT_FORMAT_STRFMON),
216 kwdef_gcc_attr( "strftime", T_AT_FORMAT_STRFTIME),
217 kwdef_gcc_attr( "string", T_AT_STRING),
218 kwdef("struct", T_STRUCT_OR_UNION, 0, STRUCT, 0, 0,0,0,0,1),
219 kwdef_keyword( "switch", T_SWITCH),
220 kwdef_token( "__symbolrename", T_SYMBOLRENAME, 0,0,0,0,1),
221 kwdef_gcc_attr( "syslog", T_AT_FORMAT_SYSLOG),
222 kwdef_tqual( "__thread", THREAD, 0,0,1,0,1),
223 kwdef_tqual( "_Thread_local", THREAD, 0,1,0,0,1),
224 kwdef_gcc_attr( "tls_model", T_AT_TLS_MODEL),
225 kwdef_gcc_attr( "transparent_union", T_AT_TUNION),
226 kwdef_sclass( "typedef", TYPEDEF, 0,0,0,0,1),
227 kwdef_token( "typeof", T_TYPEOF, 0,0,1,0,7),
228 #ifdef INT128_SIZE
229 kwdef_type( "__uint128_t", UINT128, 0,1,0,0,1),
230 #endif
231 kwdef("union", T_STRUCT_OR_UNION, 0, UNION, 0, 0,0,0,0,1),
232 kwdef_type( "unsigned", UNSIGN, 0,0,0,0,1),
233 kwdef_gcc_attr( "unused", T_AT_UNUSED),
234 kwdef_gcc_attr( "used", T_AT_USED),
235 kwdef_gcc_attr( "visibility", T_AT_VISIBILITY),
236 kwdef_type( "void", VOID, 0,0,0,0,1),
237 kwdef_tqual( "volatile", VOLATILE, 1,0,0,0,7),
238 kwdef_gcc_attr( "warn_unused_result", T_AT_WARN_UNUSED_RESULT),
239 kwdef_gcc_attr( "weak", T_AT_WEAK),
240 kwdef_keyword( "while", T_WHILE),
241 kwdef(NULL, 0, 0, 0, 0, 0, 0, 0, 0, 0),
242 #undef kwdef
243 #undef kwdef_token
244 #undef kwdef_sclass
245 #undef kwdef_type
246 #undef kwdef_tqual
247 #undef kwdef_keyword
248 #undef kwdef_gcc_attr
249 };
250
251 /* Symbol table */
252 static sym_t *symtab[HSHSIZ1];
253
254 /* free list for sbuf structures */
255 static sbuf_t *sbfrlst;
256
257 /* type of next expected symbol */
258 symt_t symtyp;
259
260
261 static void
262 symtab_add(sym_t *sym)
263 {
264 size_t h;
265
266 h = hash(sym->s_name);
267 if ((sym->s_link = symtab[h]) != NULL)
268 symtab[h]->s_rlink = &sym->s_link;
269 sym->s_rlink = &symtab[h];
270 symtab[h] = sym;
271 }
272
273 static void
274 symtab_remove(sym_t *sym)
275 {
276
277 if ((*sym->s_rlink = sym->s_link) != NULL)
278 sym->s_link->s_rlink = sym->s_rlink;
279 sym->s_link = NULL;
280 }
281
282
283 static void
284 add_keyword(const struct kwtab *kw, u_int deco)
285 {
286 sym_t *sym;
287 char buf[256];
288 const char *name;
289
290 if ((kw->kw_deco & deco) == 0)
291 return;
292
293 switch (deco) {
294 case 1:
295 name = kw->kw_name;
296 break;
297 case 2:
298 snprintf(buf, sizeof(buf), "__%s", kw->kw_name);
299 name = strdup(buf);
300 break;
301 default:
302 lint_assert(deco == 4);
303 snprintf(buf, sizeof(buf), "__%s__", kw->kw_name);
304 name = strdup(buf);
305 break;
306 }
307
308 if (name == NULL)
309 err(1, "Can't init symbol table");
310
311 sym = getblk(sizeof(*sym));
312 sym->s_name = name;
313 sym->s_keyword = kw;
314 sym->s_value.v_quad = kw->kw_token;
315 if (kw->kw_token == T_TYPE || kw->kw_token == T_STRUCT_OR_UNION) {
316 sym->s_tspec = kw->kw_tspec;
317 } else if (kw->kw_token == T_SCLASS) {
318 sym->s_scl = kw->kw_scl;
319 } else if (kw->kw_token == T_QUAL) {
320 sym->s_tqual = kw->kw_tqual;
321 }
322
323 symtab_add(sym);
324 }
325
326 /*
327 * All keywords are written to the symbol table. This saves us looking
328 * in a extra table for each name we found.
329 */
330 void
331 initscan(void)
332 {
333 struct kwtab *kw;
334
335 for (kw = kwtab; kw->kw_name != NULL; kw++) {
336 if ((kw->kw_c89 || kw->kw_c99) && tflag)
337 continue;
338 if (kw->kw_c99 && !(Sflag || gflag))
339 continue;
340 if (kw->kw_gcc && !gflag)
341 continue;
342 add_keyword(kw, 1);
343 add_keyword(kw, 2);
344 add_keyword(kw, 4);
345 }
346 }
347
348 /*
349 * Get a free sbuf structure, if possible from the free list
350 */
351 static sbuf_t *
352 allocsb(void)
353 {
354 sbuf_t *sb;
355
356 if ((sb = sbfrlst) != NULL) {
357 sbfrlst = sb->sb_next;
358 #ifdef BLKDEBUG
359 (void)memset(sb, 0, sizeof(*sb));
360 #else
361 sb->sb_next = NULL;
362 #endif
363 } else {
364 sb = xmalloc(sizeof(*sb));
365 (void)memset(sb, 0, sizeof(*sb));
366 }
367 return sb;
368 }
369
370 /*
371 * Put a sbuf structure to the free list
372 */
373 static void
374 freesb(sbuf_t *sb)
375 {
376
377 (void)memset(sb, ZERO, sizeof(*sb));
378 sb->sb_next = sbfrlst;
379 sbfrlst = sb;
380 }
381
382 /*
383 * Read a character and ensure that it is positive (except EOF).
384 * Increment line count(s) if necessary.
385 */
386 static int
387 inpc(void)
388 {
389 int c;
390
391 if ((c = lex_input()) == EOF)
392 return c;
393 c &= CHAR_MASK;
394 if (c == '\0')
395 return EOF; /* lex returns 0 on EOF. */
396 if (c == '\n')
397 lex_next_line();
398 return c;
399 }
400
401 static int
402 hash(const char *s)
403 {
404 u_int v;
405 const u_char *us;
406
407 v = 0;
408 for (us = (const u_char *)s; *us != '\0'; us++) {
409 v = (v << sizeof(v)) + *us;
410 v ^= v >> (sizeof(v) * CHAR_BIT - sizeof(v));
411 }
412 return v % HSHSIZ1;
413 }
414
415 /*
416 * Lex has found a letter followed by zero or more letters or digits.
417 * It looks for a symbol in the symbol table with the same name. This
418 * symbol must either be a keyword or a symbol of the type required by
419 * symtyp (label, member, tag, ...).
420 *
421 * If it is a keyword, the token is returned. In some cases it is described
422 * more deeply by data written to yylval.
423 *
424 * If it is a symbol, T_NAME is returned and the pointer to a sbuf struct
425 * is stored in yylval. This struct contains the name of the symbol, its
426 * length and hash value. If there is already a symbol of the same name
427 * and type in the symbol table, the sbuf struct also contains a pointer
428 * to the symbol table entry.
429 */
430 extern int
431 lex_name(const char *yytext, size_t yyleng)
432 {
433 char *s;
434 sbuf_t *sb;
435 sym_t *sym;
436 int tok;
437
438 sb = allocsb();
439 sb->sb_name = yytext;
440 sb->sb_len = yyleng;
441 if ((sym = search(sb)) != NULL && sym->s_keyword != NULL) {
442 freesb(sb);
443 return keyw(sym);
444 }
445
446 sb->sb_sym = sym;
447
448 if (sym != NULL) {
449 lint_assert(block_level >= sym->s_block_level);
450 sb->sb_name = sym->s_name;
451 sb->sb_len = strlen(sym->s_name);
452 tok = sym->s_scl == TYPEDEF ? T_TYPENAME : T_NAME;
453 } else {
454 s = getblk(yyleng + 1);
455 (void)memcpy(s, yytext, yyleng + 1);
456 sb->sb_name = s;
457 sb->sb_len = yyleng;
458 tok = T_NAME;
459 }
460
461 yylval.y_name = sb;
462 return tok;
463 }
464
465 static sym_t *
466 search(sbuf_t *sb)
467 {
468 int h;
469 sym_t *sym;
470 const struct kwtab *kw;
471
472 h = hash(sb->sb_name);
473 for (sym = symtab[h]; sym != NULL; sym = sym->s_link) {
474 if (strcmp(sym->s_name, sb->sb_name) != 0)
475 continue;
476 kw = sym->s_keyword;
477
478 if (kw != NULL && !kw->kw_attr)
479 return sym;
480 if (kw != NULL && attron)
481 return sym;
482 if (kw == NULL && !attron && sym->s_kind == symtyp)
483 return sym;
484 }
485
486 return NULL;
487 }
488
489 static int
490 keyw(sym_t *sym)
491 {
492 int t;
493
494 if ((t = (int)sym->s_value.v_quad) == T_SCLASS) {
495 yylval.y_scl = sym->s_scl;
496 } else if (t == T_TYPE || t == T_STRUCT_OR_UNION) {
497 yylval.y_tspec = sym->s_tspec;
498 } else if (t == T_QUAL) {
499 yylval.y_tqual = sym->s_tqual;
500 }
501 return t;
502 }
503
504 /*
505 * Convert a string representing an integer into internal representation.
506 * Return T_CON, storing the numeric value in yylval, for yylex.
507 */
508 int
509 lex_integer_constant(const char *yytext, size_t yyleng, int base)
510 {
511 int l_suffix, u_suffix;
512 int len;
513 const char *cp;
514 char c, *eptr;
515 tspec_t typ;
516 bool ansiu;
517 bool warned = false;
518 #ifdef TARG_INT128_MAX
519 __uint128_t uq = 0;
520 static tspec_t contypes[2][4] = {
521 { INT, LONG, QUAD, INT128, },
522 { UINT, ULONG, UQUAD, UINT128, }
523 };
524 #else
525 uint64_t uq = 0;
526 static tspec_t contypes[2][3] = {
527 { INT, LONG, QUAD, },
528 { UINT, ULONG, UQUAD, }
529 };
530 #endif
531
532 cp = yytext;
533 len = yyleng;
534
535 /* skip 0[xX] or 0[bB] */
536 if (base == 16 || base == 2) {
537 cp += 2;
538 len -= 2;
539 }
540
541 /* read suffixes */
542 l_suffix = u_suffix = 0;
543 for (;;) {
544 if ((c = cp[len - 1]) == 'l' || c == 'L') {
545 l_suffix++;
546 } else if (c == 'u' || c == 'U') {
547 u_suffix++;
548 } else {
549 break;
550 }
551 len--;
552 }
553 if (l_suffix > 2 || u_suffix > 1) {
554 /* malformed integer constant */
555 warning(251);
556 if (l_suffix > 2)
557 l_suffix = 2;
558 if (u_suffix > 1)
559 u_suffix = 1;
560 }
561 if (tflag && u_suffix != 0) {
562 /* suffix U is illegal in traditional C */
563 warning(97);
564 }
565 typ = contypes[u_suffix][l_suffix];
566
567 errno = 0;
568
569 uq = strtouq(cp, &eptr, base);
570 lint_assert(eptr == cp + len);
571 if (errno != 0) {
572 /* integer constant out of range */
573 warning(252);
574 warned = true;
575 }
576
577 /*
578 * If the value is too big for the current type, we must choose
579 * another type.
580 */
581 ansiu = false;
582 switch (typ) {
583 case INT:
584 if (uq <= TARG_INT_MAX) {
585 /* ok */
586 } else if (uq <= TARG_UINT_MAX && base != 10) {
587 typ = UINT;
588 } else if (uq <= TARG_LONG_MAX) {
589 typ = LONG;
590 } else {
591 typ = ULONG;
592 if (uq > TARG_ULONG_MAX && !warned) {
593 /* integer constant out of range */
594 warning(252);
595 }
596 }
597 if (typ == UINT || typ == ULONG) {
598 if (tflag) {
599 typ = LONG;
600 } else if (!sflag) {
601 /*
602 * Remember that the constant is unsigned
603 * only in ANSI C
604 */
605 ansiu = true;
606 }
607 }
608 break;
609 case UINT:
610 if (uq > TARG_UINT_MAX) {
611 typ = ULONG;
612 if (uq > TARG_ULONG_MAX && !warned) {
613 /* integer constant out of range */
614 warning(252);
615 }
616 }
617 break;
618 case LONG:
619 if (uq > TARG_LONG_MAX && !tflag) {
620 typ = ULONG;
621 if (!sflag)
622 ansiu = true;
623 if (uq > TARG_ULONG_MAX && !warned) {
624 /* integer constant out of range */
625 warning(252);
626 }
627 }
628 break;
629 case ULONG:
630 if (uq > TARG_ULONG_MAX && !warned) {
631 /* integer constant out of range */
632 warning(252);
633 }
634 break;
635 case QUAD:
636 if (uq > TARG_QUAD_MAX && !tflag) {
637 typ = UQUAD;
638 if (!sflag)
639 ansiu = true;
640 }
641 break;
642 case UQUAD:
643 if (uq > TARG_UQUAD_MAX && !warned) {
644 /* integer constant out of range */
645 warning(252);
646 }
647 break;
648 #ifdef INT128_SIZE
649 case INT128:
650 #ifdef TARG_INT128_MAX
651 if (uq > TARG_INT128_MAX && !tflag) {
652 typ = UINT128;
653 if (!sflag)
654 ansiu = true;
655 }
656 #endif
657 break;
658 case UINT128:
659 #ifdef TARG_INT128_MAX
660 if (uq > TARG_UINT128_MAX && !warned) {
661 /* integer constant out of range */
662 warning(252);
663 }
664 #endif
665 break;
666 #endif
667 /* LINTED206: (enumeration values not handled in switch) */
668 case STRUCT:
669 case VOID:
670 case LDOUBLE:
671 case FUNC:
672 case ARRAY:
673 case PTR:
674 case ENUM:
675 case UNION:
676 case SIGNED:
677 case NOTSPEC:
678 case DOUBLE:
679 case FLOAT:
680 case USHORT:
681 case SHORT:
682 case UCHAR:
683 case SCHAR:
684 case CHAR:
685 case BOOL:
686 case UNSIGN:
687 case FCOMPLEX:
688 case DCOMPLEX:
689 case LCOMPLEX:
690 case COMPLEX:
691 break;
692 }
693
694 uq = (uint64_t)convert_integer((int64_t)uq, typ, -1);
695
696 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
697 yylval.y_val->v_tspec = typ;
698 yylval.y_val->v_unsigned_since_c90 = ansiu;
699 yylval.y_val->v_quad = (int64_t)uq;
700
701 return T_CON;
702 }
703
704 int
705 msb(int64_t q, tspec_t t, int len)
706 {
707
708 if (len <= 0)
709 len = size_in_bits(t);
710 return (q & bit(len - 1)) != 0 ? 1 : 0;
711 }
712
713 /*
714 * Extend or truncate q to match t. If t is signed, sign-extend.
715 *
716 * len is the number of significant bits. If len is -1, len is set
717 * to the width of type t.
718 */
719 int64_t
720 convert_integer(int64_t q, tspec_t t, int len)
721 {
722 uint64_t vbits;
723
724 if (len <= 0)
725 len = size_in_bits(t);
726
727 vbits = value_bits(len);
728 return t == PTR || is_uinteger(t) || msb(q, t, len) == 0
729 ? q & vbits
730 : q | ~vbits;
731 }
732
733 /*
734 * Convert a string representing a floating point value into its integral
735 * representation. Type and value are returned in yylval. fcon()
736 * (and yylex()) returns T_CON.
737 * XXX Currently it is not possible to convert constants of type
738 * long double which are greater than DBL_MAX.
739 */
740 int
741 lex_floating_constant(const char *yytext, size_t yyleng)
742 {
743 const char *cp;
744 int len;
745 tspec_t typ;
746 char c, *eptr;
747 double d;
748 float f = 0;
749
750 cp = yytext;
751 len = yyleng;
752
753 if (cp[len - 1] == 'i') {
754 /* imaginary, do nothing for now */
755 len--;
756 }
757 if ((c = cp[len - 1]) == 'f' || c == 'F') {
758 typ = FLOAT;
759 len--;
760 } else if (c == 'l' || c == 'L') {
761 typ = LDOUBLE;
762 len--;
763 } else {
764 if (c == 'd' || c == 'D')
765 len--;
766 typ = DOUBLE;
767 }
768
769 if (tflag && typ != DOUBLE) {
770 /* suffixes F and L are illegal in traditional C */
771 warning(98);
772 }
773
774 errno = 0;
775 d = strtod(cp, &eptr);
776 if (eptr != cp + len) {
777 switch (*eptr) {
778 /*
779 * XXX: non-native non-current strtod() may not handle hex
780 * floats, ignore the rest if we find traces of hex float
781 * syntax...
782 */
783 case 'p':
784 case 'P':
785 case 'x':
786 case 'X':
787 d = 0;
788 errno = 0;
789 break;
790 default:
791 INTERNAL_ERROR("fcon(%s->%s)", cp, eptr);
792 }
793 }
794 if (errno != 0)
795 /* floating-point constant out of range */
796 warning(248);
797
798 if (typ == FLOAT) {
799 f = (float)d;
800 if (finite(f) == 0) {
801 /* floating-point constant out of range */
802 warning(248);
803 f = f > 0 ? FLT_MAX : -FLT_MAX;
804 }
805 }
806
807 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
808 yylval.y_val->v_tspec = typ;
809 if (typ == FLOAT) {
810 yylval.y_val->v_ldbl = f;
811 } else {
812 yylval.y_val->v_ldbl = d;
813 }
814
815 return T_CON;
816 }
817
818 int
819 lex_operator(int t, op_t o)
820 {
821
822 yylval.y_op = o;
823 return t;
824 }
825
826 /*
827 * Called if lex found a leading \'.
828 */
829 int
830 lex_character_constant(void)
831 {
832 size_t n;
833 int val, c;
834
835 n = 0;
836 val = 0;
837 while ((c = get_escaped_char('\'')) >= 0) {
838 val = (val << CHAR_SIZE) + c;
839 n++;
840 }
841 if (c == -2) {
842 /* unterminated character constant */
843 error(253);
844 } else if (n > sizeof(int) || (n > 1 && (pflag || hflag))) {
845 /* XXX: should rather be sizeof(TARG_INT) */
846
847 /* too many characters in character constant */
848 error(71);
849 } else if (n > 1) {
850 /* multi-character character constant */
851 warning(294);
852 } else if (n == 0) {
853 /* empty character constant */
854 error(73);
855 }
856 if (n == 1) {
857 /*
858 * XXX: use the target platform's 'char' instead of the
859 * 'char' from the execution environment, to be able to
860 * run lint for powerpc on x86_64.
861 */
862 val = (char)val;
863 }
864
865 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
866 yylval.y_val->v_tspec = INT;
867 yylval.y_val->v_quad = val;
868
869 return T_CON;
870 }
871
872 /*
873 * Called if lex found a leading L\'
874 */
875 int
876 lex_wide_character_constant(void)
877 {
878 static char buf[MB_LEN_MAX + 1];
879 size_t n, nmax;
880 int c;
881 wchar_t wc;
882
883 nmax = MB_CUR_MAX;
884
885 n = 0;
886 while ((c = get_escaped_char('\'')) >= 0) {
887 if (n < nmax)
888 buf[n] = (char)c;
889 n++;
890 }
891
892 wc = 0;
893
894 if (c == -2) {
895 /* unterminated character constant */
896 error(253);
897 } else if (n == 0) {
898 /* empty character constant */
899 error(73);
900 } else if (n > nmax) {
901 n = nmax;
902 /* too many characters in character constant */
903 error(71);
904 } else {
905 buf[n] = '\0';
906 (void)mbtowc(NULL, NULL, 0);
907 if (mbtowc(&wc, buf, nmax) < 0)
908 /* invalid multibyte character */
909 error(291);
910 }
911
912 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
913 yylval.y_val->v_tspec = WCHAR;
914 yylval.y_val->v_quad = wc;
915
916 return T_CON;
917 }
918
919 /*
920 * Read a character which is part of a character constant or of a string
921 * and handle escapes.
922 *
923 * The argument is the character which delimits the character constant or
924 * string.
925 *
926 * Returns -1 if the end of the character constant or string is reached,
927 * -2 if the EOF is reached, and the character otherwise.
928 */
929 static int
930 get_escaped_char(int delim)
931 {
932 static int pbc = -1;
933 int n, c, v;
934
935 if (pbc == -1) {
936 c = inpc();
937 } else {
938 c = pbc;
939 pbc = -1;
940 }
941 if (c == delim)
942 return -1;
943 switch (c) {
944 case '\n':
945 if (tflag) {
946 /* newline in string or char constant */
947 error(254);
948 return -2;
949 }
950 return c;
951 case 0:
952 /* syntax error '%s' */
953 error(249, "EOF or null byte in literal");
954 return -2;
955 case EOF:
956 return -2;
957 case '\\':
958 switch (c = inpc()) {
959 case '"':
960 if (tflag && delim == '\'')
961 /* \" inside character constants undef... */
962 warning(262);
963 return '"';
964 case '\'':
965 return '\'';
966 case '?':
967 if (tflag)
968 /* \? undefined in traditional C */
969 warning(263);
970 return '?';
971 case '\\':
972 return '\\';
973 case 'a':
974 if (tflag)
975 /* \a undefined in traditional C */
976 warning(81);
977 return '\a';
978 case 'b':
979 return '\b';
980 case 'f':
981 return '\f';
982 case 'n':
983 return '\n';
984 case 'r':
985 return '\r';
986 case 't':
987 return '\t';
988 case 'v':
989 if (tflag)
990 /* \v undefined in traditional C */
991 warning(264);
992 return '\v';
993 case '8': case '9':
994 /* bad octal digit %c */
995 warning(77, c);
996 /* FALLTHROUGH */
997 case '0': case '1': case '2': case '3':
998 case '4': case '5': case '6': case '7':
999 n = 3;
1000 v = 0;
1001 do {
1002 v = (v << 3) + (c - '0');
1003 c = inpc();
1004 } while (--n > 0 && '0' <= c && c <= '7');
1005 pbc = c;
1006 if (v > TARG_UCHAR_MAX) {
1007 /* character escape does not fit in character */
1008 warning(76);
1009 v &= CHAR_MASK;
1010 }
1011 return v;
1012 case 'x':
1013 if (tflag)
1014 /* \x undefined in traditional C */
1015 warning(82);
1016 v = 0;
1017 n = 0;
1018 while ((c = inpc()) >= 0 && isxdigit(c)) {
1019 c = isdigit(c) ?
1020 c - '0' : toupper(c) - 'A' + 10;
1021 v = (v << 4) + c;
1022 if (n >= 0) {
1023 if ((v & ~CHAR_MASK) != 0) {
1024 /* overflow in hex escape */
1025 warning(75);
1026 n = -1;
1027 } else {
1028 n++;
1029 }
1030 }
1031 }
1032 pbc = c;
1033 if (n == 0) {
1034 /* no hex digits follow \x */
1035 error(74);
1036 } if (n == -1) {
1037 v &= CHAR_MASK;
1038 }
1039 return v;
1040 case '\n':
1041 return get_escaped_char(delim);
1042 case EOF:
1043 return -2;
1044 default:
1045 if (isprint(c)) {
1046 /* dubious escape \%c */
1047 warning(79, c);
1048 } else {
1049 /* dubious escape \%o */
1050 warning(80, c);
1051 }
1052 }
1053 }
1054 return c;
1055 }
1056
1057 /* See https://gcc.gnu.org/onlinedocs/cpp/Preprocessor-Output.html */
1058 static void
1059 parse_line_directive_flags(const char *p,
1060 bool *is_begin, bool *is_end, bool *is_system)
1061 {
1062
1063 *is_begin = false;
1064 *is_end = false;
1065 *is_system = false;
1066
1067 while (*p != '\0') {
1068 while (ch_isspace(*p))
1069 p++;
1070
1071 const char *word_start = p;
1072 while (*p != '\0' && !ch_isspace(*p))
1073 p++;
1074 const char *word_end = p;
1075
1076 if (word_end - word_start == 1 && word_start[0] == '1')
1077 *is_begin = true;
1078 if (word_end - word_start == 1 && word_start[0] == '2')
1079 *is_end = true;
1080 if (word_end - word_start == 1 && word_start[0] == '3')
1081 *is_system = true;
1082 /* Flag '4' would only be interesting if lint handled C++. */
1083 }
1084
1085 #if 0
1086 if (*p != '\0') {
1087 /* syntax error '%s' */
1088 warning(249, "extra character(s) after directive");
1089 }
1090 #endif
1091 }
1092
1093 /*
1094 * Called for preprocessor directives. Currently implemented are:
1095 * # lineno
1096 * # lineno "filename"
1097 * # lineno "filename" GCC-flag...
1098 */
1099 void
1100 lex_directive(const char *yytext)
1101 {
1102 const char *cp, *fn;
1103 char c, *eptr;
1104 size_t fnl;
1105 long ln;
1106 bool is_begin, is_end, is_system;
1107
1108 static bool first = true;
1109
1110 /* Go to first non-whitespace after # */
1111 for (cp = yytext + 1; (c = *cp) == ' ' || c == '\t'; cp++)
1112 continue;
1113
1114 if (!ch_isdigit(c)) {
1115 if (strncmp(cp, "pragma", 6) == 0 && ch_isspace(cp[6]))
1116 return;
1117 error:
1118 /* undefined or invalid # directive */
1119 warning(255);
1120 return;
1121 }
1122 ln = strtol(--cp, &eptr, 10);
1123 if (cp == eptr)
1124 goto error;
1125 if ((c = *(cp = eptr)) != ' ' && c != '\t' && c != '\0')
1126 goto error;
1127 while ((c = *cp++) == ' ' || c == '\t')
1128 continue;
1129 if (c != '\0') {
1130 if (c != '"')
1131 goto error;
1132 fn = cp;
1133 while ((c = *cp) != '"' && c != '\0')
1134 cp++;
1135 if (c != '"')
1136 goto error;
1137 if ((fnl = cp++ - fn) > PATH_MAX)
1138 goto error;
1139 /* empty string means stdin */
1140 if (fnl == 0) {
1141 fn = "{standard input}";
1142 fnl = 16; /* strlen (fn) */
1143 }
1144 curr_pos.p_file = record_filename(fn, fnl);
1145 /*
1146 * If this is the first directive, the name is the name
1147 * of the C source file as specified at the command line.
1148 * It is written to the output file.
1149 */
1150 if (first) {
1151 csrc_pos.p_file = curr_pos.p_file;
1152 outsrc(transform_filename(curr_pos.p_file,
1153 strlen(curr_pos.p_file)));
1154 first = false;
1155 }
1156
1157 parse_line_directive_flags(cp, &is_begin, &is_end, &is_system);
1158 update_location(curr_pos.p_file, (int)ln, is_begin, is_end);
1159 in_system_header = is_system;
1160 }
1161 curr_pos.p_line = (int)ln - 1;
1162 curr_pos.p_uniq = 0;
1163 if (curr_pos.p_file == csrc_pos.p_file) {
1164 csrc_pos.p_line = (int)ln - 1;
1165 csrc_pos.p_uniq = 0;
1166 }
1167 }
1168
1169 /*
1170 * Handle lint comments such as ARGSUSED.
1171 *
1172 * If one of these comments is recognized, the argument, if any, is
1173 * parsed and a function which handles this comment is called.
1174 */
1175 void
1176 lex_comment(void)
1177 {
1178 int c, lc;
1179 static const struct {
1180 const char *keywd;
1181 bool arg;
1182 void (*func)(int);
1183 } keywtab[] = {
1184 { "ARGSUSED", true, argsused },
1185 { "BITFIELDTYPE", false, bitfieldtype },
1186 { "CONSTCOND", false, constcond },
1187 { "CONSTANTCOND", false, constcond },
1188 { "CONSTANTCONDITION", false, constcond },
1189 { "FALLTHRU", false, fallthru },
1190 { "FALLTHROUGH", false, fallthru },
1191 { "LINTLIBRARY", false, lintlib },
1192 { "LINTED", true, linted },
1193 { "LONGLONG", false, longlong },
1194 { "NOSTRICT", true, linted },
1195 { "NOTREACHED", false, not_reached },
1196 { "PRINTFLIKE", true, printflike },
1197 { "PROTOLIB", true, protolib },
1198 { "SCANFLIKE", true, scanflike },
1199 { "VARARGS", true, varargs },
1200 };
1201 char keywd[32];
1202 char arg[32];
1203 size_t l, i;
1204 int a;
1205 bool eoc;
1206
1207 eoc = false;
1208
1209 /* Skip whitespace after the start of the comment */
1210 while ((c = inpc()) != EOF && isspace(c))
1211 continue;
1212
1213 /* Read the potential keyword to keywd */
1214 l = 0;
1215 while (c != EOF && isupper(c) && l < sizeof(keywd) - 1) {
1216 keywd[l++] = (char)c;
1217 c = inpc();
1218 }
1219 keywd[l] = '\0';
1220
1221 /* look for the keyword */
1222 for (i = 0; i < sizeof(keywtab) / sizeof(keywtab[0]); i++) {
1223 if (strcmp(keywtab[i].keywd, keywd) == 0)
1224 break;
1225 }
1226 if (i == sizeof(keywtab) / sizeof(keywtab[0]))
1227 goto skip_rest;
1228
1229 /* skip whitespace after the keyword */
1230 while (c != EOF && isspace(c))
1231 c = inpc();
1232
1233 /* read the argument, if the keyword accepts one and there is one */
1234 l = 0;
1235 if (keywtab[i].arg) {
1236 while (c != EOF && isdigit(c) && l < sizeof(arg) - 1) {
1237 arg[l++] = (char)c;
1238 c = inpc();
1239 }
1240 }
1241 arg[l] = '\0';
1242 a = l != 0 ? atoi(arg) : -1;
1243
1244 /* skip whitespace after the argument */
1245 while (c != EOF && isspace(c))
1246 c = inpc();
1247
1248 if (c != '*' || (c = inpc()) != '/') {
1249 if (keywtab[i].func != linted)
1250 /* extra characters in lint comment */
1251 warning(257);
1252 } else {
1253 /*
1254 * remember that we have already found the end of the
1255 * comment
1256 */
1257 eoc = true;
1258 }
1259
1260 if (keywtab[i].func != NULL)
1261 (*keywtab[i].func)(a);
1262
1263 skip_rest:
1264 while (!eoc) {
1265 lc = c;
1266 if ((c = inpc()) == EOF) {
1267 /* unterminated comment */
1268 error(256);
1269 break;
1270 }
1271 if (lc == '*' && c == '/')
1272 eoc = true;
1273 }
1274 }
1275
1276 /*
1277 * Handle // style comments
1278 */
1279 void
1280 lex_slash_slash_comment(void)
1281 {
1282 int c;
1283
1284 if (!Sflag && !gflag)
1285 /* %s C does not support // comments */
1286 gnuism(312, tflag ? "traditional" : "ANSI");
1287
1288 while ((c = inpc()) != EOF && c != '\n')
1289 continue;
1290 }
1291
1292 /*
1293 * Clear flags for lint comments LINTED, LONGLONG and CONSTCOND.
1294 * clear_warn_flags() is called after function definitions and global and
1295 * local declarations and definitions. It is also called between
1296 * the controlling expression and the body of control statements
1297 * (if, switch, for, while).
1298 */
1299 void
1300 clear_warn_flags(void)
1301 {
1302
1303 lwarn = LWARN_ALL;
1304 quadflg = false;
1305 constcond_flag = false;
1306 }
1307
1308 /*
1309 * Strings are stored in a dynamically allocated buffer and passed
1310 * in yylval.y_xstrg to the parser. The parser or the routines called
1311 * by the parser are responsible for freeing this buffer.
1312 */
1313 int
1314 lex_string(void)
1315 {
1316 u_char *s;
1317 int c;
1318 size_t len, max;
1319 strg_t *strg;
1320
1321 s = xmalloc(max = 64);
1322
1323 len = 0;
1324 while ((c = get_escaped_char('"')) >= 0) {
1325 /* +1 to reserve space for a trailing NUL character */
1326 if (len + 1 == max)
1327 s = xrealloc(s, max *= 2);
1328 s[len++] = (char)c;
1329 }
1330 s[len] = '\0';
1331 if (c == -2)
1332 /* unterminated string constant */
1333 error(258);
1334
1335 strg = xcalloc(1, sizeof(*strg));
1336 strg->st_tspec = CHAR;
1337 strg->st_len = len;
1338 strg->st_cp = s;
1339
1340 yylval.y_string = strg;
1341 return T_STRING;
1342 }
1343
1344 int
1345 lex_wide_string(void)
1346 {
1347 char *s;
1348 int c, n;
1349 size_t i, wi;
1350 size_t len, max, wlen;
1351 wchar_t *ws;
1352 strg_t *strg;
1353
1354 s = xmalloc(max = 64);
1355 len = 0;
1356 while ((c = get_escaped_char('"')) >= 0) {
1357 /* +1 to save space for a trailing NUL character */
1358 if (len + 1 >= max)
1359 s = xrealloc(s, max *= 2);
1360 s[len++] = (char)c;
1361 }
1362 s[len] = '\0';
1363 if (c == -2)
1364 /* unterminated string constant */
1365 error(258);
1366
1367 /* get length of wide-character string */
1368 (void)mblen(NULL, 0);
1369 for (i = 0, wlen = 0; i < len; i += n, wlen++) {
1370 if ((n = mblen(&s[i], MB_CUR_MAX)) == -1) {
1371 /* invalid multibyte character */
1372 error(291);
1373 break;
1374 }
1375 if (n == 0)
1376 n = 1;
1377 }
1378
1379 ws = xmalloc((wlen + 1) * sizeof(*ws));
1380
1381 /* convert from multibyte to wide char */
1382 (void)mbtowc(NULL, NULL, 0);
1383 for (i = 0, wi = 0; i < len; i += n, wi++) {
1384 if ((n = mbtowc(&ws[wi], &s[i], MB_CUR_MAX)) == -1)
1385 break;
1386 if (n == 0)
1387 n = 1;
1388 }
1389 ws[wi] = 0;
1390 free(s);
1391
1392 strg = xcalloc(1, sizeof(*strg));
1393 strg->st_tspec = WCHAR;
1394 strg->st_len = wlen;
1395 strg->st_wcp = ws;
1396
1397 yylval.y_string = strg;
1398 return T_STRING;
1399 }
1400
1401 /*
1402 * As noted above, the scanner does not create new symbol table entries
1403 * for symbols it cannot find in the symbol table. This is to avoid
1404 * putting undeclared symbols into the symbol table if a syntax error
1405 * occurs.
1406 *
1407 * getsym() is called as soon as it is probably ok to put the symbol in the
1408 * symbol table. It is still possible that symbols are put in the symbol
1409 * table that are not completely declared due to syntax errors. To avoid too
1410 * many problems in this case, symbols get type 'int' in getsym().
1411 *
1412 * XXX calls to getsym() should be delayed until decl1*() is called.
1413 */
1414 sym_t *
1415 getsym(sbuf_t *sb)
1416 {
1417 dinfo_t *di;
1418 char *s;
1419 sym_t *sym;
1420
1421 sym = sb->sb_sym;
1422
1423 /*
1424 * During member declaration it is possible that name() looked
1425 * for symbols of type FVFT, although it should have looked for
1426 * symbols of type FTAG. Same can happen for labels. Both cases
1427 * are compensated here.
1428 */
1429 if (symtyp == FMEMBER || symtyp == FLABEL) {
1430 if (sym == NULL || sym->s_kind == FVFT)
1431 sym = search(sb);
1432 }
1433
1434 if (sym != NULL) {
1435 if (sym->s_kind != symtyp)
1436 INTERNAL_ERROR("getsym(%d, %d)", sym->s_kind, symtyp);
1437 symtyp = FVFT;
1438 freesb(sb);
1439 return sym;
1440 }
1441
1442 /* create a new symbol table entry */
1443
1444 /* labels must always be allocated at level 1 (outermost block) */
1445 if (symtyp == FLABEL) {
1446 sym = getlblk(1, sizeof(*sym));
1447 s = getlblk(1, sb->sb_len + 1);
1448 (void)memcpy(s, sb->sb_name, sb->sb_len + 1);
1449 sym->s_name = s;
1450 sym->s_block_level = 1;
1451 di = dcs;
1452 while (di->d_next != NULL && di->d_next->d_next != NULL)
1453 di = di->d_next;
1454 lint_assert(di->d_ctx == AUTO);
1455 } else {
1456 sym = getblk(sizeof(*sym));
1457 sym->s_name = sb->sb_name;
1458 sym->s_block_level = block_level;
1459 di = dcs;
1460 }
1461
1462 UNIQUE_CURR_POS(sym->s_def_pos);
1463 if ((sym->s_kind = symtyp) != FLABEL)
1464 sym->s_type = gettyp(INT);
1465
1466 symtyp = FVFT;
1467
1468 symtab_add(sym);
1469
1470 *di->d_ldlsym = sym;
1471 di->d_ldlsym = &sym->s_dlnxt;
1472
1473 freesb(sb);
1474 return sym;
1475 }
1476
1477 /*
1478 * Construct a temporary symbol. The symbol name starts with a digit, making
1479 * the name illegal.
1480 */
1481 sym_t *
1482 mktempsym(type_t *t)
1483 {
1484 static int n = 0;
1485 char *s = getlblk(block_level, 64);
1486 sym_t *sym = getblk(sizeof(*sym));
1487 scl_t scl;
1488
1489 (void)snprintf(s, 64, "%.8d_tmp", n++);
1490
1491 scl = dcs->d_scl;
1492 if (scl == NOSCL)
1493 scl = block_level > 0 ? AUTO : EXTERN;
1494
1495 sym->s_name = s;
1496 sym->s_type = t;
1497 sym->s_block_level = block_level;
1498 sym->s_scl = scl;
1499 sym->s_kind = FVFT;
1500 sym->s_used = true;
1501 sym->s_set = true;
1502
1503 symtab_add(sym);
1504
1505 *dcs->d_ldlsym = sym;
1506 dcs->d_ldlsym = &sym->s_dlnxt;
1507
1508 return sym;
1509 }
1510
1511 /* Remove a symbol forever from the symbol table. */
1512 void
1513 rmsym(sym_t *sym)
1514 {
1515
1516 debug_step("rmsym '%s' %d '%s'",
1517 sym->s_name, (int)sym->s_kind, type_name(sym->s_type));
1518 symtab_remove(sym);
1519
1520 /* avoid that the symbol will later be put back to the symbol table */
1521 sym->s_block_level = -1;
1522 }
1523
1524 /*
1525 * Remove a list of symbols declared at one level from the symbol
1526 * table.
1527 */
1528 void
1529 rmsyms(sym_t *syms)
1530 {
1531 sym_t *sym;
1532
1533 for (sym = syms; sym != NULL; sym = sym->s_dlnxt) {
1534 if (sym->s_block_level != -1) {
1535 debug_step("rmsyms '%s' %d '%s'",
1536 sym->s_name, (int)sym->s_kind,
1537 type_name(sym->s_type));
1538 symtab_remove(sym);
1539 sym->s_rlink = NULL;
1540 }
1541 }
1542 }
1543
1544 /*
1545 * Put a symbol into the symbol table.
1546 */
1547 void
1548 inssym(int bl, sym_t *sym)
1549 {
1550
1551 debug_step("inssym '%s' %d '%s'",
1552 sym->s_name, sym->s_kind, type_name(sym->s_type));
1553 symtab_add(sym);
1554 sym->s_block_level = bl;
1555 lint_assert(sym->s_link == NULL ||
1556 sym->s_block_level >= sym->s_link->s_block_level);
1557 }
1558
1559 /*
1560 * Called at level 0 after syntax errors.
1561 *
1562 * Removes all symbols which are not declared at level 0 from the
1563 * symbol table. Also frees all memory which is not associated with
1564 * level 0.
1565 */
1566 void
1567 cleanup(void)
1568 {
1569 sym_t *sym, *nsym;
1570 int i;
1571
1572 for (i = 0; i < HSHSIZ1; i++) {
1573 for (sym = symtab[i]; sym != NULL; sym = nsym) {
1574 nsym = sym->s_link;
1575 if (sym->s_block_level >= 1)
1576 symtab_remove(sym);
1577 }
1578 }
1579
1580 for (i = mem_block_level; i > 0; i--)
1581 freelblk(i);
1582 }
1583
1584 /*
1585 * Create a new symbol with the name of an existing symbol.
1586 */
1587 sym_t *
1588 pushdown(const sym_t *sym)
1589 {
1590 sym_t *nsym;
1591
1592 debug_step("pushdown '%s' %d '%s'",
1593 sym->s_name, (int)sym->s_kind, type_name(sym->s_type));
1594 nsym = getblk(sizeof(*nsym));
1595 lint_assert(sym->s_block_level <= block_level);
1596 nsym->s_name = sym->s_name;
1597 UNIQUE_CURR_POS(nsym->s_def_pos);
1598 nsym->s_kind = sym->s_kind;
1599 nsym->s_block_level = block_level;
1600
1601 symtab_add(nsym);
1602
1603 *dcs->d_ldlsym = nsym;
1604 dcs->d_ldlsym = &nsym->s_dlnxt;
1605
1606 return nsym;
1607 }
1608
1609 /*
1610 * Free any dynamically allocated memory referenced by
1611 * the value stack or yylval.
1612 * The type of information in yylval is described by tok.
1613 */
1614 void
1615 freeyyv(void *sp, int tok)
1616 {
1617 if (tok == T_NAME || tok == T_TYPENAME) {
1618 sbuf_t *sb = *(sbuf_t **)sp;
1619 freesb(sb);
1620 } else if (tok == T_CON) {
1621 val_t *val = *(val_t **)sp;
1622 free(val);
1623 } else if (tok == T_STRING) {
1624 strg_t *strg = *(strg_t **)sp;
1625 if (strg->st_tspec == CHAR) {
1626 free(strg->st_cp);
1627 } else {
1628 lint_assert(strg->st_tspec == WCHAR);
1629 free(strg->st_wcp);
1630 }
1631 free(strg);
1632 }
1633 }
1634