1 /*- 2 * Copyright (c) 1991, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * This code is derived from software contributed to Berkeley by 6 * Kenneth Almquist. 7 * 8 * %sccs.include.redist.c% 9 */ 10 11 #ifndef lint 12 static char sccsid[] = "@(#)parser.c 8.1 (Berkeley) 05/31/93"; 13 #endif /* not lint */ 14 15 #include "shell.h" 16 #include "parser.h" 17 #include "nodes.h" 18 #include "expand.h" /* defines rmescapes() */ 19 #include "redir.h" /* defines copyfd() */ 20 #include "syntax.h" 21 #include "options.h" 22 #include "input.h" 23 #include "output.h" 24 #include "var.h" 25 #include "error.h" 26 #include "memalloc.h" 27 #include "mystring.h" 28 #include "alias.h" 29 #include "myhistedit.h" 30 31 32 /* 33 * Shell command parser. 34 */ 35 36 #define EOFMARKLEN 79 37 38 /* values returned by readtoken */ 39 #include "token.def" 40 41 42 43 struct heredoc { 44 struct heredoc *next; /* next here document in list */ 45 union node *here; /* redirection node */ 46 char *eofmark; /* string indicating end of input */ 47 int striptabs; /* if set, strip leading tabs */ 48 }; 49 50 51 52 struct heredoc *heredoclist; /* list of here documents to read */ 53 int parsebackquote; /* nonzero if we are inside backquotes */ 54 int doprompt; /* if set, prompt the user */ 55 int needprompt; /* true if interactive and at start of line */ 56 int lasttoken; /* last token read */ 57 MKINIT int tokpushback; /* last token pushed back */ 58 char *wordtext; /* text of last word returned by readtoken */ 59 MKINIT int checkkwd; /* 1 == check for kwds, 2 == also eat newlines */ 60 struct nodelist *backquotelist; 61 union node *redirnode; 62 struct heredoc *heredoc; 63 int quoteflag; /* set if (part of) last token was quoted */ 64 int startlinno; /* line # where last token started */ 65 66 67 #define GDB_HACK 1 /* avoid local declarations which gdb can't handle */ 68 #ifdef GDB_HACK 69 static const char argvars[5] = {CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'}; 70 static const char types[] = "}-+?="; 71 #endif 72 73 74 STATIC union node *list __P((int)); 75 STATIC union node *andor __P((void)); 76 STATIC union node *pipeline __P((void)); 77 STATIC union node *command __P((void)); 78 STATIC union node *simplecmd __P((union node **, union node *)); 79 STATIC void parsefname __P((void)); 80 STATIC void parseheredoc __P((void)); 81 STATIC int readtoken __P((void)); 82 STATIC int readtoken1 __P((int, char const *, char *, int)); 83 STATIC void attyline __P((void)); 84 STATIC int noexpand __P((char *)); 85 STATIC void synexpect __P((int)); 86 STATIC void synerror __P((char *)); 87 STATIC void setprompt __P((int)); 88 89 /* 90 * Read and parse a command. Returns NEOF on end of file. (NULL is a 91 * valid parse tree indicating a blank line.) 92 */ 93 94 union node * 95 parsecmd(interact) { 96 int t; 97 98 doprompt = interact; 99 if (doprompt) 100 setprompt(1); 101 else 102 setprompt(0); 103 needprompt = 0; 104 t = readtoken(); 105 if (t == TEOF) 106 return NEOF; 107 if (t == TNL) 108 return NULL; 109 tokpushback++; 110 return list(1); 111 } 112 113 114 STATIC union node * 115 list(nlflag) { 116 union node *n1, *n2, *n3; 117 118 checkkwd = 2; 119 if (nlflag == 0 && tokendlist[peektoken()]) 120 return NULL; 121 n1 = andor(); 122 for (;;) { 123 switch (readtoken()) { 124 case TBACKGND: 125 if (n1->type == NCMD || n1->type == NPIPE) { 126 n1->ncmd.backgnd = 1; 127 } else if (n1->type == NREDIR) { 128 n1->type = NBACKGND; 129 } else { 130 n3 = (union node *)stalloc(sizeof (struct nredir)); 131 n3->type = NBACKGND; 132 n3->nredir.n = n1; 133 n3->nredir.redirect = NULL; 134 n1 = n3; 135 } 136 goto tsemi; 137 case TNL: 138 tokpushback++; 139 /* fall through */ 140 tsemi: case TSEMI: 141 if (readtoken() == TNL) { 142 parseheredoc(); 143 if (nlflag) 144 return n1; 145 } else { 146 tokpushback++; 147 } 148 checkkwd = 2; 149 if (tokendlist[peektoken()]) 150 return n1; 151 n2 = andor(); 152 n3 = (union node *)stalloc(sizeof (struct nbinary)); 153 n3->type = NSEMI; 154 n3->nbinary.ch1 = n1; 155 n3->nbinary.ch2 = n2; 156 n1 = n3; 157 break; 158 case TEOF: 159 if (heredoclist) 160 parseheredoc(); 161 else 162 pungetc(); /* push back EOF on input */ 163 return n1; 164 default: 165 if (nlflag) 166 synexpect(-1); 167 tokpushback++; 168 return n1; 169 } 170 } 171 } 172 173 174 175 STATIC union node * 176 andor() { 177 union node *n1, *n2, *n3; 178 int t; 179 180 n1 = pipeline(); 181 for (;;) { 182 if ((t = readtoken()) == TAND) { 183 t = NAND; 184 } else if (t == TOR) { 185 t = NOR; 186 } else { 187 tokpushback++; 188 return n1; 189 } 190 n2 = pipeline(); 191 n3 = (union node *)stalloc(sizeof (struct nbinary)); 192 n3->type = t; 193 n3->nbinary.ch1 = n1; 194 n3->nbinary.ch2 = n2; 195 n1 = n3; 196 } 197 } 198 199 200 201 STATIC union node * 202 pipeline() { 203 union node *n1, *pipenode, *notnode; 204 struct nodelist *lp, *prev; 205 int negate = 0; 206 207 TRACE(("pipeline: entered\n")); 208 while (readtoken() == TNOT) { 209 TRACE(("pipeline: TNOT recognized\n")); 210 negate = !negate; 211 } 212 tokpushback++; 213 n1 = command(); 214 if (readtoken() == TPIPE) { 215 pipenode = (union node *)stalloc(sizeof (struct npipe)); 216 pipenode->type = NPIPE; 217 pipenode->npipe.backgnd = 0; 218 lp = (struct nodelist *)stalloc(sizeof (struct nodelist)); 219 pipenode->npipe.cmdlist = lp; 220 lp->n = n1; 221 do { 222 prev = lp; 223 lp = (struct nodelist *)stalloc(sizeof (struct nodelist)); 224 lp->n = command(); 225 prev->next = lp; 226 } while (readtoken() == TPIPE); 227 lp->next = NULL; 228 n1 = pipenode; 229 } 230 tokpushback++; 231 if (negate) { 232 notnode = (union node *)stalloc(sizeof (struct nnot)); 233 notnode->type = NNOT; 234 notnode->nnot.com = n1; 235 n1 = notnode; 236 } 237 return n1; 238 } 239 240 241 242 STATIC union node * 243 command() { 244 union node *n1, *n2; 245 union node *ap, **app; 246 union node *cp, **cpp; 247 union node *redir, **rpp; 248 int t; 249 250 checkkwd = 2; 251 redir = 0; 252 rpp = &redir; 253 /* Check for redirection which may precede command */ 254 while (readtoken() == TREDIR) { 255 *rpp = n2 = redirnode; 256 rpp = &n2->nfile.next; 257 parsefname(); 258 } 259 tokpushback++; 260 261 switch (readtoken()) { 262 case TIF: 263 n1 = (union node *)stalloc(sizeof (struct nif)); 264 n1->type = NIF; 265 n1->nif.test = list(0); 266 if (readtoken() != TTHEN) 267 synexpect(TTHEN); 268 n1->nif.ifpart = list(0); 269 n2 = n1; 270 while (readtoken() == TELIF) { 271 n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif)); 272 n2 = n2->nif.elsepart; 273 n2->type = NIF; 274 n2->nif.test = list(0); 275 if (readtoken() != TTHEN) 276 synexpect(TTHEN); 277 n2->nif.ifpart = list(0); 278 } 279 if (lasttoken == TELSE) 280 n2->nif.elsepart = list(0); 281 else { 282 n2->nif.elsepart = NULL; 283 tokpushback++; 284 } 285 if (readtoken() != TFI) 286 synexpect(TFI); 287 checkkwd = 1; 288 break; 289 case TWHILE: 290 case TUNTIL: { 291 int got; 292 n1 = (union node *)stalloc(sizeof (struct nbinary)); 293 n1->type = (lasttoken == TWHILE)? NWHILE : NUNTIL; 294 n1->nbinary.ch1 = list(0); 295 if ((got=readtoken()) != TDO) { 296 TRACE(("expecting DO got %s %s\n", tokname[got], got == TWORD ? wordtext : "")); 297 synexpect(TDO); 298 } 299 n1->nbinary.ch2 = list(0); 300 if (readtoken() != TDONE) 301 synexpect(TDONE); 302 checkkwd = 1; 303 break; 304 } 305 case TFOR: 306 if (readtoken() != TWORD || quoteflag || ! goodname(wordtext)) 307 synerror("Bad for loop variable"); 308 n1 = (union node *)stalloc(sizeof (struct nfor)); 309 n1->type = NFOR; 310 n1->nfor.var = wordtext; 311 if (readtoken() == TWORD && ! quoteflag && equal(wordtext, "in")) { 312 app = ≈ 313 while (readtoken() == TWORD) { 314 n2 = (union node *)stalloc(sizeof (struct narg)); 315 n2->type = NARG; 316 n2->narg.text = wordtext; 317 n2->narg.backquote = backquotelist; 318 *app = n2; 319 app = &n2->narg.next; 320 } 321 *app = NULL; 322 n1->nfor.args = ap; 323 if (lasttoken != TNL && lasttoken != TSEMI) 324 synexpect(-1); 325 } else { 326 #ifndef GDB_HACK 327 static const char argvars[5] = {CTLVAR, VSNORMAL|VSQUOTE, 328 '@', '=', '\0'}; 329 #endif 330 n2 = (union node *)stalloc(sizeof (struct narg)); 331 n2->type = NARG; 332 n2->narg.text = (char *)argvars; 333 n2->narg.backquote = NULL; 334 n2->narg.next = NULL; 335 n1->nfor.args = n2; 336 /* 337 * Newline or semicolon here is optional (but note 338 * that the original Bourne shell only allowed NL). 339 */ 340 if (lasttoken != TNL && lasttoken != TSEMI) 341 tokpushback++; 342 } 343 checkkwd = 2; 344 if ((t = readtoken()) == TDO) 345 t = TDONE; 346 else if (t == TBEGIN) 347 t = TEND; 348 else 349 synexpect(-1); 350 n1->nfor.body = list(0); 351 if (readtoken() != t) 352 synexpect(t); 353 checkkwd = 1; 354 break; 355 case TCASE: 356 n1 = (union node *)stalloc(sizeof (struct ncase)); 357 n1->type = NCASE; 358 if (readtoken() != TWORD) 359 synexpect(TWORD); 360 n1->ncase.expr = n2 = (union node *)stalloc(sizeof (struct narg)); 361 n2->type = NARG; 362 n2->narg.text = wordtext; 363 n2->narg.backquote = backquotelist; 364 n2->narg.next = NULL; 365 while (readtoken() == TNL); 366 if (lasttoken != TWORD || ! equal(wordtext, "in")) 367 synerror("expecting \"in\""); 368 cpp = &n1->ncase.cases; 369 while (checkkwd = 2, readtoken() == TWORD) { 370 *cpp = cp = (union node *)stalloc(sizeof (struct nclist)); 371 cp->type = NCLIST; 372 app = &cp->nclist.pattern; 373 for (;;) { 374 *app = ap = (union node *)stalloc(sizeof (struct narg)); 375 ap->type = NARG; 376 ap->narg.text = wordtext; 377 ap->narg.backquote = backquotelist; 378 if (readtoken() != TPIPE) 379 break; 380 app = &ap->narg.next; 381 if (readtoken() != TWORD) 382 synexpect(TWORD); 383 } 384 ap->narg.next = NULL; 385 if (lasttoken != TRP) 386 synexpect(TRP); 387 cp->nclist.body = list(0); 388 if ((t = readtoken()) == TESAC) 389 tokpushback++; 390 else if (t != TENDCASE) 391 synexpect(TENDCASE); 392 cpp = &cp->nclist.next; 393 } 394 *cpp = NULL; 395 if (lasttoken != TESAC) 396 synexpect(TESAC); 397 checkkwd = 1; 398 break; 399 case TLP: 400 n1 = (union node *)stalloc(sizeof (struct nredir)); 401 n1->type = NSUBSHELL; 402 n1->nredir.n = list(0); 403 n1->nredir.redirect = NULL; 404 if (readtoken() != TRP) 405 synexpect(TRP); 406 checkkwd = 1; 407 break; 408 case TBEGIN: 409 n1 = list(0); 410 if (readtoken() != TEND) 411 synexpect(TEND); 412 checkkwd = 1; 413 break; 414 /* Handle an empty command like other simple commands. */ 415 case TNL: 416 case TWORD: 417 tokpushback++; 418 return simplecmd(rpp, redir); 419 default: 420 synexpect(-1); 421 } 422 423 /* Now check for redirection which may follow command */ 424 while (readtoken() == TREDIR) { 425 *rpp = n2 = redirnode; 426 rpp = &n2->nfile.next; 427 parsefname(); 428 } 429 tokpushback++; 430 *rpp = NULL; 431 if (redir) { 432 if (n1->type != NSUBSHELL) { 433 n2 = (union node *)stalloc(sizeof (struct nredir)); 434 n2->type = NREDIR; 435 n2->nredir.n = n1; 436 n1 = n2; 437 } 438 n1->nredir.redirect = redir; 439 } 440 return n1; 441 } 442 443 444 STATIC union node * 445 simplecmd(rpp, redir) 446 union node **rpp, *redir; 447 { 448 union node *args, **app; 449 union node **orig_rpp = rpp; 450 union node *n; 451 452 /* If we don't have any redirections already, then we must reset */ 453 /* rpp to be the address of the local redir variable. */ 454 if (redir == 0) 455 rpp = &redir; 456 457 args = NULL; 458 app = &args; 459 /* 460 * We save the incoming value, because we need this for shell 461 * functions. There can not be a redirect or an argument between 462 * the function name and the open parenthesis. 463 */ 464 orig_rpp = rpp; 465 466 for (;;) { 467 if (readtoken() == TWORD) { 468 n = (union node *)stalloc(sizeof (struct narg)); 469 n->type = NARG; 470 n->narg.text = wordtext; 471 n->narg.backquote = backquotelist; 472 *app = n; 473 app = &n->narg.next; 474 } else if (lasttoken == TREDIR) { 475 *rpp = n = redirnode; 476 rpp = &n->nfile.next; 477 parsefname(); /* read name of redirection file */ 478 } else if (lasttoken == TLP && app == &args->narg.next 479 && rpp == orig_rpp) { 480 /* We have a function */ 481 if (readtoken() != TRP) 482 synexpect(TRP); 483 #ifdef notdef 484 if (! goodname(n->narg.text)) 485 synerror("Bad function name"); 486 #endif 487 n->type = NDEFUN; 488 n->narg.next = command(); 489 return n; 490 } else { 491 tokpushback++; 492 break; 493 } 494 } 495 *app = NULL; 496 *rpp = NULL; 497 n = (union node *)stalloc(sizeof (struct ncmd)); 498 n->type = NCMD; 499 n->ncmd.backgnd = 0; 500 n->ncmd.args = args; 501 n->ncmd.redirect = redir; 502 return n; 503 } 504 505 506 STATIC void 507 parsefname() { 508 union node *n = redirnode; 509 510 if (readtoken() != TWORD) 511 synexpect(-1); 512 if (n->type == NHERE) { 513 struct heredoc *here = heredoc; 514 struct heredoc *p; 515 int i; 516 517 if (quoteflag == 0) 518 n->type = NXHERE; 519 TRACE(("Here document %d\n", n->type)); 520 if (here->striptabs) { 521 while (*wordtext == '\t') 522 wordtext++; 523 } 524 if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN) 525 synerror("Illegal eof marker for << redirection"); 526 rmescapes(wordtext); 527 here->eofmark = wordtext; 528 here->next = NULL; 529 if (heredoclist == NULL) 530 heredoclist = here; 531 else { 532 for (p = heredoclist ; p->next ; p = p->next); 533 p->next = here; 534 } 535 } else if (n->type == NTOFD || n->type == NFROMFD) { 536 if (is_digit(wordtext[0])) 537 n->ndup.dupfd = digit_val(wordtext[0]); 538 else if (wordtext[0] == '-') 539 n->ndup.dupfd = -1; 540 else 541 goto bad; 542 if (wordtext[1] != '\0') { 543 bad: 544 synerror("Bad fd number"); 545 } 546 } else { 547 n->nfile.fname = (union node *)stalloc(sizeof (struct narg)); 548 n = n->nfile.fname; 549 n->type = NARG; 550 n->narg.next = NULL; 551 n->narg.text = wordtext; 552 n->narg.backquote = backquotelist; 553 } 554 } 555 556 557 /* 558 * Input any here documents. 559 */ 560 561 STATIC void 562 parseheredoc() { 563 struct heredoc *here; 564 union node *n; 565 566 while (heredoclist) { 567 here = heredoclist; 568 heredoclist = here->next; 569 if (needprompt) { 570 setprompt(2); 571 needprompt = 0; 572 } 573 readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX, 574 here->eofmark, here->striptabs); 575 n = (union node *)stalloc(sizeof (struct narg)); 576 n->narg.type = NARG; 577 n->narg.next = NULL; 578 n->narg.text = wordtext; 579 n->narg.backquote = backquotelist; 580 here->here->nhere.doc = n; 581 } 582 } 583 584 STATIC int 585 peektoken() { 586 int t; 587 588 t = readtoken(); 589 tokpushback++; 590 return (t); 591 } 592 593 STATIC int xxreadtoken(); 594 595 STATIC int 596 readtoken() { 597 int t; 598 int savecheckkwd = checkkwd; 599 struct alias *ap; 600 #ifdef DEBUG 601 int alreadyseen = tokpushback; 602 #endif 603 604 top: 605 t = xxreadtoken(); 606 607 if (checkkwd) { 608 /* 609 * eat newlines 610 */ 611 if (checkkwd == 2) { 612 checkkwd = 0; 613 while (t == TNL) { 614 parseheredoc(); 615 t = xxreadtoken(); 616 } 617 } else 618 checkkwd = 0; 619 /* 620 * check for keywords and aliases 621 */ 622 if (t == TWORD && !quoteflag) { 623 register char * const *pp, *s; 624 625 for (pp = (char **)parsekwd; *pp; pp++) { 626 if (**pp == *wordtext && equal(*pp, wordtext)) { 627 lasttoken = t = pp - parsekwd + KWDOFFSET; 628 TRACE(("keyword %s recognized\n", tokname[t])); 629 goto out; 630 } 631 } 632 if (ap = lookupalias(wordtext, 1)) { 633 pushstring(ap->val, strlen(ap->val), ap); 634 checkkwd = savecheckkwd; 635 goto top; 636 } 637 } 638 out: 639 checkkwd = 0; 640 } 641 #ifdef DEBUG 642 if (!alreadyseen) 643 TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : "")); 644 else 645 TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : "")); 646 #endif 647 return (t); 648 } 649 650 651 /* 652 * Read the next input token. 653 * If the token is a word, we set backquotelist to the list of cmds in 654 * backquotes. We set quoteflag to true if any part of the word was 655 * quoted. 656 * If the token is TREDIR, then we set redirnode to a structure containing 657 * the redirection. 658 * In all cases, the variable startlinno is set to the number of the line 659 * on which the token starts. 660 * 661 * [Change comment: here documents and internal procedures] 662 * [Readtoken shouldn't have any arguments. Perhaps we should make the 663 * word parsing code into a separate routine. In this case, readtoken 664 * doesn't need to have any internal procedures, but parseword does. 665 * We could also make parseoperator in essence the main routine, and 666 * have parseword (readtoken1?) handle both words and redirection.] 667 */ 668 669 #define RETURN(token) return lasttoken = token 670 671 STATIC int 672 xxreadtoken() { 673 register c; 674 675 if (tokpushback) { 676 tokpushback = 0; 677 return lasttoken; 678 } 679 if (needprompt) { 680 setprompt(2); 681 needprompt = 0; 682 } 683 startlinno = plinno; 684 for (;;) { /* until token or start of word found */ 685 c = pgetc_macro(); 686 if (c == ' ' || c == '\t') 687 continue; /* quick check for white space first */ 688 switch (c) { 689 case ' ': case '\t': 690 continue; 691 case '#': 692 while ((c = pgetc()) != '\n' && c != PEOF); 693 pungetc(); 694 continue; 695 case '\\': 696 if (pgetc() == '\n') { 697 startlinno = ++plinno; 698 if (doprompt) 699 setprompt(2); 700 else 701 setprompt(0); 702 continue; 703 } 704 pungetc(); 705 goto breakloop; 706 case '\n': 707 plinno++; 708 needprompt = doprompt; 709 RETURN(TNL); 710 case PEOF: 711 RETURN(TEOF); 712 case '&': 713 if (pgetc() == '&') 714 RETURN(TAND); 715 pungetc(); 716 RETURN(TBACKGND); 717 case '|': 718 if (pgetc() == '|') 719 RETURN(TOR); 720 pungetc(); 721 RETURN(TPIPE); 722 case ';': 723 if (pgetc() == ';') 724 RETURN(TENDCASE); 725 pungetc(); 726 RETURN(TSEMI); 727 case '(': 728 RETURN(TLP); 729 case ')': 730 RETURN(TRP); 731 default: 732 goto breakloop; 733 } 734 } 735 breakloop: 736 return readtoken1(c, BASESYNTAX, (char *)NULL, 0); 737 #undef RETURN 738 } 739 740 741 742 /* 743 * If eofmark is NULL, read a word or a redirection symbol. If eofmark 744 * is not NULL, read a here document. In the latter case, eofmark is the 745 * word which marks the end of the document and striptabs is true if 746 * leading tabs should be stripped from the document. The argument firstc 747 * is the first character of the input token or document. 748 * 749 * Because C does not have internal subroutines, I have simulated them 750 * using goto's to implement the subroutine linkage. The following macros 751 * will run code that appears at the end of readtoken1. 752 */ 753 754 #define CHECKEND() {goto checkend; checkend_return:;} 755 #define PARSEREDIR() {goto parseredir; parseredir_return:;} 756 #define PARSESUB() {goto parsesub; parsesub_return:;} 757 #define PARSEBACKQOLD() {oldstyle = 1; goto parsebackq; parsebackq_oldreturn:;} 758 #define PARSEBACKQNEW() {oldstyle = 0; goto parsebackq; parsebackq_newreturn:;} 759 #define PARSEARITH() {goto parsearith; parsearith_return:;} 760 761 STATIC int 762 readtoken1(firstc, syntax, eofmark, striptabs) 763 int firstc; 764 char const *syntax; 765 char *eofmark; 766 int striptabs; 767 { 768 register c = firstc; 769 register char *out; 770 int len; 771 char line[EOFMARKLEN + 1]; 772 struct nodelist *bqlist; 773 int quotef; 774 int dblquote; 775 int varnest; /* levels of variables expansion */ 776 int arinest; /* levels of arithmetic expansion */ 777 int parenlevel; /* levels of parens in arithmetic */ 778 int oldstyle; 779 char const *prevsyntax; /* syntax before arithmetic */ 780 781 startlinno = plinno; 782 dblquote = 0; 783 if (syntax == DQSYNTAX) 784 dblquote = 1; 785 quotef = 0; 786 bqlist = NULL; 787 varnest = 0; 788 arinest = 0; 789 parenlevel = 0; 790 791 STARTSTACKSTR(out); 792 loop: { /* for each line, until end of word */ 793 #if ATTY 794 if (c == '\034' && doprompt 795 && attyset() && ! equal(termval(), "emacs")) { 796 attyline(); 797 if (syntax == BASESYNTAX) 798 return readtoken(); 799 c = pgetc(); 800 goto loop; 801 } 802 #endif 803 CHECKEND(); /* set c to PEOF if at end of here document */ 804 for (;;) { /* until end of line or end of word */ 805 CHECKSTRSPACE(3, out); /* permit 3 calls to USTPUTC */ 806 if (parsebackquote && c == '\\') { 807 c = pgetc(); /* XXX - compat with old /bin/sh */ 808 if (c != '\\' && c != '`' && c != '$') { 809 pungetc(); 810 c = '\\'; 811 } 812 } 813 switch(syntax[c]) { 814 case CNL: /* '\n' */ 815 if (syntax == BASESYNTAX) 816 goto endword; /* exit outer loop */ 817 USTPUTC(c, out); 818 plinno++; 819 if (doprompt) 820 setprompt(2); 821 else 822 setprompt(0); 823 c = pgetc(); 824 goto loop; /* continue outer loop */ 825 case CWORD: 826 USTPUTC(c, out); 827 break; 828 case CCTL: 829 if (eofmark == NULL || dblquote) 830 USTPUTC(CTLESC, out); 831 USTPUTC(c, out); 832 break; 833 case CBACK: /* backslash */ 834 c = pgetc(); 835 if (c == PEOF) { 836 USTPUTC('\\', out); 837 pungetc(); 838 } else if (c == '\n') { 839 if (doprompt) 840 setprompt(2); 841 else 842 setprompt(0); 843 } else { 844 if (dblquote && c != '\\' && c != '`' && c != '$' 845 && (c != '"' || eofmark != NULL)) 846 USTPUTC('\\', out); 847 if (SQSYNTAX[c] == CCTL) 848 USTPUTC(CTLESC, out); 849 USTPUTC(c, out); 850 quotef++; 851 } 852 break; 853 case CSQUOTE: 854 syntax = SQSYNTAX; 855 break; 856 case CDQUOTE: 857 syntax = DQSYNTAX; 858 dblquote = 1; 859 break; 860 case CENDQUOTE: 861 if (eofmark) { 862 USTPUTC(c, out); 863 } else { 864 if (arinest) 865 syntax = ARISYNTAX; 866 else 867 syntax = BASESYNTAX; 868 quotef++; 869 dblquote = 0; 870 } 871 break; 872 case CVAR: /* '$' */ 873 PARSESUB(); /* parse substitution */ 874 break; 875 case CENDVAR: /* '}' */ 876 if (varnest > 0) { 877 varnest--; 878 USTPUTC(CTLENDVAR, out); 879 } else { 880 USTPUTC(c, out); 881 } 882 break; 883 case CLP: /* '(' in arithmetic */ 884 parenlevel++; 885 USTPUTC(c, out); 886 break; 887 case CRP: /* ')' in arithmetic */ 888 if (parenlevel > 0) { 889 USTPUTC(c, out); 890 --parenlevel; 891 } else { 892 if (pgetc() == ')') { 893 if (--arinest == 0) { 894 USTPUTC(CTLENDARI, out); 895 syntax = prevsyntax; 896 } else 897 USTPUTC(')', out); 898 } else { 899 /* 900 * unbalanced parens 901 * (don't 2nd guess - no error) 902 */ 903 pungetc(); 904 USTPUTC(')', out); 905 } 906 } 907 break; 908 case CBQUOTE: /* '`' */ 909 PARSEBACKQOLD(); 910 break; 911 case CEOF: 912 goto endword; /* exit outer loop */ 913 default: 914 if (varnest == 0) 915 goto endword; /* exit outer loop */ 916 USTPUTC(c, out); 917 } 918 c = pgetc_macro(); 919 } 920 } 921 endword: 922 if (syntax == ARISYNTAX) 923 synerror("Missing '))'"); 924 if (syntax != BASESYNTAX && ! parsebackquote && eofmark == NULL) 925 synerror("Unterminated quoted string"); 926 if (varnest != 0) { 927 startlinno = plinno; 928 synerror("Missing '}'"); 929 } 930 USTPUTC('\0', out); 931 len = out - stackblock(); 932 out = stackblock(); 933 if (eofmark == NULL) { 934 if ((c == '>' || c == '<') 935 && quotef == 0 936 && len <= 2 937 && (*out == '\0' || is_digit(*out))) { 938 PARSEREDIR(); 939 return lasttoken = TREDIR; 940 } else { 941 pungetc(); 942 } 943 } 944 quoteflag = quotef; 945 backquotelist = bqlist; 946 grabstackblock(len); 947 wordtext = out; 948 return lasttoken = TWORD; 949 /* end of readtoken routine */ 950 951 952 953 /* 954 * Check to see whether we are at the end of the here document. When this 955 * is called, c is set to the first character of the next input line. If 956 * we are at the end of the here document, this routine sets the c to PEOF. 957 */ 958 959 checkend: { 960 if (eofmark) { 961 if (striptabs) { 962 while (c == '\t') 963 c = pgetc(); 964 } 965 if (c == *eofmark) { 966 if (pfgets(line, sizeof line) != NULL) { 967 register char *p, *q; 968 969 p = line; 970 for (q = eofmark + 1 ; *q && *p == *q ; p++, q++); 971 if (*p == '\n' && *q == '\0') { 972 c = PEOF; 973 plinno++; 974 needprompt = doprompt; 975 } else { 976 pushstring(line, strlen(line), NULL); 977 } 978 } 979 } 980 } 981 goto checkend_return; 982 } 983 984 985 /* 986 * Parse a redirection operator. The variable "out" points to a string 987 * specifying the fd to be redirected. The variable "c" contains the 988 * first character of the redirection operator. 989 */ 990 991 parseredir: { 992 char fd = *out; 993 union node *np; 994 995 np = (union node *)stalloc(sizeof (struct nfile)); 996 if (c == '>') { 997 np->nfile.fd = 1; 998 c = pgetc(); 999 if (c == '>') 1000 np->type = NAPPEND; 1001 else if (c == '&') 1002 np->type = NTOFD; 1003 else { 1004 np->type = NTO; 1005 pungetc(); 1006 } 1007 } else { /* c == '<' */ 1008 np->nfile.fd = 0; 1009 c = pgetc(); 1010 if (c == '<') { 1011 if (sizeof (struct nfile) != sizeof (struct nhere)) { 1012 np = (union node *)stalloc(sizeof (struct nhere)); 1013 np->nfile.fd = 0; 1014 } 1015 np->type = NHERE; 1016 heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc)); 1017 heredoc->here = np; 1018 if ((c = pgetc()) == '-') { 1019 heredoc->striptabs = 1; 1020 } else { 1021 heredoc->striptabs = 0; 1022 pungetc(); 1023 } 1024 } else if (c == '&') 1025 np->type = NFROMFD; 1026 else { 1027 np->type = NFROM; 1028 pungetc(); 1029 } 1030 } 1031 if (fd != '\0') 1032 np->nfile.fd = digit_val(fd); 1033 redirnode = np; 1034 goto parseredir_return; 1035 } 1036 1037 1038 /* 1039 * Parse a substitution. At this point, we have read the dollar sign 1040 * and nothing else. 1041 */ 1042 1043 parsesub: { 1044 int subtype; 1045 int typeloc; 1046 int flags; 1047 char *p; 1048 #ifndef GDB_HACK 1049 static const char types[] = "}-+?="; 1050 #endif 1051 1052 c = pgetc(); 1053 if (c != '(' && c != '{' && !is_name(c) && !is_special(c)) { 1054 USTPUTC('$', out); 1055 pungetc(); 1056 } else if (c == '(') { /* $(command) or $((arith)) */ 1057 if (pgetc() == '(') { 1058 PARSEARITH(); 1059 } else { 1060 pungetc(); 1061 PARSEBACKQNEW(); 1062 } 1063 } else { 1064 USTPUTC(CTLVAR, out); 1065 typeloc = out - stackblock(); 1066 USTPUTC(VSNORMAL, out); 1067 subtype = VSNORMAL; 1068 if (c == '{') { 1069 c = pgetc(); 1070 subtype = 0; 1071 } 1072 if (is_name(c)) { 1073 do { 1074 STPUTC(c, out); 1075 c = pgetc(); 1076 } while (is_in_name(c)); 1077 } else { 1078 if (! is_special(c)) 1079 badsub: synerror("Bad substitution"); 1080 USTPUTC(c, out); 1081 c = pgetc(); 1082 } 1083 STPUTC('=', out); 1084 flags = 0; 1085 if (subtype == 0) { 1086 if (c == ':') { 1087 flags = VSNUL; 1088 c = pgetc(); 1089 } 1090 p = strchr(types, c); 1091 if (p == NULL) 1092 goto badsub; 1093 subtype = p - types + VSNORMAL; 1094 } else { 1095 pungetc(); 1096 } 1097 if (dblquote || arinest) 1098 flags |= VSQUOTE; 1099 *(stackblock() + typeloc) = subtype | flags; 1100 if (subtype != VSNORMAL) 1101 varnest++; 1102 } 1103 goto parsesub_return; 1104 } 1105 1106 1107 /* 1108 * Called to parse command substitutions. Newstyle is set if the command 1109 * is enclosed inside $(...); nlpp is a pointer to the head of the linked 1110 * list of commands (passed by reference), and savelen is the number of 1111 * characters on the top of the stack which must be preserved. 1112 */ 1113 1114 parsebackq: { 1115 struct nodelist **nlpp; 1116 int savepbq; 1117 union node *n; 1118 char *volatile str; 1119 struct jmploc jmploc; 1120 struct jmploc *volatile savehandler; 1121 int savelen; 1122 1123 savepbq = parsebackquote; 1124 if (setjmp(jmploc.loc)) { 1125 if (str) 1126 ckfree(str); 1127 parsebackquote = 0; 1128 handler = savehandler; 1129 longjmp(handler->loc, 1); 1130 } 1131 INTOFF; 1132 str = NULL; 1133 savelen = out - stackblock(); 1134 if (savelen > 0) { 1135 str = ckmalloc(savelen); 1136 bcopy(stackblock(), str, savelen); 1137 } 1138 savehandler = handler; 1139 handler = &jmploc; 1140 INTON; 1141 if (oldstyle) { 1142 /* We must read until the closing backquote, giving special 1143 treatment to some slashes, and then push the string and 1144 reread it as input, interpreting it normally. */ 1145 register char *out; 1146 register c; 1147 int savelen; 1148 char *str; 1149 1150 STARTSTACKSTR(out); 1151 while ((c = pgetc ()) != '`') { 1152 if (c == '\\') { 1153 c = pgetc (); 1154 if (c != '\\' && c != '`' && c != '$' 1155 && (!dblquote || c != '"')) 1156 STPUTC('\\', out); 1157 } 1158 STPUTC(c, out); 1159 } 1160 STPUTC('\0', out); 1161 savelen = out - stackblock(); 1162 if (savelen > 0) { 1163 str = ckmalloc(savelen); 1164 bcopy(stackblock(), str, savelen); 1165 } 1166 setinputstring(str, 1); 1167 } 1168 nlpp = &bqlist; 1169 while (*nlpp) 1170 nlpp = &(*nlpp)->next; 1171 *nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist)); 1172 (*nlpp)->next = NULL; 1173 parsebackquote = oldstyle; 1174 n = list(0); 1175 if (!oldstyle && (readtoken() != TRP)) 1176 synexpect(TRP); 1177 (*nlpp)->n = n; 1178 /* Start reading from old file again. */ 1179 if (oldstyle) 1180 popfile(); 1181 while (stackblocksize() <= savelen) 1182 growstackblock(); 1183 STARTSTACKSTR(out); 1184 if (str) { 1185 bcopy(str, out, savelen); 1186 STADJUST(savelen, out); 1187 INTOFF; 1188 ckfree(str); 1189 str = NULL; 1190 INTON; 1191 } 1192 parsebackquote = savepbq; 1193 handler = savehandler; 1194 if (arinest || dblquote) 1195 USTPUTC(CTLBACKQ | CTLQUOTE, out); 1196 else 1197 USTPUTC(CTLBACKQ, out); 1198 if (oldstyle) 1199 goto parsebackq_oldreturn; 1200 else 1201 goto parsebackq_newreturn; 1202 } 1203 1204 /* 1205 * Parse an arithmetic expansion (indicate start of one and set state) 1206 */ 1207 parsearith: { 1208 1209 if (++arinest == 1) { 1210 prevsyntax = syntax; 1211 syntax = ARISYNTAX; 1212 USTPUTC(CTLARI, out); 1213 } else { 1214 /* 1215 * we collapse embedded arithmetic expansion to 1216 * parenthesis, which should be equivalent 1217 */ 1218 USTPUTC('(', out); 1219 } 1220 goto parsearith_return; 1221 } 1222 1223 } /* end of readtoken */ 1224 1225 1226 1227 #ifdef mkinit 1228 RESET { 1229 tokpushback = 0; 1230 checkkwd = 0; 1231 } 1232 #endif 1233 1234 /* 1235 * Returns true if the text contains nothing to expand (no dollar signs 1236 * or backquotes). 1237 */ 1238 1239 STATIC int 1240 noexpand(text) 1241 char *text; 1242 { 1243 register char *p; 1244 register char c; 1245 1246 p = text; 1247 while ((c = *p++) != '\0') { 1248 if (c == CTLESC) 1249 p++; 1250 else if (BASESYNTAX[c] == CCTL) 1251 return 0; 1252 } 1253 return 1; 1254 } 1255 1256 1257 /* 1258 * Return true if the argument is a legal variable name (a letter or 1259 * underscore followed by zero or more letters, underscores, and digits). 1260 */ 1261 1262 int 1263 goodname(name) 1264 char *name; 1265 { 1266 register char *p; 1267 1268 p = name; 1269 if (! is_name(*p)) 1270 return 0; 1271 while (*++p) { 1272 if (! is_in_name(*p)) 1273 return 0; 1274 } 1275 return 1; 1276 } 1277 1278 1279 /* 1280 * Called when an unexpected token is read during the parse. The argument 1281 * is the token that is expected, or -1 if more than one type of token can 1282 * occur at this point. 1283 */ 1284 1285 STATIC void 1286 synexpect(token) { 1287 char msg[64]; 1288 1289 if (token >= 0) { 1290 fmtstr(msg, 64, "%s unexpected (expecting %s)", 1291 tokname[lasttoken], tokname[token]); 1292 } else { 1293 fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]); 1294 } 1295 synerror(msg); 1296 } 1297 1298 1299 STATIC void 1300 synerror(msg) 1301 char *msg; 1302 { 1303 if (commandname) 1304 outfmt(&errout, "%s: %d: ", commandname, startlinno); 1305 outfmt(&errout, "Syntax error: %s\n", msg); 1306 error((char *)NULL); 1307 } 1308 1309 STATIC void 1310 setprompt(which) 1311 int which; 1312 { 1313 whichprompt = which; 1314 1315 if (!el) 1316 out2str(getprompt(NULL)); 1317 } 1318 1319 /* 1320 * called by editline -- any expansions to the prompt 1321 * should be added here. 1322 */ 1323 char * 1324 getprompt(unused) 1325 void *unused; 1326 { 1327 switch (whichprompt) { 1328 case 0: 1329 return ""; 1330 case 1: 1331 return ps1val(); 1332 case 2: 1333 return ps2val(); 1334 default: 1335 return "<internal prompt error>"; 1336 } 1337 } 1338