scan.l revision 1.35 1 %{
2 /* $NetBSD: scan.l,v 1.35 2006/10/16 13:33:57 he Exp $ */
3
4 /*
5 * Copyright (c) 1996 Christopher G. Demetriou. All Rights Reserved.
6 * Copyright (c) 1994, 1995 Jochen Pohl
7 * All Rights Reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by Jochen Pohl for
20 * The NetBSD Project.
21 * 4. The name of the author may not be used to endorse or promote products
22 * derived from this software without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
25 * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
26 * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
27 * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
28 * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
29 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
30 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
31 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
32 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
33 * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
34 */
35
36 #include <sys/cdefs.h>
37 #if defined(__RCSID) && !defined(lint)
38 __RCSID("$NetBSD: scan.l,v 1.35 2006/10/16 13:33:57 he Exp $");
39 #endif
40
41 #include <stdlib.h>
42 #include <string.h>
43 #include <limits.h>
44 #include <float.h>
45 #include <ctype.h>
46 #include <errno.h>
47 #include <math.h>
48
49 #include "lint1.h"
50 #include "cgram.h"
51
52 #define CHAR_MASK (~(~0 << CHAR_BIT))
53 #define YY_NO_UNPUT
54
55 /* Current position (its also updated when an included file is parsed) */
56 pos_t curr_pos = { 1, "", 0 };
57
58 /*
59 * Current position in C source (not updated when an included file is
60 * parsed).
61 */
62 pos_t csrc_pos = { 1, "", 0 };
63
64 static void incline(void);
65 static void badchar(int);
66 static sbuf_t *allocsb(void);
67 static void freesb(sbuf_t *);
68 static int inpc(void);
69 static int hash(const char *);
70 static sym_t *search(sbuf_t *);
71 static int name(void);
72 static int keyw(sym_t *);
73 static int icon(int);
74 static int fcon(void);
75 static int operator(int, op_t);
76 static int ccon(void);
77 static int wccon(void);
78 static int getescc(int);
79 static void directive(void);
80 static void comment(void);
81 static void slashslashcomment(void);
82 static int string(void);
83 static int wcstrg(void);
84
85 %}
86
87 L [_A-Za-z]
88 D [0-9]
89 NZD [1-9]
90 OD [0-7]
91 HD [0-9A-Fa-f]
92 EX ([eE][+-]?[0-9]+)
93
94 %%
95
96 {L}({L}|{D})* return (name());
97 0{OD}*[lLuU]* return (icon(8));
98 {NZD}{D}*[lLuU]* return (icon(10));
99 0[xX]{HD}+[lLuU]* return (icon(16));
100 {D}+\.{D}*{EX}?[fFlL]? |
101 {D}+{EX}[fFlL]? |
102 \.{D}+{EX}?[fFlL]? return (fcon());
103 "=" return (operator(T_ASSIGN, ASSIGN));
104 "*=" return (operator(T_OPASS, MULASS));
105 "/=" return (operator(T_OPASS, DIVASS));
106 "%=" return (operator(T_OPASS, MODASS));
107 "+=" return (operator(T_OPASS, ADDASS));
108 "-=" return (operator(T_OPASS, SUBASS));
109 "<<=" return (operator(T_OPASS, SHLASS));
110 ">>=" return (operator(T_OPASS, SHRASS));
111 "&=" return (operator(T_OPASS, ANDASS));
112 "^=" return (operator(T_OPASS, XORASS));
113 "|=" return (operator(T_OPASS, ORASS));
114 "||" return (operator(T_LOGOR, LOGOR));
115 "&&" return (operator(T_LOGAND, LOGAND));
116 "|" return (operator(T_OR, OR));
117 "&" return (operator(T_AND, AND));
118 "^" return (operator(T_XOR, XOR));
119 "==" return (operator(T_EQOP, EQ));
120 "!=" return (operator(T_EQOP, NE));
121 "<" return (operator(T_RELOP, LT));
122 ">" return (operator(T_RELOP, GT));
123 "<=" return (operator(T_RELOP, LE));
124 ">=" return (operator(T_RELOP, GE));
125 "<<" return (operator(T_SHFTOP, SHL));
126 ">>" return (operator(T_SHFTOP, SHR));
127 "++" return (operator(T_INCDEC, INC));
128 "--" return (operator(T_INCDEC, DEC));
129 "->" return (operator(T_STROP, ARROW));
130 "." return (operator(T_STROP, POINT));
131 "+" return (operator(T_ADDOP, PLUS));
132 "-" return (operator(T_ADDOP, MINUS));
133 "*" return (operator(T_MULT, MULT));
134 "/" return (operator(T_DIVOP, DIV));
135 "%" return (operator(T_DIVOP, MOD));
136 "!" return (operator(T_UNOP, NOT));
137 "~" return (operator(T_UNOP, COMPL));
138 "\"" return (string());
139 "L\"" return (wcstrg());
140 ";" return (T_SEMI);
141 "{" return (T_LBRACE);
142 "}" return (T_RBRACE);
143 "," return (T_COMMA);
144 ":" return (T_COLON);
145 "?" return (T_QUEST);
146 "[" return (T_LBRACK);
147 "]" return (T_RBRACK);
148 "(" return (T_LPARN);
149 ")" return (T_RPARN);
150 "..." return (T_ELLIPSE);
151 "'" return (ccon());
152 "L'" return (wccon());
153 ^#.*$ directive();
154 \n incline();
155 \t|" "|\f|\v ;
156 "/*" comment();
157 "//" slashslashcomment();
158 . badchar(yytext[0]);
159
160 %%
161
162 static void
163 incline(void)
164 {
165 curr_pos.p_line++;
166 curr_pos.p_uniq = 0;
167 if (curr_pos.p_file == csrc_pos.p_file) {
168 csrc_pos.p_line++;
169 csrc_pos.p_uniq = 0;
170 }
171 }
172
173 static void
174 badchar(int c)
175 {
176
177 /* unknown character \%o */
178 error(250, c);
179 }
180
181 /*
182 * Keywords.
183 * During initialisation they are written to the symbol table.
184 */
185 static struct kwtab {
186 const char *kw_name; /* keyword */
187 int kw_token; /* token returned by yylex() */
188 scl_t kw_scl; /* storage class if kw_token T_SCLASS */
189 tspec_t kw_tspec; /* type spec. if kw_token T_TYPE or T_SOU */
190 tqual_t kw_tqual; /* type qual. fi kw_token T_QUAL */
191 u_int kw_c89; /* c89 keyword */
192 u_int kw_c99; /* c99 keyword */
193 u_int kw_gcc; /* GCC keyword */
194 } kwtab[] = {
195 { "asm", T_ASM, 0, 0, 0, 0, 0, 1 },
196 { "__asm", T_ASM, 0, 0, 0, 0, 0, 0 },
197 { "__asm__", T_ASM, 0, 0, 0, 0, 0, 0 },
198 { "auto", T_SCLASS, AUTO, 0, 0, 0, 0, 0 },
199 { "break", T_BREAK, 0, 0, 0, 0, 0, 0 },
200 { "_Bool", T_TYPE, 0, BOOL, 0, 0, 1, 0 },
201 { "case", T_CASE, 0, 0, 0, 0, 0, 0 },
202 { "char", T_TYPE, 0, CHAR, 0, 0, 0, 0 },
203 { "const", T_QUAL, 0, 0, CONST, 1, 0, 0 },
204 { "__const__", T_QUAL, 0, 0, CONST, 0, 0, 0 },
205 { "__const", T_QUAL, 0, 0, CONST, 0, 0, 0 },
206 { "continue", T_CONTINUE, 0, 0, 0, 0, 0, 0 },
207 { "default", T_DEFAULT, 0, 0, 0, 0, 0, 0 },
208 { "do", T_DO, 0, 0, 0, 0, 0, 0 },
209 { "double", T_TYPE, 0, DOUBLE, 0, 0, 0, 0 },
210 { "else", T_ELSE, 0, 0, 0, 0, 0, 0 },
211 { "enum", T_ENUM, 0, 0, 0, 0, 0, 0 },
212 { "extern", T_SCLASS, EXTERN, 0, 0, 0, 0, 0 },
213 { "float", T_TYPE, 0, FLOAT, 0, 0, 0, 0 },
214 { "for", T_FOR, 0, 0, 0, 0, 0, 0 },
215 { "goto", T_GOTO, 0, 0, 0, 0, 0, 0 },
216 { "if", T_IF, 0, 0, 0, 0, 0, 0 },
217 { "inline", T_SCLASS, INLINE, 0, 0, 0, 1, 0 },
218 { "__inline__", T_SCLASS, INLINE, 0, 0, 0, 0, 0 },
219 { "__inline", T_SCLASS, INLINE, 0, 0, 0, 0, 0 },
220 { "int", T_TYPE, 0, INT, 0, 0, 0, 0 },
221 { "__symbolrename", T_SYMBOLRENAME, 0, 0, 0, 0, 0, 0 },
222 { "long", T_TYPE, 0, LONG, 0, 0, 0, 0 },
223 { "register", T_SCLASS, REG, 0, 0, 0, 0, 0 },
224 { "return", T_RETURN, 0, 0, 0, 0, 0, 0 },
225 { "short", T_TYPE, 0, SHORT, 0, 0, 0, 0 },
226 { "signed", T_TYPE, 0, SIGNED, 0, 1, 0, 0 },
227 { "__signed__", T_TYPE, 0, SIGNED, 0, 0, 0, 0 },
228 { "__signed", T_TYPE, 0, SIGNED, 0, 0, 0, 0 },
229 { "sizeof", T_SIZEOF, 0, 0, 0, 0, 0, 0 },
230 { "static", T_SCLASS, STATIC, 0, 0, 0, 0, 0 },
231 { "struct", T_SOU, 0, STRUCT, 0, 0, 0, 0 },
232 { "switch", T_SWITCH, 0, 0, 0, 0, 0, 0 },
233 { "typedef", T_SCLASS, TYPEDEF, 0, 0, 0, 0, 0 },
234 { "union", T_SOU, 0, UNION, 0, 0, 0, 0 },
235 { "unsigned", T_TYPE, 0, UNSIGN, 0, 0, 0, 0 },
236 { "void", T_TYPE, 0, VOID, 0, 0, 0, 0 },
237 { "volatile", T_QUAL, 0, 0, VOLATILE, 1, 0, 0 },
238 { "__volatile__", T_QUAL, 0, 0, VOLATILE, 0, 0, 0 },
239 { "__volatile", T_QUAL, 0, 0, VOLATILE, 0, 0, 0 },
240 { "while", T_WHILE, 0, 0, 0, 0, 0, 0 },
241 { NULL, 0, 0, 0, 0, 0, 0, 0 }
242 };
243
244 /* Symbol table */
245 static sym_t *symtab[HSHSIZ1];
246
247 /* bit i of the entry with index i is set */
248 uint64_t qbmasks[sizeof(uint64_t) * CHAR_BIT];
249
250 /* least significant i bits are set in the entry with index i */
251 uint64_t qlmasks[sizeof(uint64_t) * CHAR_BIT + 1];
252
253 /* least significant i bits are not set in the entry with index i */
254 uint64_t qumasks[sizeof(uint64_t) * CHAR_BIT + 1];
255
256 /* free list for sbuf structures */
257 static sbuf_t *sbfrlst;
258
259 /* Typ of next expected symbol */
260 symt_t symtyp;
261
262
263 /*
264 * All keywords are written to the symbol table. This saves us looking
265 * in a extra table for each name we found.
266 */
267 void
268 initscan(void)
269 {
270 struct kwtab *kw;
271 sym_t *sym;
272 int h, i;
273 uint64_t uq;
274
275 for (kw = kwtab; kw->kw_name != NULL; kw++) {
276 if ((kw->kw_c89 || kw->kw_c99) && tflag)
277 continue;
278 if (kw->kw_c99 && !(Sflag || gflag))
279 continue;
280 if (kw->kw_gcc && !gflag)
281 continue;
282 sym = getblk(sizeof (sym_t));
283 sym->s_name = kw->kw_name;
284 sym->s_keyw = 1;
285 sym->s_value.v_quad = kw->kw_token;
286 if (kw->kw_token == T_TYPE || kw->kw_token == T_SOU) {
287 sym->s_tspec = kw->kw_tspec;
288 } else if (kw->kw_token == T_SCLASS) {
289 sym->s_scl = kw->kw_scl;
290 } else if (kw->kw_token == T_QUAL) {
291 sym->s_tqual = kw->kw_tqual;
292 }
293 h = hash(sym->s_name);
294 if ((sym->s_link = symtab[h]) != NULL)
295 symtab[h]->s_rlink = &sym->s_link;
296 (symtab[h] = sym)->s_rlink = &symtab[h];
297 }
298
299 /* initialize bit-masks for quads */
300 for (i = 0; i < sizeof (uint64_t) * CHAR_BIT; i++) {
301 qbmasks[i] = (uint64_t)1 << i;
302 uq = ~(uint64_t)0 << i;
303 qumasks[i] = uq;
304 qlmasks[i] = ~uq;
305 }
306 qumasks[i] = 0;
307 qlmasks[i] = ~(uint64_t)0;
308 }
309
310 /*
311 * Get a free sbuf structure, if possible from the free list
312 */
313 static sbuf_t *
314 allocsb(void)
315 {
316 sbuf_t *sb;
317
318 if ((sb = sbfrlst) != NULL) {
319 sbfrlst = sb->sb_nxt;
320 } else {
321 sb = xmalloc(sizeof (sbuf_t));
322 }
323 (void)memset(sb, 0, sizeof (sb));
324 return (sb);
325 }
326
327 /*
328 * Put a sbuf structure to the free list
329 */
330 static void
331 freesb(sbuf_t *sb)
332 {
333
334 sb->sb_nxt = sbfrlst;
335 sbfrlst = sb;
336 }
337
338 /*
339 * Read a character and ensure that it is positive (except EOF).
340 * Increment line count(s) if necessary.
341 */
342 static int
343 inpc(void)
344 {
345 int c;
346
347 if ((c = input()) != EOF && (c &= CHAR_MASK) == '\n')
348 incline();
349 return (c);
350 }
351
352 static int
353 hash(const char *s)
354 {
355 u_int v;
356 const u_char *us;
357
358 v = 0;
359 for (us = (const u_char *)s; *us != '\0'; us++) {
360 v = (v << sizeof (v)) + *us;
361 v ^= v >> (sizeof (v) * CHAR_BIT - sizeof (v));
362 }
363 return (v % HSHSIZ1);
364 }
365
366 /*
367 * Lex has found a letter followed by zero or more letters or digits.
368 * It looks for a symbol in the symbol table with the same name. This
369 * symbol must either be a keyword or a symbol of the type required by
370 * symtyp (label, member, tag, ...).
371 *
372 * If it is a keyword, the token is returned. In some cases it is described
373 * more deeply by data written to yylval.
374 *
375 * If it is a symbol, T_NAME is returned and the pointer to a sbuf struct
376 * is stored in yylval. This struct contains the name of the symbol, it's
377 * length and hash value. If there is already a symbol of the same name
378 * and type in the symbol table, the sbuf struct also contains a pointer
379 * to the symbol table entry.
380 */
381 static int
382 name(void)
383 {
384 char *s;
385 sbuf_t *sb;
386 sym_t *sym;
387 int tok;
388
389 sb = allocsb();
390 sb->sb_name = yytext;
391 sb->sb_len = yyleng;
392 sb->sb_hash = hash(yytext);
393
394 if ((sym = search(sb)) != NULL && sym->s_keyw) {
395 freesb(sb);
396 return (keyw(sym));
397 }
398
399 sb->sb_sym = sym;
400
401 if (sym != NULL) {
402 if (blklev < sym->s_blklev)
403 LERROR("name()");
404 sb->sb_name = sym->s_name;
405 sb->sb_len = strlen(sym->s_name);
406 tok = sym->s_scl == TYPEDEF ? T_TYPENAME : T_NAME;
407 } else {
408 s = getblk(yyleng + 1);
409 (void)memcpy(s, yytext, yyleng + 1);
410 sb->sb_name = s;
411 sb->sb_len = yyleng;
412 tok = T_NAME;
413 }
414
415 yylval.y_sb = sb;
416 return (tok);
417 }
418
419 static sym_t *
420 search(sbuf_t *sb)
421 {
422 sym_t *sym;
423
424 for (sym = symtab[sb->sb_hash]; sym != NULL; sym = sym->s_link) {
425 if (strcmp(sym->s_name, sb->sb_name) == 0) {
426 if (sym->s_keyw || sym->s_kind == symtyp)
427 return (sym);
428 }
429 }
430
431 return (NULL);
432 }
433
434 static int
435 keyw(sym_t *sym)
436 {
437 int t;
438
439 if ((t = (int)sym->s_value.v_quad) == T_SCLASS) {
440 yylval.y_scl = sym->s_scl;
441 } else if (t == T_TYPE || t == T_SOU) {
442 yylval.y_tspec = sym->s_tspec;
443 } else if (t == T_QUAL) {
444 yylval.y_tqual = sym->s_tqual;
445 }
446 return (t);
447 }
448
449 /*
450 * Convert a string representing an integer into internal representation.
451 * The value is returned in yylval. icon() (and yylex()) returns T_CON.
452 */
453 static int
454 icon(int base)
455 {
456 int l_suffix, u_suffix;
457 int len;
458 const char *cp;
459 char c, *eptr;
460 tspec_t typ;
461 uint64_t uq = 0;
462 int ansiu;
463 static tspec_t contypes[2][3] = {
464 { INT, LONG, QUAD },
465 { UINT, ULONG, UQUAD }
466 };
467
468 cp = yytext;
469 len = yyleng;
470
471 /* skip 0x */
472 if (base == 16) {
473 cp += 2;
474 len -= 2;
475 }
476
477 /* read suffixes */
478 l_suffix = u_suffix = 0;
479 for ( ; ; ) {
480 if ((c = cp[len - 1]) == 'l' || c == 'L') {
481 l_suffix++;
482 } else if (c == 'u' || c == 'U') {
483 u_suffix++;
484 } else {
485 break;
486 }
487 len--;
488 }
489 if (l_suffix > 2 || u_suffix > 1) {
490 /* malformed integer constant */
491 warning(251);
492 if (l_suffix > 2)
493 l_suffix = 2;
494 if (u_suffix > 1)
495 u_suffix = 1;
496 }
497 if (tflag && u_suffix != 0) {
498 /* suffix U is illegal in traditional C */
499 warning(97);
500 }
501 typ = contypes[u_suffix][l_suffix];
502
503 errno = 0;
504
505 uq = strtouq(cp, &eptr, base);
506 if (eptr != cp + len)
507 LERROR("icon()");
508 if (errno != 0)
509 /* integer constant out of range */
510 warning(252);
511
512 /*
513 * If the value is too big for the current type, we must choose
514 * another type.
515 */
516 ansiu = 0;
517 switch (typ) {
518 case INT:
519 if (uq <= TARG_INT_MAX) {
520 /* ok */
521 } else if (uq <= TARG_UINT_MAX && base != 10) {
522 typ = UINT;
523 } else if (uq <= TARG_LONG_MAX) {
524 typ = LONG;
525 } else {
526 typ = ULONG;
527 if (uq > TARG_ULONG_MAX) {
528 /* integer constant out of range */
529 warning(252);
530 }
531 }
532 if (typ == UINT || typ == ULONG) {
533 if (tflag) {
534 typ = LONG;
535 } else if (!sflag) {
536 /*
537 * Remember that the constant is unsigned
538 * only in ANSI C
539 */
540 ansiu = 1;
541 }
542 }
543 break;
544 case UINT:
545 if (uq > TARG_UINT_MAX) {
546 typ = ULONG;
547 if (uq > TARG_ULONG_MAX) {
548 /* integer constant out of range */
549 warning(252);
550 }
551 }
552 break;
553 case LONG:
554 if (uq > TARG_LONG_MAX && !tflag) {
555 typ = ULONG;
556 if (!sflag)
557 ansiu = 1;
558 if (uq > TARG_ULONG_MAX) {
559 /* integer constant out of range */
560 warning(252);
561 }
562 }
563 break;
564 case QUAD:
565 if (uq > TARG_QUAD_MAX && !tflag) {
566 typ = UQUAD;
567 if (!sflag)
568 ansiu = 1;
569 }
570 break;
571 /* LINTED (enumeration values not handled in switch) */
572 case STRUCT:
573 case VOID:
574 case LDOUBLE:
575 case FUNC:
576 case ARRAY:
577 case PTR:
578 case ENUM:
579 case UNION:
580 case SIGNED:
581 case NOTSPEC:
582 case DOUBLE:
583 case FLOAT:
584 case UQUAD:
585 case ULONG:
586 case USHORT:
587 case SHORT:
588 case UCHAR:
589 case SCHAR:
590 case CHAR:
591 case BOOL:
592 case UNSIGN:
593 break;
594
595 case NTSPEC: /* this value unused */
596 break;
597 }
598
599 uq = (uint64_t)xsign((int64_t)uq, typ, -1);
600
601 (yylval.y_val = xcalloc(1, sizeof (val_t)))->v_tspec = typ;
602 yylval.y_val->v_ansiu = ansiu;
603 yylval.y_val->v_quad = (int64_t)uq;
604
605 return (T_CON);
606 }
607
608 /*
609 * Returns 1 if t is a signed type and the value is negative.
610 *
611 * len is the number of significant bits. If len is -1, len is set
612 * to the width of type t.
613 */
614 int
615 sign(int64_t q, tspec_t t, int len)
616 {
617
618 if (t == PTR || isutyp(t))
619 return (0);
620 return (msb(q, t, len));
621 }
622
623 int
624 msb(int64_t q, tspec_t t, int len)
625 {
626
627 if (len <= 0)
628 len = size(t);
629 return ((q & qbmasks[len - 1]) != 0);
630 }
631
632 /*
633 * Extends the sign of q.
634 */
635 int64_t
636 xsign(int64_t q, tspec_t t, int len)
637 {
638
639 if (len <= 0)
640 len = size(t);
641
642 if (t == PTR || isutyp(t) || !sign(q, t, len)) {
643 q &= qlmasks[len];
644 } else {
645 q |= qumasks[len];
646 }
647 return (q);
648 }
649
650 /*
651 * Convert a string representing a floating point value into its interal
652 * representation. Type and value are returned in yylval. fcon()
653 * (and yylex()) returns T_CON.
654 * XXX Currently it is not possible to convert constants of type
655 * long double which are greater than DBL_MAX.
656 */
657 static int
658 fcon(void)
659 {
660 const char *cp;
661 int len;
662 tspec_t typ;
663 char c, *eptr;
664 double d;
665 float f = 0;
666
667 cp = yytext;
668 len = yyleng;
669
670 if ((c = cp[len - 1]) == 'f' || c == 'F') {
671 typ = FLOAT;
672 len--;
673 } else if (c == 'l' || c == 'L') {
674 typ = LDOUBLE;
675 len--;
676 } else {
677 typ = DOUBLE;
678 }
679
680 if (tflag && typ != DOUBLE) {
681 /* suffixes F and L are illegal in traditional C */
682 warning(98);
683 }
684
685 errno = 0;
686 d = strtod(cp, &eptr);
687 if (eptr != cp + len)
688 LERROR("fcon()");
689 if (errno != 0)
690 /* floating-point constant out of range */
691 warning(248);
692
693 if (typ == FLOAT) {
694 f = (float)d;
695 if (!finite(f)) {
696 /* floating-point constant out of range */
697 warning(248);
698 f = f > 0 ? FLT_MAX : -FLT_MAX;
699 }
700 }
701
702 (yylval.y_val = xcalloc(1, sizeof (val_t)))->v_tspec = typ;
703 if (typ == FLOAT) {
704 yylval.y_val->v_ldbl = f;
705 } else {
706 yylval.y_val->v_ldbl = d;
707 }
708
709 return (T_CON);
710 }
711
712 static int
713 operator(int t, op_t o)
714 {
715
716 yylval.y_op = o;
717 return (t);
718 }
719
720 /*
721 * Called if lex found a leading \'.
722 */
723 static int
724 ccon(void)
725 {
726 int n, val, c;
727 char cv;
728
729 n = 0;
730 val = 0;
731 while ((c = getescc('\'')) >= 0) {
732 val = (val << CHAR_BIT) + c;
733 n++;
734 }
735 if (c == -2) {
736 /* unterminated character constant */
737 error(253);
738 } else {
739 if (n > sizeof (int) || (n > 1 && (pflag || hflag))) {
740 /* too many characters in character constant */
741 error(71);
742 } else if (n > 1) {
743 /* multi-character character constant */
744 warning(294);
745 } else if (n == 0) {
746 /* empty character constant */
747 error(73);
748 }
749 }
750 if (n == 1) {
751 cv = (char)val;
752 val = cv;
753 }
754
755 yylval.y_val = xcalloc(1, sizeof (val_t));
756 yylval.y_val->v_tspec = INT;
757 yylval.y_val->v_quad = val;
758
759 return (T_CON);
760 }
761
762 /*
763 * Called if lex found a leading L\'
764 */
765 static int
766 wccon(void)
767 {
768 static char buf[MB_LEN_MAX + 1];
769 int i, c;
770 wchar_t wc;
771
772 i = 0;
773 while ((c = getescc('\'')) >= 0) {
774 if (i < MB_CUR_MAX)
775 buf[i] = (char)c;
776 i++;
777 }
778
779 wc = 0;
780
781 if (c == -2) {
782 /* unterminated character constant */
783 error(253);
784 } else if (c == 0) {
785 /* empty character constant */
786 error(73);
787 } else {
788 if (i > MB_CUR_MAX) {
789 i = MB_CUR_MAX;
790 /* too many characters in character constant */
791 error(71);
792 } else {
793 buf[i] = '\0';
794 (void)mbtowc(NULL, NULL, 0);
795 if (mbtowc(&wc, buf, MB_CUR_MAX) < 0)
796 /* invalid multibyte character */
797 error(291);
798 }
799 }
800
801 yylval.y_val = xcalloc(1, sizeof (val_t));
802 yylval.y_val->v_tspec = WCHAR;
803 yylval.y_val->v_quad = wc;
804
805 return (T_CON);
806 }
807
808 /*
809 * Read a character which is part of a character constant or of a string
810 * and handle escapes.
811 *
812 * The Argument is the character which delimits the character constant or
813 * string.
814 *
815 * Returns -1 if the end of the character constant or string is reached,
816 * -2 if the EOF is reached, and the character otherwise.
817 */
818 static int
819 getescc(int d)
820 {
821 static int pbc = -1;
822 int n, c, v;
823
824 if (pbc == -1) {
825 c = inpc();
826 } else {
827 c = pbc;
828 pbc = -1;
829 }
830 if (c == d)
831 return (-1);
832 switch (c) {
833 case '\n':
834 if (tflag) {
835 /* newline in string or char constant */
836 error(254);
837 return (-2);
838 }
839 return (c);
840 case EOF:
841 return (-2);
842 case '\\':
843 switch (c = inpc()) {
844 case '"':
845 if (tflag && d == '\'')
846 /* \" inside character constant undef. ... */
847 warning(262);
848 return ('"');
849 case '\'':
850 return ('\'');
851 case '?':
852 if (tflag)
853 /* \? undefined in traditional C */
854 warning(263);
855 return ('?');
856 case '\\':
857 return ('\\');
858 case 'a':
859 if (tflag)
860 /* \a undefined in traditional C */
861 warning(81);
862 return ('\a');
863 case 'b':
864 return ('\b');
865 case 'f':
866 return ('\f');
867 case 'n':
868 return ('\n');
869 case 'r':
870 return ('\r');
871 case 't':
872 return ('\t');
873 case 'v':
874 if (tflag)
875 /* \v undefined in traditional C */
876 warning(264);
877 return ('\v');
878 case '8': case '9':
879 /* bad octal digit %c */
880 warning(77, c);
881 /* FALLTHROUGH */
882 case '0': case '1': case '2': case '3':
883 case '4': case '5': case '6': case '7':
884 n = 3;
885 v = 0;
886 do {
887 v = (v << 3) + (c - '0');
888 c = inpc();
889 } while (--n && isdigit(c) && (tflag || c <= '7'));
890 if (tflag && n > 0 && isdigit(c))
891 /* bad octal digit %c */
892 warning(77, c);
893 pbc = c;
894 if (v > UCHAR_MAX) {
895 /* character escape does not fit in char. */
896 warning(76);
897 v &= CHAR_MASK;
898 }
899 return (v);
900 case 'x':
901 if (tflag)
902 /* \x undefined in traditional C */
903 warning(82);
904 v = 0;
905 n = 0;
906 while ((c = inpc()) >= 0 && isxdigit(c)) {
907 c = isdigit(c) ?
908 c - '0' : toupper(c) - 'A' + 10;
909 v = (v << 4) + c;
910 if (n >= 0) {
911 if ((v & ~CHAR_MASK) != 0) {
912 /* overflow in hex escape */
913 warning(75);
914 n = -1;
915 } else {
916 n++;
917 }
918 }
919 }
920 pbc = c;
921 if (n == 0) {
922 /* no hex digits follow \x */
923 error(74);
924 } if (n == -1) {
925 v &= CHAR_MASK;
926 }
927 return (v);
928 case '\n':
929 return (getescc(d));
930 case EOF:
931 return (-2);
932 default:
933 if (isprint(c)) {
934 /* dubious escape \%c */
935 warning(79, c);
936 } else {
937 /* dubious escape \%o */
938 warning(80, c);
939 }
940 }
941 }
942 return (c);
943 }
944
945 /*
946 * Called for preprocessor directives. Currently implemented are:
947 * # lineno
948 * # lineno "filename"
949 */
950 static void
951 directive(void)
952 {
953 const char *cp, *fn;
954 char c, *eptr;
955 size_t fnl;
956 long ln;
957 static int first = 1;
958
959 /* Go to first non-whitespace after # */
960 for (cp = yytext + 1; (c = *cp) == ' ' || c == '\t'; cp++)
961 continue;
962
963 if (!isdigit((unsigned char)c)) {
964 error:
965 /* undefined or invalid # directive */
966 warning(255);
967 return;
968 }
969 ln = strtol(--cp, &eptr, 10);
970 if (cp == eptr)
971 goto error;
972 if ((c = *(cp = eptr)) != ' ' && c != '\t' && c != '\0')
973 goto error;
974 while ((c = *cp++) == ' ' || c == '\t')
975 continue;
976 if (c != '\0') {
977 if (c != '"')
978 goto error;
979 fn = cp;
980 while ((c = *cp) != '"' && c != '\0')
981 cp++;
982 if (c != '"')
983 goto error;
984 if ((fnl = cp++ - fn) > PATH_MAX)
985 goto error;
986 while ((c = *cp++) == ' ' || c == '\t')
987 continue;
988 #if 0
989 if (c != '\0')
990 warning("extra character(s) after directive");
991 #endif
992
993 /* empty string means stdin */
994 if (fnl == 0) {
995 fn = "{standard input}";
996 fnl = 16; /* strlen (fn) */
997 }
998 curr_pos.p_file = fnnalloc(fn, fnl);
999 /*
1000 * If this is the first directive, the name is the name
1001 * of the C source file as specified at the command line.
1002 * It is written to the output file.
1003 */
1004 if (first) {
1005 csrc_pos.p_file = curr_pos.p_file;
1006 outsrc(curr_pos.p_file);
1007 first = 0;
1008 }
1009 }
1010 curr_pos.p_line = (int)ln - 1;
1011 curr_pos.p_uniq = 0;
1012 if (curr_pos.p_file == csrc_pos.p_file) {
1013 csrc_pos.p_line = (int)ln - 1;
1014 csrc_pos.p_uniq = 0;
1015 }
1016 }
1017
1018 /*
1019 * Handle lint comments. Following comments are currently understood:
1020 * ARGSUSEDn
1021 * BITFIELDTYPE
1022 * CONSTCOND CONSTANTCOND CONSTANTCONDITION
1023 * FALLTHRU FALLTHROUGH
1024 * LINTLIBRARY
1025 * LINTED NOSTRICT
1026 * LONGLONG
1027 * NOTREACHED
1028 * PRINTFLIKEn
1029 * PROTOLIB
1030 * SCANFLIKEn
1031 * VARARGSn
1032 * If one of this comments is recognized, the arguments, if any, are
1033 * parsed and a function which handles this comment is called.
1034 */
1035 static void
1036 comment(void)
1037 {
1038 int c, lc;
1039 static struct {
1040 const char *keywd;
1041 int arg;
1042 void (*func)(int);
1043 } keywtab[] = {
1044 { "ARGSUSED", 1, argsused },
1045 { "BITFIELDTYPE", 0, bitfieldtype },
1046 { "CONSTCOND", 0, constcond },
1047 { "CONSTANTCOND", 0, constcond },
1048 { "CONSTANTCONDITION", 0, constcond },
1049 { "FALLTHRU", 0, fallthru },
1050 { "FALLTHROUGH", 0, fallthru },
1051 { "LINTLIBRARY", 0, lintlib },
1052 { "LINTED", 0, linted },
1053 { "LONGLONG", 0, longlong },
1054 { "NOSTRICT", 0, linted },
1055 { "NOTREACHED", 0, notreach },
1056 { "PRINTFLIKE", 1, printflike },
1057 { "PROTOLIB", 1, protolib },
1058 { "SCANFLIKE", 1, scanflike },
1059 { "VARARGS", 1, varargs },
1060 };
1061 char keywd[32];
1062 char arg[32];
1063 int l, i, a;
1064 int eoc;
1065
1066 eoc = 0;
1067
1068 /* Skip white spaces after the start of the comment */
1069 while ((c = inpc()) != EOF && isspace(c))
1070 continue;
1071
1072 /* Read the potential keyword to keywd */
1073 l = 0;
1074 while (c != EOF && isupper(c) && l < sizeof (keywd) - 1) {
1075 keywd[l++] = (char)c;
1076 c = inpc();
1077 }
1078 keywd[l] = '\0';
1079
1080 /* look for the keyword */
1081 for (i = 0; i < sizeof (keywtab) / sizeof (keywtab[0]); i++) {
1082 if (strcmp(keywtab[i].keywd, keywd) == 0)
1083 break;
1084 }
1085 if (i == sizeof (keywtab) / sizeof (keywtab[0]))
1086 goto skip_rest;
1087
1088 /* skip white spaces after the keyword */
1089 while (c != EOF && isspace(c))
1090 c = inpc();
1091
1092 /* read the argument, if the keyword accepts one and there is one */
1093 l = 0;
1094 if (keywtab[i].arg) {
1095 while (c != EOF && isdigit(c) && l < sizeof (arg) - 1) {
1096 arg[l++] = (char)c;
1097 c = inpc();
1098 }
1099 }
1100 arg[l] = '\0';
1101 a = l != 0 ? atoi(arg) : -1;
1102
1103 /* skip white spaces after the argument */
1104 while (c != EOF && isspace(c))
1105 c = inpc();
1106
1107 if (c != '*' || (c = inpc()) != '/') {
1108 if (keywtab[i].func != linted)
1109 /* extra characters in lint comment */
1110 warning(257);
1111 } else {
1112 /*
1113 * remember that we have already found the end of the
1114 * comment
1115 */
1116 eoc = 1;
1117 }
1118
1119 if (keywtab[i].func != NULL)
1120 (*keywtab[i].func)(a);
1121
1122 skip_rest:
1123 while (!eoc) {
1124 lc = c;
1125 if ((c = inpc()) == EOF) {
1126 /* unterminated comment */
1127 error(256);
1128 break;
1129 }
1130 if (lc == '*' && c == '/')
1131 eoc = 1;
1132 }
1133 }
1134
1135 /*
1136 * Handle // style comments
1137 */
1138 static void
1139 slashslashcomment(void)
1140 {
1141 int c;
1142
1143 if (!Sflag && !gflag)
1144 /* // comments only supported in C99 */
1145 (void)gnuism(312, tflag ? "traditional" : "ANSI");
1146
1147 while ((c = inpc()) != EOF && c != '\n')
1148 continue;
1149 }
1150
1151 /*
1152 * Clear flags for lint comments LINTED, LONGLONG and CONSTCOND.
1153 * clrwflgs() is called after function definitions and global and
1154 * local declarations and definitions. It is also called between
1155 * the controlling expression and the body of control statements
1156 * (if, switch, for, while).
1157 */
1158 void
1159 clrwflgs(void)
1160 {
1161
1162 nowarn = 0;
1163 quadflg = 0;
1164 ccflg = 0;
1165 }
1166
1167 /*
1168 * Strings are stored in a dynamically alloceted buffer and passed
1169 * in yylval.y_xstrg to the parser. The parser or the routines called
1170 * by the parser are responsible for freeing this buffer.
1171 */
1172 static int
1173 string(void)
1174 {
1175 u_char *s;
1176 int c;
1177 size_t len, max;
1178 strg_t *strg;
1179
1180 s = xmalloc(max = 64);
1181
1182 len = 0;
1183 while ((c = getescc('"')) >= 0) {
1184 /* +1 to reserve space for a trailing NUL character */
1185 if (len + 1 == max)
1186 s = xrealloc(s, max *= 2);
1187 s[len++] = (char)c;
1188 }
1189 s[len] = '\0';
1190 if (c == -2)
1191 /* unterminated string constant */
1192 error(258);
1193
1194 strg = xcalloc(1, sizeof (strg_t));
1195 strg->st_tspec = CHAR;
1196 strg->st_len = len;
1197 strg->st_cp = s;
1198
1199 yylval.y_strg = strg;
1200 return (T_STRING);
1201 }
1202
1203 static int
1204 wcstrg(void)
1205 {
1206 char *s;
1207 int c, i, n, wi;
1208 size_t len, max, wlen;
1209 wchar_t *ws;
1210 strg_t *strg;
1211
1212 s = xmalloc(max = 64);
1213 len = 0;
1214 while ((c = getescc('"')) >= 0) {
1215 /* +1 to save space for a trailing NUL character */
1216 if (len + 1 >= max)
1217 s = xrealloc(s, max *= 2);
1218 s[len++] = (char)c;
1219 }
1220 s[len] = '\0';
1221 if (c == -2)
1222 /* unterminated string constant */
1223 error(258);
1224
1225 /* get length of wide character string */
1226 (void)mblen(NULL, 0);
1227 for (i = 0, wlen = 0; i < len; i += n, wlen++) {
1228 if ((n = mblen(&s[i], MB_CUR_MAX)) == -1) {
1229 /* invalid multibyte character */
1230 error(291);
1231 break;
1232 }
1233 if (n == 0)
1234 n = 1;
1235 }
1236
1237 ws = xmalloc((wlen + 1) * sizeof (wchar_t));
1238
1239 /* convert from multibyte to wide char */
1240 (void)mbtowc(NULL, NULL, 0);
1241 for (i = 0, wi = 0; i < len; i += n, wi++) {
1242 if ((n = mbtowc(&ws[wi], &s[i], MB_CUR_MAX)) == -1)
1243 break;
1244 if (n == 0)
1245 n = 1;
1246 }
1247 ws[wi] = 0;
1248 free(s);
1249
1250 strg = xcalloc(1, sizeof (strg_t));
1251 strg->st_tspec = WCHAR;
1252 strg->st_len = wlen;
1253 strg->st_wcp = ws;
1254
1255 yylval.y_strg = strg;
1256 return (T_STRING);
1257 }
1258
1259 /*
1260 * As noted above the scanner does not create new symbol table entries
1261 * for symbols it cannot find in the symbol table. This is to avoid
1262 * putting undeclared symbols into the symbol table if a syntax error
1263 * occurs.
1264 *
1265 * getsym() is called as soon as it is probably ok to put the symbol to
1266 * the symbol table. This does not mean that it is not possible that
1267 * symbols are put to the symbol table which are than not completely
1268 * declared due to syntax errors. To avoid too many problems in this
1269 * case symbols get type int in getsym().
1270 *
1271 * XXX calls to getsym() should be delayed until decl1*() is called
1272 */
1273 sym_t *
1274 getsym(sbuf_t *sb)
1275 {
1276 dinfo_t *di;
1277 char *s;
1278 sym_t *sym;
1279
1280 sym = sb->sb_sym;
1281
1282 /*
1283 * During member declaration it is possible that name() looked
1284 * for symbols of type FVFT, although it should have looked for
1285 * symbols of type FTAG. Same can happen for labels. Both cases
1286 * are compensated here.
1287 */
1288 if (symtyp == FMOS || symtyp == FLAB) {
1289 if (sym == NULL || sym->s_kind == FVFT)
1290 sym = search(sb);
1291 }
1292
1293 if (sym != NULL) {
1294 if (sym->s_kind != symtyp)
1295 LERROR("storesym()");
1296 symtyp = FVFT;
1297 freesb(sb);
1298 return (sym);
1299 }
1300
1301 /* create a new symbol table entry */
1302
1303 /* labels must always be allocated at level 1 (outhermost block) */
1304 if (symtyp == FLAB) {
1305 sym = getlblk(1, sizeof (sym_t));
1306 s = getlblk(1, sb->sb_len + 1);
1307 (void)memcpy(s, sb->sb_name, sb->sb_len + 1);
1308 sym->s_name = s;
1309 sym->s_blklev = 1;
1310 di = dcs;
1311 while (di->d_nxt != NULL && di->d_nxt->d_nxt != NULL)
1312 di = di->d_nxt;
1313 if (di->d_ctx != AUTO)
1314 LERROR("storesym()");
1315 } else {
1316 sym = getblk(sizeof (sym_t));
1317 sym->s_name = sb->sb_name;
1318 sym->s_blklev = blklev;
1319 di = dcs;
1320 }
1321
1322 UNIQUE_CURR_POS(sym->s_dpos);
1323 if ((sym->s_kind = symtyp) != FLAB)
1324 sym->s_type = gettyp(INT);
1325
1326 symtyp = FVFT;
1327
1328 if ((sym->s_link = symtab[sb->sb_hash]) != NULL)
1329 symtab[sb->sb_hash]->s_rlink = &sym->s_link;
1330 (symtab[sb->sb_hash] = sym)->s_rlink = &symtab[sb->sb_hash];
1331
1332 *di->d_ldlsym = sym;
1333 di->d_ldlsym = &sym->s_dlnxt;
1334
1335 freesb(sb);
1336 return (sym);
1337 }
1338
1339 /*
1340 * Construct a temporary symbol. The symbol starts with a digit, so that
1341 * it is illegal.
1342 */
1343 sym_t *
1344 mktempsym(type_t *t)
1345 {
1346 static int n = 0;
1347 int h;
1348 char *s = getlblk(blklev, 64);
1349 sym_t *sym = getblk(sizeof (sym_t));
1350
1351 (void)snprintf(s, 64, "%.8d_tmp", n++);
1352 h = hash(s);
1353
1354 sym->s_name = s;
1355 sym->s_type = t;
1356 sym->s_blklev = blklev;
1357 sym->s_scl = AUTO;
1358 sym->s_kind = FVFT;
1359 sym->s_used = 1;
1360 sym->s_set = 1;
1361
1362 if ((sym->s_link = symtab[h]) != NULL)
1363 symtab[h]->s_rlink = &sym->s_link;
1364 (symtab[h] = sym)->s_rlink = &symtab[h];
1365
1366 *dcs->d_ldlsym = sym;
1367 dcs->d_ldlsym = &sym->s_dlnxt;
1368
1369 return sym;
1370 }
1371
1372 /*
1373 * Remove a symbol forever from the symbol table. s_blklev
1374 * is set to -1 to avoid that the symbol will later be put
1375 * back to the symbol table.
1376 */
1377 void
1378 rmsym(sym_t *sym)
1379 {
1380
1381 if ((*sym->s_rlink = sym->s_link) != NULL)
1382 sym->s_link->s_rlink = sym->s_rlink;
1383 sym->s_blklev = -1;
1384 sym->s_link = NULL;
1385 }
1386
1387 /*
1388 * Remove a list of symbols declared at one level from the symbol
1389 * table.
1390 */
1391 void
1392 rmsyms(sym_t *syms)
1393 {
1394 sym_t *sym;
1395
1396 for (sym = syms; sym != NULL; sym = sym->s_dlnxt) {
1397 if (sym->s_blklev != -1) {
1398 if ((*sym->s_rlink = sym->s_link) != NULL)
1399 sym->s_link->s_rlink = sym->s_rlink;
1400 sym->s_link = NULL;
1401 sym->s_rlink = NULL;
1402 }
1403 }
1404 }
1405
1406 /*
1407 * Put a symbol into the symbol table
1408 */
1409 void
1410 inssym(int bl, sym_t *sym)
1411 {
1412 int h;
1413
1414 h = hash(sym->s_name);
1415 if ((sym->s_link = symtab[h]) != NULL)
1416 symtab[h]->s_rlink = &sym->s_link;
1417 (symtab[h] = sym)->s_rlink = &symtab[h];
1418 sym->s_blklev = bl;
1419 if (sym->s_link != NULL && sym->s_blklev < sym->s_link->s_blklev)
1420 LERROR("inssym()");
1421 }
1422
1423 /*
1424 * Called at level 0 after syntax errors
1425 * Removes all symbols which are not declared at level 0 from the
1426 * symbol table. Also frees all memory which is not associated with
1427 * level 0.
1428 */
1429 void
1430 cleanup(void)
1431 {
1432 sym_t *sym, *nsym;
1433 int i;
1434
1435 for (i = 0; i < HSHSIZ1; i++) {
1436 for (sym = symtab[i]; sym != NULL; sym = nsym) {
1437 nsym = sym->s_link;
1438 if (sym->s_blklev >= 1) {
1439 if ((*sym->s_rlink = nsym) != NULL)
1440 nsym->s_rlink = sym->s_rlink;
1441 }
1442 }
1443 }
1444
1445 for (i = mblklev; i > 0; i--)
1446 freelblk(i);
1447 }
1448
1449 /*
1450 * Create a new symbol with the name of an existing symbol.
1451 */
1452 sym_t *
1453 pushdown(sym_t *sym)
1454 {
1455 int h;
1456 sym_t *nsym;
1457
1458 h = hash(sym->s_name);
1459 nsym = getblk(sizeof (sym_t));
1460 if (sym->s_blklev > blklev)
1461 LERROR("pushdown()");
1462 nsym->s_name = sym->s_name;
1463 UNIQUE_CURR_POS(nsym->s_dpos);
1464 nsym->s_kind = sym->s_kind;
1465 nsym->s_blklev = blklev;
1466
1467 if ((nsym->s_link = symtab[h]) != NULL)
1468 symtab[h]->s_rlink = &nsym->s_link;
1469 (symtab[h] = nsym)->s_rlink = &symtab[h];
1470
1471 *dcs->d_ldlsym = nsym;
1472 dcs->d_ldlsym = &nsym->s_dlnxt;
1473
1474 return (nsym);
1475 }
1476
1477 /*
1478 * Free any dynamically allocated memory referenced by
1479 * the value stack or yylval.
1480 * The type of information in yylval is described by tok.
1481 */
1482 void
1483 freeyyv(void *sp, int tok)
1484 {
1485 if (tok == T_NAME || tok == T_TYPENAME) {
1486 sbuf_t *sb = *(sbuf_t **)sp;
1487 freesb(sb);
1488 } else if (tok == T_CON) {
1489 val_t *val = *(val_t **)sp;
1490 free(val);
1491 } else if (tok == T_STRING) {
1492 strg_t *strg = *(strg_t **)sp;
1493 if (strg->st_tspec == CHAR) {
1494 free(strg->st_cp);
1495 } else if (strg->st_tspec == WCHAR) {
1496 free(strg->st_wcp);
1497 } else {
1498 LERROR("fryylv()");
1499 }
1500 free(strg);
1501 }
1502 }
1503