lex.c revision 1.30 1 /* $NetBSD: lex.c,v 1.30 2021/04/30 19:46:24 christos 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 1.30 2021/04/30 19:46:24 christos 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 #ifdef DEBUG
86 printf("parsing %s:%d\n", curr_pos.p_file, curr_pos.p_line);
87 #endif
88 if (curr_pos.p_file == csrc_pos.p_file) {
89 csrc_pos.p_line++;
90 csrc_pos.p_uniq = 0;
91 }
92 }
93
94 void
95 lex_unknown_character(int c)
96 {
97
98 /* unknown character \%o */
99 error(250, c);
100 }
101
102 #define kwdef(name, token, scl, tspec, tqual, c89, c99, gcc, attr, deco) \
103 { \
104 name, token, scl, tspec, tqual, \
105 (c89) > 0, (c99) > 0, (gcc) > 0, (attr) > 0, deco, \
106 }
107 #define kwdef_token(name, token, c89, c99, gcc, attr, deco) \
108 kwdef(name, token, 0, 0, 0, c89, c99, gcc, attr, deco)
109 #define kwdef_sclass(name, sclass, c89, c99, gcc, attr, deco) \
110 kwdef(name, T_SCLASS, sclass, 0, 0, c89, c99, gcc, attr, deco)
111 #define kwdef_type(name, tspec, c89, c99, gcc, attr, deco) \
112 kwdef(name, T_TYPE, 0, tspec, 0, c89, c99, gcc, attr, deco)
113 #define kwdef_tqual(name, tqual, c89, c99, gcc, attr, deco) \
114 kwdef(name, T_QUAL, 0, 0, tqual, c89, c99, gcc, attr, deco)
115
116 /*
117 * Keywords.
118 * During initialization they are written to the symbol table.
119 */
120 static struct kwtab {
121 const char *kw_name; /* keyword */
122 int kw_token; /* token returned by yylex() */
123 scl_t kw_scl; /* storage class if kw_token T_SCLASS */
124 tspec_t kw_tspec; /* type spec. if kw_token
125 * T_TYPE or T_STRUCT_OR_UNION */
126 tqual_t kw_tqual; /* type qual. if kw_token T_QUAL */
127 bool kw_c89 : 1; /* C89 keyword */
128 bool kw_c99 : 1; /* C99 keyword */
129 bool kw_gcc : 1; /* GCC keyword */
130 bool kw_attr : 1; /* GCC attribute, keyword */
131 u_int kw_deco : 3; /* 1 = name, 2 = __name, 4 = __name__ */
132 } kwtab[] = {
133 #ifdef INT128_SIZE
134 kwdef_type( "__int128_t", INT128, 0,1,0,0,1),
135 kwdef_type( "__uint128_t", UINT128, 0,1,0,0,1),
136 #endif
137 kwdef_tqual( "__thread", THREAD, 0,0,1,0,1),
138 kwdef_token( "_Alignas", T_ALIGNAS, 0,0,0,0,1),
139 kwdef_token( "_Alignof", T_ALIGNOF, 0,0,0,0,1),
140 kwdef_type( "_Bool", BOOL, 0,1,0,0,1),
141 kwdef_type( "_Complex", COMPLEX, 0,1,0,0,1),
142 kwdef_token( "_Generic", T_GENERIC, 0,1,0,0,1),
143 kwdef_token( "_Noreturn", T_NORETURN, 0,1,0,0,1),
144 kwdef_tqual( "_Thread_local", THREAD, 0,1,0,0,1),
145 kwdef_token( "alias", T_AT_ALIAS, 0,0,1,1,5),
146 kwdef_token( "aligned", T_AT_ALIGNED, 0,0,1,1,5),
147 kwdef_token( "alignof", T_ALIGNOF, 0,0,0,0,4),
148 kwdef_token( "alloc_size", T_AT_ALLOC_SIZE, 0,0,1,1,5),
149 kwdef_token( "always_inline", T_AT_ALWAYS_INLINE, 0,0,1,1,5),
150 kwdef_token( "asm", T_ASM, 0,0,1,0,7),
151 kwdef_token( "attribute", T_ATTRIBUTE, 0,0,1,0,6),
152 kwdef_sclass( "auto", AUTO, 0,0,0,0,1),
153 kwdef_token( "bounded", T_AT_BOUNDED, 0,0,1,1,5),
154 kwdef_token( "break", T_BREAK, 0,0,0,0,1),
155 kwdef_token( "buffer", T_AT_BUFFER, 0,0,1,1,5),
156 kwdef_token( "builtin_offsetof", T_BUILTIN_OFFSETOF, 0,0,1,0,2),
157 kwdef_token( "case", T_CASE, 0,0,0,0,1),
158 kwdef_type( "char", CHAR, 0,0,0,0,1),
159 kwdef_token( "cold", T_AT_COLD, 0,0,1,1,5),
160 kwdef_token( "common", T_AT_COMMON, 0,0,1,1,5),
161 kwdef_tqual( "const", CONST, 1,0,0,0,7),
162 kwdef_token( "constructor", T_AT_CONSTRUCTOR, 0,0,1,1,5),
163 kwdef_token( "continue", T_CONTINUE, 0,0,0,0,1),
164 kwdef_token( "default", T_DEFAULT, 0,0,0,0,1),
165 kwdef_token( "deprecated", T_AT_DEPRECATED, 0,0,1,1,5),
166 kwdef_token( "destructor", T_AT_DESTRUCTOR, 0,0,1,1,5),
167 kwdef_token( "do", T_DO, 0,0,0,0,1),
168 kwdef_type( "double", DOUBLE, 0,0,0,0,1),
169 kwdef_token( "else", T_ELSE, 0,0,0,0,1),
170 kwdef_token( "enum", T_ENUM, 0,0,0,0,1),
171 kwdef_token( "extension", T_EXTENSION, 0,0,1,0,4),
172 kwdef_sclass( "extern", EXTERN, 0,0,0,0,1),
173 kwdef_token( "fallthrough", T_AT_FALLTHROUGH, 0,0,1,1,5),
174 kwdef_type( "float", FLOAT, 0,0,0,0,1),
175 kwdef_token( "for", T_FOR, 0,0,0,0,1),
176 kwdef_token( "format", T_AT_FORMAT, 0,0,1,1,5),
177 kwdef_token( "format_arg", T_AT_FORMAT_ARG, 0,0,1,1,5),
178 kwdef_token( "gnu_inline", T_AT_GNU_INLINE, 0,0,1,1,5),
179 kwdef_token( "gnu_printf", T_AT_FORMAT_GNU_PRINTF, 0,0,1,1,5),
180 kwdef_token( "goto", T_GOTO, 0,0,0,0,1),
181 kwdef_token( "if", T_IF, 0,0,0,0,1),
182 kwdef_token( "imag", T_IMAG, 0,1,0,0,4),
183 kwdef_sclass( "inline", INLINE, 0,1,0,0,7),
184 kwdef_type( "int", INT, 0,0,0,0,1),
185 kwdef_type( "long", LONG, 0,0,0,0,1),
186 kwdef_token( "malloc", T_AT_MALLOC, 0,0,1,1,5),
187 kwdef_token( "may_alias", T_AT_MAY_ALIAS, 0,0,1,1,5),
188 kwdef_token( "minbytes", T_AT_MINBYTES, 0,0,1,1,5),
189 kwdef_token( "mode", T_AT_MODE, 0,0,1,1,5),
190 kwdef_token( "no_instrument_function",
191 T_AT_NO_INSTRUMENT_FUNCTION, 0,0,1,1,5),
192 kwdef_token( "noinline", T_AT_NOINLINE, 0,0,1,1,5),
193 kwdef_token( "nonstring", T_AT_NONSTRING, 0,0,1,1,5),
194 kwdef_token( "noreturn", T_AT_NORETURN, 0,0,1,1,5),
195 kwdef_token( "nothrow", T_AT_NOTHROW, 0,0,1,1,5),
196 kwdef_token( "optimize", T_AT_OPTIMIZE, 0,0,1,1,5),
197 kwdef_token( "packed", T_AT_PACKED, 0,0,1,1,5),
198 kwdef_token( "packed", T_PACKED, 0,0,0,0,2),
199 kwdef_token( "pcs", T_AT_PCS, 0,0,0,0,5),
200 kwdef_token( "printf", T_AT_FORMAT_PRINTF, 0,0,1,1,5),
201 kwdef_token( "pure", T_AT_PURE, 0,0,1,1,5),
202 kwdef_token( "real", T_REAL, 0,1,0,0,4),
203 kwdef_sclass( "register", REG, 0,0,0,0,1),
204 kwdef_tqual( "restrict", RESTRICT, 0,1,0,0,5),
205 kwdef_token( "return", T_RETURN, 0,0,0,0,1),
206 kwdef_token( "returns_twice", T_AT_RETURNS_TWICE, 0,0,1,1,5),
207 kwdef_token( "scanf", T_AT_FORMAT_SCANF, 0,0,1,1,5),
208 kwdef_token( "section", T_AT_SECTION, 0,0,1,1,7),
209 kwdef_token( "sentinel", T_AT_SENTINEL, 0,0,1,1,5),
210 kwdef_type( "short", SHORT, 0,0,0,0,1),
211 kwdef_type( "signed", SIGNED, 1,0,0,0,3),
212 kwdef_token( "sizeof", T_SIZEOF, 0,0,0,0,1),
213 kwdef_sclass( "static", STATIC, 0,0,0,0,1),
214 kwdef_token( "strfmon", T_AT_FORMAT_STRFMON, 0,0,1,1,5),
215 kwdef_token( "strftime", T_AT_FORMAT_STRFTIME, 0,0,1,1,5),
216 kwdef_token( "string", T_AT_STRING, 0,0,1,1,5),
217 kwdef("struct", T_STRUCT_OR_UNION, 0, STRUCT, 0, 0,0,0,0,1),
218 kwdef_token( "switch", T_SWITCH, 0,0,0,0,1),
219 kwdef_token( "symbolrename", T_SYMBOLRENAME, 0,0,0,0,2),
220 kwdef_token( "syslog", T_AT_FORMAT_SYSLOG, 0,0,1,1,5),
221 kwdef_token( "transparent_union", T_AT_TUNION, 0,0,1,1,5),
222 kwdef_token( "tls_model", T_AT_TLS_MODEL, 0,0,1,1,5),
223 kwdef_sclass( "typedef", TYPEDEF, 0,0,0,0,1),
224 kwdef_token( "typeof", T_TYPEOF, 0,0,1,0,7),
225 kwdef("union", T_STRUCT_OR_UNION, 0, UNION, 0, 0,0,0,0,1),
226 kwdef_type( "unsigned", UNSIGN, 0,0,0,0,1),
227 kwdef_token( "unused", T_AT_UNUSED, 0,0,1,1,5),
228 kwdef_token( "used", T_AT_USED, 0,0,1,1,5),
229 kwdef_token( "visibility", T_AT_VISIBILITY, 0,0,1,1,5),
230 kwdef_type( "void", VOID, 0,0,0,0,1),
231 kwdef_tqual( "volatile", VOLATILE, 1,0,0,0,7),
232 kwdef_token("warn_unused_result", T_AT_WARN_UNUSED_RESULT, 0,0,1,1,5),
233 kwdef_token( "weak", T_AT_WEAK, 0,0,1,1,5),
234 kwdef_token( "while", T_WHILE, 0,0,0,0,1),
235 kwdef(NULL, 0, 0, 0, 0, 0, 0, 0, 0, 0),
236 #undef kwdef
237 #undef kwdef_token
238 #undef kwdef_sclass
239 #undef kwdef_type
240 #undef kwdef_tqual
241 };
242
243 /* Symbol table */
244 static sym_t *symtab[HSHSIZ1];
245
246 /* bit i of the entry with index i is set */
247 uint64_t qbmasks[64];
248
249 /* least significant i bits are set in the entry with index i */
250 uint64_t qlmasks[64 + 1];
251
252 /* least significant i bits are not set in the entry with index i */
253 uint64_t qumasks[64 + 1];
254
255 /* free list for sbuf structures */
256 static sbuf_t *sbfrlst;
257
258 /* type of next expected symbol */
259 symt_t symtyp;
260
261
262 static void
263 add_keyword(struct kwtab *kw, int deco)
264 {
265 sym_t *sym;
266 size_t h;
267 char buf[256];
268 const char *name;
269
270 if ((kw->kw_deco & deco) == 0)
271 return;
272
273 switch (deco) {
274 case 1:
275 name = kw->kw_name;
276 break;
277 case 2:
278 snprintf(buf, sizeof(buf), "__%s", kw->kw_name);
279 name = strdup(buf);
280 break;
281 default:
282 lint_assert(deco == 4);
283 snprintf(buf, sizeof(buf), "__%s__", kw->kw_name);
284 name = strdup(buf);
285 break;
286 }
287
288 if (name == NULL)
289 err(1, "Can't init symbol table");
290
291 sym = getblk(sizeof(*sym));
292 sym->s_name = name;
293 sym->s_keyword = kw;
294 sym->s_value.v_quad = kw->kw_token;
295 if (kw->kw_token == T_TYPE || kw->kw_token == T_STRUCT_OR_UNION) {
296 sym->s_tspec = kw->kw_tspec;
297 } else if (kw->kw_token == T_SCLASS) {
298 sym->s_scl = kw->kw_scl;
299 } else if (kw->kw_token == T_QUAL) {
300 sym->s_tqual = kw->kw_tqual;
301 }
302 h = hash(sym->s_name);
303 if ((sym->s_link = symtab[h]) != NULL)
304 symtab[h]->s_rlink = &sym->s_link;
305 sym->s_rlink = &symtab[h];
306 symtab[h] = sym;
307 }
308
309 /*
310 * All keywords are written to the symbol table. This saves us looking
311 * in a extra table for each name we found.
312 */
313 void
314 initscan(void)
315 {
316 struct kwtab *kw;
317 size_t i;
318 uint64_t uq;
319
320 for (kw = kwtab; kw->kw_name != NULL; kw++) {
321 if ((kw->kw_c89 || kw->kw_c99) && tflag)
322 continue;
323 if (kw->kw_c99 && !(Sflag || gflag))
324 continue;
325 if (kw->kw_gcc && !gflag)
326 continue;
327 add_keyword(kw, 1);
328 add_keyword(kw, 2);
329 add_keyword(kw, 4);
330 }
331
332 /* initialize bit-masks for quads */
333 for (i = 0; i < 64; i++) {
334 qbmasks[i] = (uint64_t)1 << i;
335 uq = ~(uint64_t)0 << i;
336 qumasks[i] = uq;
337 qlmasks[i] = ~uq;
338 }
339 qumasks[i] = 0;
340 qlmasks[i] = ~(uint64_t)0;
341 }
342
343 /*
344 * Get a free sbuf structure, if possible from the free list
345 */
346 static sbuf_t *
347 allocsb(void)
348 {
349 sbuf_t *sb;
350
351 if ((sb = sbfrlst) != NULL) {
352 sbfrlst = sb->sb_next;
353 #ifdef BLKDEBUG
354 (void)memset(sb, 0, sizeof(*sb));
355 #else
356 sb->sb_next = NULL;
357 #endif
358 } else {
359 sb = xmalloc(sizeof(*sb));
360 (void)memset(sb, 0, sizeof(*sb));
361 }
362 return sb;
363 }
364
365 /*
366 * Put a sbuf structure to the free list
367 */
368 static void
369 freesb(sbuf_t *sb)
370 {
371
372 (void)memset(sb, ZERO, sizeof(*sb));
373 sb->sb_next = sbfrlst;
374 sbfrlst = sb;
375 }
376
377 /*
378 * Read a character and ensure that it is positive (except EOF).
379 * Increment line count(s) if necessary.
380 */
381 static int
382 inpc(void)
383 {
384 int c;
385
386 if ((c = lex_input()) != EOF && (c &= CHAR_MASK) == '\n')
387 lex_next_line();
388 return c;
389 }
390
391 static int
392 hash(const char *s)
393 {
394 u_int v;
395 const u_char *us;
396
397 v = 0;
398 for (us = (const u_char *)s; *us != '\0'; us++) {
399 v = (v << sizeof(v)) + *us;
400 v ^= v >> (sizeof(v) * CHAR_BIT - sizeof(v));
401 }
402 return v % HSHSIZ1;
403 }
404
405 /*
406 * Lex has found a letter followed by zero or more letters or digits.
407 * It looks for a symbol in the symbol table with the same name. This
408 * symbol must either be a keyword or a symbol of the type required by
409 * symtyp (label, member, tag, ...).
410 *
411 * If it is a keyword, the token is returned. In some cases it is described
412 * more deeply by data written to yylval.
413 *
414 * If it is a symbol, T_NAME is returned and the pointer to a sbuf struct
415 * is stored in yylval. This struct contains the name of the symbol, its
416 * length and hash value. If there is already a symbol of the same name
417 * and type in the symbol table, the sbuf struct also contains a pointer
418 * to the symbol table entry.
419 */
420 extern int
421 lex_name(const char *yytext, size_t yyleng)
422 {
423 char *s;
424 sbuf_t *sb;
425 sym_t *sym;
426 int tok;
427
428 sb = allocsb();
429 sb->sb_name = yytext;
430 sb->sb_len = yyleng;
431 sb->sb_hash = hash(yytext);
432 if ((sym = search(sb)) != NULL && sym->s_keyword != NULL) {
433 freesb(sb);
434 return keyw(sym);
435 }
436
437 sb->sb_sym = sym;
438
439 if (sym != NULL) {
440 lint_assert(block_level >= sym->s_block_level);
441 sb->sb_name = sym->s_name;
442 sb->sb_len = strlen(sym->s_name);
443 tok = sym->s_scl == TYPEDEF ? T_TYPENAME : T_NAME;
444 } else {
445 s = getblk(yyleng + 1);
446 (void)memcpy(s, yytext, yyleng + 1);
447 sb->sb_name = s;
448 sb->sb_len = yyleng;
449 tok = T_NAME;
450 }
451
452 yylval.y_sb = sb;
453 return tok;
454 }
455
456 static sym_t *
457 search(sbuf_t *sb)
458 {
459 sym_t *sym;
460
461 for (sym = symtab[sb->sb_hash]; sym != NULL; sym = sym->s_link) {
462 if (strcmp(sym->s_name, sb->sb_name) == 0) {
463 if (sym->s_keyword != NULL) {
464 struct kwtab *kw = sym->s_keyword;
465 if (!kw->kw_attr || attron)
466 return sym;
467 } else if (!attron && sym->s_kind == symtyp)
468 return sym;
469 }
470 }
471
472 return NULL;
473 }
474
475 static int
476 keyw(sym_t *sym)
477 {
478 int t;
479
480 if ((t = (int)sym->s_value.v_quad) == T_SCLASS) {
481 yylval.y_scl = sym->s_scl;
482 } else if (t == T_TYPE || t == T_STRUCT_OR_UNION) {
483 yylval.y_tspec = sym->s_tspec;
484 } else if (t == T_QUAL) {
485 yylval.y_tqual = sym->s_tqual;
486 }
487 return t;
488 }
489
490 /*
491 * Convert a string representing an integer into internal representation.
492 * Return T_CON, storing the numeric value in yylval, for yylex.
493 */
494 int
495 lex_integer_constant(const char *yytext, size_t yyleng, int base)
496 {
497 int l_suffix, u_suffix;
498 int len;
499 const char *cp;
500 char c, *eptr;
501 tspec_t typ;
502 bool ansiu;
503 bool warned = false;
504 #ifdef TARG_INT128_MAX
505 __uint128_t uq = 0;
506 static tspec_t contypes[2][4] = {
507 { INT, LONG, QUAD, INT128, },
508 { UINT, ULONG, UQUAD, UINT128, }
509 };
510 #else
511 uint64_t uq = 0;
512 static tspec_t contypes[2][3] = {
513 { INT, LONG, QUAD, },
514 { UINT, ULONG, UQUAD, }
515 };
516 #endif
517
518 cp = yytext;
519 len = yyleng;
520
521 /* skip 0[xX] or 0[bB] */
522 if (base == 16 || base == 2) {
523 cp += 2;
524 len -= 2;
525 }
526
527 /* read suffixes */
528 l_suffix = u_suffix = 0;
529 for (;;) {
530 if ((c = cp[len - 1]) == 'l' || c == 'L') {
531 l_suffix++;
532 } else if (c == 'u' || c == 'U') {
533 u_suffix++;
534 } else {
535 break;
536 }
537 len--;
538 }
539 if (l_suffix > 2 || u_suffix > 1) {
540 /* malformed integer constant */
541 warning(251);
542 if (l_suffix > 2)
543 l_suffix = 2;
544 if (u_suffix > 1)
545 u_suffix = 1;
546 }
547 if (tflag && u_suffix != 0) {
548 /* suffix U is illegal in traditional C */
549 warning(97);
550 }
551 typ = contypes[u_suffix][l_suffix];
552
553 errno = 0;
554
555 uq = strtouq(cp, &eptr, base);
556 lint_assert(eptr == cp + len);
557 if (errno != 0) {
558 /* integer constant out of range */
559 warning(252);
560 warned = true;
561 }
562
563 /*
564 * If the value is too big for the current type, we must choose
565 * another type.
566 */
567 ansiu = false;
568 switch (typ) {
569 case INT:
570 if (uq <= TARG_INT_MAX) {
571 /* ok */
572 } else if (uq <= TARG_UINT_MAX && base != 10) {
573 typ = UINT;
574 } else if (uq <= TARG_LONG_MAX) {
575 typ = LONG;
576 } else {
577 typ = ULONG;
578 if (uq > TARG_ULONG_MAX && !warned) {
579 /* integer constant out of range */
580 warning(252);
581 }
582 }
583 if (typ == UINT || typ == ULONG) {
584 if (tflag) {
585 typ = LONG;
586 } else if (!sflag) {
587 /*
588 * Remember that the constant is unsigned
589 * only in ANSI C
590 */
591 ansiu = true;
592 }
593 }
594 break;
595 case UINT:
596 if (uq > TARG_UINT_MAX) {
597 typ = ULONG;
598 if (uq > TARG_ULONG_MAX && !warned) {
599 /* integer constant out of range */
600 warning(252);
601 }
602 }
603 break;
604 case LONG:
605 if (uq > TARG_LONG_MAX && !tflag) {
606 typ = ULONG;
607 if (!sflag)
608 ansiu = true;
609 if (uq > TARG_ULONG_MAX && !warned) {
610 /* integer constant out of range */
611 warning(252);
612 }
613 }
614 break;
615 case ULONG:
616 if (uq > TARG_ULONG_MAX && !warned) {
617 /* integer constant out of range */
618 warning(252);
619 }
620 break;
621 case QUAD:
622 if (uq > TARG_QUAD_MAX && !tflag) {
623 typ = UQUAD;
624 if (!sflag)
625 ansiu = true;
626 }
627 break;
628 case UQUAD:
629 if (uq > TARG_UQUAD_MAX && !warned) {
630 /* integer constant out of range */
631 warning(252);
632 }
633 break;
634 #ifdef INT128_SIZE
635 case INT128:
636 #ifdef TARG_INT128_MAX
637 if (uq > TARG_INT128_MAX && !tflag) {
638 typ = UINT128;
639 if (!sflag)
640 ansiu = true;
641 }
642 #endif
643 break;
644 case UINT128:
645 #ifdef TARG_INT128_MAX
646 if (uq > TARG_UINT128_MAX && !warned) {
647 /* integer constant out of range */
648 warning(252);
649 }
650 #endif
651 break;
652 #endif
653 /* LINTED206: (enumeration values not handled in switch) */
654 case STRUCT:
655 case VOID:
656 case LDOUBLE:
657 case FUNC:
658 case ARRAY:
659 case PTR:
660 case ENUM:
661 case UNION:
662 case SIGNED:
663 case NOTSPEC:
664 case DOUBLE:
665 case FLOAT:
666 case USHORT:
667 case SHORT:
668 case UCHAR:
669 case SCHAR:
670 case CHAR:
671 case BOOL:
672 case UNSIGN:
673 case FCOMPLEX:
674 case DCOMPLEX:
675 case LCOMPLEX:
676 case COMPLEX:
677 break;
678 }
679
680 uq = (uint64_t)xsign((int64_t)uq, typ, -1);
681
682 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
683 yylval.y_val->v_tspec = typ;
684 yylval.y_val->v_ansiu = ansiu;
685 yylval.y_val->v_quad = (int64_t)uq;
686
687 return T_CON;
688 }
689
690 /*
691 * Returns whether t is a signed type and the value is negative.
692 *
693 * len is the number of significant bits. If len is -1, len is set
694 * to the width of type t.
695 */
696 static bool
697 sign(int64_t q, tspec_t t, int len)
698 {
699
700 if (t == PTR || is_uinteger(t))
701 return false;
702 return msb(q, t, len) != 0;
703 }
704
705 int
706 msb(int64_t q, tspec_t t, int len)
707 {
708
709 if (len <= 0)
710 len = size_in_bits(t);
711 return (q & qbmasks[len - 1]) != 0 ? 1 : 0;
712 }
713
714 /*
715 * Extends the sign of q.
716 */
717 int64_t
718 xsign(int64_t q, tspec_t t, int len)
719 {
720
721 if (len <= 0)
722 len = size_in_bits(t);
723
724 if (t == PTR || is_uinteger(t) || !sign(q, t, len)) {
725 q &= qlmasks[len];
726 } else {
727 q |= qumasks[len];
728 }
729 return q;
730 }
731
732 /*
733 * Convert a string representing a floating point value into its integral
734 * representation. Type and value are returned in yylval. fcon()
735 * (and yylex()) returns T_CON.
736 * XXX Currently it is not possible to convert constants of type
737 * long double which are greater than DBL_MAX.
738 */
739 int
740 lex_floating_constant(const char *yytext, size_t yyleng)
741 {
742 const char *cp;
743 int len;
744 tspec_t typ;
745 char c, *eptr;
746 double d;
747 float f = 0;
748
749 cp = yytext;
750 len = yyleng;
751
752 if (cp[len - 1] == 'i') {
753 /* imaginary, do nothing for now */
754 len--;
755 }
756 if ((c = cp[len - 1]) == 'f' || c == 'F') {
757 typ = FLOAT;
758 len--;
759 } else if (c == 'l' || c == 'L') {
760 typ = LDOUBLE;
761 len--;
762 } else {
763 if (c == 'd' || c == 'D')
764 len--;
765 typ = DOUBLE;
766 }
767
768 if (tflag && typ != DOUBLE) {
769 /* suffixes F and L are illegal in traditional C */
770 warning(98);
771 }
772
773 errno = 0;
774 d = strtod(cp, &eptr);
775 if (eptr != cp + len) {
776 switch (*eptr) {
777 /*
778 * XXX: non-native non-current strtod() may not handle hex
779 * floats, ignore the rest if we find traces of hex float
780 * syntax...
781 */
782 case 'p':
783 case 'P':
784 case 'x':
785 case 'X':
786 d = 0;
787 errno = 0;
788 break;
789 default:
790 INTERNAL_ERROR("fcon(%s->%s)", cp, eptr);
791 }
792 }
793 if (errno != 0)
794 /* floating-point constant out of range */
795 warning(248);
796
797 if (typ == FLOAT) {
798 f = (float)d;
799 if (finite(f) == 0) {
800 /* floating-point constant out of range */
801 warning(248);
802 f = f > 0 ? FLT_MAX : -FLT_MAX;
803 }
804 }
805
806 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
807 yylval.y_val->v_tspec = typ;
808 if (typ == FLOAT) {
809 yylval.y_val->v_ldbl = f;
810 } else {
811 yylval.y_val->v_ldbl = d;
812 }
813
814 return T_CON;
815 }
816
817 int
818 lex_operator(int t, op_t o)
819 {
820
821 yylval.y_op = o;
822 return t;
823 }
824
825 /*
826 * Called if lex found a leading \'.
827 */
828 int
829 lex_character_constant(void)
830 {
831 size_t n;
832 int val, c;
833 char cv;
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 {
845 /* XXX: should rather be sizeof(TARG_INT) */
846 if (n > sizeof(int) || (n > 1 && (pflag || hflag))) {
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 }
857 if (n == 1) {
858 cv = (char)val;
859 val = cv;
860 }
861
862 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
863 yylval.y_val->v_tspec = INT;
864 yylval.y_val->v_quad = val;
865
866 return T_CON;
867 }
868
869 /*
870 * Called if lex found a leading L\'
871 */
872 int
873 lex_wide_character_constant(void)
874 {
875 static char buf[MB_LEN_MAX + 1];
876 size_t i, imax;
877 int c;
878 wchar_t wc;
879
880 imax = MB_CUR_MAX;
881
882 i = 0;
883 while ((c = get_escaped_char('\'')) >= 0) {
884 if (i < imax)
885 buf[i] = (char)c;
886 i++;
887 }
888
889 wc = 0;
890
891 if (c == -2) {
892 /* unterminated character constant */
893 error(253);
894 } else if (c == 0) {
895 /* empty character constant */
896 error(73);
897 } else {
898 if (i > imax) {
899 i = imax;
900 /* too many characters in character constant */
901 error(71);
902 } else {
903 buf[i] = '\0';
904 (void)mbtowc(NULL, NULL, 0);
905 if (mbtowc(&wc, buf, imax) < 0)
906 /* invalid multibyte character */
907 error(291);
908 }
909 }
910
911 yylval.y_val = xcalloc(1, sizeof(*yylval.y_val));
912 yylval.y_val->v_tspec = WCHAR;
913 yylval.y_val->v_quad = wc;
914
915 return T_CON;
916 }
917
918 /*
919 * Read a character which is part of a character constant or of a string
920 * and handle escapes.
921 *
922 * The argument is the character which delimits the character constant or
923 * string.
924 *
925 * Returns -1 if the end of the character constant or string is reached,
926 * -2 if the EOF is reached, and the character otherwise.
927 */
928 static int
929 get_escaped_char(int delim)
930 {
931 static int pbc = -1;
932 int n, c, v;
933
934 if (pbc == -1) {
935 c = inpc();
936 } else {
937 c = pbc;
938 pbc = -1;
939 }
940 if (c == delim)
941 return -1;
942 switch (c) {
943 case '\n':
944 if (tflag) {
945 /* newline in string or char constant */
946 error(254);
947 return -2;
948 }
949 return c;
950 case EOF:
951 return -2;
952 case '\\':
953 switch (c = inpc()) {
954 case '"':
955 if (tflag && delim == '\'')
956 /* \" inside character constants undef... */
957 warning(262);
958 return '"';
959 case '\'':
960 return '\'';
961 case '?':
962 if (tflag)
963 /* \? undefined in traditional C */
964 warning(263);
965 return '?';
966 case '\\':
967 return '\\';
968 case 'a':
969 if (tflag)
970 /* \a undefined in traditional C */
971 warning(81);
972 return '\a';
973 case 'b':
974 return '\b';
975 case 'f':
976 return '\f';
977 case 'n':
978 return '\n';
979 case 'r':
980 return '\r';
981 case 't':
982 return '\t';
983 case 'v':
984 if (tflag)
985 /* \v undefined in traditional C */
986 warning(264);
987 return '\v';
988 case '8': case '9':
989 /* bad octal digit %c */
990 warning(77, c);
991 /* FALLTHROUGH */
992 case '0': case '1': case '2': case '3':
993 case '4': case '5': case '6': case '7':
994 n = 3;
995 v = 0;
996 do {
997 v = (v << 3) + (c - '0');
998 c = inpc();
999 } while (--n > 0 && isdigit(c) && (tflag || c <= '7'));
1000 if (tflag && n > 0 && isdigit(c))
1001 /* bad octal digit %c */
1002 warning(77, c);
1003 pbc = c;
1004 if (v > TARG_UCHAR_MAX) {
1005 /* character escape does not fit in character */
1006 warning(76);
1007 v &= CHAR_MASK;
1008 }
1009 return v;
1010 case 'x':
1011 if (tflag)
1012 /* \x undefined in traditional C */
1013 warning(82);
1014 v = 0;
1015 n = 0;
1016 while ((c = inpc()) >= 0 && isxdigit(c)) {
1017 c = isdigit(c) ?
1018 c - '0' : toupper(c) - 'A' + 10;
1019 v = (v << 4) + c;
1020 if (n >= 0) {
1021 if ((v & ~CHAR_MASK) != 0) {
1022 /* overflow in hex escape */
1023 warning(75);
1024 n = -1;
1025 } else {
1026 n++;
1027 }
1028 }
1029 }
1030 pbc = c;
1031 if (n == 0) {
1032 /* no hex digits follow \x */
1033 error(74);
1034 } if (n == -1) {
1035 v &= CHAR_MASK;
1036 }
1037 return v;
1038 case '\n':
1039 return get_escaped_char(delim);
1040 case EOF:
1041 return -2;
1042 default:
1043 if (isprint(c)) {
1044 /* dubious escape \%c */
1045 warning(79, c);
1046 } else {
1047 /* dubious escape \%o */
1048 warning(80, c);
1049 }
1050 }
1051 }
1052 return c;
1053 }
1054
1055 /* See https://gcc.gnu.org/onlinedocs/cpp/Preprocessor-Output.html */
1056 static void
1057 parse_line_directive_flags(const char *p,
1058 bool *is_begin, bool *is_end, bool *is_system)
1059 {
1060
1061 *is_begin = false;
1062 *is_end = false;
1063 *is_system = false;
1064
1065 while (*p != '\0') {
1066 while (ch_isspace(*p))
1067 p++;
1068
1069 const char *word_start = p;
1070 while (*p != '\0' && !ch_isspace(*p))
1071 p++;
1072 const char *word_end = p;
1073
1074 if (word_end - word_start == 1 && word_start[0] == '1')
1075 *is_begin = true;
1076 if (word_end - word_start == 1 && word_start[0] == '2')
1077 *is_end = true;
1078 if (word_end - word_start == 1 && word_start[0] == '3')
1079 *is_system = true;
1080 /* Flag '4' would only be interesting if lint handled C++. */
1081 }
1082
1083 #if 0
1084 if (*p != '\0') {
1085 /* syntax error '%s' */
1086 warning(249, "extra character(s) after directive");
1087 }
1088 #endif
1089 }
1090
1091 /*
1092 * Called for preprocessor directives. Currently implemented are:
1093 * # lineno
1094 * # lineno "filename"
1095 * # lineno "filename" GCC-flag...
1096 */
1097 void
1098 lex_directive(const char *yytext)
1099 {
1100 const char *cp, *fn;
1101 char c, *eptr;
1102 size_t fnl;
1103 long ln;
1104 bool is_begin, is_end, is_system;
1105
1106 static bool first = true;
1107
1108 /* Go to first non-whitespace after # */
1109 for (cp = yytext + 1; (c = *cp) == ' ' || c == '\t'; cp++)
1110 continue;
1111
1112 if (!ch_isdigit(c)) {
1113 if (strncmp(cp, "pragma", 6) == 0 && ch_isspace(cp[6]))
1114 return;
1115 error:
1116 /* undefined or invalid # directive */
1117 warning(255);
1118 return;
1119 }
1120 ln = strtol(--cp, &eptr, 10);
1121 if (cp == eptr)
1122 goto error;
1123 if ((c = *(cp = eptr)) != ' ' && c != '\t' && c != '\0')
1124 goto error;
1125 while ((c = *cp++) == ' ' || c == '\t')
1126 continue;
1127 if (c != '\0') {
1128 if (c != '"')
1129 goto error;
1130 fn = cp;
1131 while ((c = *cp) != '"' && c != '\0')
1132 cp++;
1133 if (c != '"')
1134 goto error;
1135 if ((fnl = cp++ - fn) > PATH_MAX)
1136 goto error;
1137 /* empty string means stdin */
1138 if (fnl == 0) {
1139 fn = "{standard input}";
1140 fnl = 16; /* strlen (fn) */
1141 }
1142 curr_pos.p_file = record_filename(fn, fnl);
1143 /*
1144 * If this is the first directive, the name is the name
1145 * of the C source file as specified at the command line.
1146 * It is written to the output file.
1147 */
1148 if (first) {
1149 csrc_pos.p_file = curr_pos.p_file;
1150 outsrc(transform_filename(curr_pos.p_file,
1151 strlen(curr_pos.p_file)));
1152 first = false;
1153 }
1154
1155 parse_line_directive_flags(cp, &is_begin, &is_end, &is_system);
1156 update_location(curr_pos.p_file, (int)ln, is_begin, is_end);
1157 in_system_header = is_system;
1158 }
1159 curr_pos.p_line = (int)ln - 1;
1160 curr_pos.p_uniq = 0;
1161 if (curr_pos.p_file == csrc_pos.p_file) {
1162 csrc_pos.p_line = (int)ln - 1;
1163 csrc_pos.p_uniq = 0;
1164 }
1165 }
1166
1167 /*
1168 * Handle lint comments such as ARGSUSED.
1169 *
1170 * If one of these comments is recognized, the argument, if any, is
1171 * parsed and a function which handles this comment is called.
1172 */
1173 void
1174 lex_comment(void)
1175 {
1176 int c, lc;
1177 static const struct {
1178 const char *keywd;
1179 bool arg;
1180 void (*func)(int);
1181 } keywtab[] = {
1182 { "ARGSUSED", true, argsused },
1183 { "BITFIELDTYPE", false, bitfieldtype },
1184 { "CONSTCOND", false, constcond },
1185 { "CONSTANTCOND", false, constcond },
1186 { "CONSTANTCONDITION", false, constcond },
1187 { "FALLTHRU", false, fallthru },
1188 { "FALLTHROUGH", false, fallthru },
1189 { "LINTLIBRARY", false, lintlib },
1190 { "LINTED", true, linted },
1191 { "LONGLONG", false, longlong },
1192 { "NOSTRICT", true, linted },
1193 { "NOTREACHED", false, not_reached },
1194 { "PRINTFLIKE", true, printflike },
1195 { "PROTOLIB", true, protolib },
1196 { "SCANFLIKE", true, scanflike },
1197 { "VARARGS", true, varargs },
1198 };
1199 char keywd[32];
1200 char arg[32];
1201 size_t l, i;
1202 int a;
1203 bool eoc;
1204
1205 eoc = false;
1206
1207 /* Skip whitespace after the start of the comment */
1208 while ((c = inpc()) != EOF && isspace(c))
1209 continue;
1210
1211 /* Read the potential keyword to keywd */
1212 l = 0;
1213 while (c != EOF && isupper(c) && l < sizeof(keywd) - 1) {
1214 keywd[l++] = (char)c;
1215 c = inpc();
1216 }
1217 keywd[l] = '\0';
1218
1219 /* look for the keyword */
1220 for (i = 0; i < sizeof(keywtab) / sizeof(keywtab[0]); i++) {
1221 if (strcmp(keywtab[i].keywd, keywd) == 0)
1222 break;
1223 }
1224 if (i == sizeof(keywtab) / sizeof(keywtab[0]))
1225 goto skip_rest;
1226
1227 /* skip whitespace after the keyword */
1228 while (c != EOF && isspace(c))
1229 c = inpc();
1230
1231 /* read the argument, if the keyword accepts one and there is one */
1232 l = 0;
1233 if (keywtab[i].arg) {
1234 while (c != EOF && isdigit(c) && l < sizeof(arg) - 1) {
1235 arg[l++] = (char)c;
1236 c = inpc();
1237 }
1238 }
1239 arg[l] = '\0';
1240 a = l != 0 ? atoi(arg) : -1;
1241
1242 /* skip whitespace after the argument */
1243 while (c != EOF && isspace(c))
1244 c = inpc();
1245
1246 if (c != '*' || (c = inpc()) != '/') {
1247 if (keywtab[i].func != linted)
1248 /* extra characters in lint comment */
1249 warning(257);
1250 } else {
1251 /*
1252 * remember that we have already found the end of the
1253 * comment
1254 */
1255 eoc = true;
1256 }
1257
1258 if (keywtab[i].func != NULL)
1259 (*keywtab[i].func)(a);
1260
1261 skip_rest:
1262 while (!eoc) {
1263 lc = c;
1264 if ((c = inpc()) == EOF) {
1265 /* unterminated comment */
1266 error(256);
1267 break;
1268 }
1269 if (lc == '*' && c == '/')
1270 eoc = true;
1271 }
1272 }
1273
1274 /*
1275 * Handle // style comments
1276 */
1277 void
1278 lex_slash_slash_comment(void)
1279 {
1280 int c;
1281
1282 if (!Sflag && !gflag)
1283 /* %s C does not support // comments */
1284 gnuism(312, tflag ? "traditional" : "ANSI");
1285
1286 while ((c = inpc()) != EOF && c != '\n')
1287 continue;
1288 }
1289
1290 /*
1291 * Clear flags for lint comments LINTED, LONGLONG and CONSTCOND.
1292 * clear_warn_flags() is called after function definitions and global and
1293 * local declarations and definitions. It is also called between
1294 * the controlling expression and the body of control statements
1295 * (if, switch, for, while).
1296 */
1297 void
1298 clear_warn_flags(void)
1299 {
1300
1301 lwarn = LWARN_ALL;
1302 quadflg = false;
1303 constcond_flag = false;
1304 }
1305
1306 /*
1307 * Strings are stored in a dynamically allocated buffer and passed
1308 * in yylval.y_xstrg to the parser. The parser or the routines called
1309 * by the parser are responsible for freeing this buffer.
1310 */
1311 int
1312 lex_string(void)
1313 {
1314 u_char *s;
1315 int c;
1316 size_t len, max;
1317 strg_t *strg;
1318
1319 s = xmalloc(max = 64);
1320
1321 len = 0;
1322 while ((c = get_escaped_char('"')) >= 0) {
1323 /* +1 to reserve space for a trailing NUL character */
1324 if (len + 1 == max)
1325 s = xrealloc(s, max *= 2);
1326 s[len++] = (char)c;
1327 }
1328 s[len] = '\0';
1329 if (c == -2)
1330 /* unterminated string constant */
1331 error(258);
1332
1333 strg = xcalloc(1, sizeof(*strg));
1334 strg->st_tspec = CHAR;
1335 strg->st_len = len;
1336 strg->st_cp = s;
1337
1338 yylval.y_string = strg;
1339 return T_STRING;
1340 }
1341
1342 int
1343 lex_wide_string(void)
1344 {
1345 char *s;
1346 int c, n;
1347 size_t i, wi;
1348 size_t len, max, wlen;
1349 wchar_t *ws;
1350 strg_t *strg;
1351
1352 s = xmalloc(max = 64);
1353 len = 0;
1354 while ((c = get_escaped_char('"')) >= 0) {
1355 /* +1 to save space for a trailing NUL character */
1356 if (len + 1 >= max)
1357 s = xrealloc(s, max *= 2);
1358 s[len++] = (char)c;
1359 }
1360 s[len] = '\0';
1361 if (c == -2)
1362 /* unterminated string constant */
1363 error(258);
1364
1365 /* get length of wide-character string */
1366 (void)mblen(NULL, 0);
1367 for (i = 0, wlen = 0; i < len; i += n, wlen++) {
1368 if ((n = mblen(&s[i], MB_CUR_MAX)) == -1) {
1369 /* invalid multibyte character */
1370 error(291);
1371 break;
1372 }
1373 if (n == 0)
1374 n = 1;
1375 }
1376
1377 ws = xmalloc((wlen + 1) * sizeof(*ws));
1378
1379 /* convert from multibyte to wide char */
1380 (void)mbtowc(NULL, NULL, 0);
1381 for (i = 0, wi = 0; i < len; i += n, wi++) {
1382 if ((n = mbtowc(&ws[wi], &s[i], MB_CUR_MAX)) == -1)
1383 break;
1384 if (n == 0)
1385 n = 1;
1386 }
1387 ws[wi] = 0;
1388 free(s);
1389
1390 strg = xcalloc(1, sizeof(*strg));
1391 strg->st_tspec = WCHAR;
1392 strg->st_len = wlen;
1393 strg->st_wcp = ws;
1394
1395 yylval.y_string = strg;
1396 return T_STRING;
1397 }
1398
1399 /*
1400 * As noted above the scanner does not create new symbol table entries
1401 * for symbols it cannot find in the symbol table. This is to avoid
1402 * putting undeclared symbols into the symbol table if a syntax error
1403 * occurs.
1404 *
1405 * getsym() is called as soon as it is probably ok to put the symbol to
1406 * the symbol table. This does not mean that it is not possible that
1407 * symbols are put to the symbol table which are not completely
1408 * declared due to syntax errors. To avoid too many problems in this
1409 * case, symbols get type int in getsym().
1410 *
1411 * XXX calls to getsym() should be delayed until decl1*() is called.
1412 */
1413 sym_t *
1414 getsym(sbuf_t *sb)
1415 {
1416 dinfo_t *di;
1417 char *s;
1418 sym_t *sym;
1419
1420 sym = sb->sb_sym;
1421
1422 /*
1423 * During member declaration it is possible that name() looked
1424 * for symbols of type FVFT, although it should have looked for
1425 * symbols of type FTAG. Same can happen for labels. Both cases
1426 * are compensated here.
1427 */
1428 if (symtyp == FMEMBER || symtyp == FLABEL) {
1429 if (sym == NULL || sym->s_kind == FVFT)
1430 sym = search(sb);
1431 }
1432
1433 if (sym != NULL) {
1434 if (sym->s_kind != symtyp)
1435 INTERNAL_ERROR("getsym(%d, %d)", sym->s_kind, symtyp);
1436 symtyp = FVFT;
1437 freesb(sb);
1438 return sym;
1439 }
1440
1441 /* create a new symbol table entry */
1442
1443 /* labels must always be allocated at level 1 (outermost block) */
1444 if (symtyp == FLABEL) {
1445 sym = getlblk(1, sizeof(*sym));
1446 s = getlblk(1, sb->sb_len + 1);
1447 (void)memcpy(s, sb->sb_name, sb->sb_len + 1);
1448 sym->s_name = s;
1449 sym->s_block_level = 1;
1450 di = dcs;
1451 while (di->d_next != NULL && di->d_next->d_next != NULL)
1452 di = di->d_next;
1453 lint_assert(di->d_ctx == AUTO);
1454 } else {
1455 sym = getblk(sizeof(*sym));
1456 sym->s_name = sb->sb_name;
1457 sym->s_block_level = block_level;
1458 di = dcs;
1459 }
1460
1461 UNIQUE_CURR_POS(sym->s_def_pos);
1462 if ((sym->s_kind = symtyp) != FLABEL)
1463 sym->s_type = gettyp(INT);
1464
1465 symtyp = FVFT;
1466
1467 if ((sym->s_link = symtab[sb->sb_hash]) != NULL)
1468 symtab[sb->sb_hash]->s_rlink = &sym->s_link;
1469 sym->s_rlink = &symtab[sb->sb_hash];
1470 symtab[sb->sb_hash] = sym;
1471
1472 *di->d_ldlsym = sym;
1473 di->d_ldlsym = &sym->s_dlnxt;
1474
1475 freesb(sb);
1476 return sym;
1477 }
1478
1479 /*
1480 * Construct a temporary symbol. The symbol starts with a digit, so that
1481 * it is illegal.
1482 */
1483 sym_t *
1484 mktempsym(type_t *t)
1485 {
1486 static int n = 0;
1487 int h;
1488 char *s = getlblk(block_level, 64);
1489 sym_t *sym = getblk(sizeof(*sym));
1490 scl_t scl;
1491
1492 (void)snprintf(s, 64, "%.8d_tmp", n++);
1493 h = hash(s);
1494
1495 scl = dcs->d_scl;
1496 if (scl == NOSCL)
1497 scl = block_level > 0 ? AUTO : EXTERN;
1498
1499 sym->s_name = s;
1500 sym->s_type = t;
1501 sym->s_block_level = block_level;
1502 sym->s_scl = scl;
1503 sym->s_kind = FVFT;
1504 sym->s_used = true;
1505 sym->s_set = true;
1506
1507 if ((sym->s_link = symtab[h]) != NULL)
1508 symtab[h]->s_rlink = &sym->s_link;
1509 sym->s_rlink = &symtab[h];
1510 symtab[h] = sym;
1511
1512 *dcs->d_ldlsym = sym;
1513 dcs->d_ldlsym = &sym->s_dlnxt;
1514
1515 return sym;
1516 }
1517
1518 /*
1519 * Remove a symbol forever from the symbol table. s_block_level
1520 * is set to -1 to avoid that the symbol will later be put
1521 * back to the symbol table.
1522 */
1523 void
1524 rmsym(sym_t *sym)
1525 {
1526
1527 if ((*sym->s_rlink = sym->s_link) != NULL)
1528 sym->s_link->s_rlink = sym->s_rlink;
1529 sym->s_block_level = -1;
1530 sym->s_link = NULL;
1531 }
1532
1533 /*
1534 * Remove a list of symbols declared at one level from the symbol
1535 * table.
1536 */
1537 void
1538 rmsyms(sym_t *syms)
1539 {
1540 sym_t *sym;
1541
1542 for (sym = syms; sym != NULL; sym = sym->s_dlnxt) {
1543 if (sym->s_block_level != -1) {
1544 if ((*sym->s_rlink = sym->s_link) != NULL)
1545 sym->s_link->s_rlink = sym->s_rlink;
1546 sym->s_link = NULL;
1547 sym->s_rlink = NULL;
1548 }
1549 }
1550 }
1551
1552 /*
1553 * Put a symbol into the symbol table.
1554 */
1555 void
1556 inssym(int bl, sym_t *sym)
1557 {
1558 int h;
1559
1560 h = hash(sym->s_name);
1561 if ((sym->s_link = symtab[h]) != NULL)
1562 symtab[h]->s_rlink = &sym->s_link;
1563 sym->s_rlink = &symtab[h];
1564 symtab[h] = sym;
1565 sym->s_block_level = bl;
1566 lint_assert(sym->s_link == NULL ||
1567 sym->s_block_level >= sym->s_link->s_block_level);
1568 }
1569
1570 /*
1571 * Called at level 0 after syntax errors.
1572 *
1573 * Removes all symbols which are not declared at level 0 from the
1574 * symbol table. Also frees all memory which is not associated with
1575 * level 0.
1576 */
1577 void
1578 cleanup(void)
1579 {
1580 sym_t *sym, *nsym;
1581 int i;
1582
1583 for (i = 0; i < HSHSIZ1; i++) {
1584 for (sym = symtab[i]; sym != NULL; sym = nsym) {
1585 nsym = sym->s_link;
1586 if (sym->s_block_level >= 1) {
1587 if ((*sym->s_rlink = nsym) != NULL)
1588 nsym->s_rlink = sym->s_rlink;
1589 }
1590 }
1591 }
1592
1593 for (i = mem_block_level; i > 0; i--)
1594 freelblk(i);
1595 }
1596
1597 /*
1598 * Create a new symbol with the name of an existing symbol.
1599 */
1600 sym_t *
1601 pushdown(const sym_t *sym)
1602 {
1603 int h;
1604 sym_t *nsym;
1605
1606 h = hash(sym->s_name);
1607 nsym = getblk(sizeof(*nsym));
1608 lint_assert(sym->s_block_level <= block_level);
1609 nsym->s_name = sym->s_name;
1610 UNIQUE_CURR_POS(nsym->s_def_pos);
1611 nsym->s_kind = sym->s_kind;
1612 nsym->s_block_level = block_level;
1613
1614 if ((nsym->s_link = symtab[h]) != NULL)
1615 symtab[h]->s_rlink = &nsym->s_link;
1616 nsym->s_rlink = &symtab[h];
1617 symtab[h] = nsym;
1618
1619 *dcs->d_ldlsym = nsym;
1620 dcs->d_ldlsym = &nsym->s_dlnxt;
1621
1622 return nsym;
1623 }
1624
1625 /*
1626 * Free any dynamically allocated memory referenced by
1627 * the value stack or yylval.
1628 * The type of information in yylval is described by tok.
1629 */
1630 void
1631 freeyyv(void *sp, int tok)
1632 {
1633 if (tok == T_NAME || tok == T_TYPENAME) {
1634 sbuf_t *sb = *(sbuf_t **)sp;
1635 freesb(sb);
1636 } else if (tok == T_CON) {
1637 val_t *val = *(val_t **)sp;
1638 free(val);
1639 } else if (tok == T_STRING) {
1640 strg_t *strg = *(strg_t **)sp;
1641 if (strg->st_tspec == CHAR) {
1642 free(strg->st_cp);
1643 } else {
1644 lint_assert(strg->st_tspec == WCHAR);
1645 free(strg->st_wcp);
1646 }
1647 free(strg);
1648 }
1649 }
1650