scan.l revision 204872
1%{ 2/* $NetBSD: scan.l,v 1.26 2002/01/31 22:30:21 tv 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 1.26 2002/01/31 22:30:21 tv Exp $"); 39#endif 40__FBSDID("$FreeBSD: head/usr.bin/xlint/lint1/scan.l 204872 2010-03-08 16:23:32Z jh $"); 41 42#include <stdlib.h> 43#include <string.h> 44#include <limits.h> 45#include <float.h> 46#include <ctype.h> 47#include <errno.h> 48#include <err.h> 49#include <math.h> 50 51#include "lint1.h" 52#include "cgram.h" 53 54#define CHAR_MASK (~(~0 << CHAR_BIT)) 55#define YY_NO_UNPUT 56 57/* Current position (its also updated when an included file is parsed) */ 58pos_t curr_pos = { 1, "", 0 }; 59 60/* 61 * Current position in C source (not updated when an included file is 62 * parsed). 63 */ 64pos_t csrc_pos = { 1, "", 0 }; 65 66static void incline(void); 67static void badchar(int); 68static sbuf_t *allocsb(void); 69static void freesb(sbuf_t *); 70static int inpc(void); 71static int hash(const char *); 72static sym_t *search(sbuf_t *); 73static int name(void); 74static int keyw(sym_t *); 75static int icon(int); 76static int fcon(void); 77static int operator(int, op_t); 78static int ccon(void); 79static int wccon(void); 80static int getescc(int); 81static void directive(void); 82static void comment(void); 83static void slashslashcomment(void); 84static int string(void); 85static int wcstrg(void); 86 87%} 88 89L [_A-Za-z] 90D [0-9] 91NZD [1-9] 92OD [0-7] 93HD [0-9A-Fa-f] 94EX ([eE][+-]?[0-9]+) 95 96%% 97 98{L}({L}|{D})* return (name()); 990{OD}*[lLuU]* return (icon(8)); 100{NZD}{D}*[lLuU]* return (icon(10)); 1010[xX]{HD}+[lLuU]* return (icon(16)); 102{D}+\.{D}*{EX}?[fFlL]? | 103{D}+{EX}[fFlL]? | 104\.{D}+{EX}?[fFlL]? return (fcon()); 105"=" return (operator(T_ASSIGN, ASSIGN)); 106"*=" return (operator(T_OPASS, MULASS)); 107"/=" return (operator(T_OPASS, DIVASS)); 108"%=" return (operator(T_OPASS, MODASS)); 109"+=" return (operator(T_OPASS, ADDASS)); 110"-=" return (operator(T_OPASS, SUBASS)); 111"<<=" return (operator(T_OPASS, SHLASS)); 112">>=" return (operator(T_OPASS, SHRASS)); 113"&=" return (operator(T_OPASS, ANDASS)); 114"^=" return (operator(T_OPASS, XORASS)); 115"|=" return (operator(T_OPASS, ORASS)); 116"||" return (operator(T_LOGOR, LOGOR)); 117"&&" return (operator(T_LOGAND, LOGAND)); 118"|" return (operator(T_OR, OR)); 119"&" return (operator(T_AND, AND)); 120"^" return (operator(T_XOR, XOR)); 121"==" return (operator(T_EQOP, EQ)); 122"!=" return (operator(T_EQOP, NE)); 123"<" return (operator(T_RELOP, LT)); 124">" return (operator(T_RELOP, GT)); 125"<=" return (operator(T_RELOP, LE)); 126">=" return (operator(T_RELOP, GE)); 127"<<" return (operator(T_SHFTOP, SHL)); 128">>" return (operator(T_SHFTOP, SHR)); 129"++" return (operator(T_INCDEC, INC)); 130"--" return (operator(T_INCDEC, DEC)); 131"->" return (operator(T_STROP, ARROW)); 132"." return (operator(T_STROP, POINT)); 133"+" return (operator(T_ADDOP, PLUS)); 134"-" return (operator(T_ADDOP, MINUS)); 135"*" return (operator(T_MULT, MULT)); 136"/" return (operator(T_DIVOP, DIV)); 137"%" return (operator(T_DIVOP, MOD)); 138"!" return (operator(T_UNOP, NOT)); 139"~" return (operator(T_UNOP, COMPL)); 140"\"" return (string()); 141"L\"" return (wcstrg()); 142";" return (T_SEMI); 143"{" return (T_LBRACE); 144"}" return (T_RBRACE); 145"," return (T_COMMA); 146":" return (T_COLON); 147"?" return (T_QUEST); 148"[" return (T_LBRACK); 149"]" return (T_RBRACK); 150"(" return (T_LPARN); 151")" return (T_RPARN); 152"..." return (T_ELLIPSE); 153"'" return (ccon()); 154"L'" return (wccon()); 155^#.*$ directive(); 156\n incline(); 157\t|" "|\f|\v ; 158"/*" comment(); 159"//" slashslashcomment(); 160. badchar(yytext[0]); 161 162%% 163 164static void 165incline(void) 166{ 167 curr_pos.p_line++; 168 curr_pos.p_uniq = 0; 169 if (curr_pos.p_file == csrc_pos.p_file) { 170 csrc_pos.p_line++; 171 csrc_pos.p_uniq = 0; 172 } 173} 174 175static void 176badchar(int c) 177{ 178 179 /* unknown character \%o */ 180 error(250, c); 181} 182 183/* 184 * Keywords. 185 * During initialisation they are written to the symbol table. 186 */ 187static struct kwtab { 188 const char *kw_name; /* keyword */ 189 int kw_token; /* token returned by yylex() */ 190 scl_t kw_scl; /* storage class if kw_token T_SCLASS */ 191 tspec_t kw_tspec; /* type spec. if kw_token T_TYPE or T_SOU */ 192 tqual_t kw_tqual; /* type qual. fi kw_token T_QUAL */ 193 u_int kw_stdc : 1; /* STDC keyword */ 194 u_int kw_gcc : 1; /* GCC keyword */ 195} kwtab[] = { 196 { "asm", T_ASM, 0, 0, 0, 0, 1 }, 197 { "__asm", T_ASM, 0, 0, 0, 0, 0 }, 198 { "__asm__", T_ASM, 0, 0, 0, 0, 0 }, 199 { "auto", T_SCLASS, AUTO, 0, 0, 0, 0 }, 200 { "break", T_BREAK, 0, 0, 0, 0, 0 }, 201 { "case", T_CASE, 0, 0, 0, 0, 0 }, 202 { "char", T_TYPE, 0, CHAR, 0, 0, 0 }, 203 { "const", T_QUAL, 0, 0, CONST, 1, 0 }, 204 { "__const__", T_QUAL, 0, 0, CONST, 0, 0 }, 205 { "__const", T_QUAL, 0, 0, CONST, 0, 0 }, 206 { "continue", T_CONTINUE, 0, 0, 0, 0, 0 }, 207 { "default", T_DEFAULT, 0, 0, 0, 0, 0 }, 208 { "do", T_DO, 0, 0, 0, 0, 0 }, 209 { "double", T_TYPE, 0, DOUBLE, 0, 0, 0 }, 210 { "else", T_ELSE, 0, 0, 0, 0, 0 }, 211 { "enum", T_ENUM, 0, 0, 0, 0, 0 }, 212 { "extern", T_SCLASS, EXTERN, 0, 0, 0, 0 }, 213 { "float", T_TYPE, 0, FLOAT, 0, 0, 0 }, 214 { "for", T_FOR, 0, 0, 0, 0, 0 }, 215 { "goto", T_GOTO, 0, 0, 0, 0, 0 }, 216 { "if", T_IF, 0, 0, 0, 0, 0 }, 217 { "inline", T_SCLASS, INLINE, 0, 0, 0, 1 }, 218 { "__inline__", T_SCLASS, INLINE, 0, 0, 0, 0 }, 219 { "__inline", T_SCLASS, INLINE, 0, 0, 0, 0 }, 220 { "int", T_TYPE, 0, INT, 0, 0, 0 }, 221 { "__symbolrename", T_SYMBOLRENAME, 0, 0, 0, 0, 0 }, 222 { "long", T_TYPE, 0, LONG, 0, 0, 0 }, 223 { "register", T_SCLASS, REG, 0, 0, 0, 0 }, 224 { "return", T_RETURN, 0, 0, 0, 0, 0 }, 225 { "short", T_TYPE, 0, SHORT, 0, 0, 0 }, 226 { "signed", T_TYPE, 0, SIGNED, 0, 1, 0 }, 227 { "__signed__", T_TYPE, 0, SIGNED, 0, 0, 0 }, 228 { "__signed", T_TYPE, 0, SIGNED, 0, 0, 0 }, 229 { "sizeof", T_SIZEOF, 0, 0, 0, 0, 0 }, 230 { "static", T_SCLASS, STATIC, 0, 0, 0, 0 }, 231 { "struct", T_SOU, 0, STRUCT, 0, 0, 0 }, 232 { "switch", T_SWITCH, 0, 0, 0, 0, 0 }, 233 { "typedef", T_SCLASS, TYPEDEF, 0, 0, 0, 0 }, 234 { "union", T_SOU, 0, UNION, 0, 0, 0 }, 235 { "unsigned", T_TYPE, 0, UNSIGN, 0, 0, 0 }, 236 { "void", T_TYPE, 0, VOID, 0, 0, 0 }, 237 { "volatile", T_QUAL, 0, 0, VOLATILE, 1, 0 }, 238 { "__volatile__", T_QUAL, 0, 0, VOLATILE, 0, 0 }, 239 { "__volatile", T_QUAL, 0, 0, VOLATILE, 0, 0 }, 240 { "while", T_WHILE, 0, 0, 0, 0, 0 }, 241 { NULL, 0, 0, 0, 0, 0, 0 } 242}; 243 244/* Symbol table */ 245static sym_t *symtab[HSHSIZ1]; 246 247/* bit i of the entry with index i is set */ 248uint64_t qbmasks[sizeof(uint64_t) * CHAR_BIT]; 249 250/* least significant i bits are set in the entry with index i */ 251uint64_t qlmasks[sizeof(uint64_t) * CHAR_BIT + 1]; 252 253/* least significant i bits are not set in the entry with index i */ 254uint64_t qumasks[sizeof(uint64_t) * CHAR_BIT + 1]; 255 256/* free list for sbuf structures */ 257static sbuf_t *sbfrlst; 258 259/* Typ of next expected symbol */ 260symt_t symtyp; 261 262 263/* 264 * All keywords are written to the symbol table. This saves us looking 265 * in an extra table for each name we found. 266 */ 267void 268initscan(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_stdc && tflag) 277 continue; 278 if (kw->kw_gcc && !gflag) 279 continue; 280 sym = getblk(sizeof (sym_t)); 281 sym->s_name = kw->kw_name; 282 sym->s_keyw = 1; 283 sym->s_value.v_quad = kw->kw_token; 284 if (kw->kw_token == T_TYPE || kw->kw_token == T_SOU) { 285 sym->s_tspec = kw->kw_tspec; 286 } else if (kw->kw_token == T_SCLASS) { 287 sym->s_scl = kw->kw_scl; 288 } else if (kw->kw_token == T_QUAL) { 289 sym->s_tqual = kw->kw_tqual; 290 } 291 h = hash(sym->s_name); 292 if ((sym->s_link = symtab[h]) != NULL) 293 symtab[h]->s_rlink = &sym->s_link; 294 (symtab[h] = sym)->s_rlink = &symtab[h]; 295 } 296 297 /* initialize bit-masks for quads */ 298 for (i = 0; i < sizeof (uint64_t) * CHAR_BIT; i++) { 299 qbmasks[i] = (uint64_t)1 << i; 300 uq = ~(uint64_t)0 << i; 301 qumasks[i] = uq; 302 qlmasks[i] = ~uq; 303 } 304 qumasks[i] = 0; 305 qlmasks[i] = ~(uint64_t)0; 306} 307 308/* 309 * Get a free sbuf structure, if possible from the free list 310 */ 311static sbuf_t * 312allocsb(void) 313{ 314 sbuf_t *sb; 315 316 if ((sb = sbfrlst) != NULL) { 317 sbfrlst = sb->sb_nxt; 318 } else { 319 if ((sb = malloc(sizeof (sbuf_t))) == NULL) 320 nomem(); 321 } 322 (void)memset(sb, 0, sizeof (*sb)); 323 return (sb); 324} 325 326/* 327 * Put a sbuf structure to the free list 328 */ 329static void 330freesb(sbuf_t *sb) 331{ 332 333 sb->sb_nxt = sbfrlst; 334 sbfrlst = sb; 335} 336 337/* 338 * Read a character and ensure that it is positive (except EOF). 339 * Increment line count(s) if necessary. 340 */ 341static int 342inpc(void) 343{ 344 int c; 345 346 if ((c = input()) != EOF && (c &= CHAR_MASK) == '\n') 347 incline(); 348 return (c); 349} 350 351static int 352hash(const char *s) 353{ 354 u_int v; 355 const u_char *us; 356 357 v = 0; 358 for (us = (const u_char *)s; *us != '\0'; us++) { 359 v = (v << sizeof (v)) + *us; 360 v ^= v >> (sizeof (v) * CHAR_BIT - sizeof (v)); 361 } 362 return (v % HSHSIZ1); 363} 364 365/* 366 * Lex has found a letter followed by zero or more letters or digits. 367 * It looks for a symbol in the symbol table with the same name. This 368 * symbol must either be a keyword or a symbol of the type required by 369 * symtyp (label, member, tag, ...). 370 * 371 * If it is a keyword, the token is returned. In some cases it is described 372 * more deeply by data written to yylval. 373 * 374 * If it is a symbol, T_NAME is returned and the pointer to a sbuf struct 375 * is stored in yylval. This struct contains the name of the symbol, it's 376 * length and hash value. If there is already a symbol of the same name 377 * and type in the symbol table, the sbuf struct also contains a pointer 378 * to the symbol table entry. 379 */ 380static int 381name(void) 382{ 383 char *s; 384 sbuf_t *sb; 385 sym_t *sym; 386 int tok; 387 388 sb = allocsb(); 389 sb->sb_name = yytext; 390 sb->sb_len = yyleng; 391 sb->sb_hash = hash(yytext); 392 393 if ((sym = search(sb)) != NULL && sym->s_keyw) { 394 freesb(sb); 395 return (keyw(sym)); 396 } 397 398 sb->sb_sym = sym; 399 400 if (sym != NULL) { 401 if (blklev < sym->s_blklev) 402 lerror("name() 1"); 403 sb->sb_name = sym->s_name; 404 sb->sb_len = strlen(sym->s_name); 405 tok = sym->s_scl == TYPEDEF ? T_TYPENAME : T_NAME; 406 } else { 407 s = getblk(yyleng + 1); 408 (void)memcpy(s, yytext, yyleng + 1); 409 sb->sb_name = s; 410 sb->sb_len = yyleng; 411 tok = T_NAME; 412 } 413 414 yylval.y_sb = sb; 415 return (tok); 416} 417 418static sym_t * 419search(sbuf_t *sb) 420{ 421 sym_t *sym; 422 423 for (sym = symtab[sb->sb_hash]; sym != NULL; sym = sym->s_link) { 424 if (strcmp(sym->s_name, sb->sb_name) == 0) { 425 if (sym->s_keyw || sym->s_kind == symtyp) 426 return (sym); 427 } 428 } 429 430 return (NULL); 431} 432 433static int 434keyw(sym_t *sym) 435{ 436 int t; 437 438 if ((t = (int)sym->s_value.v_quad) == T_SCLASS) { 439 yylval.y_scl = sym->s_scl; 440 } else if (t == T_TYPE || t == T_SOU) { 441 yylval.y_tspec = sym->s_tspec; 442 } else if (t == T_QUAL) { 443 yylval.y_tqual = sym->s_tqual; 444 } 445 return (t); 446} 447 448/* 449 * Convert a string representing an integer into internal representation. 450 * The value is returned in yylval. icon() (and yylex()) returns T_CON. 451 */ 452static int 453icon(int base) 454{ 455 int l_suffix, u_suffix; 456 int len; 457 const char *cp; 458 char c, *eptr; 459 tspec_t typ; 460 u_long ul = 0; 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 if (l_suffix < 2) { 505 ul = strtoul(cp, &eptr, base); 506 } else { 507 uq = strtouq(cp, &eptr, base); 508 } 509 if (eptr != cp + len) 510 lerror("icon() 1"); 511 if (errno != 0) 512 /* integer constant out of range */ 513 warning(252); 514 515 /* 516 * If the value is to big for the current type, we must choose 517 * another type. 518 */ 519 ansiu = 0; 520 switch (typ) { 521 case INT: 522 if (ul <= INT_MAX) { 523 /* ok */ 524 } else if (ul <= (unsigned)UINT_MAX && base != 10) { 525 typ = UINT; 526 } else if (ul <= LONG_MAX) { 527 typ = LONG; 528 } else { 529 typ = ULONG; 530 } 531 if (typ == UINT || typ == ULONG) { 532 if (tflag) { 533 typ = LONG; 534 } else if (!sflag) { 535 /* 536 * Remember that the constant is unsigned 537 * only in ANSI C 538 */ 539 ansiu = 1; 540 } 541 } 542 break; 543 case UINT: 544 if (ul > (u_int)UINT_MAX) 545 typ = ULONG; 546 break; 547 case LONG: 548 if (ul > LONG_MAX && !tflag) { 549 typ = ULONG; 550 if (!sflag) 551 ansiu = 1; 552 } 553 break; 554 case QUAD: 555 if (uq > QUAD_MAX && !tflag) { 556 typ = UQUAD; 557 if (!sflag) 558 ansiu = 1; 559 } 560 break; 561 /* LINTED (enumeration values not handled in switch) */ 562 case STRUCT: 563 case VOID: 564 case LDOUBLE: 565 case FUNC: 566 case ARRAY: 567 case PTR: 568 case ENUM: 569 case UNION: 570 case SIGNED: 571 case NOTSPEC: 572 case DOUBLE: 573 case FLOAT: 574 case UQUAD: 575 case ULONG: 576 case USHORT: 577 case SHORT: 578 case UCHAR: 579 case SCHAR: 580 case CHAR: 581 case UNSIGN: 582 break; 583 584 case NTSPEC: /* this value unused */ 585 break; 586 } 587 588 if (typ != QUAD && typ != UQUAD) { 589 if (isutyp(typ)) { 590 uq = ul; 591 } else { 592 uq = (int64_t)(long)ul; 593 } 594 } 595 596 uq = (uint64_t)xsign((int64_t)uq, typ, -1); 597 598 if ((yylval.y_val = calloc(1, sizeof(val_t))) == NULL) 599 nomem(); 600 yylval.y_val->v_tspec = typ; 601 yylval.y_val->v_ansiu = ansiu; 602 yylval.y_val->v_quad = (int64_t)uq; 603 604 return (T_CON); 605} 606 607/* 608 * Returns 1 if t is a signed type and the value is negative. 609 * 610 * len is the number of significant bits. If len is -1, len is set 611 * to the width of type t. 612 */ 613int 614sign(int64_t q, tspec_t t, int len) 615{ 616 617 if (t == PTR || isutyp(t)) 618 return (0); 619 return (msb(q, t, len)); 620} 621 622int 623msb(int64_t q, tspec_t t, int len) 624{ 625 626 if (len <= 0) 627 len = size(t); 628 return ((q & qbmasks[len - 1]) != 0); 629} 630 631/* 632 * Extends the sign of q. 633 */ 634int64_t 635xsign(int64_t q, tspec_t t, int len) 636{ 637 638 if (len <= 0) 639 len = size(t); 640 641 if (t == PTR || isutyp(t) || !sign(q, t, len)) { 642 q &= qlmasks[len]; 643 } else { 644 q |= qumasks[len]; 645 } 646 return (q); 647} 648 649/* 650 * Convert a string representing a floating point value into its interal 651 * representation. Type and value are returned in yylval. fcon() 652 * (and yylex()) returns T_CON. 653 * XXX Currently it is not possible to convert constants of type 654 * long double which are greater than DBL_MAX. 655 */ 656static int 657fcon(void) 658{ 659 const char *cp; 660 int len; 661 tspec_t typ; 662 char c, *eptr; 663 double d; 664 float f = 0; 665 666 cp = yytext; 667 len = yyleng; 668 669 if ((c = cp[len - 1]) == 'f' || c == 'F') { 670 typ = FLOAT; 671 len--; 672 } else if (c == 'l' || c == 'L') { 673 typ = LDOUBLE; 674 len--; 675 } else { 676 typ = DOUBLE; 677 } 678 679 if (tflag && typ != DOUBLE) { 680 /* suffixes F and L are illegal in traditional C */ 681 warning(98); 682 } 683 684 errno = 0; 685 d = strtod(cp, &eptr); 686 if (eptr != cp + len) 687 lerror("fcon() 1"); 688 if (errno != 0) 689 /* floating-point constant out of range */ 690 warning(248); 691 692 if (typ == FLOAT) { 693 f = (float)d; 694 if (!finite(f)) { 695 /* floating-point constant out of range */ 696 warning(248); 697 f = f > 0 ? FLT_MAX : -FLT_MAX; 698 } 699 } 700 701 if ((yylval.y_val = calloc(1, sizeof (val_t))) == NULL) 702 nomem(); 703 yylval.y_val->v_tspec = typ; 704 if (typ == FLOAT) { 705 yylval.y_val->v_ldbl = f; 706 } else { 707 yylval.y_val->v_ldbl = d; 708 } 709 710 return (T_CON); 711} 712 713static int 714operator(int t, op_t o) 715{ 716 717 yylval.y_op = o; 718 return (t); 719} 720 721/* 722 * Called if lex found a leading \'. 723 */ 724static int 725ccon(void) 726{ 727 int n, val, c; 728 char cv; 729 730 n = 0; 731 val = 0; 732 while ((c = getescc('\'')) >= 0) { 733 val = (val << CHAR_BIT) + c; 734 n++; 735 } 736 if (c == -2) { 737 /* unterminated character constant */ 738 error(253); 739 } else { 740 if (n > sizeof (int) || (n > 1 && (pflag || hflag))) { 741 /* too many characters in character constant */ 742 error(71); 743 } else if (n > 1) { 744 /* multi-character character constant */ 745 warning(294); 746 } else if (n == 0) { 747 /* empty character constant */ 748 error(73); 749 } 750 } 751 if (n == 1) { 752 cv = (char)val; 753 val = cv; 754 } 755 756 yylval.y_val = xcalloc(1, sizeof (val_t)); 757 yylval.y_val->v_tspec = INT; 758 yylval.y_val->v_quad = val; 759 760 return (T_CON); 761} 762 763/* 764 * Called if lex found a leading L\' 765 */ 766static int 767wccon(void) 768{ 769 static char buf[MB_LEN_MAX + 1]; 770 int i, c; 771 wchar_t wc; 772 773 i = 0; 774 while ((c = getescc('\'')) >= 0) { 775 if (i < MB_CUR_MAX) 776 buf[i] = (char)c; 777 i++; 778 } 779 780 wc = 0; 781 782 if (c == -2) { 783 /* unterminated character constant */ 784 error(253); 785 } else if (c == 0) { 786 /* empty character constant */ 787 error(73); 788 } else { 789 if (i > MB_CUR_MAX) { 790 i = MB_CUR_MAX; 791 /* too many characters in character constant */ 792 error(71); 793 } else { 794 buf[i] = '\0'; 795 (void)mbtowc(NULL, NULL, 0); 796 if (mbtowc(&wc, buf, MB_CUR_MAX) < 0) 797 /* invalid multibyte character */ 798 error(291); 799 } 800 } 801 802 if ((yylval.y_val = calloc(1, sizeof (val_t))) == NULL) 803 nomem(); 804 yylval.y_val->v_tspec = WCHAR; 805 yylval.y_val->v_quad = wc; 806 807 return (T_CON); 808} 809 810/* 811 * Read a character which is part of a character constant or of a string 812 * and handle escapes. 813 * 814 * The Argument is the character which delimits the character constant or 815 * string. 816 * 817 * Returns -1 if the end of the character constant or string is reached, 818 * -2 if the EOF is reached, and the character otherwise. 819 */ 820static int 821getescc(int d) 822{ 823 static int pbc = -1; 824 int n, c, v; 825 826 if (pbc == -1) { 827 c = inpc(); 828 } else { 829 c = pbc; 830 pbc = -1; 831 } 832 if (c == d) 833 return (-1); 834 switch (c) { 835 case '\n': 836 if (tflag) { 837 /* newline in string or char constant */ 838 error(254); 839 return (-2); 840 } 841 return (c); 842 case EOF: 843 return (-2); 844 case '\\': 845 switch (c = inpc()) { 846 case '"': 847 if (tflag && d == '\'') 848 /* \" inside character constant undef. ... */ 849 warning(262); 850 return ('"'); 851 case '\'': 852 return ('\''); 853 case '?': 854 if (tflag) 855 /* \? undefined in traditional C */ 856 warning(263); 857 return ('?'); 858 case '\\': 859 return ('\\'); 860 case 'a': 861 if (tflag) 862 /* \a undefined in traditional C */ 863 warning(81); 864 return ('\a'); 865 case 'b': 866 return ('\b'); 867 case 'f': 868 return ('\f'); 869 case 'n': 870 return ('\n'); 871 case 'r': 872 return ('\r'); 873 case 't': 874 return ('\t'); 875 case 'v': 876 if (tflag) 877 /* \v undefined in traditional C */ 878 warning(264); 879 return ('\v'); 880 case '8': case '9': 881 /* bad octal digit %c */ 882 warning(77, c); 883 /* FALLTHROUGH */ 884 case '0': case '1': case '2': case '3': 885 case '4': case '5': case '6': case '7': 886 n = 3; 887 v = 0; 888 do { 889 v = (v << 3) + (c - '0'); 890 c = inpc(); 891 } while (--n && isdigit(c) && (tflag || c <= '7')); 892 if (tflag && n > 0 && isdigit(c)) 893 /* bad octal digit %c */ 894 warning(77, c); 895 pbc = c; 896 if (v > UCHAR_MAX) { 897 /* character escape does not fit in char. */ 898 warning(76); 899 v &= CHAR_MASK; 900 } 901 return (v); 902 case 'x': 903 if (tflag) 904 /* \x undefined in traditional C */ 905 warning(82); 906 v = 0; 907 n = 0; 908 while ((c = inpc()) >= 0 && isxdigit(c)) { 909 c = isdigit(c) ? 910 c - '0' : toupper(c) - 'A' + 10; 911 v = (v << 4) + c; 912 if (n >= 0) { 913 if ((v & ~CHAR_MASK) != 0) { 914 /* overflow in hex escape */ 915 warning(75); 916 n = -1; 917 } else { 918 n++; 919 } 920 } 921 } 922 pbc = c; 923 if (n == 0) { 924 /* no hex digits follow \x */ 925 error(74); 926 } if (n == -1) { 927 v &= CHAR_MASK; 928 } 929 return (v); 930 case '\n': 931 return (getescc(d)); 932 case EOF: 933 return (-2); 934 default: 935 if (isprint(c)) { 936 /* dubious escape \%c */ 937 warning(79, c); 938 } else { 939 /* dubious escape \%o */ 940 warning(80, c); 941 } 942 } 943 } 944 return (c); 945} 946 947/* 948 * Called for preprocessor directives. Currently implemented are: 949 * # lineno 950 * # lineno "filename" 951 */ 952static void 953directive(void) 954{ 955 const char *cp, *fn; 956 char c, *eptr; 957 size_t fnl; 958 long ln; 959 static int first = 1; 960 961 /* Go to first non-whitespace after # */ 962 for (cp = yytext + 1; (c = *cp) == ' ' || c == '\t'; cp++) 963 continue; 964 965 if (!isdigit((unsigned char)c)) { 966 error: 967 /* undefined or invalid # directive */ 968 warning(255); 969 return; 970 } 971 ln = strtol(--cp, &eptr, 10); 972 if (cp == eptr) 973 goto error; 974 if ((c = *(cp = eptr)) != ' ' && c != '\t' && c != '\0') 975 goto error; 976 while ((c = *cp++) == ' ' || c == '\t') 977 continue; 978 if (c != '\0') { 979 if (c != '"') 980 goto error; 981 fn = cp; 982 while ((c = *cp) != '"' && c != '\0') 983 cp++; 984 if (c != '"') 985 goto error; 986 if ((fnl = cp++ - fn) > PATH_MAX) 987 goto error; 988 while ((c = *cp++) == ' ' || c == '\t') 989 continue; 990#if 0 991 if (c != '\0') 992 warning("extra character(s) after directive"); 993#endif 994 995 /* empty string means stdin */ 996 if (fnl == 0) { 997 fn = "{standard input}"; 998 fnl = 16; /* strlen (fn) */ 999 } 1000 curr_pos.p_file = fnnalloc(fn, fnl); 1001 /* 1002 * If this is the first directive, the name is the name 1003 * of the C source file as specified at the command line. 1004 * It is written to the output file. 1005 */ 1006 if (first) { 1007 csrc_pos.p_file = curr_pos.p_file; 1008 outsrc(curr_pos.p_file); 1009 first = 0; 1010 } 1011 } 1012 curr_pos.p_line = (int)ln - 1; 1013 curr_pos.p_uniq = 0; 1014 if (curr_pos.p_file == csrc_pos.p_file) { 1015 csrc_pos.p_line = (int)ln - 1; 1016 csrc_pos.p_uniq = 0; 1017 } 1018} 1019 1020/* 1021 * Handle lint comments. Following comments are currently understood: 1022 * ARGSUSEDn 1023 * BITFIELDTYPE 1024 * CONSTCOND CONSTANTCOND CONSTANTCONDITION 1025 * FALLTHRU FALLTHROUGH 1026 * LINTLIBRARY 1027 * LINTED NOSTRICT 1028 * LONGLONG 1029 * NOTREACHED 1030 * PRINTFLIKEn 1031 * PROTOLIB 1032 * SCANFLIKEn 1033 * VARARGSn 1034 * If one of this comments is recognized, the arguments, if any, are 1035 * parsed and a function which handles this comment is called. 1036 */ 1037static void 1038comment(void) 1039{ 1040 int c, lc; 1041 static struct { 1042 const char *keywd; 1043 int arg; 1044 void (*func)(int); 1045 } keywtab[] = { 1046 { "ARGSUSED", 1, argsused }, 1047 { "BITFIELDTYPE", 0, bitfieldtype }, 1048 { "CONSTCOND", 0, constcond }, 1049 { "CONSTANTCOND", 0, constcond }, 1050 { "CONSTANTCONDITION", 0, constcond }, 1051 { "FALLTHRU", 0, fallthru }, 1052 { "FALLTHROUGH", 0, fallthru }, 1053 { "LINTLIBRARY", 0, lintlib }, 1054 { "LINTED", 0, linted }, 1055 { "LONGLONG", 0, longlong }, 1056 { "NOSTRICT", 0, linted }, 1057 { "NOTREACHED", 0, notreach }, 1058 { "PRINTFLIKE", 1, printflike }, 1059 { "PROTOLIB", 1, protolib }, 1060 { "SCANFLIKE", 1, scanflike }, 1061 { "VARARGS", 1, varargs }, 1062 }; 1063 char keywd[32]; 1064 char arg[32]; 1065 int l, i, a; 1066 int eoc; 1067 1068 eoc = 0; 1069 1070 /* Skip white spaces after the start of the comment */ 1071 while ((c = inpc()) != EOF && isspace(c)) 1072 continue; 1073 1074 /* Read the potential keyword to keywd */ 1075 l = 0; 1076 while (c != EOF && isupper(c) && l < sizeof (keywd) - 1) { 1077 keywd[l++] = (char)c; 1078 c = inpc(); 1079 } 1080 keywd[l] = '\0'; 1081 1082 /* look for the keyword */ 1083 for (i = 0; i < sizeof (keywtab) / sizeof (keywtab[0]); i++) { 1084 if (strcmp(keywtab[i].keywd, keywd) == 0) 1085 break; 1086 } 1087 if (i == sizeof (keywtab) / sizeof (keywtab[0])) 1088 goto skip_rest; 1089 1090 /* skip white spaces after the keyword */ 1091 while (c != EOF && isspace(c)) 1092 c = inpc(); 1093 1094 /* read the argument, if the keyword accepts one and there is one */ 1095 l = 0; 1096 if (keywtab[i].arg) { 1097 while (c != EOF && isdigit(c) && l < sizeof (arg) - 1) { 1098 arg[l++] = (char)c; 1099 c = inpc(); 1100 } 1101 } 1102 arg[l] = '\0'; 1103 a = l != 0 ? atoi(arg) : -1; 1104 1105 /* skip white spaces after the argument */ 1106 while (c != EOF && isspace(c)) 1107 c = inpc(); 1108 1109 if (c != '*' || (c = inpc()) != '/') { 1110 if (keywtab[i].func != linted) 1111 /* extra characters in lint comment */ 1112 warning(257); 1113 } else { 1114 /* 1115 * remember that we have already found the end of the 1116 * comment 1117 */ 1118 eoc = 1; 1119 } 1120 1121 if (keywtab[i].func != NULL) 1122 (*keywtab[i].func)(a); 1123 1124 skip_rest: 1125 while (!eoc) { 1126 lc = c; 1127 if ((c = inpc()) == EOF) { 1128 /* unterminated comment */ 1129 error(256); 1130 break; 1131 } 1132 if (lc == '*' && c == '/') 1133 eoc = 1; 1134 } 1135} 1136 1137/* 1138 * Handle // style comments 1139 */ 1140static void 1141slashslashcomment(void) 1142{ 1143 int c; 1144 1145 if (sflag < 2 && !gflag) 1146 /* // comments only supported in C99 */ 1147 (void)gnuism(312, tflag ? "traditional" : "ANSI"); 1148 1149 while ((c = inpc()) != EOF && c != '\n') 1150 continue; 1151} 1152 1153/* 1154 * Clear flags for lint comments LINTED, LONGLONG and CONSTCOND. 1155 * clrwflgs() is called after function definitions and global and 1156 * local declarations and definitions. It is also called between 1157 * the controlling expression and the body of control statements 1158 * (if, switch, for, while). 1159 */ 1160void 1161clrwflgs(void) 1162{ 1163 1164 nowarn = 0; 1165 quadflg = 0; 1166 ccflg = 0; 1167} 1168 1169/* 1170 * Strings are stored in a dynamically alloceted buffer and passed 1171 * in yylval.y_xstrg to the parser. The parser or the routines called 1172 * by the parser are responsible for freeing this buffer. 1173 */ 1174static int 1175string(void) 1176{ 1177 u_char *s; 1178 int c; 1179 size_t len, max; 1180 strg_t *strg; 1181 1182 if ((s = malloc(max = 64)) == NULL) 1183 nomem(); 1184 1185 len = 0; 1186 while ((c = getescc('"')) >= 0) { 1187 /* +1 to reserve space for a trailing NUL character */ 1188 if (len + 1 == max) 1189 if ((s = realloc(s, max *= 2)) == NULL) 1190 nomem(); 1191 s[len++] = (char)c; 1192 } 1193 s[len] = '\0'; 1194 if (c == -2) 1195 /* unterminated string constant */ 1196 error(258); 1197 1198 if ((strg = calloc(1, sizeof (strg_t))) == NULL) 1199 nomem(); 1200 strg->st_tspec = CHAR; 1201 strg->st_len = len; 1202 strg->st_cp = s; 1203 1204 yylval.y_strg = strg; 1205 return (T_STRING); 1206} 1207 1208static int 1209wcstrg(void) 1210{ 1211 char *s; 1212 int c, i, n, wi; 1213 size_t len, max, wlen; 1214 wchar_t *ws; 1215 strg_t *strg; 1216 1217 if ((s = malloc(max = 64)) == NULL) 1218 nomem(); 1219 len = 0; 1220 while ((c = getescc('"')) >= 0) { 1221 /* +1 to save space for a trailing NUL character */ 1222 if (len + 1 >= max) 1223 if ((s = realloc(s, max *= 2)) == NULL) 1224 nomem(); 1225 s[len++] = (char)c; 1226 } 1227 s[len] = '\0'; 1228 if (c == -2) 1229 /* unterminated string constant */ 1230 error(258); 1231 1232 /* get length of wide character string */ 1233 (void)mblen(NULL, 0); 1234 for (i = 0, wlen = 0; i < len; i += n, wlen++) { 1235 if ((n = mblen(&s[i], MB_CUR_MAX)) == -1) { 1236 /* invalid multibyte character */ 1237 error(291); 1238 break; 1239 } 1240 if (n == 0) 1241 n = 1; 1242 } 1243 1244 if ((ws = malloc((wlen + 1) * sizeof (wchar_t))) == NULL) 1245 nomem(); 1246 1247 /* convert from multibyte to wide char */ 1248 (void)mbtowc(NULL, NULL, 0); 1249 for (i = 0, wi = 0; i < len; i += n, wi++) { 1250 if ((n = mbtowc(&ws[wi], &s[i], MB_CUR_MAX)) == -1) 1251 break; 1252 if (n == 0) 1253 n = 1; 1254 } 1255 ws[wi] = 0; 1256 free(s); 1257 1258 if ((strg = calloc(1, sizeof (strg_t))) == NULL) 1259 nomem(); 1260 strg->st_tspec = WCHAR; 1261 strg->st_len = wlen; 1262 strg->st_wcp = ws; 1263 1264 yylval.y_strg = strg; 1265 return (T_STRING); 1266} 1267 1268/* 1269 * As noted above the scanner does not create new symbol table entries 1270 * for symbols it cannot find in the symbol table. This is to avoid 1271 * putting undeclared symbols into the symbol table if a syntax error 1272 * occurs. 1273 * 1274 * getsym() is called as soon as it is probably ok to put the symbol to 1275 * the symbol table. This does not mean that it is not possible that 1276 * symbols are put to the symbol table which are than not completely 1277 * declared due to syntax errors. To avoid too many problems in this 1278 * case symbols get type int in getsym(). 1279 * 1280 * XXX calls to getsym() should be delayed until decl1*() is called 1281 */ 1282sym_t * 1283getsym(sbuf_t *sb) 1284{ 1285 dinfo_t *di; 1286 char *s; 1287 sym_t *sym; 1288 1289 sym = sb->sb_sym; 1290 1291 /* 1292 * During member declaration it is possible that name() looked 1293 * for symbols of type FVFT, although it should have looked for 1294 * symbols of type FTAG. Same can happen for labels. Both cases 1295 * are compensated here. 1296 */ 1297 if (symtyp == FMOS || symtyp == FLAB) { 1298 if (sym == NULL || sym->s_kind == FVFT) 1299 sym = search(sb); 1300 } 1301 1302 if (sym != NULL) { 1303 if (sym->s_kind != symtyp) 1304 lerror("storesym() 1"); 1305 symtyp = FVFT; 1306 freesb(sb); 1307 return (sym); 1308 } 1309 1310 /* create a new symbol table entry */ 1311 1312 /* labels must always be allocated at level 1 (outhermost block) */ 1313 if (symtyp == FLAB) { 1314 sym = getlblk(1, sizeof (sym_t)); 1315 s = getlblk(1, sb->sb_len + 1); 1316 (void)memcpy(s, sb->sb_name, sb->sb_len + 1); 1317 sym->s_name = s; 1318 sym->s_blklev = 1; 1319 di = dcs; 1320 while (di->d_nxt != NULL && di->d_nxt->d_nxt != NULL) 1321 di = di->d_nxt; 1322 if (di->d_ctx != AUTO) 1323 lerror("storesym() 2"); 1324 } else { 1325 sym = getblk(sizeof (sym_t)); 1326 sym->s_name = sb->sb_name; 1327 sym->s_blklev = blklev; 1328 di = dcs; 1329 } 1330 1331 UNIQUE_CURR_POS(sym->s_dpos); 1332 if ((sym->s_kind = symtyp) != FLAB) 1333 sym->s_type = gettyp(INT); 1334 1335 symtyp = FVFT; 1336 1337 if ((sym->s_link = symtab[sb->sb_hash]) != NULL) 1338 symtab[sb->sb_hash]->s_rlink = &sym->s_link; 1339 (symtab[sb->sb_hash] = sym)->s_rlink = &symtab[sb->sb_hash]; 1340 1341 *di->d_ldlsym = sym; 1342 di->d_ldlsym = &sym->s_dlnxt; 1343 1344 freesb(sb); 1345 return (sym); 1346} 1347 1348/* 1349 * Remove a symbol forever from the symbol table. s_blklev 1350 * is set to -1 to avoid that the symbol will later be put 1351 * back to the symbol table. 1352 */ 1353void 1354rmsym(sym_t *sym) 1355{ 1356 1357 if ((*sym->s_rlink = sym->s_link) != NULL) 1358 sym->s_link->s_rlink = sym->s_rlink; 1359 sym->s_blklev = -1; 1360 sym->s_link = NULL; 1361} 1362 1363/* 1364 * Remove a list of symbols declared at one level from the symbol 1365 * table. 1366 */ 1367void 1368rmsyms(sym_t *syms) 1369{ 1370 sym_t *sym; 1371 1372 for (sym = syms; sym != NULL; sym = sym->s_dlnxt) { 1373 if (sym->s_blklev != -1) { 1374 if ((*sym->s_rlink = sym->s_link) != NULL) 1375 sym->s_link->s_rlink = sym->s_rlink; 1376 sym->s_link = NULL; 1377 sym->s_rlink = NULL; 1378 } 1379 } 1380} 1381 1382/* 1383 * Put a symbol into the symbol table 1384 */ 1385void 1386inssym(int bl, sym_t *sym) 1387{ 1388 int h; 1389 1390 h = hash(sym->s_name); 1391 if ((sym->s_link = symtab[h]) != NULL) 1392 symtab[h]->s_rlink = &sym->s_link; 1393 (symtab[h] = sym)->s_rlink = &symtab[h]; 1394 sym->s_blklev = bl; 1395 if (sym->s_link != NULL && sym->s_blklev < sym->s_link->s_blklev) 1396 lerror("inssym()"); 1397} 1398 1399/* 1400 * Called at level 0 after syntax errors 1401 * Removes all symbols which are not declared at level 0 from the 1402 * symbol table. Also frees all memory which is not associated with 1403 * level 0. 1404 */ 1405void 1406cleanup(void) 1407{ 1408 sym_t *sym, *nsym; 1409 int i; 1410 1411 for (i = 0; i < HSHSIZ1; i++) { 1412 for (sym = symtab[i]; sym != NULL; sym = nsym) { 1413 nsym = sym->s_link; 1414 if (sym->s_blklev >= 1) { 1415 if ((*sym->s_rlink = nsym) != NULL) 1416 nsym->s_rlink = sym->s_rlink; 1417 } 1418 } 1419 } 1420 1421 for (i = mblklev; i > 0; i--) 1422 freelblk(i); 1423} 1424 1425/* 1426 * Create a new symbol with the name of an existing symbol. 1427 */ 1428sym_t * 1429pushdown(sym_t *sym) 1430{ 1431 int h; 1432 sym_t *nsym; 1433 1434 h = hash(sym->s_name); 1435 nsym = getblk(sizeof (sym_t)); 1436 if (sym->s_blklev > blklev) 1437 lerror("pushdown()"); 1438 nsym->s_name = sym->s_name; 1439 UNIQUE_CURR_POS(nsym->s_dpos); 1440 nsym->s_kind = sym->s_kind; 1441 nsym->s_blklev = blklev; 1442 1443 if ((nsym->s_link = symtab[h]) != NULL) 1444 symtab[h]->s_rlink = &nsym->s_link; 1445 (symtab[h] = nsym)->s_rlink = &symtab[h]; 1446 1447 *dcs->d_ldlsym = nsym; 1448 dcs->d_ldlsym = &nsym->s_dlnxt; 1449 1450 return (nsym); 1451} 1452 1453/* 1454 * Free any dynamically allocated memory referenced by 1455 * the value stack or yylval. 1456 * The type of information in yylval is described by tok. 1457 */ 1458void 1459freeyyv(void *sp, int tok) 1460{ 1461 if (tok == T_NAME || tok == T_TYPENAME) { 1462 sbuf_t *sb = *(sbuf_t **)sp; 1463 freesb(sb); 1464 } else if (tok == T_CON) { 1465 val_t *val = *(val_t **)sp; 1466 free(val); 1467 } else if (tok == T_STRING) { 1468 strg_t *strg = *(strg_t **)sp; 1469 if (strg->st_tspec == CHAR) { 1470 free(strg->st_cp); 1471 } else if (strg->st_tspec == WCHAR) { 1472 free(strg->st_wcp); 1473 } else { 1474 lerror("fryylv() 1"); 1475 } 1476 free(strg); 1477 } 1478} 1479