scan.l revision 1.87 1 %{
2 /* $NetBSD: scan.l,v 1.87 2019/03/03 23:06:08 christos 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 1.87 2019/03/03 23:06:08 christos 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 ((int)(~(~0U << CHAR_BIT)))
53
54 /* Current position (its also updated when an included file is parsed) */
55 pos_t curr_pos = { 1, "", 0 };
56
57 /*
58 * Current position in C source (not updated when an included file is
59 * parsed).
60 */
61 pos_t csrc_pos = { 1, "", 0 };
62
63 /* Are we parsing a gcc attribute? */
64 int attron;
65
66 static void incline(void);
67 static void badchar(int);
68 static sbuf_t *allocsb(void);
69 static void freesb(sbuf_t *);
70 static int inpc(void);
71 static int hash(const char *);
72 static sym_t *search(sbuf_t *);
73 static int name(void);
74 static int keyw(sym_t *);
75 static int icon(int);
76 static int fcon(void);
77 static int operator(int, op_t);
78 static int ccon(void);
79 static int wccon(void);
80 static int getescc(int);
81 static void directive(void);
82 static void comment(void);
83 static void slashslashcomment(void);
84 static int string(void);
85 static int wcstrg(void);
86
87 %}
88
89
90 L [_A-Za-z]
91 D [0-9]
92 NZD [1-9]
93 BD [0-1]
94 OD [0-7]
95 HD [0-9A-Fa-f]
96 EX ([eE][+-]?[0-9]+)
97 HX (p[+-]?[0-9A-Fa-f]+)
98 TL ([fFlL]?[i]?)
99
100 %option nounput
101
102 %%
103
104 {L}({L}|{D})* return (name());
105 0[bB]{BD}+[lLuU]* return (icon(2));
106 0{OD}*[lLuU]* return (icon(8));
107 {NZD}{D}*[lLuU]* return (icon(10));
108 0[xX]{HD}+[lLuU]* return (icon(16));
109 {D}+\.{D}*{EX}?{TL} |
110 {D}+{EX}{TL} |
111 0[xX]{HD}+\.{HD}*{HX}{TL} |
112 0[xX]{HD}+{HX}{TL} |
113 \.{D}+{EX}?{TL} return (fcon());
114 "=" return (operator(T_ASSIGN, ASSIGN));
115 "*=" return (operator(T_OPASS, MULASS));
116 "/=" return (operator(T_OPASS, DIVASS));
117 "%=" return (operator(T_OPASS, MODASS));
118 "+=" return (operator(T_OPASS, ADDASS));
119 "-=" return (operator(T_OPASS, SUBASS));
120 "<<=" return (operator(T_OPASS, SHLASS));
121 ">>=" return (operator(T_OPASS, SHRASS));
122 "&=" return (operator(T_OPASS, ANDASS));
123 "^=" return (operator(T_OPASS, XORASS));
124 "|=" return (operator(T_OPASS, ORASS));
125 "||" return (operator(T_LOGOR, LOGOR));
126 "&&" return (operator(T_LOGAND, LOGAND));
127 "|" return (operator(T_OR, OR));
128 "&" return (operator(T_AND, AND));
129 "^" return (operator(T_XOR, XOR));
130 "==" return (operator(T_EQOP, EQ));
131 "!=" return (operator(T_EQOP, NE));
132 "<" return (operator(T_RELOP, LT));
133 ">" return (operator(T_RELOP, GT));
134 "<=" return (operator(T_RELOP, LE));
135 ">=" return (operator(T_RELOP, GE));
136 "<<" return (operator(T_SHFTOP, SHL));
137 ">>" return (operator(T_SHFTOP, SHR));
138 "++" return (operator(T_INCDEC, INC));
139 "--" return (operator(T_INCDEC, DEC));
140 "->" return (operator(T_STROP, ARROW));
141 "." return (operator(T_STROP, POINT));
142 "+" return (operator(T_ADDOP, PLUS));
143 "-" return (operator(T_ADDOP, MINUS));
144 "*" return (operator(T_MULT, MULT));
145 "/" return (operator(T_DIVOP, DIV));
146 "%" return (operator(T_DIVOP, MOD));
147 "!" return (operator(T_UNOP, NOT));
148 "~" return (operator(T_UNOP, COMPL));
149 "\"" return (string());
150 "L\"" return (wcstrg());
151 ";" return (T_SEMI);
152 "{" return (T_LBRACE);
153 "}" return (T_RBRACE);
154 "," return (T_COMMA);
155 ":" return (T_COLON);
156 "?" return (T_QUEST);
157 "[" return (T_LBRACK);
158 "]" return (T_RBRACK);
159 "(" return (T_LPARN);
160 ")" return (T_RPARN);
161 "..." return (T_ELLIPSE);
162 "'" return (ccon());
163 "L'" return (wccon());
164 ^#.*$ directive();
165 \n incline();
166 \t|" "|\f|\v ;
167 "/*" comment();
168 "//" slashslashcomment();
169 . badchar(yytext[0]);
170
171 %%
172
173 static void
174 incline(void)
175 {
176 curr_pos.p_line++;
177 curr_pos.p_uniq = 0;
178 if (curr_pos.p_file == csrc_pos.p_file) {
179 csrc_pos.p_line++;
180 csrc_pos.p_uniq = 0;
181 }
182 }
183
184 static void
185 badchar(int c)
186 {
187
188 /* unknown character \%o */
189 error(250, c);
190 }
191
192 /*
193 * Keywords.
194 * During initialisation they are written to the symbol table.
195 */
196 static struct kwtab {
197 const char *kw_name; /* keyword */
198 int kw_token; /* token returned by yylex() */
199 scl_t kw_scl; /* storage class if kw_token T_SCLASS */
200 tspec_t kw_tspec; /* type spec. if kw_token T_TYPE or T_SOU */
201 tqual_t kw_tqual; /* type qual. fi kw_token T_QUAL */
202 u_int kw_c89 : 1; /* c89 keyword */
203 u_int kw_c99 : 1; /* c99 keyword */
204 u_int kw_gcc : 1; /* GCC keyword */
205 u_int kw_attr : 1; /* GCC attribute, keyword */
206 u_int kw_deco : 3; /* name[1] __name[2] __name__[4] */
207 } kwtab[] = {
208 #ifdef INT128_SIZE
209 { "__int128_t", T_TYPE, 0, INT128, 0, 0,1,0,0,1 },
210 { "__uint128_t",T_TYPE, 0, UINT128,0, 0,1,0,0,1 },
211 #endif
212 { "_Bool", T_TYPE, 0, BOOL, 0, 0,1,0,0,1 },
213 { "_Complex", T_TYPE, 0, COMPLEX,0, 0,1,0,0,1 },
214 { "_Generic", T_GENERIC, 0, 0, 0, 0,1,0,0,1 },
215 { "_Noreturn", T_NORETURN, 0, 0, 0, 0,1,0,0,1 },
216 { "alias", T_AT_ALIAS, 0, 0, 0, 0,0,1,1,5 },
217 { "aligned", T_AT_ALIGNED, 0, 0, 0, 0,0,1,1,5 },
218 { "alignof", T_ALIGNOF, 0, 0, 0, 0,0,0,0,4 },
219 { "alloc_size", T_AT_ALLOC_SIZE,0, 0, 0, 0,0,1,1,5 },
220 { "always_inline", T_AT_ALWAYS_INLINE, 0,0, 0, 0,0,1,1,5 },
221 { "asm", T_ASM, 0, 0, 0, 0,0,1,0,7 },
222 { "attribute", T_ATTRIBUTE, 0, 0, 0, 0,0,1,0,6 },
223 { "auto", T_SCLASS, AUTO, 0, 0, 0,0,0,0,1 },
224 { "bounded", T_AT_BOUNDED, 0, 0, 0, 0,0,1,1,5 },
225 { "break", T_BREAK, 0, 0, 0, 0,0,0,0,1 },
226 { "buffer", T_AT_BUFFER, 0, 0, 0, 0,0,1,1,5 },
227 { "builtin_offsetof", T_BUILTIN_OFFSETOF, 0, 0, 0, 0,0,1,0,2 },
228 { "case", T_CASE, 0, 0, 0, 0,0,0,0,1 },
229 { "char", T_TYPE, 0, CHAR, 0, 0,0,0,0,1 },
230 { "cold", T_AT_COLD, 0, 0, 0, 0,0,1,1,5 },
231 { "const", T_QUAL, 0, 0, CONST, 1,0,0,0,7 },
232 { "constructor",T_AT_CONSTRUCTOR,0, 0, 0, 0,0,1,1,5 },
233 { "continue", T_CONTINUE, 0, 0, 0, 0,0,0,0,1 },
234 { "default", T_DEFAULT, 0, 0, 0, 0,0,0,0,1 },
235 { "deprecated", T_AT_DEPRECATED,0, 0, 0, 0,0,1,1,5 },
236 { "destructor", T_AT_DESTRUCTOR,0, 0, 0, 0,0,1,1,5 },
237 { "do", T_DO, 0, 0, 0, 0,0,0,0,1 },
238 { "double", T_TYPE, 0, DOUBLE, 0, 0,0,0,0,1 },
239 { "else", T_ELSE, 0, 0, 0, 0,0,0,0,1 },
240 { "enum", T_ENUM, 0, 0, 0, 0,0,0,0,1 },
241 { "extension", T_EXTENSION, 0, 0, 0, 0,0,1,0,4 },
242 { "extern", T_SCLASS, EXTERN, 0, 0, 0,0,0,0,1 },
243 { "float", T_TYPE, 0, FLOAT, 0, 0,0,0,0,1 },
244 { "for", T_FOR, 0, 0, 0, 0,0,0,0,1 },
245 { "format", T_AT_FORMAT, 0, 0, 0, 0,0,1,1,5 },
246 { "format_arg", T_AT_FORMAT_ARG,0, 0, 0, 0,0,1,1,5 },
247 { "gnu_inline", T_AT_GNU_INLINE,0, 0, 0, 0,0,1,1,5 },
248 { "goto", T_GOTO, 0, 0, 0, 0,0,0,0,1 },
249 { "if", T_IF, 0, 0, 0, 0,0,0,0,1 },
250 { "imag", T_IMAG, 0, 0, 0, 0,1,0,0,4 },
251 { "inline", T_SCLASS, INLINE, 0, 0, 0,1,0,0,7 },
252 { "int", T_TYPE, 0, INT, 0, 0,0,0,0,1 },
253 { "long", T_TYPE, 0, LONG, 0, 0,0,0,0,1 },
254 { "malloc", T_AT_MALLOC, 0, 0, 0, 0,0,1,1,5 },
255 { "may_alias", T_AT_MAY_ALIAS, 0, 0, 0, 0,0,1,1,5 },
256 { "minbytes", T_AT_MINBYTES, 0, 0, 0, 0,0,1,1,5 },
257 { "mode", T_AT_MODE, 0, 0, 0, 0,0,1,1,5 },
258 { "no_instrument_function", T_AT_NO_INSTRUMENT_FUNCTION,
259 0, 0, 0, 0,0,1,1,5 },
260 { "nonnull", T_AT_NONNULL, 0, 0, 0, 0,0,1,1,5 },
261 { "noinline", T_AT_NOINLINE, 0, 0, 0, 0,0,1,1,5 },
262 { "noreturn", T_AT_NORETURN, 0, 0, 0, 0,0,1,1,5 },
263 { "nothrow", T_AT_NOTHROW, 0, 0, 0, 0,0,1,1,5 },
264 { "packed", T_AT_PACKED, 0, 0, 0, 0,0,1,1,5 },
265 { "packed", T_PACKED, 0, 0, 0, 0,0,0,0,2 },
266 { "pcs", T_AT_PCS, 0, 0, 0, 0,0,0,0,5 },
267 { "printf", T_AT_FORMAT_PRINTF,0, 0, 0, 0,0,1,1,5 },
268 { "pure", T_AT_PURE, 0, 0, 0, 0,0,1,1,5 },
269 { "real", T_REAL, 0, 0, 0, 0,1,0,0,4 },
270 { "register", T_SCLASS, REG, 0, 0, 0,0,0,0,1 },
271 { "restrict", T_QUAL, 0, 0, RESTRICT, 0,1,0,0,5 },
272 { "return", T_RETURN, 0, 0, 0, 0,0,0,0,1 },
273 { "returns_twice", T_AT_RETURNS_TWICE,0,0, 0, 0,0,1,1,5 },
274 { "scanf", T_AT_FORMAT_SCANF,0, 0, 0, 0,0,1,1,5 },
275 { "section", T_AT_SECTION, 0, 0, 0, 0,0,1,1,7 },
276 { "sentinel", T_AT_SENTINEL, 0, 0, 0, 0,0,1,1,5 },
277 { "short", T_TYPE, 0, SHORT, 0, 0,0,0,0,1 },
278 { "signed", T_TYPE, 0, SIGNED, 0, 1,0,0,0,3 },
279 { "sizeof", T_SIZEOF, 0, 0, 0, 0,0,0,0,1 },
280 { "static", T_SCLASS, STATIC, 0, 0, 0,0,0,0,1 },
281 { "strfmon", T_AT_FORMAT_STRFMON,0, 0, 0, 0,0,1,1,5 },
282 { "strftime", T_AT_FORMAT_STRFTIME,0, 0, 0, 0,0,1,1,5 },
283 { "string", T_AT_STRING, 0, 0, 0, 0,0,1,1,5 },
284 { "struct", T_SOU, 0, STRUCT, 0, 0,0,0,0,1 },
285 { "switch", T_SWITCH, 0, 0, 0, 0,0,0,0,1 },
286 { "symbolrename", T_SYMBOLRENAME, 0, 0, 0, 0,0,0,0,2 },
287 { "syslog", T_AT_FORMAT_SYSLOG,0, 0, 0, 0,0,1,1,5 },
288 { "transparent_union",T_AT_TUNION,0, 0, 0, 0,0,1,1,5 },
289 { "typedef", T_SCLASS, TYPEDEF, 0, 0, 0,0,0,0,1 },
290 { "typeof", T_TYPEOF, 0, 0, 0, 0,0,1,0,7 },
291 { "union", T_SOU, 0, UNION, 0, 0,0,0,0,1 },
292 { "unsigned", T_TYPE, 0, UNSIGN, 0, 0,0,0,0,1 },
293 { "unused", T_AT_UNUSED, 0, 0, 0, 0,0,1,1,5 },
294 { "used", T_AT_USED, 0, 0, 0, 0,0,1,1,5 },
295 { "visibility", T_AT_VISIBILITY,0, 0, 0, 0,0,1,1,5 },
296 { "void", T_TYPE, 0, VOID, 0, 0,0,0,0,1 },
297 { "volatile", T_QUAL, 0, 0, VOLATILE, 1,0,0,0,7 },
298 { "warn_unused_result", T_AT_WARN_UNUSED_RESULT, 0, 0, 0, 0,0,1,1,5 },
299 { "weak", T_AT_WEAK, 0, 0, 0, 0,0,1,1,5 },
300 { "while", T_WHILE, 0, 0, 0, 0,0,0,0,1 },
301 { NULL, 0, 0, 0, 0, 0,0,0,0,0 }
302 };
303
304 /* Symbol table */
305 static sym_t *symtab[HSHSIZ1];
306
307 /* bit i of the entry with index i is set */
308 uint64_t qbmasks[sizeof(uint64_t) * CHAR_BIT];
309
310 /* least significant i bits are set in the entry with index i */
311 uint64_t qlmasks[sizeof(uint64_t) * CHAR_BIT + 1];
312
313 /* least significant i bits are not set in the entry with index i */
314 uint64_t qumasks[sizeof(uint64_t) * CHAR_BIT + 1];
315
316 /* free list for sbuf structures */
317 static sbuf_t *sbfrlst;
318
319 /* Typ of next expected symbol */
320 symt_t symtyp;
321
322
323 static void
324 addkw(struct kwtab *kw, int deco)
325 {
326 sym_t *sym;
327 size_t h;
328 char buf[256];
329 const char *name;
330
331 if (!(kw->kw_deco & deco))
332 return;
333
334 switch (deco) {
335 case 1:
336 name = kw->kw_name;
337 break;
338 case 2:
339 snprintf(buf, sizeof(buf), "__%s", kw->kw_name);
340 name = strdup(buf);
341 break;
342 case 4:
343 snprintf(buf, sizeof(buf), "__%s__", kw->kw_name);
344 name = strdup(buf);
345 break;
346 default:
347 abort();
348 }
349
350 if (name == NULL)
351 err(1, "Can't init symbol table");
352
353 sym = getblk(sizeof (sym_t));
354 sym->s_name = name;
355 sym->s_keyw = kw;
356 sym->s_value.v_quad = kw->kw_token;
357 if (kw->kw_token == T_TYPE || kw->kw_token == T_SOU) {
358 sym->s_tspec = kw->kw_tspec;
359 } else if (kw->kw_token == T_SCLASS) {
360 sym->s_scl = kw->kw_scl;
361 } else if (kw->kw_token == T_QUAL) {
362 sym->s_tqual = kw->kw_tqual;
363 }
364 h = hash(sym->s_name);
365 if ((sym->s_link = symtab[h]) != NULL)
366 symtab[h]->s_rlink = &sym->s_link;
367 sym->s_rlink = &symtab[h];
368 symtab[h] = sym;
369 }
370
371 /*
372 * All keywords are written to the symbol table. This saves us looking
373 * in a extra table for each name we found.
374 */
375 void
376 initscan(void)
377 {
378 struct kwtab *kw;
379 size_t i;
380 uint64_t uq;
381
382 for (kw = kwtab; kw->kw_name != NULL; kw++) {
383 if ((kw->kw_c89 || kw->kw_c99) && tflag)
384 continue;
385 if (kw->kw_c99 && !(Sflag || gflag))
386 continue;
387 if (kw->kw_gcc && !gflag)
388 continue;
389 addkw(kw, 1);
390 addkw(kw, 2);
391 addkw(kw, 4);
392 }
393
394 /* initialize bit-masks for quads */
395 for (i = 0; i < sizeof (uint64_t) * CHAR_BIT; i++) {
396 qbmasks[i] = (uint64_t)1 << i;
397 uq = ~(uint64_t)0 << i;
398 qumasks[i] = uq;
399 qlmasks[i] = ~uq;
400 }
401 qumasks[i] = 0;
402 qlmasks[i] = ~(uint64_t)0;
403 }
404
405 /*
406 * Get a free sbuf structure, if possible from the free list
407 */
408 static sbuf_t *
409 allocsb(void)
410 {
411 sbuf_t *sb;
412
413 if ((sb = sbfrlst) != NULL) {
414 sbfrlst = sb->sb_nxt;
415 #ifdef BLKDEBUG
416 (void)memset(sb, 0, sizeof (*sb));
417 #else
418 sb->sb_nxt = NULL;
419 #endif
420 } else {
421 sb = xmalloc(sizeof (sbuf_t));
422 (void)memset(sb, 0, sizeof (*sb));
423 }
424 return (sb);
425 }
426
427 /*
428 * Put a sbuf structure to the free list
429 */
430 static void
431 freesb(sbuf_t *sb)
432 {
433
434 (void)memset(sb, ZERO, sizeof (*sb));
435 sb->sb_nxt = sbfrlst;
436 sbfrlst = sb;
437 }
438
439 /*
440 * Read a character and ensure that it is positive (except EOF).
441 * Increment line count(s) if necessary.
442 */
443 static int
444 inpc(void)
445 {
446 int c;
447
448 if ((c = input()) != EOF && (c &= CHAR_MASK) == '\n')
449 incline();
450 return (c);
451 }
452
453 static int
454 hash(const char *s)
455 {
456 u_int v;
457 const u_char *us;
458
459 v = 0;
460 for (us = (const u_char *)s; *us != '\0'; us++) {
461 v = (v << sizeof (v)) + *us;
462 v ^= v >> (sizeof (v) * CHAR_BIT - sizeof (v));
463 }
464 return (v % HSHSIZ1);
465 }
466
467 /*
468 * Lex has found a letter followed by zero or more letters or digits.
469 * It looks for a symbol in the symbol table with the same name. This
470 * symbol must either be a keyword or a symbol of the type required by
471 * symtyp (label, member, tag, ...).
472 *
473 * If it is a keyword, the token is returned. In some cases it is described
474 * more deeply by data written to yylval.
475 *
476 * If it is a symbol, T_NAME is returned and the pointer to a sbuf struct
477 * is stored in yylval. This struct contains the name of the symbol, its
478 * length and hash value. If there is already a symbol of the same name
479 * and type in the symbol table, the sbuf struct also contains a pointer
480 * to the symbol table entry.
481 */
482 static int
483 name(void)
484 {
485 char *s;
486 sbuf_t *sb;
487 sym_t *sym;
488 int tok;
489
490 sb = allocsb();
491 sb->sb_name = yytext;
492 sb->sb_len = yyleng;
493 sb->sb_hash = hash(yytext);
494 if ((sym = search(sb)) != NULL && sym->s_keyw) {
495 freesb(sb);
496 return (keyw(sym));
497 }
498
499 sb->sb_sym = sym;
500
501 if (sym != NULL) {
502 if (blklev < sym->s_blklev)
503 LERROR("name()");
504 sb->sb_name = sym->s_name;
505 sb->sb_len = strlen(sym->s_name);
506 tok = sym->s_scl == TYPEDEF ? T_TYPENAME : T_NAME;
507 } else {
508 s = getblk(yyleng + 1);
509 (void)memcpy(s, yytext, yyleng + 1);
510 sb->sb_name = s;
511 sb->sb_len = yyleng;
512 tok = T_NAME;
513 }
514
515 yylval.y_sb = sb;
516 return (tok);
517 }
518
519 static sym_t *
520 search(sbuf_t *sb)
521 {
522 sym_t *sym;
523
524 for (sym = symtab[sb->sb_hash]; sym != NULL; sym = sym->s_link) {
525 if (strcmp(sym->s_name, sb->sb_name) == 0) {
526 if (sym->s_keyw) {
527 struct kwtab *kw = sym->s_keyw;
528 if (!kw->kw_attr || attron)
529 return (sym);
530 } else if (!attron && sym->s_kind == symtyp)
531 return sym;
532 }
533 }
534
535 return (NULL);
536 }
537
538 static int
539 keyw(sym_t *sym)
540 {
541 int t;
542
543 if ((t = (int)sym->s_value.v_quad) == T_SCLASS) {
544 yylval.y_scl = sym->s_scl;
545 } else if (t == T_TYPE || t == T_SOU) {
546 yylval.y_tspec = sym->s_tspec;
547 } else if (t == T_QUAL) {
548 yylval.y_tqual = sym->s_tqual;
549 }
550 return (t);
551 }
552
553 /*
554 * Convert a string representing an integer into internal representation.
555 * The value is returned in yylval. icon() (and yylex()) returns T_CON.
556 */
557 static int
558 icon(int base)
559 {
560 int l_suffix, u_suffix;
561 int len;
562 const char *cp;
563 char c, *eptr;
564 tspec_t typ;
565 int ansiu;
566 #ifdef TARG_INT128_MAX
567 __uint128_t uq = 0;
568 static tspec_t contypes[2][4] = {
569 { INT, LONG, QUAD, INT128, },
570 { UINT, ULONG, UQUAD, UINT128, }
571 };
572 #else
573 uint64_t uq = 0;
574 static tspec_t contypes[2][3] = {
575 { INT, LONG, QUAD, },
576 { UINT, ULONG, UQUAD, }
577 };
578 #endif
579
580 cp = yytext;
581 len = yyleng;
582
583 /* skip 0[xX] or 0[bB] */
584 if (base == 16 || base == 2) {
585 cp += 2;
586 len -= 2;
587 }
588
589 /* read suffixes */
590 l_suffix = u_suffix = 0;
591 for ( ; ; ) {
592 if ((c = cp[len - 1]) == 'l' || c == 'L') {
593 l_suffix++;
594 } else if (c == 'u' || c == 'U') {
595 u_suffix++;
596 } else {
597 break;
598 }
599 len--;
600 }
601 if (l_suffix > 2 || u_suffix > 1) {
602 /* malformed integer constant */
603 warning(251);
604 if (l_suffix > 2)
605 l_suffix = 2;
606 if (u_suffix > 1)
607 u_suffix = 1;
608 }
609 if (tflag && u_suffix != 0) {
610 /* suffix U is illegal in traditional C */
611 warning(97);
612 }
613 typ = contypes[u_suffix][l_suffix];
614
615 errno = 0;
616
617 uq = strtouq(cp, &eptr, base);
618 if (eptr != cp + len)
619 LERROR("icon()");
620 if (errno != 0)
621 /* integer constant out of range */
622 warning(252);
623
624 /*
625 * If the value is too big for the current type, we must choose
626 * another type.
627 */
628 ansiu = 0;
629 switch (typ) {
630 case INT:
631 if (uq <= TARG_INT_MAX) {
632 /* ok */
633 } else if (uq <= TARG_UINT_MAX && base != 10) {
634 typ = UINT;
635 } else if (uq <= TARG_LONG_MAX) {
636 typ = LONG;
637 } else {
638 typ = ULONG;
639 if (uq > TARG_ULONG_MAX) {
640 /* integer constant out of range */
641 warning(252);
642 }
643 }
644 if (typ == UINT || typ == ULONG) {
645 if (tflag) {
646 typ = LONG;
647 } else if (!sflag) {
648 /*
649 * Remember that the constant is unsigned
650 * only in ANSI C
651 */
652 ansiu = 1;
653 }
654 }
655 break;
656 case UINT:
657 if (uq > TARG_UINT_MAX) {
658 typ = ULONG;
659 if (uq > TARG_ULONG_MAX) {
660 /* integer constant out of range */
661 warning(252);
662 }
663 }
664 break;
665 case LONG:
666 if (uq > TARG_LONG_MAX && !tflag) {
667 typ = ULONG;
668 if (!sflag)
669 ansiu = 1;
670 if (uq > TARG_ULONG_MAX) {
671 /* integer constant out of range */
672 warning(252);
673 }
674 }
675 break;
676 case ULONG:
677 if (uq > TARG_ULONG_MAX) {
678 /* integer constant out of range */
679 warning(252);
680 }
681 break;
682 case QUAD:
683 if (uq > TARG_QUAD_MAX && !tflag) {
684 typ = UQUAD;
685 if (!sflag)
686 ansiu = 1;
687 }
688 break;
689 case UQUAD:
690 if (uq > TARG_UQUAD_MAX) {
691 /* integer constant out of range */
692 warning(252);
693 }
694 break;
695 #ifdef INT128_SIZE
696 case INT128:
697 #ifdef TARG_INT128_MAX
698 if (uq > TARG_INT128_MAX && !tflag) {
699 typ = UINT128;
700 if (!sflag)
701 ansiu = 1;
702 }
703 #endif
704 break;
705 case UINT128:
706 #ifdef TARG_INT128_MAX
707 if (uq > TARG_UINT128_MAX) {
708 /* integer constant out of range */
709 warning(252);
710 }
711 #endif
712 break;
713 #endif
714 /* LINTED206: (enumeration values not handled in switch) */
715 case STRUCT:
716 case VOID:
717 case LDOUBLE:
718 case FUNC:
719 case ARRAY:
720 case PTR:
721 case ENUM:
722 case UNION:
723 case SIGNED:
724 case NOTSPEC:
725 case DOUBLE:
726 case FLOAT:
727 case USHORT:
728 case SHORT:
729 case UCHAR:
730 case SCHAR:
731 case CHAR:
732 case BOOL:
733 case UNSIGN:
734 case FCOMPLEX:
735 case DCOMPLEX:
736 case LCOMPLEX:
737 case COMPLEX:
738 break;
739
740 case NTSPEC: /* this value unused */
741 break;
742 }
743
744 uq = (uint64_t)xsign((int64_t)uq, typ, -1);
745
746 (yylval.y_val = xcalloc(1, sizeof (val_t)))->v_tspec = typ;
747 yylval.y_val->v_ansiu = ansiu;
748 yylval.y_val->v_quad = (int64_t)uq;
749
750 return (T_CON);
751 }
752
753 /*
754 * Returns 1 if t is a signed type and the value is negative.
755 *
756 * len is the number of significant bits. If len is -1, len is set
757 * to the width of type t.
758 */
759 int
760 sign(int64_t q, tspec_t t, int len)
761 {
762
763 if (t == PTR || isutyp(t))
764 return (0);
765 return (msb(q, t, len));
766 }
767
768 int
769 msb(int64_t q, tspec_t t, int len)
770 {
771
772 if (len <= 0)
773 len = size(t);
774 return ((q & qbmasks[len - 1]) != 0);
775 }
776
777 /*
778 * Extends the sign of q.
779 */
780 int64_t
781 xsign(int64_t q, tspec_t t, int len)
782 {
783
784 if (len <= 0)
785 len = size(t);
786
787 if (t == PTR || isutyp(t) || !sign(q, t, len)) {
788 q &= qlmasks[len];
789 } else {
790 q |= qumasks[len];
791 }
792 return (q);
793 }
794
795 /*
796 * Convert a string representing a floating point value into its interal
797 * representation. Type and value are returned in yylval. fcon()
798 * (and yylex()) returns T_CON.
799 * XXX Currently it is not possible to convert constants of type
800 * long double which are greater than DBL_MAX.
801 */
802 static int
803 fcon(void)
804 {
805 const char *cp;
806 int len;
807 tspec_t typ;
808 char c, *eptr;
809 double d;
810 float f = 0;
811
812 cp = yytext;
813 len = yyleng;
814
815 if (cp[len - 1] == 'i') {
816 /* imaginary, do nothing for now */
817 len--;
818 }
819 if ((c = cp[len - 1]) == 'f' || c == 'F') {
820 typ = FLOAT;
821 len--;
822 } else if (c == 'l' || c == 'L') {
823 typ = LDOUBLE;
824 len--;
825 } else {
826 if (c == 'd' || c == 'D')
827 len--;
828 typ = DOUBLE;
829 }
830
831 if (tflag && typ != DOUBLE) {
832 /* suffixes F and L are illegal in traditional C */
833 warning(98);
834 }
835
836 errno = 0;
837 d = strtod(cp, &eptr);
838 if (eptr != cp + len) {
839 switch (*eptr) {
840 /*
841 * XXX: non-native non-current strtod() may not handle hex
842 * floats, ignore the rest if we find traces of hex float
843 * syntax...
844 */
845 case 'p':
846 case 'P':
847 case 'x':
848 case 'X':
849 d = 0;
850 errno = 0;
851 break;
852 default:
853 LERROR("fcon(%s->%s)", cp, eptr);
854 }
855 }
856 if (errno != 0)
857 /* floating-point constant out of range */
858 warning(248);
859
860 if (typ == FLOAT) {
861 f = (float)d;
862 if (!finite(f)) {
863 /* floating-point constant out of range */
864 warning(248);
865 f = f > 0 ? FLT_MAX : -FLT_MAX;
866 }
867 }
868
869 (yylval.y_val = xcalloc(1, sizeof (val_t)))->v_tspec = typ;
870 if (typ == FLOAT) {
871 yylval.y_val->v_ldbl = f;
872 } else {
873 yylval.y_val->v_ldbl = d;
874 }
875
876 return (T_CON);
877 }
878
879 static int
880 operator(int t, op_t o)
881 {
882
883 yylval.y_op = o;
884 return (t);
885 }
886
887 /*
888 * Called if lex found a leading \'.
889 */
890 static int
891 ccon(void)
892 {
893 size_t n;
894 int val, c;
895 char cv;
896
897 n = 0;
898 val = 0;
899 while ((c = getescc('\'')) >= 0) {
900 val = (val << CHAR_BIT) + c;
901 n++;
902 }
903 if (c == -2) {
904 /* unterminated character constant */
905 error(253);
906 } else {
907 if (n > sizeof (int) || (n > 1 && (pflag || hflag))) {
908 /* too many characters in character constant */
909 error(71);
910 } else if (n > 1) {
911 /* multi-character character constant */
912 warning(294);
913 } else if (n == 0) {
914 /* empty character constant */
915 error(73);
916 }
917 }
918 if (n == 1) {
919 cv = (char)val;
920 val = cv;
921 }
922
923 yylval.y_val = xcalloc(1, sizeof (val_t));
924 yylval.y_val->v_tspec = INT;
925 yylval.y_val->v_quad = val;
926
927 return (T_CON);
928 }
929
930 /*
931 * Called if lex found a leading L\'
932 */
933 static int
934 wccon(void)
935 {
936 static char buf[MB_LEN_MAX + 1];
937 size_t i;
938 int c;
939 wchar_t wc;
940
941 i = 0;
942 while ((c = getescc('\'')) >= 0) {
943 if (i < MB_CUR_MAX)
944 buf[i] = (char)c;
945 i++;
946 }
947
948 wc = 0;
949
950 if (c == -2) {
951 /* unterminated character constant */
952 error(253);
953 } else if (c == 0) {
954 /* empty character constant */
955 error(73);
956 } else {
957 if (i > MB_CUR_MAX) {
958 i = MB_CUR_MAX;
959 /* too many characters in character constant */
960 error(71);
961 } else {
962 buf[i] = '\0';
963 (void)mbtowc(NULL, NULL, 0);
964 if (mbtowc(&wc, buf, MB_CUR_MAX) < 0)
965 /* invalid multibyte character */
966 error(291);
967 }
968 }
969
970 yylval.y_val = xcalloc(1, sizeof (val_t));
971 yylval.y_val->v_tspec = WCHAR;
972 yylval.y_val->v_quad = wc;
973
974 return (T_CON);
975 }
976
977 /*
978 * Read a character which is part of a character constant or of a string
979 * and handle escapes.
980 *
981 * The Argument is the character which delimits the character constant or
982 * string.
983 *
984 * Returns -1 if the end of the character constant or string is reached,
985 * -2 if the EOF is reached, and the character otherwise.
986 */
987 static int
988 getescc(int d)
989 {
990 static int pbc = -1;
991 int n, c, v;
992
993 if (pbc == -1) {
994 c = inpc();
995 } else {
996 c = pbc;
997 pbc = -1;
998 }
999 if (c == d)
1000 return (-1);
1001 switch (c) {
1002 case '\n':
1003 if (tflag) {
1004 /* newline in string or char constant */
1005 error(254);
1006 return (-2);
1007 }
1008 return (c);
1009 case EOF:
1010 return (-2);
1011 case '\\':
1012 switch (c = inpc()) {
1013 case '"':
1014 if (tflag && d == '\'')
1015 /* \" inside character constant undef. ... */
1016 warning(262);
1017 return ('"');
1018 case '\'':
1019 return ('\'');
1020 case '?':
1021 if (tflag)
1022 /* \? undefined in traditional C */
1023 warning(263);
1024 return ('?');
1025 case '\\':
1026 return ('\\');
1027 case 'a':
1028 if (tflag)
1029 /* \a undefined in traditional C */
1030 warning(81);
1031 return ('\a');
1032 case 'b':
1033 return ('\b');
1034 case 'f':
1035 return ('\f');
1036 case 'n':
1037 return ('\n');
1038 case 'r':
1039 return ('\r');
1040 case 't':
1041 return ('\t');
1042 case 'v':
1043 if (tflag)
1044 /* \v undefined in traditional C */
1045 warning(264);
1046 return ('\v');
1047 case '8': case '9':
1048 /* bad octal digit %c */
1049 warning(77, c);
1050 /* FALLTHROUGH */
1051 case '0': case '1': case '2': case '3':
1052 case '4': case '5': case '6': case '7':
1053 n = 3;
1054 v = 0;
1055 do {
1056 v = (v << 3) + (c - '0');
1057 c = inpc();
1058 } while (--n && isdigit(c) && (tflag || c <= '7'));
1059 if (tflag && n > 0 && isdigit(c))
1060 /* bad octal digit %c */
1061 warning(77, c);
1062 pbc = c;
1063 if (v > TARG_UCHAR_MAX) {
1064 /* character escape does not fit in char. */
1065 warning(76);
1066 v &= CHAR_MASK;
1067 }
1068 return (v);
1069 case 'x':
1070 if (tflag)
1071 /* \x undefined in traditional C */
1072 warning(82);
1073 v = 0;
1074 n = 0;
1075 while ((c = inpc()) >= 0 && isxdigit(c)) {
1076 c = isdigit(c) ?
1077 c - '0' : toupper(c) - 'A' + 10;
1078 v = (v << 4) + c;
1079 if (n >= 0) {
1080 if ((v & ~CHAR_MASK) != 0) {
1081 /* overflow in hex escape */
1082 warning(75);
1083 n = -1;
1084 } else {
1085 n++;
1086 }
1087 }
1088 }
1089 pbc = c;
1090 if (n == 0) {
1091 /* no hex digits follow \x */
1092 error(74);
1093 } if (n == -1) {
1094 v &= CHAR_MASK;
1095 }
1096 return (v);
1097 case '\n':
1098 return (getescc(d));
1099 case EOF:
1100 return (-2);
1101 default:
1102 if (isprint(c)) {
1103 /* dubious escape \%c */
1104 warning(79, c);
1105 } else {
1106 /* dubious escape \%o */
1107 warning(80, c);
1108 }
1109 }
1110 }
1111 return (c);
1112 }
1113
1114 /*
1115 * Called for preprocessor directives. Currently implemented are:
1116 * # lineno
1117 * # lineno "filename"
1118 */
1119 static void
1120 directive(void)
1121 {
1122 const char *cp, *fn;
1123 char c, *eptr;
1124 size_t fnl;
1125 long ln;
1126 static int first = 1;
1127
1128 /* Go to first non-whitespace after # */
1129 for (cp = yytext + 1; (c = *cp) == ' ' || c == '\t'; cp++)
1130 continue;
1131
1132 if (!isdigit((unsigned char)c)) {
1133 if (strncmp(cp, "pragma", 6) == 0
1134 && isspace((unsigned char)cp[6]))
1135 return;
1136 error:
1137 /* undefined or invalid # directive */
1138 warning(255);
1139 return;
1140 }
1141 ln = strtol(--cp, &eptr, 10);
1142 if (cp == eptr)
1143 goto error;
1144 if ((c = *(cp = eptr)) != ' ' && c != '\t' && c != '\0')
1145 goto error;
1146 while ((c = *cp++) == ' ' || c == '\t')
1147 continue;
1148 if (c != '\0') {
1149 if (c != '"')
1150 goto error;
1151 fn = cp;
1152 while ((c = *cp) != '"' && c != '\0')
1153 cp++;
1154 if (c != '"')
1155 goto error;
1156 if ((fnl = cp++ - fn) > PATH_MAX)
1157 goto error;
1158 while ((c = *cp++) == ' ' || c == '\t')
1159 continue;
1160 #if 0
1161 if (c != '\0')
1162 warning("extra character(s) after directive");
1163 #endif
1164
1165 /* empty string means stdin */
1166 if (fnl == 0) {
1167 fn = "{standard input}";
1168 fnl = 16; /* strlen (fn) */
1169 }
1170 curr_pos.p_file = fnnalloc(fn, fnl);
1171 /*
1172 * If this is the first directive, the name is the name
1173 * of the C source file as specified at the command line.
1174 * It is written to the output file.
1175 */
1176 if (first) {
1177 csrc_pos.p_file = curr_pos.p_file;
1178 outsrc(fnxform(curr_pos.p_file,
1179 strlen(curr_pos.p_file)));
1180 first = 0;
1181 }
1182 }
1183 curr_pos.p_line = (int)ln - 1;
1184 curr_pos.p_uniq = 0;
1185 if (curr_pos.p_file == csrc_pos.p_file) {
1186 csrc_pos.p_line = (int)ln - 1;
1187 csrc_pos.p_uniq = 0;
1188 }
1189 }
1190
1191 /*
1192 * Handle lint comments. Following comments are currently understood:
1193 * ARGSUSEDn
1194 * BITFIELDTYPE
1195 * CONSTCOND CONSTANTCOND CONSTANTCONDITION
1196 * FALLTHRU FALLTHROUGH
1197 * LINTLIBRARY
1198 * LINTEDn NOSTRICTn
1199 * LONGLONG
1200 * NOTREACHED
1201 * PRINTFLIKEn
1202 * PROTOLIB
1203 * SCANFLIKEn
1204 * VARARGSn
1205 * If one of this comments is recognized, the arguments, if any, are
1206 * parsed and a function which handles this comment is called.
1207 */
1208 static void
1209 comment(void)
1210 {
1211 int c, lc;
1212 static struct {
1213 const char *keywd;
1214 int arg;
1215 void (*func)(int);
1216 } keywtab[] = {
1217 { "ARGSUSED", 1, argsused },
1218 { "BITFIELDTYPE", 0, bitfieldtype },
1219 { "CONSTCOND", 0, constcond },
1220 { "CONSTANTCOND", 0, constcond },
1221 { "CONSTANTCONDITION", 0, constcond },
1222 { "FALLTHRU", 0, fallthru },
1223 { "FALLTHROUGH", 0, fallthru },
1224 { "LINTLIBRARY", 0, lintlib },
1225 { "LINTED", 1, linted },
1226 { "LONGLONG", 0, longlong },
1227 { "NOSTRICT", 1, linted },
1228 { "NOTREACHED", 0, notreach },
1229 { "PRINTFLIKE", 1, printflike },
1230 { "PROTOLIB", 1, protolib },
1231 { "SCANFLIKE", 1, scanflike },
1232 { "VARARGS", 1, varargs },
1233 };
1234 char keywd[32];
1235 char arg[32];
1236 size_t l, i;
1237 int a;
1238 int eoc;
1239
1240 eoc = 0;
1241
1242 /* Skip white spaces after the start of the comment */
1243 while ((c = inpc()) != EOF && isspace(c))
1244 continue;
1245
1246 /* Read the potential keyword to keywd */
1247 l = 0;
1248 while (c != EOF && isupper(c) && l < sizeof (keywd) - 1) {
1249 keywd[l++] = (char)c;
1250 c = inpc();
1251 }
1252 keywd[l] = '\0';
1253
1254 /* look for the keyword */
1255 for (i = 0; i < sizeof (keywtab) / sizeof (keywtab[0]); i++) {
1256 if (strcmp(keywtab[i].keywd, keywd) == 0)
1257 break;
1258 }
1259 if (i == sizeof (keywtab) / sizeof (keywtab[0]))
1260 goto skip_rest;
1261
1262 /* skip white spaces after the keyword */
1263 while (c != EOF && isspace(c))
1264 c = inpc();
1265
1266 /* read the argument, if the keyword accepts one and there is one */
1267 l = 0;
1268 if (keywtab[i].arg) {
1269 while (c != EOF && isdigit(c) && l < sizeof (arg) - 1) {
1270 arg[l++] = (char)c;
1271 c = inpc();
1272 }
1273 }
1274 arg[l] = '\0';
1275 a = l != 0 ? atoi(arg) : -1;
1276
1277 /* skip white spaces after the argument */
1278 while (c != EOF && isspace(c))
1279 c = inpc();
1280
1281 if (c != '*' || (c = inpc()) != '/') {
1282 if (keywtab[i].func != linted)
1283 /* extra characters in lint comment */
1284 warning(257);
1285 } else {
1286 /*
1287 * remember that we have already found the end of the
1288 * comment
1289 */
1290 eoc = 1;
1291 }
1292
1293 if (keywtab[i].func != NULL)
1294 (*keywtab[i].func)(a);
1295
1296 skip_rest:
1297 while (!eoc) {
1298 lc = c;
1299 if ((c = inpc()) == EOF) {
1300 /* unterminated comment */
1301 error(256);
1302 break;
1303 }
1304 if (lc == '*' && c == '/')
1305 eoc = 1;
1306 }
1307 }
1308
1309 /*
1310 * Handle // style comments
1311 */
1312 static void
1313 slashslashcomment(void)
1314 {
1315 int c;
1316
1317 if (!Sflag && !gflag)
1318 /* // comments only supported in C99 */
1319 (void)gnuism(312, tflag ? "traditional" : "ANSI");
1320
1321 while ((c = inpc()) != EOF && c != '\n')
1322 continue;
1323 }
1324
1325 /*
1326 * Clear flags for lint comments LINTED, LONGLONG and CONSTCOND.
1327 * clrwflgs() is called after function definitions and global and
1328 * local declarations and definitions. It is also called between
1329 * the controlling expression and the body of control statements
1330 * (if, switch, for, while).
1331 */
1332 void
1333 clrwflgs(void)
1334 {
1335
1336 lwarn = LWARN_ALL;
1337 quadflg = 0;
1338 ccflg = 0;
1339 }
1340
1341 /*
1342 * Strings are stored in a dynamically alloceted buffer and passed
1343 * in yylval.y_xstrg to the parser. The parser or the routines called
1344 * by the parser are responsible for freeing this buffer.
1345 */
1346 static int
1347 string(void)
1348 {
1349 u_char *s;
1350 int c;
1351 size_t len, max;
1352 strg_t *strg;
1353
1354 s = xmalloc(max = 64);
1355
1356 len = 0;
1357 while ((c = getescc('"')) >= 0) {
1358 /* +1 to reserve space for a trailing NUL character */
1359 if (len + 1 == max)
1360 s = xrealloc(s, max *= 2);
1361 s[len++] = (char)c;
1362 }
1363 s[len] = '\0';
1364 if (c == -2)
1365 /* unterminated string constant */
1366 error(258);
1367
1368 strg = xcalloc(1, sizeof (strg_t));
1369 strg->st_tspec = CHAR;
1370 strg->st_len = len;
1371 strg->st_cp = s;
1372
1373 yylval.y_strg = strg;
1374 return (T_STRING);
1375 }
1376
1377 static int
1378 wcstrg(void)
1379 {
1380 char *s;
1381 int c, n;
1382 size_t i, wi;
1383 size_t len, max, wlen;
1384 wchar_t *ws;
1385 strg_t *strg;
1386
1387 s = xmalloc(max = 64);
1388 len = 0;
1389 while ((c = getescc('"')) >= 0) {
1390 /* +1 to save space for a trailing NUL character */
1391 if (len + 1 >= max)
1392 s = xrealloc(s, max *= 2);
1393 s[len++] = (char)c;
1394 }
1395 s[len] = '\0';
1396 if (c == -2)
1397 /* unterminated string constant */
1398 error(258);
1399
1400 /* get length of wide-character string */
1401 (void)mblen(NULL, 0);
1402 for (i = 0, wlen = 0; i < len; i += n, wlen++) {
1403 if ((n = mblen(&s[i], MB_CUR_MAX)) == -1) {
1404 /* invalid multibyte character */
1405 error(291);
1406 break;
1407 }
1408 if (n == 0)
1409 n = 1;
1410 }
1411
1412 ws = xmalloc((wlen + 1) * sizeof (wchar_t));
1413
1414 /* convert from multibyte to wide char */
1415 (void)mbtowc(NULL, NULL, 0);
1416 for (i = 0, wi = 0; i < len; i += n, wi++) {
1417 if ((n = mbtowc(&ws[wi], &s[i], MB_CUR_MAX)) == -1)
1418 break;
1419 if (n == 0)
1420 n = 1;
1421 }
1422 ws[wi] = 0;
1423 free(s);
1424
1425 strg = xcalloc(1, sizeof (strg_t));
1426 strg->st_tspec = WCHAR;
1427 strg->st_len = wlen;
1428 strg->st_wcp = ws;
1429
1430 yylval.y_strg = strg;
1431 return (T_STRING);
1432 }
1433
1434 /*
1435 * As noted above the scanner does not create new symbol table entries
1436 * for symbols it cannot find in the symbol table. This is to avoid
1437 * putting undeclared symbols into the symbol table if a syntax error
1438 * occurs.
1439 *
1440 * getsym() is called as soon as it is probably ok to put the symbol to
1441 * the symbol table. This does not mean that it is not possible that
1442 * symbols are put to the symbol table which are than not completely
1443 * declared due to syntax errors. To avoid too many problems in this
1444 * case symbols get type int in getsym().
1445 *
1446 * XXX calls to getsym() should be delayed until decl1*() is called
1447 */
1448 sym_t *
1449 getsym(sbuf_t *sb)
1450 {
1451 dinfo_t *di;
1452 char *s;
1453 sym_t *sym;
1454
1455 sym = sb->sb_sym;
1456
1457 /*
1458 * During member declaration it is possible that name() looked
1459 * for symbols of type FVFT, although it should have looked for
1460 * symbols of type FTAG. Same can happen for labels. Both cases
1461 * are compensated here.
1462 */
1463 if (symtyp == FMOS || symtyp == FLAB) {
1464 if (sym == NULL || sym->s_kind == FVFT)
1465 sym = search(sb);
1466 }
1467
1468 if (sym != NULL) {
1469 if (sym->s_kind != symtyp)
1470 LERROR("storesym(%d, %d)", sym->s_kind, symtyp);
1471 symtyp = FVFT;
1472 freesb(sb);
1473 return (sym);
1474 }
1475
1476 /* create a new symbol table entry */
1477
1478 /* labels must always be allocated at level 1 (outhermost block) */
1479 if (symtyp == FLAB) {
1480 sym = getlblk(1, sizeof (sym_t));
1481 s = getlblk(1, sb->sb_len + 1);
1482 (void)memcpy(s, sb->sb_name, sb->sb_len + 1);
1483 sym->s_name = s;
1484 sym->s_blklev = 1;
1485 di = dcs;
1486 while (di->d_nxt != NULL && di->d_nxt->d_nxt != NULL)
1487 di = di->d_nxt;
1488 if (di->d_ctx != AUTO)
1489 LERROR("storesym()");
1490 } else {
1491 sym = getblk(sizeof (sym_t));
1492 sym->s_name = sb->sb_name;
1493 sym->s_blklev = blklev;
1494 di = dcs;
1495 }
1496
1497 UNIQUE_CURR_POS(sym->s_dpos);
1498 if ((sym->s_kind = symtyp) != FLAB)
1499 sym->s_type = gettyp(INT);
1500
1501 symtyp = FVFT;
1502
1503 if ((sym->s_link = symtab[sb->sb_hash]) != NULL)
1504 symtab[sb->sb_hash]->s_rlink = &sym->s_link;
1505 sym->s_rlink = &symtab[sb->sb_hash];
1506 symtab[sb->sb_hash] = sym;
1507
1508 *di->d_ldlsym = sym;
1509 di->d_ldlsym = &sym->s_dlnxt;
1510
1511 freesb(sb);
1512 return (sym);
1513 }
1514
1515 /*
1516 * Construct a temporary symbol. The symbol starts with a digit, so that
1517 * it is illegal.
1518 */
1519 sym_t *
1520 mktempsym(type_t *t)
1521 {
1522 static int n = 0;
1523 int h;
1524 char *s = getlblk(blklev, 64);
1525 sym_t *sym = getblk(sizeof (sym_t));
1526
1527 (void)snprintf(s, 64, "%.8d_tmp", n++);
1528 h = hash(s);
1529
1530 sym->s_name = s;
1531 sym->s_type = t;
1532 sym->s_blklev = blklev;
1533 sym->s_scl = AUTO;
1534 sym->s_kind = FVFT;
1535 sym->s_used = 1;
1536 sym->s_set = 1;
1537
1538 if ((sym->s_link = symtab[h]) != NULL)
1539 symtab[h]->s_rlink = &sym->s_link;
1540 sym->s_rlink = &symtab[h];
1541 symtab[h] = sym;
1542
1543 *dcs->d_ldlsym = sym;
1544 dcs->d_ldlsym = &sym->s_dlnxt;
1545
1546 return sym;
1547 }
1548
1549 /*
1550 * Remove a symbol forever from the symbol table. s_blklev
1551 * is set to -1 to avoid that the symbol will later be put
1552 * back to the symbol table.
1553 */
1554 void
1555 rmsym(sym_t *sym)
1556 {
1557
1558 if ((*sym->s_rlink = sym->s_link) != NULL)
1559 sym->s_link->s_rlink = sym->s_rlink;
1560 sym->s_blklev = -1;
1561 sym->s_link = NULL;
1562 }
1563
1564 /*
1565 * Remove a list of symbols declared at one level from the symbol
1566 * table.
1567 */
1568 void
1569 rmsyms(sym_t *syms)
1570 {
1571 sym_t *sym;
1572
1573 for (sym = syms; sym != NULL; sym = sym->s_dlnxt) {
1574 if (sym->s_blklev != -1) {
1575 if ((*sym->s_rlink = sym->s_link) != NULL)
1576 sym->s_link->s_rlink = sym->s_rlink;
1577 sym->s_link = NULL;
1578 sym->s_rlink = NULL;
1579 }
1580 }
1581 }
1582
1583 /*
1584 * Put a symbol into the symbol table
1585 */
1586 void
1587 inssym(int bl, sym_t *sym)
1588 {
1589 int h;
1590
1591 h = hash(sym->s_name);
1592 if ((sym->s_link = symtab[h]) != NULL)
1593 symtab[h]->s_rlink = &sym->s_link;
1594 sym->s_rlink = &symtab[h];
1595 symtab[h] = sym;
1596 sym->s_blklev = bl;
1597 if (sym->s_link != NULL && sym->s_blklev < sym->s_link->s_blklev)
1598 LERROR("inssym()");
1599 }
1600
1601 /*
1602 * Called at level 0 after syntax errors
1603 * Removes all symbols which are not declared at level 0 from the
1604 * symbol table. Also frees all memory which is not associated with
1605 * level 0.
1606 */
1607 void
1608 cleanup(void)
1609 {
1610 sym_t *sym, *nsym;
1611 int i;
1612
1613 for (i = 0; i < HSHSIZ1; i++) {
1614 for (sym = symtab[i]; sym != NULL; sym = nsym) {
1615 nsym = sym->s_link;
1616 if (sym->s_blklev >= 1) {
1617 if ((*sym->s_rlink = nsym) != NULL)
1618 nsym->s_rlink = sym->s_rlink;
1619 }
1620 }
1621 }
1622
1623 for (i = mblklev; i > 0; i--)
1624 freelblk(i);
1625 }
1626
1627 /*
1628 * Create a new symbol with the name of an existing symbol.
1629 */
1630 sym_t *
1631 pushdown(sym_t *sym)
1632 {
1633 int h;
1634 sym_t *nsym;
1635
1636 h = hash(sym->s_name);
1637 nsym = getblk(sizeof (sym_t));
1638 if (sym->s_blklev > blklev)
1639 LERROR("pushdown()");
1640 nsym->s_name = sym->s_name;
1641 UNIQUE_CURR_POS(nsym->s_dpos);
1642 nsym->s_kind = sym->s_kind;
1643 nsym->s_blklev = blklev;
1644
1645 if ((nsym->s_link = symtab[h]) != NULL)
1646 symtab[h]->s_rlink = &nsym->s_link;
1647 nsym->s_rlink = &symtab[h];
1648 symtab[h] = nsym;
1649
1650 *dcs->d_ldlsym = nsym;
1651 dcs->d_ldlsym = &nsym->s_dlnxt;
1652
1653 return (nsym);
1654 }
1655
1656 /*
1657 * Free any dynamically allocated memory referenced by
1658 * the value stack or yylval.
1659 * The type of information in yylval is described by tok.
1660 */
1661 void
1662 freeyyv(void *sp, int tok)
1663 {
1664 if (tok == T_NAME || tok == T_TYPENAME) {
1665 sbuf_t *sb = *(sbuf_t **)sp;
1666 freesb(sb);
1667 } else if (tok == T_CON) {
1668 val_t *val = *(val_t **)sp;
1669 free(val);
1670 } else if (tok == T_STRING) {
1671 strg_t *strg = *(strg_t **)sp;
1672 if (strg->st_tspec == CHAR) {
1673 free(strg->st_cp);
1674 } else if (strg->st_tspec == WCHAR) {
1675 free(strg->st_wcp);
1676 } else {
1677 LERROR("fryylv()");
1678 }
1679 free(strg);
1680 }
1681 }
1682