1 /*- 2 * Copyright (c) 1992 Diomidis Spinellis. 3 * Copyright (c) 1992 The Regents of the University of California. 4 * All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * Diomidis Spinellis of Imperial College, University of London. 8 * 9 * %sccs.include.redist.c% 10 */ 11 12 #ifndef lint 13 static char sccsid[] = "@(#)compile.c 5.4 (Berkeley) 08/28/92"; 14 #endif /* not lint */ 15 16 #include <sys/types.h> 17 #include <sys/stat.h> 18 19 #include <ctype.h> 20 #include <errno.h> 21 #include <fcntl.h> 22 #include <limits.h> 23 #include <regex.h> 24 #include <stdio.h> 25 #include <stdlib.h> 26 #include <string.h> 27 28 #include "defs.h" 29 #include "extern.h" 30 31 static char *compile_addr __P((char *, struct s_addr *)); 32 static char *compile_delimited __P((char *, char *)); 33 static char *compile_flags __P((char *, struct s_subst *)); 34 static char *compile_re __P((char *, regex_t **)); 35 static char *compile_subst __P((char *, struct s_subst *)); 36 static char *compile_text __P((void)); 37 static char *compile_tr __P((char *, char **)); 38 static struct s_command 39 **compile_stream __P((char *, struct s_command **, char *)); 40 static char *duptoeol __P((char *)); 41 static struct s_command 42 *findlabel __P((struct s_command *, struct s_command *)); 43 static void fixuplabel __P((struct s_command *, struct s_command *)); 44 45 /* 46 * Command specification. This is used to drive the command parser. 47 */ 48 struct s_format { 49 char code; /* Command code */ 50 int naddr; /* Number of address args */ 51 enum e_args args; /* Argument type */ 52 }; 53 54 static struct s_format cmd_fmts[] = { 55 {'{', 2, GROUP}, 56 {'a', 1, TEXT}, 57 {'b', 2, BRANCH}, 58 {'c', 2, TEXT}, 59 {'d', 2, EMPTY}, 60 {'D', 2, EMPTY}, 61 {'g', 2, EMPTY}, 62 {'G', 2, EMPTY}, 63 {'h', 2, EMPTY}, 64 {'H', 2, EMPTY}, 65 {'i', 1, TEXT}, 66 {'l', 2, EMPTY}, 67 {'n', 2, EMPTY}, 68 {'N', 2, EMPTY}, 69 {'p', 2, EMPTY}, 70 {'P', 2, EMPTY}, 71 {'q', 1, EMPTY}, 72 {'r', 1, RFILE}, 73 {'s', 2, SUBST}, 74 {'t', 2, BRANCH}, 75 {'w', 2, WFILE}, 76 {'x', 2, EMPTY}, 77 {'y', 2, TR}, 78 {'!', 2, NONSEL}, 79 {':', 0, LABEL}, 80 {'#', 0, COMMENT}, 81 {'=', 1, EMPTY}, 82 {'\0', 0, COMMENT}, 83 }; 84 85 /* The compiled program. */ 86 struct s_command *prog; 87 88 /* 89 * Compile the program into prog. 90 * Initialise appends. 91 */ 92 void 93 compile() 94 { 95 *compile_stream(NULL, &prog, NULL) = NULL; 96 fixuplabel(prog, prog); 97 appends = xmalloc(sizeof(struct s_appends) * appendnum); 98 match = xmalloc((maxnsub + 1) * sizeof(regmatch_t)); 99 } 100 101 #define EATSPACE() do { \ 102 if (p) \ 103 while (*p && isascii(*p) && isspace(*p)) \ 104 p++; \ 105 } while (0) 106 107 static struct s_command ** 108 compile_stream(terminator, link, p) 109 char *terminator; 110 struct s_command **link; 111 register char *p; 112 { 113 static char lbuf[_POSIX2_LINE_MAX + 1]; /* To save stack */ 114 struct s_command *cmd, *cmd2; 115 struct s_format *fp; 116 int naddr; /* Number of addresses */ 117 118 if (p != NULL) 119 goto semicolon; 120 for (;;) { 121 if ((p = cu_fgets(lbuf, sizeof(lbuf))) == NULL) { 122 if (terminator != NULL) 123 err(COMPILE, "unexpected EOF (pending }'s)"); 124 return (link); 125 } 126 127 semicolon: EATSPACE(); 128 if (p && (*p == '#' || *p == '\0')) 129 continue; 130 if (*p == '}') { 131 if (terminator == NULL) 132 err(COMPILE, "unexpected }"); 133 return (link); 134 } 135 *link = cmd = xmalloc(sizeof(struct s_command)); 136 link = &cmd->next; 137 cmd->nonsel = cmd->inrange = 0; 138 /* First parse the addresses */ 139 naddr = 0; 140 cmd->a1 = cmd->a2 = NULL; 141 142 /* Valid characters to start an address */ 143 #define addrchar(c) (strchr("0123456789/\\$", (c))) 144 if (addrchar(*p)) { 145 naddr++; 146 cmd->a1 = xmalloc(sizeof(struct s_addr)); 147 p = compile_addr(p, cmd->a1); 148 EATSPACE(); /* EXTENSION */ 149 if (*p == ',') { 150 naddr++; 151 p++; 152 EATSPACE(); /* EXTENSION */ 153 cmd->a2 = xmalloc(sizeof(struct s_addr)); 154 p = compile_addr(p, cmd->a2); 155 } 156 } 157 158 nonsel: /* Now parse the command */ 159 EATSPACE(); 160 if (!*p) 161 err(COMPILE, "command expected"); 162 cmd->code = *p; 163 for (fp = cmd_fmts; fp->code; fp++) 164 if (fp->code == *p) 165 break; 166 if (!fp->code) 167 err(COMPILE, "invalid command code %c", *p); 168 if (naddr > fp->naddr) 169 err(COMPILE, 170 "command %c expects up to %d address(es), found %d", *p, fp->naddr, naddr); 171 switch (fp->args) { 172 case NONSEL: /* ! */ 173 cmd->nonsel = ! cmd->nonsel; 174 p++; 175 goto nonsel; 176 case GROUP: /* { */ 177 p++; 178 EATSPACE(); 179 if (!*p) 180 p = NULL; 181 cmd2 = xmalloc(sizeof(struct s_command)); 182 cmd2->code = '}'; 183 *compile_stream("}", &cmd->u.c, p) = cmd2; 184 cmd->next = cmd2; 185 link = &cmd2->next; 186 break; 187 case EMPTY: /* d D g G h H l n N p P q x = \0 */ 188 p++; 189 EATSPACE(); 190 if (*p == ';') { 191 p++; 192 link = &cmd->next; 193 goto semicolon; 194 } 195 if (*p) 196 err(COMPILE, 197 "extra characters at the end of %c command", cmd->code); 198 break; 199 case TEXT: /* a c i */ 200 p++; 201 EATSPACE(); 202 if (*p != '\\') 203 err(COMPILE, 204 "command %c expects \\ followed by text", cmd->code); 205 p++; 206 EATSPACE(); 207 if (*p) 208 err(COMPILE, 209 "extra characters after \\ at the end of %c command", cmd->code); 210 cmd->t = compile_text(); 211 break; 212 case COMMENT: /* \0 # */ 213 break; 214 case WFILE: /* w */ 215 p++; 216 EATSPACE(); 217 if (*p == '\0') 218 err(COMPILE, "filename expected"); 219 cmd->t = duptoeol(p); 220 if (aflag) 221 cmd->u.fd = -1; 222 else if ((cmd->u.fd = open(p, 223 O_WRONLY|O_APPEND|O_CREAT|O_TRUNC, 224 DEFFILEMODE)) == -1) 225 err(FATAL, "%s: %s\n", p, strerror(errno)); 226 break; 227 case RFILE: /* r */ 228 p++; 229 EATSPACE(); 230 if (*p == '\0') 231 err(COMPILE, "filename expected"); 232 else 233 cmd->t = duptoeol(p); 234 break; 235 case BRANCH: /* b t */ 236 p++; 237 EATSPACE(); 238 if (*p == '\0') 239 cmd->t = NULL; 240 else 241 cmd->t = duptoeol(p); 242 break; 243 case LABEL: /* : */ 244 p++; 245 EATSPACE(); 246 cmd->t = duptoeol(p); 247 if (strlen(p) == 0) 248 err(COMPILE, "empty label"); 249 break; 250 case SUBST: /* s */ 251 p++; 252 if (*p == '\0' || *p == '\\') 253 err(COMPILE, 254 "substitute pattern can not be delimited by newline or backslash"); 255 cmd->u.s = xmalloc(sizeof(struct s_subst)); 256 p = compile_re(p, &cmd->u.s->re); 257 if (p == NULL) 258 err(COMPILE, "unterminated substitute pattern"); 259 --p; 260 p = compile_subst(p, cmd->u.s); 261 p = compile_flags(p, cmd->u.s); 262 EATSPACE(); 263 if (*p == ';') { 264 p++; 265 link = &cmd->next; 266 goto semicolon; 267 } 268 break; 269 case TR: /* y */ 270 p++; 271 p = compile_tr(p, (char **)&cmd->u.y); 272 EATSPACE(); 273 if (*p == ';') { 274 p++; 275 link = &cmd->next; 276 goto semicolon; 277 } 278 if (*p) 279 err(COMPILE, 280 "extra text at the end of a transform command"); 281 break; 282 } 283 } 284 } 285 286 /* 287 * Get a delimited string. P points to the delimeter of the string; d points 288 * to a buffer area. Newline and delimiter escapes are processed; other 289 * escapes are ignored. 290 * 291 * Returns a pointer to the first character after the final delimiter or NULL 292 * in the case of a non-terminated string. The character array d is filled 293 * with the processed string. 294 */ 295 static char * 296 compile_delimited(p, d) 297 char *p, *d; 298 { 299 char c; 300 301 c = *p++; 302 if (c == '\0') 303 return (NULL); 304 else if (c == '\\') 305 err(COMPILE, "\\ can not be used as a string delimiter"); 306 else if (c == '\n') 307 err(COMPILE, "newline can not be used as a string delimiter"); 308 while (*p) { 309 if (*p == '\\' && p[1] == c) 310 p++; 311 else if (*p == '\\' && p[1] == 'n') { 312 *d++ = '\n'; 313 p += 2; 314 continue; 315 } else if (*p == c) { 316 *d = '\0'; 317 return (p + 1); 318 } 319 *d++ = *p++; 320 } 321 return (NULL); 322 } 323 324 /* 325 * Get a regular expression. P points to the delimiter of the regular 326 * expression; repp points to the address of a regexp pointer. Newline 327 * and delimiter escapes are processed; other escapes are ignored. 328 * Returns a pointer to the first character after the final delimiter 329 * or NULL in the case of a non terminated regular expression. The regexp 330 * pointer is set to the compiled regular expression. 331 * Cflags are passed to regcomp. 332 */ 333 static char * 334 compile_re(p, repp) 335 char *p; 336 regex_t **repp; 337 { 338 int eval; 339 char re[_POSIX2_LINE_MAX + 1]; 340 341 p = compile_delimited(p, re); 342 if (p && strlen(re) == 0) { 343 *repp = NULL; 344 return (p); 345 } 346 *repp = xmalloc(sizeof(regex_t)); 347 if (p && (eval = regcomp(*repp, re, 0)) != 0) 348 err(COMPILE, "RE error: %s", strregerror(eval, *repp)); 349 if (maxnsub < (*repp)->re_nsub) 350 maxnsub = (*repp)->re_nsub; 351 return (p); 352 } 353 354 /* 355 * Compile the substitution string of a regular expression and set res to 356 * point to a saved copy of it. Nsub is the number of parenthesized regular 357 * expressions. 358 */ 359 static char * 360 compile_subst(p, s) 361 char *p; 362 struct s_subst *s; 363 { 364 static char lbuf[_POSIX2_LINE_MAX + 1]; 365 int asize, ref, size; 366 char c, *text, *op, *sp; 367 368 c = *p++; /* Terminator character */ 369 if (c == '\0') 370 return (NULL); 371 372 s->maxbref = 0; 373 s->linenum = linenum; 374 asize = 2 * _POSIX2_LINE_MAX + 1; 375 text = xmalloc(asize); 376 size = 0; 377 do { 378 op = sp = text + size; 379 for (; *p; p++) { 380 if (*p == '\\') { 381 p++; 382 if (strchr("123456789", *p) != NULL) { 383 *sp++ = '\\'; 384 ref = *p - '0'; 385 if (s->re != NULL && 386 ref > s->re->re_nsub) 387 err(COMPILE, 388 "\\%c not defined in the RE", *p); 389 if (s->maxbref < ref) 390 s->maxbref = ref; 391 } else if (*p == '&') 392 *sp++ = '\\'; 393 } else if (*p == c) { 394 p++; 395 *sp++ = '\0'; 396 size += sp - op; 397 s->new = xrealloc(text, size); 398 return (p); 399 } else if (*p == '\n') { 400 err(COMPILE, 401 "unescaped newline inside substitute pattern"); 402 /* NOTREACHED */ 403 } 404 *sp++ = *p; 405 } 406 size += sp - op; 407 if (asize - size < _POSIX2_LINE_MAX + 1) { 408 asize *= 2; 409 text = xmalloc(asize); 410 } 411 } while (cu_fgets(p = lbuf, sizeof(lbuf))); 412 err(COMPILE, "unterminated substitute in regular expression"); 413 /* NOTREACHED */ 414 } 415 416 /* 417 * Compile the flags of the s command 418 */ 419 static char * 420 compile_flags(p, s) 421 char *p; 422 struct s_subst *s; 423 { 424 int gn; /* True if we have seen g or n */ 425 char wfile[_POSIX2_LINE_MAX + 1], *q; 426 427 s->n = 1; /* Default */ 428 s->p = 0; 429 s->wfile = NULL; 430 s->wfd = -1; 431 for (gn = 0;;) { 432 EATSPACE(); /* EXTENSION */ 433 switch (*p) { 434 case 'g': 435 if (gn) 436 err(COMPILE, 437 "more than one number or 'g' in substitute flags"); 438 gn = 1; 439 s->n = 0; 440 break; 441 case '\0': 442 case '\n': 443 case ';': 444 return (p); 445 case 'p': 446 s->p = 1; 447 break; 448 case '1': case '2': case '3': 449 case '4': case '5': case '6': 450 case '7': case '8': case '9': 451 if (gn) 452 err(COMPILE, 453 "more than one number or 'g' in substitute flags"); 454 gn = 1; 455 /* XXX Check for overflow */ 456 s->n = (int)strtol(p, &p, 10); 457 break; 458 case 'w': 459 p++; 460 #ifdef HISTORIC_PRACTICE 461 if (*p != ' ') { 462 err(WARNING, "space missing before w wfile"); 463 return (p); 464 } 465 #endif 466 EATSPACE(); 467 q = wfile; 468 while (*p) { 469 if (*p == '\n') 470 break; 471 *q++ = *p++; 472 } 473 *q = '\0'; 474 if (q == wfile) 475 err(COMPILE, "no wfile specified"); 476 s->wfile = strdup(wfile); 477 if (!aflag && (s->wfd = open(wfile, 478 O_WRONLY|O_APPEND|O_CREAT|O_TRUNC, 479 DEFFILEMODE)) == -1) 480 err(FATAL, "%s: %s\n", wfile, strerror(errno)); 481 return (p); 482 default: 483 err(COMPILE, 484 "bad flag in substitute command: '%c'", *p); 485 break; 486 } 487 p++; 488 } 489 } 490 491 /* 492 * Compile a translation set of strings into a lookup table. 493 */ 494 static char * 495 compile_tr(p, transtab) 496 char *p; 497 char **transtab; 498 { 499 int i; 500 char *lt, *op, *np; 501 char old[_POSIX2_LINE_MAX + 1]; 502 char new[_POSIX2_LINE_MAX + 1]; 503 504 if (*p == '\0' || *p == '\\') 505 err(COMPILE, 506 "transform pattern can not be delimited by newline or backslash"); 507 p = compile_delimited(p, old); 508 if (p == NULL) { 509 err(COMPILE, "unterminated transform source string"); 510 return (NULL); 511 } 512 p = compile_delimited(--p, new); 513 if (p == NULL) { 514 err(COMPILE, "unterminated transform target string"); 515 return (NULL); 516 } 517 EATSPACE(); 518 if (strlen(new) != strlen(old)) { 519 err(COMPILE, "transform strings are not the same length"); 520 return (NULL); 521 } 522 /* We assume characters are 8 bits */ 523 lt = xmalloc(UCHAR_MAX); 524 for (i = 0; i <= UCHAR_MAX; i++) 525 lt[i] = (char)i; 526 for (op = old, np = new; *op; op++, np++) 527 lt[(u_char)*op] = *np; 528 *transtab = lt; 529 return (p); 530 } 531 532 /* 533 * Compile the text following an a or i command. 534 */ 535 static char * 536 compile_text() 537 { 538 int asize, size; 539 char *text, *p, *op, *s; 540 char lbuf[_POSIX2_LINE_MAX + 1]; 541 542 asize = 2 * _POSIX2_LINE_MAX + 1; 543 text = xmalloc(asize); 544 size = 0; 545 while (cu_fgets(lbuf, sizeof(lbuf))) { 546 op = s = text + size; 547 p = lbuf; 548 EATSPACE(); 549 for (; *p; p++) { 550 if (*p == '\\') 551 p++; 552 *s++ = *p; 553 } 554 size += s - op; 555 if (p[-2] != '\\') { 556 *s = '\0'; 557 break; 558 } 559 if (asize - size < _POSIX2_LINE_MAX + 1) { 560 asize *= 2; 561 text = xmalloc(asize); 562 } 563 } 564 return (xrealloc(text, size + 1)); 565 } 566 567 /* 568 * Get an address and return a pointer to the first character after 569 * it. Fill the structure pointed to according to the address. 570 */ 571 static char * 572 compile_addr(p, a) 573 char *p; 574 struct s_addr *a; 575 { 576 char *end; 577 578 switch (*p) { 579 case '\\': /* Context address */ 580 ++p; 581 /* FALLTHROUGH */ 582 case '/': /* Context address */ 583 p = compile_re(p, &a->u.r); 584 if (p == NULL) 585 err(COMPILE, "unterminated regular expression"); 586 a->type = AT_RE; 587 return (p); 588 589 case '$': /* Last line */ 590 a->type = AT_LAST; 591 return (p + 1); 592 /* Line number */ 593 case '0': case '1': case '2': case '3': case '4': 594 case '5': case '6': case '7': case '8': case '9': 595 a->type = AT_LINE; 596 a->u.l = strtol(p, &end, 10); 597 return (end); 598 default: 599 err(COMPILE, "expected context address"); 600 return (NULL); 601 } 602 } 603 604 /* 605 * Return a copy of all the characters up to \n or \0 606 */ 607 static char * 608 duptoeol(s) 609 register char *s; 610 { 611 size_t len; 612 char *start; 613 614 for (start = s; *s != '\0' && *s != '\n'; ++s); 615 *s = '\0'; 616 len = s - start + 1; 617 return (memmove(xmalloc(len), start, len)); 618 } 619 620 /* 621 * Find the label contained in the command l in the command linked list cp. 622 * L is excluded from the search. Return NULL if not found. 623 */ 624 static struct s_command * 625 findlabel(l, cp) 626 struct s_command *l, *cp; 627 { 628 struct s_command *r; 629 630 for (; cp; cp = cp->next) 631 if (cp->code == ':' && cp != l && strcmp(l->t, cp->t) == 0) 632 return (cp); 633 else if (cp->code == '{' && (r = findlabel(l, cp->u.c))) 634 return (r); 635 return (NULL); 636 } 637 638 /* 639 * Convert goto label names to addresses. 640 * Detect duplicate labels. 641 * Set appendnum to the number of a and r commands in the script. 642 * Free the memory used by labels in b and t commands (but not by :) 643 * Root is a pointer to the script linked list; cp points to the 644 * search start. 645 * TODO: Remove } nodes 646 */ 647 static void 648 fixuplabel(root, cp) 649 struct s_command *root, *cp; 650 { 651 struct s_command *cp2; 652 653 for (; cp; cp = cp->next) 654 switch (cp->code) { 655 case ':': 656 if (findlabel(cp, root)) 657 err(COMPILE2, "duplicate label %s", cp->t); 658 break; 659 case 'a': 660 case 'r': 661 appendnum++; 662 break; 663 case 'b': 664 case 't': 665 if (cp->t == NULL) { 666 cp->u.c = NULL; 667 break; 668 } 669 if ((cp2 = findlabel(cp, root)) == NULL) 670 err(COMPILE2, "undefined label '%s'", cp->t); 671 free(cp->t); 672 cp->u.c = cp2; 673 break; 674 case '{': 675 fixuplabel(root, cp->u.c); 676 break; 677 } 678 } 679