expand.c revision 216387
1/*- 2 * Copyright (c) 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * Copyright (c) 1997-2005 5 * Herbert Xu <herbert@gondor.apana.org.au>. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Kenneth Almquist. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 4. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 */ 34 35#ifndef lint 36#if 0 37static char sccsid[] = "@(#)expand.c 8.5 (Berkeley) 5/15/95"; 38#endif 39#endif /* not lint */ 40#include <sys/cdefs.h> 41__FBSDID("$FreeBSD: head/bin/sh/expand.c 216387 2010-12-12 00:07:27Z jilles $"); 42 43#include <sys/types.h> 44#include <sys/time.h> 45#include <sys/stat.h> 46#include <dirent.h> 47#include <errno.h> 48#include <inttypes.h> 49#include <limits.h> 50#include <pwd.h> 51#include <stdio.h> 52#include <stdlib.h> 53#include <string.h> 54#include <unistd.h> 55 56/* 57 * Routines to expand arguments to commands. We have to deal with 58 * backquotes, shell variables, and file metacharacters. 59 */ 60 61#include "shell.h" 62#include "main.h" 63#include "nodes.h" 64#include "eval.h" 65#include "expand.h" 66#include "syntax.h" 67#include "parser.h" 68#include "jobs.h" 69#include "options.h" 70#include "var.h" 71#include "input.h" 72#include "output.h" 73#include "memalloc.h" 74#include "error.h" 75#include "mystring.h" 76#include "arith.h" 77#include "show.h" 78 79/* 80 * Structure specifying which parts of the string should be searched 81 * for IFS characters. 82 */ 83 84struct ifsregion { 85 struct ifsregion *next; /* next region in list */ 86 int begoff; /* offset of start of region */ 87 int endoff; /* offset of end of region */ 88 int inquotes; /* search for nul bytes only */ 89}; 90 91 92static char *expdest; /* output of current string */ 93static struct nodelist *argbackq; /* list of back quote expressions */ 94static struct ifsregion ifsfirst; /* first struct in list of ifs regions */ 95static struct ifsregion *ifslastp; /* last struct in list */ 96static struct arglist exparg; /* holds expanded arg list */ 97 98static void argstr(char *, int); 99static char *exptilde(char *, int); 100static void expbackq(union node *, int, int); 101static int subevalvar(char *, char *, int, int, int, int, int); 102static char *evalvar(char *, int); 103static int varisset(char *, int); 104static void varvalue(char *, int, int, int); 105static void recordregion(int, int, int); 106static void removerecordregions(int); 107static void ifsbreakup(char *, struct arglist *); 108static void expandmeta(struct strlist *, int); 109static void expmeta(char *, char *); 110static void addfname(char *); 111static struct strlist *expsort(struct strlist *); 112static struct strlist *msort(struct strlist *, int); 113static char *cvtnum(int, char *); 114static int collate_range_cmp(int, int); 115 116static int 117collate_range_cmp(int c1, int c2) 118{ 119 static char s1[2], s2[2]; 120 121 s1[0] = c1; 122 s2[0] = c2; 123 return (strcoll(s1, s2)); 124} 125 126/* 127 * Expand shell variables and backquotes inside a here document. 128 * union node *arg the document 129 * int fd; where to write the expanded version 130 */ 131 132void 133expandhere(union node *arg, int fd) 134{ 135 expandarg(arg, (struct arglist *)NULL, 0); 136 xwrite(fd, stackblock(), expdest - stackblock()); 137} 138 139static char * 140stputs_quotes(const char *data, const char *syntax, char *p) 141{ 142 while (*data) { 143 CHECKSTRSPACE(2, p); 144 if (syntax[(int)*data] == CCTL) 145 USTPUTC(CTLESC, p); 146 USTPUTC(*data++, p); 147 } 148 return (p); 149} 150#define STPUTS_QUOTES(data, syntax, p) p = stputs_quotes((data), syntax, p) 151 152/* 153 * Perform expansions on an argument, placing the resulting list of arguments 154 * in arglist. Parameter expansion, command substitution and arithmetic 155 * expansion are always performed; additional expansions can be requested 156 * via flag (EXP_*). 157 * The result is left in the stack string. 158 * When arglist is NULL, perform here document expansion. A partial result 159 * may be written to herefd, which is then not included in the stack string. 160 * 161 * Caution: this function uses global state and is not reentrant. 162 * However, a new invocation after an interrupted invocation is safe 163 * and will reset the global state for the new call. 164 */ 165void 166expandarg(union node *arg, struct arglist *arglist, int flag) 167{ 168 struct strlist *sp; 169 char *p; 170 171 argbackq = arg->narg.backquote; 172 STARTSTACKSTR(expdest); 173 ifsfirst.next = NULL; 174 ifslastp = NULL; 175 argstr(arg->narg.text, flag); 176 if (arglist == NULL) { 177 return; /* here document expanded */ 178 } 179 STPUTC('\0', expdest); 180 p = grabstackstr(expdest); 181 exparg.lastp = &exparg.list; 182 /* 183 * TODO - EXP_REDIR 184 */ 185 if (flag & EXP_FULL) { 186 ifsbreakup(p, &exparg); 187 *exparg.lastp = NULL; 188 exparg.lastp = &exparg.list; 189 expandmeta(exparg.list, flag); 190 } else { 191 if (flag & EXP_REDIR) /*XXX - for now, just remove escapes */ 192 rmescapes(p); 193 sp = (struct strlist *)stalloc(sizeof (struct strlist)); 194 sp->text = p; 195 *exparg.lastp = sp; 196 exparg.lastp = &sp->next; 197 } 198 while (ifsfirst.next != NULL) { 199 struct ifsregion *ifsp; 200 INTOFF; 201 ifsp = ifsfirst.next->next; 202 ckfree(ifsfirst.next); 203 ifsfirst.next = ifsp; 204 INTON; 205 } 206 *exparg.lastp = NULL; 207 if (exparg.list) { 208 *arglist->lastp = exparg.list; 209 arglist->lastp = exparg.lastp; 210 } 211} 212 213 214 215/* 216 * Perform parameter expansion, command substitution and arithmetic 217 * expansion, and tilde expansion if requested via EXP_TILDE/EXP_VARTILDE. 218 * Processing ends at a CTLENDVAR character as well as '\0'. 219 * This is used to expand word in ${var+word} etc. 220 * If EXP_FULL, EXP_CASE or EXP_REDIR are set, keep and/or generate CTLESC 221 * characters to allow for further processing. 222 * If EXP_FULL is set, also preserve CTLQUOTEMARK characters. 223 */ 224static void 225argstr(char *p, int flag) 226{ 227 char c; 228 int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR); /* do CTLESC */ 229 int firsteq = 1; 230 int split_lit; 231 int lit_quoted; 232 233 split_lit = flag & EXP_SPLIT_LIT; 234 lit_quoted = flag & EXP_LIT_QUOTED; 235 flag &= ~(EXP_SPLIT_LIT | EXP_LIT_QUOTED); 236 if (*p == '~' && (flag & (EXP_TILDE | EXP_VARTILDE))) 237 p = exptilde(p, flag); 238 for (;;) { 239 CHECKSTRSPACE(2, expdest); 240 switch (c = *p++) { 241 case '\0': 242 case CTLENDVAR: 243 goto breakloop; 244 case CTLQUOTEMARK: 245 lit_quoted = 1; 246 /* "$@" syntax adherence hack */ 247 if (p[0] == CTLVAR && p[2] == '@' && p[3] == '=') 248 break; 249 if ((flag & EXP_FULL) != 0) 250 USTPUTC(c, expdest); 251 break; 252 case CTLQUOTEEND: 253 lit_quoted = 0; 254 break; 255 case CTLESC: 256 if (quotes) 257 USTPUTC(c, expdest); 258 c = *p++; 259 USTPUTC(c, expdest); 260 if (split_lit && !lit_quoted) 261 recordregion(expdest - stackblock() - 262 (quotes ? 2 : 1), 263 expdest - stackblock(), 0); 264 break; 265 case CTLVAR: 266 p = evalvar(p, flag); 267 break; 268 case CTLBACKQ: 269 case CTLBACKQ|CTLQUOTE: 270 expbackq(argbackq->n, c & CTLQUOTE, flag); 271 argbackq = argbackq->next; 272 break; 273 case CTLENDARI: 274 expari(flag); 275 break; 276 case ':': 277 case '=': 278 /* 279 * sort of a hack - expand tildes in variable 280 * assignments (after the first '=' and after ':'s). 281 */ 282 USTPUTC(c, expdest); 283 if (split_lit && !lit_quoted) 284 recordregion(expdest - stackblock() - 1, 285 expdest - stackblock(), 0); 286 if (flag & EXP_VARTILDE && *p == '~' && 287 (c != '=' || firsteq)) { 288 if (c == '=') 289 firsteq = 0; 290 p = exptilde(p, flag); 291 } 292 break; 293 default: 294 USTPUTC(c, expdest); 295 if (split_lit && !lit_quoted) 296 recordregion(expdest - stackblock() - 1, 297 expdest - stackblock(), 0); 298 } 299 } 300breakloop:; 301} 302 303/* 304 * Perform tilde expansion, placing the result in the stack string and 305 * returning the next position in the input string to process. 306 */ 307static char * 308exptilde(char *p, int flag) 309{ 310 char c, *startp = p; 311 struct passwd *pw; 312 char *home; 313 int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR); 314 315 while ((c = *p) != '\0') { 316 switch(c) { 317 case CTLESC: /* This means CTL* are always considered quoted. */ 318 case CTLVAR: 319 case CTLBACKQ: 320 case CTLBACKQ | CTLQUOTE: 321 case CTLARI: 322 case CTLENDARI: 323 case CTLQUOTEMARK: 324 return (startp); 325 case ':': 326 if (flag & EXP_VARTILDE) 327 goto done; 328 break; 329 case '/': 330 case CTLENDVAR: 331 goto done; 332 } 333 p++; 334 } 335done: 336 *p = '\0'; 337 if (*(startp+1) == '\0') { 338 if ((home = lookupvar("HOME")) == NULL) 339 goto lose; 340 } else { 341 if ((pw = getpwnam(startp+1)) == NULL) 342 goto lose; 343 home = pw->pw_dir; 344 } 345 if (*home == '\0') 346 goto lose; 347 *p = c; 348 if (quotes) 349 STPUTS_QUOTES(home, SQSYNTAX, expdest); 350 else 351 STPUTS(home, expdest); 352 return (p); 353lose: 354 *p = c; 355 return (startp); 356} 357 358 359static void 360removerecordregions(int endoff) 361{ 362 if (ifslastp == NULL) 363 return; 364 365 if (ifsfirst.endoff > endoff) { 366 while (ifsfirst.next != NULL) { 367 struct ifsregion *ifsp; 368 INTOFF; 369 ifsp = ifsfirst.next->next; 370 ckfree(ifsfirst.next); 371 ifsfirst.next = ifsp; 372 INTON; 373 } 374 if (ifsfirst.begoff > endoff) 375 ifslastp = NULL; 376 else { 377 ifslastp = &ifsfirst; 378 ifsfirst.endoff = endoff; 379 } 380 return; 381 } 382 383 ifslastp = &ifsfirst; 384 while (ifslastp->next && ifslastp->next->begoff < endoff) 385 ifslastp=ifslastp->next; 386 while (ifslastp->next != NULL) { 387 struct ifsregion *ifsp; 388 INTOFF; 389 ifsp = ifslastp->next->next; 390 ckfree(ifslastp->next); 391 ifslastp->next = ifsp; 392 INTON; 393 } 394 if (ifslastp->endoff > endoff) 395 ifslastp->endoff = endoff; 396} 397 398/* 399 * Expand arithmetic expression. Backup to start of expression, 400 * evaluate, place result in (backed up) result, adjust string position. 401 */ 402void 403expari(int flag) 404{ 405 char *p, *q, *start; 406 arith_t result; 407 int begoff; 408 int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR); 409 int quoted; 410 411 /* 412 * This routine is slightly over-complicated for 413 * efficiency. First we make sure there is 414 * enough space for the result, which may be bigger 415 * than the expression. Next we 416 * scan backwards looking for the start of arithmetic. If the 417 * next previous character is a CTLESC character, then we 418 * have to rescan starting from the beginning since CTLESC 419 * characters have to be processed left to right. 420 */ 421 CHECKSTRSPACE(DIGITS(result) - 2, expdest); 422 USTPUTC('\0', expdest); 423 start = stackblock(); 424 p = expdest - 2; 425 while (p >= start && *p != CTLARI) 426 --p; 427 if (p < start || *p != CTLARI) 428 error("missing CTLARI (shouldn't happen)"); 429 if (p > start && *(p - 1) == CTLESC) 430 for (p = start; *p != CTLARI; p++) 431 if (*p == CTLESC) 432 p++; 433 434 if (p[1] == '"') 435 quoted=1; 436 else 437 quoted=0; 438 begoff = p - start; 439 removerecordregions(begoff); 440 if (quotes) 441 rmescapes(p+2); 442 q = grabstackstr(expdest); 443 result = arith(p+2); 444 ungrabstackstr(q, expdest); 445 fmtstr(p, DIGITS(result), ARITH_FORMAT_STR, result); 446 while (*p++) 447 ; 448 if (quoted == 0) 449 recordregion(begoff, p - 1 - start, 0); 450 result = expdest - p + 1; 451 STADJUST(-result, expdest); 452} 453 454 455/* 456 * Perform command substitution. 457 */ 458static void 459expbackq(union node *cmd, int quoted, int flag) 460{ 461 struct backcmd in; 462 int i; 463 char buf[128]; 464 char *p; 465 char *dest = expdest; 466 struct ifsregion saveifs, *savelastp; 467 struct nodelist *saveargbackq; 468 char lastc; 469 int startloc = dest - stackblock(); 470 char const *syntax = quoted? DQSYNTAX : BASESYNTAX; 471 int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR); 472 int nnl; 473 474 INTOFF; 475 saveifs = ifsfirst; 476 savelastp = ifslastp; 477 saveargbackq = argbackq; 478 p = grabstackstr(dest); 479 evalbackcmd(cmd, &in); 480 ungrabstackstr(p, dest); 481 ifsfirst = saveifs; 482 ifslastp = savelastp; 483 argbackq = saveargbackq; 484 485 p = in.buf; 486 lastc = '\0'; 487 nnl = 0; 488 /* Don't copy trailing newlines */ 489 for (;;) { 490 if (--in.nleft < 0) { 491 if (in.fd < 0) 492 break; 493 while ((i = read(in.fd, buf, sizeof buf)) < 0 && errno == EINTR); 494 TRACE(("expbackq: read returns %d\n", i)); 495 if (i <= 0) 496 break; 497 p = buf; 498 in.nleft = i - 1; 499 } 500 lastc = *p++; 501 if (lastc != '\0') { 502 if (quotes && syntax[(int)lastc] == CCTL) 503 STPUTC(CTLESC, dest); 504 if (lastc == '\n') { 505 nnl++; 506 } else { 507 while (nnl > 0) { 508 nnl--; 509 STPUTC('\n', dest); 510 } 511 STPUTC(lastc, dest); 512 } 513 } 514 } 515 516 if (in.fd >= 0) 517 close(in.fd); 518 if (in.buf) 519 ckfree(in.buf); 520 if (in.jp) 521 exitstatus = waitforjob(in.jp, (int *)NULL); 522 if (quoted == 0) 523 recordregion(startloc, dest - stackblock(), 0); 524 TRACE(("expbackq: size=%td: \"%.*s\"\n", 525 ((dest - stackblock()) - startloc), 526 (int)((dest - stackblock()) - startloc), 527 stackblock() + startloc)); 528 expdest = dest; 529 INTON; 530} 531 532 533 534static int 535subevalvar(char *p, char *str, int strloc, int subtype, int startloc, 536 int varflags, int quotes) 537{ 538 char *startp; 539 char *loc = NULL; 540 char *q; 541 int c = 0; 542 struct nodelist *saveargbackq = argbackq; 543 int amount; 544 545 argstr(p, (subtype == VSTRIMLEFT || subtype == VSTRIMLEFTMAX || 546 subtype == VSTRIMRIGHT || subtype == VSTRIMRIGHTMAX ? 547 EXP_CASE : 0) | EXP_TILDE); 548 STACKSTRNUL(expdest); 549 argbackq = saveargbackq; 550 startp = stackblock() + startloc; 551 if (str == NULL) 552 str = stackblock() + strloc; 553 554 switch (subtype) { 555 case VSASSIGN: 556 setvar(str, startp, 0); 557 amount = startp - expdest; 558 STADJUST(amount, expdest); 559 varflags &= ~VSNUL; 560 if (c != 0) 561 *loc = c; 562 return 1; 563 564 case VSQUESTION: 565 if (*p != CTLENDVAR) { 566 outfmt(out2, "%s\n", startp); 567 error((char *)NULL); 568 } 569 error("%.*s: parameter %snot set", (int)(p - str - 1), 570 str, (varflags & VSNUL) ? "null or " 571 : nullstr); 572 return 0; 573 574 case VSTRIMLEFT: 575 for (loc = startp; loc < str; loc++) { 576 c = *loc; 577 *loc = '\0'; 578 if (patmatch(str, startp, quotes)) { 579 *loc = c; 580 goto recordleft; 581 } 582 *loc = c; 583 if (quotes && *loc == CTLESC) 584 loc++; 585 } 586 return 0; 587 588 case VSTRIMLEFTMAX: 589 for (loc = str - 1; loc >= startp;) { 590 c = *loc; 591 *loc = '\0'; 592 if (patmatch(str, startp, quotes)) { 593 *loc = c; 594 goto recordleft; 595 } 596 *loc = c; 597 loc--; 598 if (quotes && loc > startp && *(loc - 1) == CTLESC) { 599 for (q = startp; q < loc; q++) 600 if (*q == CTLESC) 601 q++; 602 if (q > loc) 603 loc--; 604 } 605 } 606 return 0; 607 608 case VSTRIMRIGHT: 609 for (loc = str - 1; loc >= startp;) { 610 if (patmatch(str, loc, quotes)) { 611 amount = loc - expdest; 612 STADJUST(amount, expdest); 613 return 1; 614 } 615 loc--; 616 if (quotes && loc > startp && *(loc - 1) == CTLESC) { 617 for (q = startp; q < loc; q++) 618 if (*q == CTLESC) 619 q++; 620 if (q > loc) 621 loc--; 622 } 623 } 624 return 0; 625 626 case VSTRIMRIGHTMAX: 627 for (loc = startp; loc < str - 1; loc++) { 628 if (patmatch(str, loc, quotes)) { 629 amount = loc - expdest; 630 STADJUST(amount, expdest); 631 return 1; 632 } 633 if (quotes && *loc == CTLESC) 634 loc++; 635 } 636 return 0; 637 638 639 default: 640 abort(); 641 } 642 643recordleft: 644 amount = ((str - 1) - (loc - startp)) - expdest; 645 STADJUST(amount, expdest); 646 while (loc != str - 1) 647 *startp++ = *loc++; 648 return 1; 649} 650 651 652/* 653 * Expand a variable, and return a pointer to the next character in the 654 * input string. 655 */ 656 657static char * 658evalvar(char *p, int flag) 659{ 660 int subtype; 661 int varflags; 662 char *var; 663 char *val; 664 int patloc; 665 int c; 666 int set; 667 int special; 668 int startloc; 669 int varlen; 670 int easy; 671 int quotes = flag & (EXP_FULL | EXP_CASE | EXP_REDIR); 672 673 varflags = (unsigned char)*p++; 674 subtype = varflags & VSTYPE; 675 var = p; 676 special = 0; 677 if (! is_name(*p)) 678 special = 1; 679 p = strchr(p, '=') + 1; 680again: /* jump here after setting a variable with ${var=text} */ 681 if (varflags & VSLINENO) { 682 set = 1; 683 special = 0; 684 val = var; 685 p[-1] = '\0'; /* temporarily overwrite '=' to have \0 686 terminated string */ 687 } else if (special) { 688 set = varisset(var, varflags & VSNUL); 689 val = NULL; 690 } else { 691 val = bltinlookup(var, 1); 692 if (val == NULL || ((varflags & VSNUL) && val[0] == '\0')) { 693 val = NULL; 694 set = 0; 695 } else 696 set = 1; 697 } 698 varlen = 0; 699 startloc = expdest - stackblock(); 700 if (!set && uflag && *var != '@' && *var != '*') { 701 switch (subtype) { 702 case VSNORMAL: 703 case VSTRIMLEFT: 704 case VSTRIMLEFTMAX: 705 case VSTRIMRIGHT: 706 case VSTRIMRIGHTMAX: 707 case VSLENGTH: 708 error("%.*s: parameter not set", (int)(p - var - 1), 709 var); 710 } 711 } 712 if (set && subtype != VSPLUS) { 713 /* insert the value of the variable */ 714 if (special) { 715 varvalue(var, varflags & VSQUOTE, subtype, flag); 716 if (subtype == VSLENGTH) { 717 varlen = expdest - stackblock() - startloc; 718 STADJUST(-varlen, expdest); 719 } 720 } else { 721 char const *syntax = (varflags & VSQUOTE) ? DQSYNTAX 722 : BASESYNTAX; 723 724 if (subtype == VSLENGTH) { 725 for (;*val; val++) 726 varlen++; 727 } 728 else { 729 if (quotes) 730 STPUTS_QUOTES(val, syntax, expdest); 731 else 732 STPUTS(val, expdest); 733 734 } 735 } 736 } 737 738 if (subtype == VSPLUS) 739 set = ! set; 740 741 easy = ((varflags & VSQUOTE) == 0 || 742 (*var == '@' && shellparam.nparam != 1)); 743 744 745 switch (subtype) { 746 case VSLENGTH: 747 expdest = cvtnum(varlen, expdest); 748 goto record; 749 750 case VSNORMAL: 751 if (!easy) 752 break; 753record: 754 recordregion(startloc, expdest - stackblock(), 755 varflags & VSQUOTE); 756 break; 757 758 case VSPLUS: 759 case VSMINUS: 760 if (!set) { 761 argstr(p, flag | (flag & EXP_FULL ? EXP_SPLIT_LIT : 0) | 762 (varflags & VSQUOTE ? EXP_LIT_QUOTED : 0)); 763 break; 764 } 765 if (easy) 766 goto record; 767 break; 768 769 case VSTRIMLEFT: 770 case VSTRIMLEFTMAX: 771 case VSTRIMRIGHT: 772 case VSTRIMRIGHTMAX: 773 if (!set) 774 break; 775 /* 776 * Terminate the string and start recording the pattern 777 * right after it 778 */ 779 STPUTC('\0', expdest); 780 patloc = expdest - stackblock(); 781 if (subevalvar(p, NULL, patloc, subtype, 782 startloc, varflags, quotes) == 0) { 783 int amount = (expdest - stackblock() - patloc) + 1; 784 STADJUST(-amount, expdest); 785 } 786 /* Remove any recorded regions beyond start of variable */ 787 removerecordregions(startloc); 788 goto record; 789 790 case VSASSIGN: 791 case VSQUESTION: 792 if (!set) { 793 if (subevalvar(p, var, 0, subtype, startloc, varflags, 794 quotes)) { 795 varflags &= ~VSNUL; 796 /* 797 * Remove any recorded regions beyond 798 * start of variable 799 */ 800 removerecordregions(startloc); 801 goto again; 802 } 803 break; 804 } 805 if (easy) 806 goto record; 807 break; 808 809 case VSERROR: 810 c = p - var - 1; 811 error("${%.*s%s}: Bad substitution", c, var, 812 (c > 0 && *p != CTLENDVAR) ? "..." : ""); 813 814 default: 815 abort(); 816 } 817 p[-1] = '='; /* recover overwritten '=' */ 818 819 if (subtype != VSNORMAL) { /* skip to end of alternative */ 820 int nesting = 1; 821 for (;;) { 822 if ((c = *p++) == CTLESC) 823 p++; 824 else if (c == CTLBACKQ || c == (CTLBACKQ|CTLQUOTE)) { 825 if (set) 826 argbackq = argbackq->next; 827 } else if (c == CTLVAR) { 828 if ((*p++ & VSTYPE) != VSNORMAL) 829 nesting++; 830 } else if (c == CTLENDVAR) { 831 if (--nesting == 0) 832 break; 833 } 834 } 835 } 836 return p; 837} 838 839 840 841/* 842 * Test whether a specialized variable is set. 843 */ 844 845static int 846varisset(char *name, int nulok) 847{ 848 849 if (*name == '!') 850 return backgndpidset(); 851 else if (*name == '@' || *name == '*') { 852 if (*shellparam.p == NULL) 853 return 0; 854 855 if (nulok) { 856 char **av; 857 858 for (av = shellparam.p; *av; av++) 859 if (**av != '\0') 860 return 1; 861 return 0; 862 } 863 } else if (is_digit(*name)) { 864 char *ap; 865 int num = atoi(name); 866 867 if (num > shellparam.nparam) 868 return 0; 869 870 if (num == 0) 871 ap = arg0; 872 else 873 ap = shellparam.p[num - 1]; 874 875 if (nulok && (ap == NULL || *ap == '\0')) 876 return 0; 877 } 878 return 1; 879} 880 881static void 882strtodest(const char *p, int flag, int subtype, int quoted) 883{ 884 if (flag & (EXP_FULL | EXP_CASE) && subtype != VSLENGTH) 885 STPUTS_QUOTES(p, quoted ? DQSYNTAX : BASESYNTAX, expdest); 886 else 887 STPUTS(p, expdest); 888} 889 890/* 891 * Add the value of a specialized variable to the stack string. 892 */ 893 894static void 895varvalue(char *name, int quoted, int subtype, int flag) 896{ 897 int num; 898 char *p; 899 int i; 900 char sep; 901 char **ap; 902 903 switch (*name) { 904 case '$': 905 num = rootpid; 906 goto numvar; 907 case '?': 908 num = oexitstatus; 909 goto numvar; 910 case '#': 911 num = shellparam.nparam; 912 goto numvar; 913 case '!': 914 num = backgndpidval(); 915numvar: 916 expdest = cvtnum(num, expdest); 917 break; 918 case '-': 919 for (i = 0 ; i < NOPTS ; i++) { 920 if (optlist[i].val) 921 STPUTC(optlist[i].letter, expdest); 922 } 923 break; 924 case '@': 925 if (flag & EXP_FULL && quoted) { 926 for (ap = shellparam.p ; (p = *ap++) != NULL ; ) { 927 strtodest(p, flag, subtype, quoted); 928 if (*ap) 929 STPUTC('\0', expdest); 930 } 931 break; 932 } 933 /* FALLTHROUGH */ 934 case '*': 935 if (ifsset()) 936 sep = ifsval()[0]; 937 else 938 sep = ' '; 939 for (ap = shellparam.p ; (p = *ap++) != NULL ; ) { 940 strtodest(p, flag, subtype, quoted); 941 if (*ap && sep) 942 STPUTC(sep, expdest); 943 } 944 break; 945 case '0': 946 p = arg0; 947 strtodest(p, flag, subtype, quoted); 948 break; 949 default: 950 if (is_digit(*name)) { 951 num = atoi(name); 952 if (num > 0 && num <= shellparam.nparam) { 953 p = shellparam.p[num - 1]; 954 strtodest(p, flag, subtype, quoted); 955 } 956 } 957 break; 958 } 959} 960 961 962 963/* 964 * Record the the fact that we have to scan this region of the 965 * string for IFS characters. 966 */ 967 968static void 969recordregion(int start, int end, int inquotes) 970{ 971 struct ifsregion *ifsp; 972 973 if (ifslastp == NULL) { 974 ifsp = &ifsfirst; 975 } else { 976 if (ifslastp->endoff == start 977 && ifslastp->inquotes == inquotes) { 978 /* extend previous area */ 979 ifslastp->endoff = end; 980 return; 981 } 982 ifsp = (struct ifsregion *)ckmalloc(sizeof (struct ifsregion)); 983 ifslastp->next = ifsp; 984 } 985 ifslastp = ifsp; 986 ifslastp->next = NULL; 987 ifslastp->begoff = start; 988 ifslastp->endoff = end; 989 ifslastp->inquotes = inquotes; 990} 991 992 993 994/* 995 * Break the argument string into pieces based upon IFS and add the 996 * strings to the argument list. The regions of the string to be 997 * searched for IFS characters have been stored by recordregion. 998 * CTLESC characters are preserved but have little effect in this pass 999 * other than escaping CTL* characters. In particular, they do not escape 1000 * IFS characters: that should be done with the ifsregion mechanism. 1001 * CTLQUOTEMARK characters are used to preserve empty quoted strings. 1002 * This pass treats them as a regular character, making the string non-empty. 1003 * Later, they are removed along with the other CTL* characters. 1004 */ 1005static void 1006ifsbreakup(char *string, struct arglist *arglist) 1007{ 1008 struct ifsregion *ifsp; 1009 struct strlist *sp; 1010 char *start; 1011 char *p; 1012 char *q; 1013 const char *ifs; 1014 const char *ifsspc; 1015 int had_param_ch = 0; 1016 1017 start = string; 1018 1019 if (ifslastp == NULL) { 1020 /* Return entire argument, IFS doesn't apply to any of it */ 1021 sp = (struct strlist *)stalloc(sizeof *sp); 1022 sp->text = start; 1023 *arglist->lastp = sp; 1024 arglist->lastp = &sp->next; 1025 return; 1026 } 1027 1028 ifs = ifsset() ? ifsval() : " \t\n"; 1029 1030 for (ifsp = &ifsfirst; ifsp != NULL; ifsp = ifsp->next) { 1031 p = string + ifsp->begoff; 1032 while (p < string + ifsp->endoff) { 1033 q = p; 1034 if (*p == CTLESC) 1035 p++; 1036 if (ifsp->inquotes) { 1037 /* Only NULs (should be from "$@") end args */ 1038 had_param_ch = 1; 1039 if (*p != 0) { 1040 p++; 1041 continue; 1042 } 1043 ifsspc = NULL; 1044 } else { 1045 if (!strchr(ifs, *p)) { 1046 had_param_ch = 1; 1047 p++; 1048 continue; 1049 } 1050 ifsspc = strchr(" \t\n", *p); 1051 1052 /* Ignore IFS whitespace at start */ 1053 if (q == start && ifsspc != NULL) { 1054 p++; 1055 start = p; 1056 continue; 1057 } 1058 had_param_ch = 0; 1059 } 1060 1061 /* Save this argument... */ 1062 *q = '\0'; 1063 sp = (struct strlist *)stalloc(sizeof *sp); 1064 sp->text = start; 1065 *arglist->lastp = sp; 1066 arglist->lastp = &sp->next; 1067 p++; 1068 1069 if (ifsspc != NULL) { 1070 /* Ignore further trailing IFS whitespace */ 1071 for (; p < string + ifsp->endoff; p++) { 1072 q = p; 1073 if (*p == CTLESC) 1074 p++; 1075 if (strchr(ifs, *p) == NULL) { 1076 p = q; 1077 break; 1078 } 1079 if (strchr(" \t\n", *p) == NULL) { 1080 p++; 1081 break; 1082 } 1083 } 1084 } 1085 start = p; 1086 } 1087 } 1088 1089 /* 1090 * Save anything left as an argument. 1091 * Traditionally we have treated 'IFS=':'; set -- x$IFS' as 1092 * generating 2 arguments, the second of which is empty. 1093 * Some recent clarification of the Posix spec say that it 1094 * should only generate one.... 1095 */ 1096 if (had_param_ch || *start != 0) { 1097 sp = (struct strlist *)stalloc(sizeof *sp); 1098 sp->text = start; 1099 *arglist->lastp = sp; 1100 arglist->lastp = &sp->next; 1101 } 1102} 1103 1104 1105static char expdir[PATH_MAX]; 1106#define expdir_end (expdir + sizeof(expdir)) 1107 1108/* 1109 * Perform pathname generation and remove control characters. 1110 * At this point, the only control characters should be CTLESC and CTLQUOTEMARK. 1111 * The results are stored in the list exparg. 1112 */ 1113static void 1114expandmeta(struct strlist *str, int flag __unused) 1115{ 1116 char *p; 1117 struct strlist **savelastp; 1118 struct strlist *sp; 1119 char c; 1120 /* TODO - EXP_REDIR */ 1121 1122 while (str) { 1123 if (fflag) 1124 goto nometa; 1125 p = str->text; 1126 for (;;) { /* fast check for meta chars */ 1127 if ((c = *p++) == '\0') 1128 goto nometa; 1129 if (c == '*' || c == '?' || c == '[') 1130 break; 1131 } 1132 savelastp = exparg.lastp; 1133 INTOFF; 1134 expmeta(expdir, str->text); 1135 INTON; 1136 if (exparg.lastp == savelastp) { 1137 /* 1138 * no matches 1139 */ 1140nometa: 1141 *exparg.lastp = str; 1142 rmescapes(str->text); 1143 exparg.lastp = &str->next; 1144 } else { 1145 *exparg.lastp = NULL; 1146 *savelastp = sp = expsort(*savelastp); 1147 while (sp->next != NULL) 1148 sp = sp->next; 1149 exparg.lastp = &sp->next; 1150 } 1151 str = str->next; 1152 } 1153} 1154 1155 1156/* 1157 * Do metacharacter (i.e. *, ?, [...]) expansion. 1158 */ 1159 1160static void 1161expmeta(char *enddir, char *name) 1162{ 1163 char *p; 1164 char *q; 1165 char *start; 1166 char *endname; 1167 int metaflag; 1168 struct stat statb; 1169 DIR *dirp; 1170 struct dirent *dp; 1171 int atend; 1172 int matchdot; 1173 int esc; 1174 1175 metaflag = 0; 1176 start = name; 1177 for (p = name; esc = 0, *p; p += esc + 1) { 1178 if (*p == '*' || *p == '?') 1179 metaflag = 1; 1180 else if (*p == '[') { 1181 q = p + 1; 1182 if (*q == '!' || *q == '^') 1183 q++; 1184 for (;;) { 1185 while (*q == CTLQUOTEMARK) 1186 q++; 1187 if (*q == CTLESC) 1188 q++; 1189 if (*q == '/' || *q == '\0') 1190 break; 1191 if (*++q == ']') { 1192 metaflag = 1; 1193 break; 1194 } 1195 } 1196 } else if (*p == '\0') 1197 break; 1198 else if (*p == CTLQUOTEMARK) 1199 continue; 1200 else { 1201 if (*p == CTLESC) 1202 esc++; 1203 if (p[esc] == '/') { 1204 if (metaflag) 1205 break; 1206 start = p + esc + 1; 1207 } 1208 } 1209 } 1210 if (metaflag == 0) { /* we've reached the end of the file name */ 1211 if (enddir != expdir) 1212 metaflag++; 1213 for (p = name ; ; p++) { 1214 if (*p == CTLQUOTEMARK) 1215 continue; 1216 if (*p == CTLESC) 1217 p++; 1218 *enddir++ = *p; 1219 if (*p == '\0') 1220 break; 1221 if (enddir == expdir_end) 1222 return; 1223 } 1224 if (metaflag == 0 || lstat(expdir, &statb) >= 0) 1225 addfname(expdir); 1226 return; 1227 } 1228 endname = p; 1229 if (start != name) { 1230 p = name; 1231 while (p < start) { 1232 while (*p == CTLQUOTEMARK) 1233 p++; 1234 if (*p == CTLESC) 1235 p++; 1236 *enddir++ = *p++; 1237 if (enddir == expdir_end) 1238 return; 1239 } 1240 } 1241 if (enddir == expdir) { 1242 p = "."; 1243 } else if (enddir == expdir + 1 && *expdir == '/') { 1244 p = "/"; 1245 } else { 1246 p = expdir; 1247 enddir[-1] = '\0'; 1248 } 1249 if ((dirp = opendir(p)) == NULL) 1250 return; 1251 if (enddir != expdir) 1252 enddir[-1] = '/'; 1253 if (*endname == 0) { 1254 atend = 1; 1255 } else { 1256 atend = 0; 1257 *endname = '\0'; 1258 endname += esc + 1; 1259 } 1260 matchdot = 0; 1261 p = start; 1262 while (*p == CTLQUOTEMARK) 1263 p++; 1264 if (*p == CTLESC) 1265 p++; 1266 if (*p == '.') 1267 matchdot++; 1268 while (! int_pending() && (dp = readdir(dirp)) != NULL) { 1269 if (dp->d_name[0] == '.' && ! matchdot) 1270 continue; 1271 if (patmatch(start, dp->d_name, 0)) { 1272 if (enddir + dp->d_namlen + 1 > expdir_end) 1273 continue; 1274 memcpy(enddir, dp->d_name, dp->d_namlen + 1); 1275 if (atend) 1276 addfname(expdir); 1277 else { 1278 if (enddir + dp->d_namlen + 2 > expdir_end) 1279 continue; 1280 enddir[dp->d_namlen] = '/'; 1281 enddir[dp->d_namlen + 1] = '\0'; 1282 expmeta(enddir + dp->d_namlen + 1, endname); 1283 } 1284 } 1285 } 1286 closedir(dirp); 1287 if (! atend) 1288 endname[-esc - 1] = esc ? CTLESC : '/'; 1289} 1290 1291 1292/* 1293 * Add a file name to the list. 1294 */ 1295 1296static void 1297addfname(char *name) 1298{ 1299 char *p; 1300 struct strlist *sp; 1301 1302 p = stalloc(strlen(name) + 1); 1303 scopy(name, p); 1304 sp = (struct strlist *)stalloc(sizeof *sp); 1305 sp->text = p; 1306 *exparg.lastp = sp; 1307 exparg.lastp = &sp->next; 1308} 1309 1310 1311/* 1312 * Sort the results of file name expansion. It calculates the number of 1313 * strings to sort and then calls msort (short for merge sort) to do the 1314 * work. 1315 */ 1316 1317static struct strlist * 1318expsort(struct strlist *str) 1319{ 1320 int len; 1321 struct strlist *sp; 1322 1323 len = 0; 1324 for (sp = str ; sp ; sp = sp->next) 1325 len++; 1326 return msort(str, len); 1327} 1328 1329 1330static struct strlist * 1331msort(struct strlist *list, int len) 1332{ 1333 struct strlist *p, *q = NULL; 1334 struct strlist **lpp; 1335 int half; 1336 int n; 1337 1338 if (len <= 1) 1339 return list; 1340 half = len >> 1; 1341 p = list; 1342 for (n = half ; --n >= 0 ; ) { 1343 q = p; 1344 p = p->next; 1345 } 1346 q->next = NULL; /* terminate first half of list */ 1347 q = msort(list, half); /* sort first half of list */ 1348 p = msort(p, len - half); /* sort second half */ 1349 lpp = &list; 1350 for (;;) { 1351 if (strcmp(p->text, q->text) < 0) { 1352 *lpp = p; 1353 lpp = &p->next; 1354 if ((p = *lpp) == NULL) { 1355 *lpp = q; 1356 break; 1357 } 1358 } else { 1359 *lpp = q; 1360 lpp = &q->next; 1361 if ((q = *lpp) == NULL) { 1362 *lpp = p; 1363 break; 1364 } 1365 } 1366 } 1367 return list; 1368} 1369 1370 1371 1372/* 1373 * Returns true if the pattern matches the string. 1374 */ 1375 1376int 1377patmatch(const char *pattern, const char *string, int squoted) 1378{ 1379 const char *p, *q; 1380 char c; 1381 1382 p = pattern; 1383 q = string; 1384 for (;;) { 1385 switch (c = *p++) { 1386 case '\0': 1387 goto breakloop; 1388 case CTLESC: 1389 if (squoted && *q == CTLESC) 1390 q++; 1391 if (*q++ != *p++) 1392 return 0; 1393 break; 1394 case CTLQUOTEMARK: 1395 continue; 1396 case '?': 1397 if (squoted && *q == CTLESC) 1398 q++; 1399 if (*q++ == '\0') 1400 return 0; 1401 break; 1402 case '*': 1403 c = *p; 1404 while (c == CTLQUOTEMARK || c == '*') 1405 c = *++p; 1406 if (c != CTLESC && c != CTLQUOTEMARK && 1407 c != '?' && c != '*' && c != '[') { 1408 while (*q != c) { 1409 if (squoted && *q == CTLESC && 1410 q[1] == c) 1411 break; 1412 if (*q == '\0') 1413 return 0; 1414 if (squoted && *q == CTLESC) 1415 q++; 1416 q++; 1417 } 1418 } 1419 do { 1420 if (patmatch(p, q, squoted)) 1421 return 1; 1422 if (squoted && *q == CTLESC) 1423 q++; 1424 } while (*q++ != '\0'); 1425 return 0; 1426 case '[': { 1427 const char *endp; 1428 int invert, found; 1429 char chr; 1430 1431 endp = p; 1432 if (*endp == '!' || *endp == '^') 1433 endp++; 1434 for (;;) { 1435 while (*endp == CTLQUOTEMARK) 1436 endp++; 1437 if (*endp == '\0') 1438 goto dft; /* no matching ] */ 1439 if (*endp == CTLESC) 1440 endp++; 1441 if (*++endp == ']') 1442 break; 1443 } 1444 invert = 0; 1445 if (*p == '!' || *p == '^') { 1446 invert++; 1447 p++; 1448 } 1449 found = 0; 1450 chr = *q++; 1451 if (squoted && chr == CTLESC) 1452 chr = *q++; 1453 if (chr == '\0') 1454 return 0; 1455 c = *p++; 1456 do { 1457 if (c == CTLQUOTEMARK) 1458 continue; 1459 if (c == CTLESC) 1460 c = *p++; 1461 if (*p == '-' && p[1] != ']') { 1462 p++; 1463 while (*p == CTLQUOTEMARK) 1464 p++; 1465 if (*p == CTLESC) 1466 p++; 1467 if ( collate_range_cmp(chr, c) >= 0 1468 && collate_range_cmp(chr, *p) <= 0 1469 ) 1470 found = 1; 1471 p++; 1472 } else { 1473 if (chr == c) 1474 found = 1; 1475 } 1476 } while ((c = *p++) != ']'); 1477 if (found == invert) 1478 return 0; 1479 break; 1480 } 1481dft: default: 1482 if (squoted && *q == CTLESC) 1483 q++; 1484 if (*q++ != c) 1485 return 0; 1486 break; 1487 } 1488 } 1489breakloop: 1490 if (*q != '\0') 1491 return 0; 1492 return 1; 1493} 1494 1495 1496 1497/* 1498 * Remove any CTLESC and CTLQUOTEMARK characters from a string. 1499 */ 1500 1501void 1502rmescapes(char *str) 1503{ 1504 char *p, *q; 1505 1506 p = str; 1507 while (*p != CTLESC && *p != CTLQUOTEMARK && *p != CTLQUOTEEND) { 1508 if (*p++ == '\0') 1509 return; 1510 } 1511 q = p; 1512 while (*p) { 1513 if (*p == CTLQUOTEMARK || *p == CTLQUOTEEND) { 1514 p++; 1515 continue; 1516 } 1517 if (*p == CTLESC) 1518 p++; 1519 *q++ = *p++; 1520 } 1521 *q = '\0'; 1522} 1523 1524 1525 1526/* 1527 * See if a pattern matches in a case statement. 1528 */ 1529 1530int 1531casematch(union node *pattern, const char *val) 1532{ 1533 struct stackmark smark; 1534 int result; 1535 char *p; 1536 1537 setstackmark(&smark); 1538 argbackq = pattern->narg.backquote; 1539 STARTSTACKSTR(expdest); 1540 ifslastp = NULL; 1541 argstr(pattern->narg.text, EXP_TILDE | EXP_CASE); 1542 STPUTC('\0', expdest); 1543 p = grabstackstr(expdest); 1544 result = patmatch(p, val, 0); 1545 popstackmark(&smark); 1546 return result; 1547} 1548 1549/* 1550 * Our own itoa(). 1551 */ 1552 1553static char * 1554cvtnum(int num, char *buf) 1555{ 1556 char temp[32]; 1557 int neg = num < 0; 1558 char *p = temp + 31; 1559 1560 temp[31] = '\0'; 1561 1562 do { 1563 *--p = num % 10 + '0'; 1564 } while ((num /= 10) != 0); 1565 1566 if (neg) 1567 *--p = '-'; 1568 1569 STPUTS(p, buf); 1570 return buf; 1571} 1572 1573/* 1574 * Do most of the work for wordexp(3). 1575 */ 1576 1577int 1578wordexpcmd(int argc, char **argv) 1579{ 1580 size_t len; 1581 int i; 1582 1583 out1fmt("%08x", argc - 1); 1584 for (i = 1, len = 0; i < argc; i++) 1585 len += strlen(argv[i]); 1586 out1fmt("%08x", (int)len); 1587 for (i = 1; i < argc; i++) 1588 outbin(argv[i], strlen(argv[i]) + 1, out1); 1589 return (0); 1590} 1591