xref: /freebsd/bin/sh/parser.c (revision f7c32ed617858bcd22f8d1b03199099d50125721)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1991, 1993
5  *	The Regents of the University of California.  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  * 3. 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
37 static char sccsid[] = "@(#)parser.c	8.7 (Berkeley) 5/16/95";
38 #endif
39 #endif /* not lint */
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
42 
43 #include <sys/param.h>
44 #include <pwd.h>
45 #include <stdlib.h>
46 #include <unistd.h>
47 #include <stdio.h>
48 
49 #include "shell.h"
50 #include "parser.h"
51 #include "nodes.h"
52 #include "expand.h"	/* defines rmescapes() */
53 #include "syntax.h"
54 #include "options.h"
55 #include "input.h"
56 #include "output.h"
57 #include "var.h"
58 #include "error.h"
59 #include "memalloc.h"
60 #include "mystring.h"
61 #include "alias.h"
62 #include "show.h"
63 #include "eval.h"
64 #include "exec.h"	/* to check for special builtins */
65 #ifndef NO_HISTORY
66 #include "myhistedit.h"
67 #endif
68 
69 /*
70  * Shell command parser.
71  */
72 
73 #define	PROMPTLEN	128
74 
75 /* values of checkkwd variable */
76 #define CHKALIAS	0x1
77 #define CHKKWD		0x2
78 #define CHKNL		0x4
79 
80 /* values returned by readtoken */
81 #include "token.h"
82 
83 
84 
85 struct heredoc {
86 	struct heredoc *next;	/* next here document in list */
87 	union node *here;		/* redirection node */
88 	char *eofmark;		/* string indicating end of input */
89 	int striptabs;		/* if set, strip leading tabs */
90 };
91 
92 struct parser_temp {
93 	struct parser_temp *next;
94 	void *data;
95 };
96 
97 
98 static struct heredoc *heredoclist;	/* list of here documents to read */
99 static int doprompt;		/* if set, prompt the user */
100 static int needprompt;		/* true if interactive and at start of line */
101 static int lasttoken;		/* last token read */
102 static int tokpushback;		/* last token pushed back */
103 static char *wordtext;		/* text of last word returned by readtoken */
104 static int checkkwd;
105 static struct nodelist *backquotelist;
106 static union node *redirnode;
107 static struct heredoc *heredoc;
108 static int quoteflag;		/* set if (part of) last token was quoted */
109 static int startlinno;		/* line # where last token started */
110 static int funclinno;		/* line # where the current function started */
111 static struct parser_temp *parser_temp;
112 
113 #define NOEOFMARK ((const char *)&heredoclist)
114 
115 
116 static union node *list(int);
117 static union node *andor(void);
118 static union node *pipeline(void);
119 static union node *command(void);
120 static union node *simplecmd(union node **, union node *);
121 static union node *makename(void);
122 static union node *makebinary(int type, union node *n1, union node *n2);
123 static void parsefname(void);
124 static void parseheredoc(void);
125 static int peektoken(void);
126 static int readtoken(void);
127 static int xxreadtoken(void);
128 static int readtoken1(int, const char *, const char *, int);
129 static int noexpand(char *);
130 static void consumetoken(int);
131 static void synexpect(int) __dead2;
132 static void synerror(const char *) __dead2;
133 static void setprompt(int);
134 static int pgetc_linecont(void);
135 static void getusername(char *, size_t);
136 
137 
138 static void *
139 parser_temp_alloc(size_t len)
140 {
141 	struct parser_temp *t;
142 
143 	INTOFF;
144 	t = ckmalloc(sizeof(*t));
145 	t->data = NULL;
146 	t->next = parser_temp;
147 	parser_temp = t;
148 	t->data = ckmalloc(len);
149 	INTON;
150 	return t->data;
151 }
152 
153 
154 static void *
155 parser_temp_realloc(void *ptr, size_t len)
156 {
157 	struct parser_temp *t;
158 
159 	INTOFF;
160 	t = parser_temp;
161 	if (ptr != t->data)
162 		error("bug: parser_temp_realloc misused");
163 	t->data = ckrealloc(t->data, len);
164 	INTON;
165 	return t->data;
166 }
167 
168 
169 static void
170 parser_temp_free_upto(void *ptr)
171 {
172 	struct parser_temp *t;
173 	int done = 0;
174 
175 	INTOFF;
176 	while (parser_temp != NULL && !done) {
177 		t = parser_temp;
178 		parser_temp = t->next;
179 		done = t->data == ptr;
180 		ckfree(t->data);
181 		ckfree(t);
182 	}
183 	INTON;
184 	if (!done)
185 		error("bug: parser_temp_free_upto misused");
186 }
187 
188 
189 static void
190 parser_temp_free_all(void)
191 {
192 	struct parser_temp *t;
193 
194 	INTOFF;
195 	while (parser_temp != NULL) {
196 		t = parser_temp;
197 		parser_temp = t->next;
198 		ckfree(t->data);
199 		ckfree(t);
200 	}
201 	INTON;
202 }
203 
204 
205 /*
206  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
207  * valid parse tree indicating a blank line.)
208  */
209 
210 union node *
211 parsecmd(int interact)
212 {
213 	int t;
214 
215 	/* This assumes the parser is not re-entered,
216 	 * which could happen if we add command substitution on PS1/PS2.
217 	 */
218 	parser_temp_free_all();
219 	heredoclist = NULL;
220 
221 	tokpushback = 0;
222 	checkkwd = 0;
223 	doprompt = interact;
224 	if (doprompt)
225 		setprompt(1);
226 	else
227 		setprompt(0);
228 	needprompt = 0;
229 	t = readtoken();
230 	if (t == TEOF)
231 		return NEOF;
232 	if (t == TNL)
233 		return NULL;
234 	tokpushback++;
235 	return list(1);
236 }
237 
238 
239 /*
240  * Read and parse words for wordexp.
241  * Returns a list of NARG nodes; NULL if there are no words.
242  */
243 union node *
244 parsewordexp(void)
245 {
246 	union node *n, *first = NULL, **pnext;
247 	int t;
248 
249 	/* This assumes the parser is not re-entered,
250 	 * which could happen if we add command substitution on PS1/PS2.
251 	 */
252 	parser_temp_free_all();
253 	heredoclist = NULL;
254 
255 	tokpushback = 0;
256 	checkkwd = 0;
257 	doprompt = 0;
258 	setprompt(0);
259 	needprompt = 0;
260 	pnext = &first;
261 	while ((t = readtoken()) != TEOF) {
262 		if (t != TWORD)
263 			synexpect(TWORD);
264 		n = makename();
265 		*pnext = n;
266 		pnext = &n->narg.next;
267 	}
268 	return first;
269 }
270 
271 
272 static union node *
273 list(int nlflag)
274 {
275 	union node *ntop, *n1, *n2, *n3;
276 	int tok;
277 
278 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
279 	if (!nlflag && tokendlist[peektoken()])
280 		return NULL;
281 	ntop = n1 = NULL;
282 	for (;;) {
283 		n2 = andor();
284 		tok = readtoken();
285 		if (tok == TBACKGND) {
286 			if (n2 != NULL && n2->type == NPIPE) {
287 				n2->npipe.backgnd = 1;
288 			} else if (n2 != NULL && n2->type == NREDIR) {
289 				n2->type = NBACKGND;
290 			} else {
291 				n3 = (union node *)stalloc(sizeof (struct nredir));
292 				n3->type = NBACKGND;
293 				n3->nredir.n = n2;
294 				n3->nredir.redirect = NULL;
295 				n2 = n3;
296 			}
297 		}
298 		if (ntop == NULL)
299 			ntop = n2;
300 		else if (n1 == NULL) {
301 			n1 = makebinary(NSEMI, ntop, n2);
302 			ntop = n1;
303 		}
304 		else {
305 			n3 = makebinary(NSEMI, n1->nbinary.ch2, n2);
306 			n1->nbinary.ch2 = n3;
307 			n1 = n3;
308 		}
309 		switch (tok) {
310 		case TBACKGND:
311 		case TSEMI:
312 			tok = readtoken();
313 			/* FALLTHROUGH */
314 		case TNL:
315 			if (tok == TNL) {
316 				parseheredoc();
317 				if (nlflag)
318 					return ntop;
319 			} else if (tok == TEOF && nlflag) {
320 				parseheredoc();
321 				return ntop;
322 			} else {
323 				tokpushback++;
324 			}
325 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
326 			if (!nlflag && tokendlist[peektoken()])
327 				return ntop;
328 			break;
329 		case TEOF:
330 			if (heredoclist)
331 				parseheredoc();
332 			else
333 				pungetc();		/* push back EOF on input */
334 			return ntop;
335 		default:
336 			if (nlflag)
337 				synexpect(-1);
338 			tokpushback++;
339 			return ntop;
340 		}
341 	}
342 }
343 
344 
345 
346 static union node *
347 andor(void)
348 {
349 	union node *n;
350 	int t;
351 
352 	n = pipeline();
353 	for (;;) {
354 		if ((t = readtoken()) == TAND) {
355 			t = NAND;
356 		} else if (t == TOR) {
357 			t = NOR;
358 		} else {
359 			tokpushback++;
360 			return n;
361 		}
362 		n = makebinary(t, n, pipeline());
363 	}
364 }
365 
366 
367 
368 static union node *
369 pipeline(void)
370 {
371 	union node *n1, *n2, *pipenode;
372 	struct nodelist *lp, *prev;
373 	int negate, t;
374 
375 	negate = 0;
376 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
377 	TRACE(("pipeline: entered\n"));
378 	while (readtoken() == TNOT)
379 		negate = !negate;
380 	tokpushback++;
381 	n1 = command();
382 	if (readtoken() == TPIPE) {
383 		pipenode = (union node *)stalloc(sizeof (struct npipe));
384 		pipenode->type = NPIPE;
385 		pipenode->npipe.backgnd = 0;
386 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
387 		pipenode->npipe.cmdlist = lp;
388 		lp->n = n1;
389 		do {
390 			prev = lp;
391 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
392 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
393 			t = readtoken();
394 			tokpushback++;
395 			if (t == TNOT)
396 				lp->n = pipeline();
397 			else
398 				lp->n = command();
399 			prev->next = lp;
400 		} while (readtoken() == TPIPE);
401 		lp->next = NULL;
402 		n1 = pipenode;
403 	}
404 	tokpushback++;
405 	if (negate) {
406 		n2 = (union node *)stalloc(sizeof (struct nnot));
407 		n2->type = NNOT;
408 		n2->nnot.com = n1;
409 		return n2;
410 	} else
411 		return n1;
412 }
413 
414 
415 
416 static union node *
417 command(void)
418 {
419 	union node *n1, *n2;
420 	union node *ap, **app;
421 	union node *cp, **cpp;
422 	union node *redir, **rpp;
423 	int t;
424 	int is_subshell;
425 
426 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
427 	is_subshell = 0;
428 	redir = NULL;
429 	n1 = NULL;
430 	rpp = &redir;
431 
432 	/* Check for redirection which may precede command */
433 	while (readtoken() == TREDIR) {
434 		*rpp = n2 = redirnode;
435 		rpp = &n2->nfile.next;
436 		parsefname();
437 	}
438 	tokpushback++;
439 
440 	switch (readtoken()) {
441 	case TIF:
442 		n1 = (union node *)stalloc(sizeof (struct nif));
443 		n1->type = NIF;
444 		if ((n1->nif.test = list(0)) == NULL)
445 			synexpect(-1);
446 		consumetoken(TTHEN);
447 		n1->nif.ifpart = list(0);
448 		n2 = n1;
449 		while (readtoken() == TELIF) {
450 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
451 			n2 = n2->nif.elsepart;
452 			n2->type = NIF;
453 			if ((n2->nif.test = list(0)) == NULL)
454 				synexpect(-1);
455 			consumetoken(TTHEN);
456 			n2->nif.ifpart = list(0);
457 		}
458 		if (lasttoken == TELSE)
459 			n2->nif.elsepart = list(0);
460 		else {
461 			n2->nif.elsepart = NULL;
462 			tokpushback++;
463 		}
464 		consumetoken(TFI);
465 		checkkwd = CHKKWD | CHKALIAS;
466 		break;
467 	case TWHILE:
468 	case TUNTIL:
469 		t = lasttoken;
470 		if ((n1 = list(0)) == NULL)
471 			synexpect(-1);
472 		consumetoken(TDO);
473 		n1 = makebinary((t == TWHILE)? NWHILE : NUNTIL, n1, list(0));
474 		consumetoken(TDONE);
475 		checkkwd = CHKKWD | CHKALIAS;
476 		break;
477 	case TFOR:
478 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
479 			synerror("Bad for loop variable");
480 		n1 = (union node *)stalloc(sizeof (struct nfor));
481 		n1->type = NFOR;
482 		n1->nfor.var = wordtext;
483 		checkkwd = CHKNL;
484 		if (readtoken() == TWORD && !quoteflag &&
485 		    equal(wordtext, "in")) {
486 			app = &ap;
487 			while (readtoken() == TWORD) {
488 				n2 = makename();
489 				*app = n2;
490 				app = &n2->narg.next;
491 			}
492 			*app = NULL;
493 			n1->nfor.args = ap;
494 			if (lasttoken == TNL)
495 				tokpushback++;
496 			else if (lasttoken != TSEMI)
497 				synexpect(-1);
498 		} else {
499 			static char argvars[5] = {
500 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
501 			};
502 			n2 = (union node *)stalloc(sizeof (struct narg));
503 			n2->type = NARG;
504 			n2->narg.text = argvars;
505 			n2->narg.backquote = NULL;
506 			n2->narg.next = NULL;
507 			n1->nfor.args = n2;
508 			/*
509 			 * Newline or semicolon here is optional (but note
510 			 * that the original Bourne shell only allowed NL).
511 			 */
512 			if (lasttoken != TSEMI)
513 				tokpushback++;
514 		}
515 		checkkwd = CHKNL | CHKKWD | CHKALIAS;
516 		if ((t = readtoken()) == TDO)
517 			t = TDONE;
518 		else if (t == TBEGIN)
519 			t = TEND;
520 		else
521 			synexpect(-1);
522 		n1->nfor.body = list(0);
523 		consumetoken(t);
524 		checkkwd = CHKKWD | CHKALIAS;
525 		break;
526 	case TCASE:
527 		n1 = (union node *)stalloc(sizeof (struct ncase));
528 		n1->type = NCASE;
529 		consumetoken(TWORD);
530 		n1->ncase.expr = makename();
531 		checkkwd = CHKNL;
532 		if (readtoken() != TWORD || ! equal(wordtext, "in"))
533 			synerror("expecting \"in\"");
534 		cpp = &n1->ncase.cases;
535 		checkkwd = CHKNL | CHKKWD, readtoken();
536 		while (lasttoken != TESAC) {
537 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
538 			cp->type = NCLIST;
539 			app = &cp->nclist.pattern;
540 			if (lasttoken == TLP)
541 				readtoken();
542 			for (;;) {
543 				*app = ap = makename();
544 				checkkwd = CHKNL | CHKKWD;
545 				if (readtoken() != TPIPE)
546 					break;
547 				app = &ap->narg.next;
548 				readtoken();
549 			}
550 			ap->narg.next = NULL;
551 			if (lasttoken != TRP)
552 				synexpect(TRP);
553 			cp->nclist.body = list(0);
554 
555 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
556 			if ((t = readtoken()) != TESAC) {
557 				if (t == TENDCASE)
558 					;
559 				else if (t == TFALLTHRU)
560 					cp->type = NCLISTFALLTHRU;
561 				else
562 					synexpect(TENDCASE);
563 				checkkwd = CHKNL | CHKKWD, readtoken();
564 			}
565 			cpp = &cp->nclist.next;
566 		}
567 		*cpp = NULL;
568 		checkkwd = CHKKWD | CHKALIAS;
569 		break;
570 	case TLP:
571 		n1 = (union node *)stalloc(sizeof (struct nredir));
572 		n1->type = NSUBSHELL;
573 		n1->nredir.n = list(0);
574 		n1->nredir.redirect = NULL;
575 		consumetoken(TRP);
576 		checkkwd = CHKKWD | CHKALIAS;
577 		is_subshell = 1;
578 		break;
579 	case TBEGIN:
580 		n1 = list(0);
581 		consumetoken(TEND);
582 		checkkwd = CHKKWD | CHKALIAS;
583 		break;
584 	/* A simple command must have at least one redirection or word. */
585 	case TBACKGND:
586 	case TSEMI:
587 	case TAND:
588 	case TOR:
589 	case TPIPE:
590 	case TENDCASE:
591 	case TFALLTHRU:
592 	case TEOF:
593 	case TNL:
594 	case TRP:
595 		if (!redir)
596 			synexpect(-1);
597 	case TWORD:
598 		tokpushback++;
599 		n1 = simplecmd(rpp, redir);
600 		return n1;
601 	default:
602 		synexpect(-1);
603 	}
604 
605 	/* Now check for redirection which may follow command */
606 	while (readtoken() == TREDIR) {
607 		*rpp = n2 = redirnode;
608 		rpp = &n2->nfile.next;
609 		parsefname();
610 	}
611 	tokpushback++;
612 	*rpp = NULL;
613 	if (redir) {
614 		if (!is_subshell) {
615 			n2 = (union node *)stalloc(sizeof (struct nredir));
616 			n2->type = NREDIR;
617 			n2->nredir.n = n1;
618 			n1 = n2;
619 		}
620 		n1->nredir.redirect = redir;
621 	}
622 
623 	return n1;
624 }
625 
626 
627 static union node *
628 simplecmd(union node **rpp, union node *redir)
629 {
630 	union node *args, **app;
631 	union node **orig_rpp = rpp;
632 	union node *n = NULL;
633 	int special;
634 	int savecheckkwd;
635 
636 	/* If we don't have any redirections already, then we must reset */
637 	/* rpp to be the address of the local redir variable.  */
638 	if (redir == NULL)
639 		rpp = &redir;
640 
641 	args = NULL;
642 	app = &args;
643 	/*
644 	 * We save the incoming value, because we need this for shell
645 	 * functions.  There can not be a redirect or an argument between
646 	 * the function name and the open parenthesis.
647 	 */
648 	orig_rpp = rpp;
649 
650 	savecheckkwd = CHKALIAS;
651 
652 	for (;;) {
653 		checkkwd = savecheckkwd;
654 		if (readtoken() == TWORD) {
655 			n = makename();
656 			*app = n;
657 			app = &n->narg.next;
658 			if (savecheckkwd != 0 && !isassignment(wordtext))
659 				savecheckkwd = 0;
660 		} else if (lasttoken == TREDIR) {
661 			*rpp = n = redirnode;
662 			rpp = &n->nfile.next;
663 			parsefname();	/* read name of redirection file */
664 		} else if (lasttoken == TLP && app == &args->narg.next
665 					    && rpp == orig_rpp) {
666 			/* We have a function */
667 			consumetoken(TRP);
668 			funclinno = plinno;
669 			/*
670 			 * - Require plain text.
671 			 * - Functions with '/' cannot be called.
672 			 * - Reject name=().
673 			 * - Reject ksh extended glob patterns.
674 			 */
675 			if (!noexpand(n->narg.text) || quoteflag ||
676 			    strchr(n->narg.text, '/') ||
677 			    strchr("!%*+-=?@}~",
678 				n->narg.text[strlen(n->narg.text) - 1]))
679 				synerror("Bad function name");
680 			rmescapes(n->narg.text);
681 			if (find_builtin(n->narg.text, &special) >= 0 &&
682 			    special)
683 				synerror("Cannot override a special builtin with a function");
684 			n->type = NDEFUN;
685 			n->narg.next = command();
686 			funclinno = 0;
687 			return n;
688 		} else {
689 			tokpushback++;
690 			break;
691 		}
692 	}
693 	*app = NULL;
694 	*rpp = NULL;
695 	n = (union node *)stalloc(sizeof (struct ncmd));
696 	n->type = NCMD;
697 	n->ncmd.args = args;
698 	n->ncmd.redirect = redir;
699 	return n;
700 }
701 
702 static union node *
703 makename(void)
704 {
705 	union node *n;
706 
707 	n = (union node *)stalloc(sizeof (struct narg));
708 	n->type = NARG;
709 	n->narg.next = NULL;
710 	n->narg.text = wordtext;
711 	n->narg.backquote = backquotelist;
712 	return n;
713 }
714 
715 static union node *
716 makebinary(int type, union node *n1, union node *n2)
717 {
718 	union node *n;
719 
720 	n = (union node *)stalloc(sizeof (struct nbinary));
721 	n->type = type;
722 	n->nbinary.ch1 = n1;
723 	n->nbinary.ch2 = n2;
724 	return (n);
725 }
726 
727 void
728 forcealias(void)
729 {
730 	checkkwd |= CHKALIAS;
731 }
732 
733 void
734 fixredir(union node *n, const char *text, int err)
735 {
736 	TRACE(("Fix redir %s %d\n", text, err));
737 	if (!err)
738 		n->ndup.vname = NULL;
739 
740 	if (is_digit(text[0]) && text[1] == '\0')
741 		n->ndup.dupfd = digit_val(text[0]);
742 	else if (text[0] == '-' && text[1] == '\0')
743 		n->ndup.dupfd = -1;
744 	else {
745 
746 		if (err)
747 			synerror("Bad fd number");
748 		else
749 			n->ndup.vname = makename();
750 	}
751 }
752 
753 
754 static void
755 parsefname(void)
756 {
757 	union node *n = redirnode;
758 
759 	consumetoken(TWORD);
760 	if (n->type == NHERE) {
761 		struct heredoc *here = heredoc;
762 		struct heredoc *p;
763 
764 		if (quoteflag == 0)
765 			n->type = NXHERE;
766 		TRACE(("Here document %d\n", n->type));
767 		if (here->striptabs) {
768 			while (*wordtext == '\t')
769 				wordtext++;
770 		}
771 		if (! noexpand(wordtext))
772 			synerror("Illegal eof marker for << redirection");
773 		rmescapes(wordtext);
774 		here->eofmark = wordtext;
775 		here->next = NULL;
776 		if (heredoclist == NULL)
777 			heredoclist = here;
778 		else {
779 			for (p = heredoclist ; p->next ; p = p->next);
780 			p->next = here;
781 		}
782 	} else if (n->type == NTOFD || n->type == NFROMFD) {
783 		fixredir(n, wordtext, 0);
784 	} else {
785 		n->nfile.fname = makename();
786 	}
787 }
788 
789 
790 /*
791  * Input any here documents.
792  */
793 
794 static void
795 parseheredoc(void)
796 {
797 	struct heredoc *here;
798 	union node *n;
799 
800 	while (heredoclist) {
801 		here = heredoclist;
802 		heredoclist = here->next;
803 		if (needprompt) {
804 			setprompt(2);
805 			needprompt = 0;
806 		}
807 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
808 				here->eofmark, here->striptabs);
809 		n = makename();
810 		here->here->nhere.doc = n;
811 	}
812 }
813 
814 static int
815 peektoken(void)
816 {
817 	int t;
818 
819 	t = readtoken();
820 	tokpushback++;
821 	return (t);
822 }
823 
824 static int
825 readtoken(void)
826 {
827 	int t;
828 	struct alias *ap;
829 #ifdef DEBUG
830 	int alreadyseen = tokpushback;
831 #endif
832 
833 	top:
834 	t = xxreadtoken();
835 
836 	/*
837 	 * eat newlines
838 	 */
839 	if (checkkwd & CHKNL) {
840 		while (t == TNL) {
841 			parseheredoc();
842 			t = xxreadtoken();
843 		}
844 	}
845 
846 	/*
847 	 * check for keywords and aliases
848 	 */
849 	if (t == TWORD && !quoteflag)
850 	{
851 		const char * const *pp;
852 
853 		if (checkkwd & CHKKWD)
854 			for (pp = parsekwd; *pp; pp++) {
855 				if (**pp == *wordtext && equal(*pp, wordtext))
856 				{
857 					lasttoken = t = pp - parsekwd + KWDOFFSET;
858 					TRACE(("keyword %s recognized\n", tokname[t]));
859 					goto out;
860 				}
861 			}
862 		if (checkkwd & CHKALIAS &&
863 		    (ap = lookupalias(wordtext, 1)) != NULL) {
864 			pushstring(ap->val, strlen(ap->val), ap);
865 			goto top;
866 		}
867 	}
868 out:
869 	if (t != TNOT)
870 		checkkwd = 0;
871 
872 #ifdef DEBUG
873 	if (!alreadyseen)
874 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
875 	else
876 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
877 #endif
878 	return (t);
879 }
880 
881 
882 /*
883  * Read the next input token.
884  * If the token is a word, we set backquotelist to the list of cmds in
885  *	backquotes.  We set quoteflag to true if any part of the word was
886  *	quoted.
887  * If the token is TREDIR, then we set redirnode to a structure containing
888  *	the redirection.
889  * In all cases, the variable startlinno is set to the number of the line
890  *	on which the token starts.
891  *
892  * [Change comment:  here documents and internal procedures]
893  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
894  *  word parsing code into a separate routine.  In this case, readtoken
895  *  doesn't need to have any internal procedures, but parseword does.
896  *  We could also make parseoperator in essence the main routine, and
897  *  have parseword (readtoken1?) handle both words and redirection.]
898  */
899 
900 #define RETURN(token)	return lasttoken = token
901 
902 static int
903 xxreadtoken(void)
904 {
905 	int c;
906 
907 	if (tokpushback) {
908 		tokpushback = 0;
909 		return lasttoken;
910 	}
911 	if (needprompt) {
912 		setprompt(2);
913 		needprompt = 0;
914 	}
915 	startlinno = plinno;
916 	for (;;) {	/* until token or start of word found */
917 		c = pgetc_macro();
918 		switch (c) {
919 		case ' ': case '\t':
920 			continue;
921 		case '#':
922 			while ((c = pgetc()) != '\n' && c != PEOF);
923 			pungetc();
924 			continue;
925 		case '\\':
926 			if (pgetc() == '\n') {
927 				startlinno = ++plinno;
928 				if (doprompt)
929 					setprompt(2);
930 				else
931 					setprompt(0);
932 				continue;
933 			}
934 			pungetc();
935 			/* FALLTHROUGH */
936 		default:
937 			return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
938 		case '\n':
939 			plinno++;
940 			needprompt = doprompt;
941 			RETURN(TNL);
942 		case PEOF:
943 			RETURN(TEOF);
944 		case '&':
945 			if (pgetc_linecont() == '&')
946 				RETURN(TAND);
947 			pungetc();
948 			RETURN(TBACKGND);
949 		case '|':
950 			if (pgetc_linecont() == '|')
951 				RETURN(TOR);
952 			pungetc();
953 			RETURN(TPIPE);
954 		case ';':
955 			c = pgetc_linecont();
956 			if (c == ';')
957 				RETURN(TENDCASE);
958 			else if (c == '&')
959 				RETURN(TFALLTHRU);
960 			pungetc();
961 			RETURN(TSEMI);
962 		case '(':
963 			RETURN(TLP);
964 		case ')':
965 			RETURN(TRP);
966 		}
967 	}
968 #undef RETURN
969 }
970 
971 
972 #define MAXNEST_static 8
973 struct tokenstate
974 {
975 	const char *syntax; /* *SYNTAX */
976 	int parenlevel; /* levels of parentheses in arithmetic */
977 	enum tokenstate_category
978 	{
979 		TSTATE_TOP,
980 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
981 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
982 		TSTATE_ARITH
983 	} category;
984 };
985 
986 
987 /*
988  * Check to see whether we are at the end of the here document.  When this
989  * is called, c is set to the first character of the next input line.  If
990  * we are at the end of the here document, this routine sets the c to PEOF.
991  * The new value of c is returned.
992  */
993 
994 static int
995 checkend(int c, const char *eofmark, int striptabs)
996 {
997 	if (striptabs) {
998 		while (c == '\t')
999 			c = pgetc();
1000 	}
1001 	if (c == *eofmark) {
1002 		int c2;
1003 		const char *q;
1004 
1005 		for (q = eofmark + 1; c2 = pgetc(), *q != '\0' && c2 == *q; q++)
1006 			;
1007 		if ((c2 == PEOF || c2 == '\n') && *q == '\0') {
1008 			c = PEOF;
1009 			if (c2 == '\n') {
1010 				plinno++;
1011 				needprompt = doprompt;
1012 			}
1013 		} else {
1014 			pungetc();
1015 			pushstring(eofmark + 1, q - (eofmark + 1), NULL);
1016 		}
1017 	} else if (c == '\n' && *eofmark == '\0') {
1018 		c = PEOF;
1019 		plinno++;
1020 		needprompt = doprompt;
1021 	}
1022 	return (c);
1023 }
1024 
1025 
1026 /*
1027  * Parse a redirection operator.  The variable "out" points to a string
1028  * specifying the fd to be redirected.  The variable "c" contains the
1029  * first character of the redirection operator.
1030  */
1031 
1032 static void
1033 parseredir(char *out, int c)
1034 {
1035 	char fd = *out;
1036 	union node *np;
1037 
1038 	np = (union node *)stalloc(sizeof (struct nfile));
1039 	if (c == '>') {
1040 		np->nfile.fd = 1;
1041 		c = pgetc_linecont();
1042 		if (c == '>')
1043 			np->type = NAPPEND;
1044 		else if (c == '&')
1045 			np->type = NTOFD;
1046 		else if (c == '|')
1047 			np->type = NCLOBBER;
1048 		else {
1049 			np->type = NTO;
1050 			pungetc();
1051 		}
1052 	} else {	/* c == '<' */
1053 		np->nfile.fd = 0;
1054 		c = pgetc_linecont();
1055 		if (c == '<') {
1056 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1057 				np = (union node *)stalloc(sizeof (struct nhere));
1058 				np->nfile.fd = 0;
1059 			}
1060 			np->type = NHERE;
1061 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1062 			heredoc->here = np;
1063 			if ((c = pgetc_linecont()) == '-') {
1064 				heredoc->striptabs = 1;
1065 			} else {
1066 				heredoc->striptabs = 0;
1067 				pungetc();
1068 			}
1069 		} else if (c == '&')
1070 			np->type = NFROMFD;
1071 		else if (c == '>')
1072 			np->type = NFROMTO;
1073 		else {
1074 			np->type = NFROM;
1075 			pungetc();
1076 		}
1077 	}
1078 	if (fd != '\0')
1079 		np->nfile.fd = digit_val(fd);
1080 	redirnode = np;
1081 }
1082 
1083 /*
1084  * Called to parse command substitutions.
1085  */
1086 
1087 static char *
1088 parsebackq(char *out, struct nodelist **pbqlist,
1089 		int oldstyle, int dblquote, int quoted)
1090 {
1091 	struct nodelist **nlpp;
1092 	union node *n;
1093 	char *volatile str;
1094 	struct jmploc jmploc;
1095 	struct jmploc *const savehandler = handler;
1096 	size_t savelen;
1097 	int saveprompt;
1098 	const int bq_startlinno = plinno;
1099 	char *volatile ostr = NULL;
1100 	struct parsefile *const savetopfile = getcurrentfile();
1101 	struct heredoc *const saveheredoclist = heredoclist;
1102 	struct heredoc *here;
1103 
1104 	str = NULL;
1105 	if (setjmp(jmploc.loc)) {
1106 		popfilesupto(savetopfile);
1107 		if (str)
1108 			ckfree(str);
1109 		if (ostr)
1110 			ckfree(ostr);
1111 		heredoclist = saveheredoclist;
1112 		handler = savehandler;
1113 		if (exception == EXERROR) {
1114 			startlinno = bq_startlinno;
1115 			synerror("Error in command substitution");
1116 		}
1117 		longjmp(handler->loc, 1);
1118 	}
1119 	INTOFF;
1120 	savelen = out - stackblock();
1121 	if (savelen > 0) {
1122 		str = ckmalloc(savelen);
1123 		memcpy(str, stackblock(), savelen);
1124 	}
1125 	handler = &jmploc;
1126 	heredoclist = NULL;
1127 	INTON;
1128         if (oldstyle) {
1129                 /* We must read until the closing backquote, giving special
1130                    treatment to some slashes, and then push the string and
1131                    reread it as input, interpreting it normally.  */
1132                 char *oout;
1133                 int c;
1134                 int olen;
1135 
1136 
1137                 STARTSTACKSTR(oout);
1138 		for (;;) {
1139 			if (needprompt) {
1140 				setprompt(2);
1141 				needprompt = 0;
1142 			}
1143 			CHECKSTRSPACE(2, oout);
1144 			c = pgetc_linecont();
1145 			if (c == '`')
1146 				break;
1147 			switch (c) {
1148 			case '\\':
1149 				c = pgetc();
1150                                 if (c != '\\' && c != '`' && c != '$'
1151                                     && (!dblquote || c != '"'))
1152                                         USTPUTC('\\', oout);
1153 				break;
1154 
1155 			case '\n':
1156 				plinno++;
1157 				needprompt = doprompt;
1158 				break;
1159 
1160 			case PEOF:
1161 			        startlinno = plinno;
1162 				synerror("EOF in backquote substitution");
1163  				break;
1164 
1165 			default:
1166 				break;
1167 			}
1168 			USTPUTC(c, oout);
1169                 }
1170                 USTPUTC('\0', oout);
1171                 olen = oout - stackblock();
1172 		INTOFF;
1173 		ostr = ckmalloc(olen);
1174 		memcpy(ostr, stackblock(), olen);
1175 		setinputstring(ostr, 1);
1176 		INTON;
1177         }
1178 	nlpp = pbqlist;
1179 	while (*nlpp)
1180 		nlpp = &(*nlpp)->next;
1181 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1182 	(*nlpp)->next = NULL;
1183 
1184 	if (oldstyle) {
1185 		saveprompt = doprompt;
1186 		doprompt = 0;
1187 	}
1188 
1189 	n = list(0);
1190 
1191 	if (oldstyle) {
1192 		if (peektoken() != TEOF)
1193 			synexpect(-1);
1194 		doprompt = saveprompt;
1195 	} else
1196 		consumetoken(TRP);
1197 
1198 	(*nlpp)->n = n;
1199         if (oldstyle) {
1200 		/*
1201 		 * Start reading from old file again, ignoring any pushed back
1202 		 * tokens left from the backquote parsing
1203 		 */
1204                 popfile();
1205 		tokpushback = 0;
1206 	}
1207 	STARTSTACKSTR(out);
1208 	CHECKSTRSPACE(savelen + 1, out);
1209 	INTOFF;
1210 	if (str) {
1211 		memcpy(out, str, savelen);
1212 		STADJUST(savelen, out);
1213 		ckfree(str);
1214 		str = NULL;
1215 	}
1216 	if (ostr) {
1217 		ckfree(ostr);
1218 		ostr = NULL;
1219 	}
1220 	here = saveheredoclist;
1221 	if (here != NULL) {
1222 		while (here->next != NULL)
1223 			here = here->next;
1224 		here->next = heredoclist;
1225 		heredoclist = saveheredoclist;
1226 	}
1227 	handler = savehandler;
1228 	INTON;
1229 	if (quoted)
1230 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1231 	else
1232 		USTPUTC(CTLBACKQ, out);
1233 	return out;
1234 }
1235 
1236 
1237 /*
1238  * Called to parse a backslash escape sequence inside $'...'.
1239  * The backslash has already been read.
1240  */
1241 static char *
1242 readcstyleesc(char *out)
1243 {
1244 	int c, vc, i, n;
1245 	unsigned int v;
1246 
1247 	c = pgetc();
1248 	switch (c) {
1249 	case '\0':
1250 		synerror("Unterminated quoted string");
1251 	case '\n':
1252 		plinno++;
1253 		if (doprompt)
1254 			setprompt(2);
1255 		else
1256 			setprompt(0);
1257 		return out;
1258 	case '\\':
1259 	case '\'':
1260 	case '"':
1261 		v = c;
1262 		break;
1263 	case 'a': v = '\a'; break;
1264 	case 'b': v = '\b'; break;
1265 	case 'e': v = '\033'; break;
1266 	case 'f': v = '\f'; break;
1267 	case 'n': v = '\n'; break;
1268 	case 'r': v = '\r'; break;
1269 	case 't': v = '\t'; break;
1270 	case 'v': v = '\v'; break;
1271 	case 'x':
1272 		  v = 0;
1273 		  for (;;) {
1274 			  c = pgetc();
1275 			  if (c >= '0' && c <= '9')
1276 				  v = (v << 4) + c - '0';
1277 			  else if (c >= 'A' && c <= 'F')
1278 				  v = (v << 4) + c - 'A' + 10;
1279 			  else if (c >= 'a' && c <= 'f')
1280 				  v = (v << 4) + c - 'a' + 10;
1281 			  else
1282 				  break;
1283 		  }
1284 		  pungetc();
1285 		  break;
1286 	case '0': case '1': case '2': case '3':
1287 	case '4': case '5': case '6': case '7':
1288 		  v = c - '0';
1289 		  c = pgetc();
1290 		  if (c >= '0' && c <= '7') {
1291 			  v <<= 3;
1292 			  v += c - '0';
1293 			  c = pgetc();
1294 			  if (c >= '0' && c <= '7') {
1295 				  v <<= 3;
1296 				  v += c - '0';
1297 			  } else
1298 				  pungetc();
1299 		  } else
1300 			  pungetc();
1301 		  break;
1302 	case 'c':
1303 		  c = pgetc();
1304 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1305 			  synerror("Bad escape sequence");
1306 		  if (c == '\\' && pgetc() != '\\')
1307 			  synerror("Bad escape sequence");
1308 		  if (c == '?')
1309 			  v = 127;
1310 		  else
1311 			  v = c & 0x1f;
1312 		  break;
1313 	case 'u':
1314 	case 'U':
1315 		  n = c == 'U' ? 8 : 4;
1316 		  v = 0;
1317 		  for (i = 0; i < n; i++) {
1318 			  c = pgetc();
1319 			  if (c >= '0' && c <= '9')
1320 				  v = (v << 4) + c - '0';
1321 			  else if (c >= 'A' && c <= 'F')
1322 				  v = (v << 4) + c - 'A' + 10;
1323 			  else if (c >= 'a' && c <= 'f')
1324 				  v = (v << 4) + c - 'a' + 10;
1325 			  else
1326 				  synerror("Bad escape sequence");
1327 		  }
1328 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1329 			  synerror("Bad escape sequence");
1330 		  /* We really need iconv here. */
1331 		  if (initial_localeisutf8 && v > 127) {
1332 			  CHECKSTRSPACE(4, out);
1333 			  /*
1334 			   * We cannot use wctomb() as the locale may have
1335 			   * changed.
1336 			   */
1337 			  if (v <= 0x7ff) {
1338 				  USTPUTC(0xc0 | v >> 6, out);
1339 				  USTPUTC(0x80 | (v & 0x3f), out);
1340 				  return out;
1341 			  } else if (v <= 0xffff) {
1342 				  USTPUTC(0xe0 | v >> 12, out);
1343 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1344 				  USTPUTC(0x80 | (v & 0x3f), out);
1345 				  return out;
1346 			  } else if (v <= 0x10ffff) {
1347 				  USTPUTC(0xf0 | v >> 18, out);
1348 				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1349 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1350 				  USTPUTC(0x80 | (v & 0x3f), out);
1351 				  return out;
1352 			  }
1353 		  }
1354 		  if (v > 127)
1355 			  v = '?';
1356 		  break;
1357 	default:
1358 		  synerror("Bad escape sequence");
1359 	}
1360 	vc = (char)v;
1361 	/*
1362 	 * We can't handle NUL bytes.
1363 	 * POSIX says we should skip till the closing quote.
1364 	 */
1365 	if (vc == '\0') {
1366 		while ((c = pgetc()) != '\'') {
1367 			if (c == '\\')
1368 				c = pgetc();
1369 			if (c == PEOF)
1370 				synerror("Unterminated quoted string");
1371 			if (c == '\n') {
1372 				plinno++;
1373 				if (doprompt)
1374 					setprompt(2);
1375 				else
1376 					setprompt(0);
1377 			}
1378 		}
1379 		pungetc();
1380 		return out;
1381 	}
1382 	if (SQSYNTAX[vc] == CCTL)
1383 		USTPUTC(CTLESC, out);
1384 	USTPUTC(vc, out);
1385 	return out;
1386 }
1387 
1388 
1389 /*
1390  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1391  * is not NULL, read a here document.  In the latter case, eofmark is the
1392  * word which marks the end of the document and striptabs is true if
1393  * leading tabs should be stripped from the document.  The argument firstc
1394  * is the first character of the input token or document.
1395  *
1396  * Because C does not have internal subroutines, I have simulated them
1397  * using goto's to implement the subroutine linkage.  The following macros
1398  * will run code that appears at the end of readtoken1.
1399  */
1400 
1401 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1402 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1403 
1404 static int
1405 readtoken1(int firstc, char const *initialsyntax, const char *eofmark,
1406     int striptabs)
1407 {
1408 	int c = firstc;
1409 	char *out;
1410 	int len;
1411 	struct nodelist *bqlist;
1412 	int quotef;
1413 	int newvarnest;
1414 	int level;
1415 	int synentry;
1416 	struct tokenstate state_static[MAXNEST_static];
1417 	int maxnest = MAXNEST_static;
1418 	struct tokenstate *state = state_static;
1419 	int sqiscstyle = 0;
1420 
1421 	startlinno = plinno;
1422 	quotef = 0;
1423 	bqlist = NULL;
1424 	newvarnest = 0;
1425 	level = 0;
1426 	state[level].syntax = initialsyntax;
1427 	state[level].parenlevel = 0;
1428 	state[level].category = TSTATE_TOP;
1429 
1430 	STARTSTACKSTR(out);
1431 	loop: {	/* for each line, until end of word */
1432 		if (eofmark && eofmark != NOEOFMARK)
1433 			/* set c to PEOF if at end of here document */
1434 			c = checkend(c, eofmark, striptabs);
1435 		for (;;) {	/* until end of line or end of word */
1436 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1437 
1438 			synentry = state[level].syntax[c];
1439 
1440 			switch(synentry) {
1441 			case CNL:	/* '\n' */
1442 				if (level == 0)
1443 					goto endword;	/* exit outer loop */
1444 				/* FALLTHROUGH */
1445 			case CQNL:
1446 				USTPUTC(c, out);
1447 				plinno++;
1448 				if (doprompt)
1449 					setprompt(2);
1450 				else
1451 					setprompt(0);
1452 				c = pgetc();
1453 				goto loop;		/* continue outer loop */
1454 			case CSBACK:
1455 				if (sqiscstyle) {
1456 					out = readcstyleesc(out);
1457 					break;
1458 				}
1459 				/* FALLTHROUGH */
1460 			case CWORD:
1461 				USTPUTC(c, out);
1462 				break;
1463 			case CCTL:
1464 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1465 					USTPUTC(CTLESC, out);
1466 				USTPUTC(c, out);
1467 				break;
1468 			case CBACK:	/* backslash */
1469 				c = pgetc();
1470 				if (c == PEOF) {
1471 					USTPUTC('\\', out);
1472 					pungetc();
1473 				} else if (c == '\n') {
1474 					plinno++;
1475 					if (doprompt)
1476 						setprompt(2);
1477 					else
1478 						setprompt(0);
1479 				} else {
1480 					if (state[level].syntax == DQSYNTAX &&
1481 					    c != '\\' && c != '`' && c != '$' &&
1482 					    (c != '"' || (eofmark != NULL &&
1483 						newvarnest == 0)) &&
1484 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1485 						USTPUTC('\\', out);
1486 					if ((eofmark == NULL ||
1487 					    newvarnest > 0) &&
1488 					    state[level].syntax == BASESYNTAX)
1489 						USTPUTC(CTLQUOTEMARK, out);
1490 					if (SQSYNTAX[c] == CCTL)
1491 						USTPUTC(CTLESC, out);
1492 					USTPUTC(c, out);
1493 					if ((eofmark == NULL ||
1494 					    newvarnest > 0) &&
1495 					    state[level].syntax == BASESYNTAX &&
1496 					    state[level].category == TSTATE_VAR_OLD)
1497 						USTPUTC(CTLQUOTEEND, out);
1498 					quotef++;
1499 				}
1500 				break;
1501 			case CSQUOTE:
1502 				USTPUTC(CTLQUOTEMARK, out);
1503 				state[level].syntax = SQSYNTAX;
1504 				sqiscstyle = 0;
1505 				break;
1506 			case CDQUOTE:
1507 				USTPUTC(CTLQUOTEMARK, out);
1508 				state[level].syntax = DQSYNTAX;
1509 				break;
1510 			case CENDQUOTE:
1511 				if (eofmark != NULL && newvarnest == 0)
1512 					USTPUTC(c, out);
1513 				else {
1514 					if (state[level].category == TSTATE_VAR_OLD)
1515 						USTPUTC(CTLQUOTEEND, out);
1516 					state[level].syntax = BASESYNTAX;
1517 					quotef++;
1518 				}
1519 				break;
1520 			case CVAR:	/* '$' */
1521 				PARSESUB();		/* parse substitution */
1522 				break;
1523 			case CENDVAR:	/* '}' */
1524 				if (level > 0 &&
1525 				    ((state[level].category == TSTATE_VAR_OLD &&
1526 				      state[level].syntax ==
1527 				      state[level - 1].syntax) ||
1528 				    (state[level].category == TSTATE_VAR_NEW &&
1529 				     state[level].syntax == BASESYNTAX))) {
1530 					if (state[level].category == TSTATE_VAR_NEW)
1531 						newvarnest--;
1532 					level--;
1533 					USTPUTC(CTLENDVAR, out);
1534 				} else {
1535 					USTPUTC(c, out);
1536 				}
1537 				break;
1538 			case CLP:	/* '(' in arithmetic */
1539 				state[level].parenlevel++;
1540 				USTPUTC(c, out);
1541 				break;
1542 			case CRP:	/* ')' in arithmetic */
1543 				if (state[level].parenlevel > 0) {
1544 					USTPUTC(c, out);
1545 					--state[level].parenlevel;
1546 				} else {
1547 					if (pgetc_linecont() == ')') {
1548 						if (level > 0 &&
1549 						    state[level].category == TSTATE_ARITH) {
1550 							level--;
1551 							USTPUTC(CTLENDARI, out);
1552 						} else
1553 							USTPUTC(')', out);
1554 					} else {
1555 						/*
1556 						 * unbalanced parens
1557 						 *  (don't 2nd guess - no error)
1558 						 */
1559 						pungetc();
1560 						USTPUTC(')', out);
1561 					}
1562 				}
1563 				break;
1564 			case CBQUOTE:	/* '`' */
1565 				out = parsebackq(out, &bqlist, 1,
1566 				    state[level].syntax == DQSYNTAX &&
1567 				    (eofmark == NULL || newvarnest > 0),
1568 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1569 				break;
1570 			case CEOF:
1571 				goto endword;		/* exit outer loop */
1572 			case CIGN:
1573 				break;
1574 			default:
1575 				if (level == 0)
1576 					goto endword;	/* exit outer loop */
1577 				USTPUTC(c, out);
1578 			}
1579 			c = pgetc_macro();
1580 		}
1581 	}
1582 endword:
1583 	if (state[level].syntax == ARISYNTAX)
1584 		synerror("Missing '))'");
1585 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1586 		synerror("Unterminated quoted string");
1587 	if (state[level].category == TSTATE_VAR_OLD ||
1588 	    state[level].category == TSTATE_VAR_NEW) {
1589 		startlinno = plinno;
1590 		synerror("Missing '}'");
1591 	}
1592 	if (state != state_static)
1593 		parser_temp_free_upto(state);
1594 	USTPUTC('\0', out);
1595 	len = out - stackblock();
1596 	out = stackblock();
1597 	if (eofmark == NULL) {
1598 		if ((c == '>' || c == '<')
1599 		 && quotef == 0
1600 		 && len <= 2
1601 		 && (*out == '\0' || is_digit(*out))) {
1602 			parseredir(out, c);
1603 			return lasttoken = TREDIR;
1604 		} else {
1605 			pungetc();
1606 		}
1607 	}
1608 	quoteflag = quotef;
1609 	backquotelist = bqlist;
1610 	grabstackblock(len);
1611 	wordtext = out;
1612 	return lasttoken = TWORD;
1613 /* end of readtoken routine */
1614 
1615 
1616 /*
1617  * Parse a substitution.  At this point, we have read the dollar sign
1618  * and nothing else.
1619  */
1620 
1621 parsesub: {
1622 	int subtype;
1623 	int typeloc;
1624 	int flags;
1625 	char *p;
1626 	static const char types[] = "}-+?=";
1627 	int linno;
1628 	int length;
1629 	int c1;
1630 
1631 	c = pgetc_linecont();
1632 	if (c == '(') {	/* $(command) or $((arith)) */
1633 		if (pgetc_linecont() == '(') {
1634 			PARSEARITH();
1635 		} else {
1636 			pungetc();
1637 			out = parsebackq(out, &bqlist, 0,
1638 			    state[level].syntax == DQSYNTAX &&
1639 			    (eofmark == NULL || newvarnest > 0),
1640 			    state[level].syntax == DQSYNTAX ||
1641 			    state[level].syntax == ARISYNTAX);
1642 		}
1643 	} else if (c == '{' || is_name(c) || is_special(c)) {
1644 		USTPUTC(CTLVAR, out);
1645 		typeloc = out - stackblock();
1646 		USTPUTC(VSNORMAL, out);
1647 		subtype = VSNORMAL;
1648 		flags = 0;
1649 		if (c == '{') {
1650 			c = pgetc_linecont();
1651 			subtype = 0;
1652 		}
1653 varname:
1654 		if (!is_eof(c) && is_name(c)) {
1655 			length = 0;
1656 			do {
1657 				STPUTC(c, out);
1658 				c = pgetc_linecont();
1659 				length++;
1660 			} while (!is_eof(c) && is_in_name(c));
1661 			if (length == 6 &&
1662 			    strncmp(out - length, "LINENO", length) == 0) {
1663 				/* Replace the variable name with the
1664 				 * current line number. */
1665 				STADJUST(-6, out);
1666 				CHECKSTRSPACE(11, out);
1667 				linno = plinno;
1668 				if (funclinno != 0)
1669 					linno -= funclinno - 1;
1670 				length = snprintf(out, 11, "%d", linno);
1671 				if (length > 10)
1672 					length = 10;
1673 				out += length;
1674 				flags |= VSLINENO;
1675 			}
1676 		} else if (is_digit(c)) {
1677 			if (subtype != VSNORMAL) {
1678 				do {
1679 					STPUTC(c, out);
1680 					c = pgetc_linecont();
1681 				} while (is_digit(c));
1682 			} else {
1683 				USTPUTC(c, out);
1684 				c = pgetc_linecont();
1685 			}
1686 		} else if (is_special(c)) {
1687 			c1 = c;
1688 			c = pgetc_linecont();
1689 			if (subtype == 0 && c1 == '#') {
1690 				subtype = VSLENGTH;
1691 				if (strchr(types, c) == NULL && c != ':' &&
1692 				    c != '#' && c != '%')
1693 					goto varname;
1694 				c1 = c;
1695 				c = pgetc_linecont();
1696 				if (c1 != '}' && c == '}') {
1697 					pungetc();
1698 					c = c1;
1699 					goto varname;
1700 				}
1701 				pungetc();
1702 				c = c1;
1703 				c1 = '#';
1704 				subtype = 0;
1705 			}
1706 			USTPUTC(c1, out);
1707 		} else {
1708 			subtype = VSERROR;
1709 			if (c == '}')
1710 				pungetc();
1711 			else if (c == '\n' || c == PEOF)
1712 				synerror("Unexpected end of line in substitution");
1713 			else if (BASESYNTAX[c] != CCTL)
1714 				USTPUTC(c, out);
1715 		}
1716 		if (subtype == 0) {
1717 			switch (c) {
1718 			case ':':
1719 				flags |= VSNUL;
1720 				c = pgetc_linecont();
1721 				/*FALLTHROUGH*/
1722 			default:
1723 				p = strchr(types, c);
1724 				if (p == NULL) {
1725 					if (c == '\n' || c == PEOF)
1726 						synerror("Unexpected end of line in substitution");
1727 					if (flags == VSNUL)
1728 						STPUTC(':', out);
1729 					if (BASESYNTAX[c] != CCTL)
1730 						STPUTC(c, out);
1731 					subtype = VSERROR;
1732 				} else
1733 					subtype = p - types + VSNORMAL;
1734 				break;
1735 			case '%':
1736 			case '#':
1737 				{
1738 					int cc = c;
1739 					subtype = c == '#' ? VSTRIMLEFT :
1740 							     VSTRIMRIGHT;
1741 					c = pgetc_linecont();
1742 					if (c == cc)
1743 						subtype++;
1744 					else
1745 						pungetc();
1746 					break;
1747 				}
1748 			}
1749 		} else if (subtype != VSERROR) {
1750 			if (subtype == VSLENGTH && c != '}')
1751 				subtype = VSERROR;
1752 			pungetc();
1753 		}
1754 		STPUTC('=', out);
1755 		if (state[level].syntax == DQSYNTAX ||
1756 		    state[level].syntax == ARISYNTAX)
1757 			flags |= VSQUOTE;
1758 		*(stackblock() + typeloc) = subtype | flags;
1759 		if (subtype != VSNORMAL) {
1760 			if (level + 1 >= maxnest) {
1761 				maxnest *= 2;
1762 				if (state == state_static) {
1763 					state = parser_temp_alloc(
1764 					    maxnest * sizeof(*state));
1765 					memcpy(state, state_static,
1766 					    MAXNEST_static * sizeof(*state));
1767 				} else
1768 					state = parser_temp_realloc(state,
1769 					    maxnest * sizeof(*state));
1770 			}
1771 			level++;
1772 			state[level].parenlevel = 0;
1773 			if (subtype == VSMINUS || subtype == VSPLUS ||
1774 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1775 				/*
1776 				 * For operators that were in the Bourne shell,
1777 				 * inherit the double-quote state.
1778 				 */
1779 				state[level].syntax = state[level - 1].syntax;
1780 				state[level].category = TSTATE_VAR_OLD;
1781 			} else {
1782 				/*
1783 				 * The other operators take a pattern,
1784 				 * so go to BASESYNTAX.
1785 				 * Also, ' and " are now special, even
1786 				 * in here documents.
1787 				 */
1788 				state[level].syntax = BASESYNTAX;
1789 				state[level].category = TSTATE_VAR_NEW;
1790 				newvarnest++;
1791 			}
1792 		}
1793 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1794 		/* $'cstylequotes' */
1795 		USTPUTC(CTLQUOTEMARK, out);
1796 		state[level].syntax = SQSYNTAX;
1797 		sqiscstyle = 1;
1798 	} else {
1799 		USTPUTC('$', out);
1800 		pungetc();
1801 	}
1802 	goto parsesub_return;
1803 }
1804 
1805 
1806 /*
1807  * Parse an arithmetic expansion (indicate start of one and set state)
1808  */
1809 parsearith: {
1810 
1811 	if (level + 1 >= maxnest) {
1812 		maxnest *= 2;
1813 		if (state == state_static) {
1814 			state = parser_temp_alloc(
1815 			    maxnest * sizeof(*state));
1816 			memcpy(state, state_static,
1817 			    MAXNEST_static * sizeof(*state));
1818 		} else
1819 			state = parser_temp_realloc(state,
1820 			    maxnest * sizeof(*state));
1821 	}
1822 	level++;
1823 	state[level].syntax = ARISYNTAX;
1824 	state[level].parenlevel = 0;
1825 	state[level].category = TSTATE_ARITH;
1826 	USTPUTC(CTLARI, out);
1827 	if (state[level - 1].syntax == DQSYNTAX)
1828 		USTPUTC('"',out);
1829 	else
1830 		USTPUTC(' ',out);
1831 	goto parsearith_return;
1832 }
1833 
1834 } /* end of readtoken */
1835 
1836 
1837 /*
1838  * Returns true if the text contains nothing to expand (no dollar signs
1839  * or backquotes).
1840  */
1841 
1842 static int
1843 noexpand(char *text)
1844 {
1845 	char *p;
1846 	char c;
1847 
1848 	p = text;
1849 	while ((c = *p++) != '\0') {
1850 		if ( c == CTLQUOTEMARK)
1851 			continue;
1852 		if (c == CTLESC)
1853 			p++;
1854 		else if (BASESYNTAX[(int)c] == CCTL)
1855 			return 0;
1856 	}
1857 	return 1;
1858 }
1859 
1860 
1861 /*
1862  * Return true if the argument is a legal variable name (a letter or
1863  * underscore followed by zero or more letters, underscores, and digits).
1864  */
1865 
1866 int
1867 goodname(const char *name)
1868 {
1869 	const char *p;
1870 
1871 	p = name;
1872 	if (! is_name(*p))
1873 		return 0;
1874 	while (*++p) {
1875 		if (! is_in_name(*p))
1876 			return 0;
1877 	}
1878 	return 1;
1879 }
1880 
1881 
1882 int
1883 isassignment(const char *p)
1884 {
1885 	if (!is_name(*p))
1886 		return 0;
1887 	p++;
1888 	for (;;) {
1889 		if (*p == '=')
1890 			return 1;
1891 		else if (!is_in_name(*p))
1892 			return 0;
1893 		p++;
1894 	}
1895 }
1896 
1897 
1898 static void
1899 consumetoken(int token)
1900 {
1901 	if (readtoken() != token)
1902 		synexpect(token);
1903 }
1904 
1905 
1906 /*
1907  * Called when an unexpected token is read during the parse.  The argument
1908  * is the token that is expected, or -1 if more than one type of token can
1909  * occur at this point.
1910  */
1911 
1912 static void
1913 synexpect(int token)
1914 {
1915 	char msg[64];
1916 
1917 	if (token >= 0) {
1918 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1919 			tokname[lasttoken], tokname[token]);
1920 	} else {
1921 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1922 	}
1923 	synerror(msg);
1924 }
1925 
1926 
1927 static void
1928 synerror(const char *msg)
1929 {
1930 	if (commandname)
1931 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1932 	else if (arg0)
1933 		outfmt(out2, "%s: ", arg0);
1934 	outfmt(out2, "Syntax error: %s\n", msg);
1935 	error((char *)NULL);
1936 }
1937 
1938 static void
1939 setprompt(int which)
1940 {
1941 	whichprompt = which;
1942 	if (which == 0)
1943 		return;
1944 
1945 #ifndef NO_HISTORY
1946 	if (!el)
1947 #endif
1948 	{
1949 		out2str(getprompt(NULL));
1950 		flushout(out2);
1951 	}
1952 }
1953 
1954 static int
1955 pgetc_linecont(void)
1956 {
1957 	int c;
1958 
1959 	while ((c = pgetc_macro()) == '\\') {
1960 		c = pgetc();
1961 		if (c == '\n') {
1962 			plinno++;
1963 			if (doprompt)
1964 				setprompt(2);
1965 			else
1966 				setprompt(0);
1967 		} else {
1968 			pungetc();
1969 			/* Allow the backslash to be pushed back. */
1970 			pushstring("\\", 1, NULL);
1971 			return (pgetc());
1972 		}
1973 	}
1974 	return (c);
1975 }
1976 
1977 
1978 static struct passwd *
1979 getpwlogin(void)
1980 {
1981 	const char *login;
1982 
1983 	login = getlogin();
1984 	if (login == NULL)
1985 		return (NULL);
1986 
1987 	return (getpwnam(login));
1988 }
1989 
1990 
1991 static void
1992 getusername(char *name, size_t namelen)
1993 {
1994 	static char cached_name[MAXLOGNAME];
1995 	struct passwd *pw;
1996 	uid_t euid;
1997 
1998 	if (cached_name[0] == '\0') {
1999 		euid = geteuid();
2000 
2001 		/*
2002 		 * Handle the case when there is more than one
2003 		 * login with the same UID, or when the login
2004 		 * returned by getlogin(2) does no longer match
2005 		 * the current UID.
2006 		 */
2007 		pw = getpwlogin();
2008 		if (pw == NULL || pw->pw_uid != euid)
2009 			pw = getpwuid(euid);
2010 
2011 		if (pw != NULL) {
2012 			strlcpy(cached_name, pw->pw_name,
2013 			    sizeof(cached_name));
2014 		} else {
2015 			snprintf(cached_name, sizeof(cached_name),
2016 			    "%u", euid);
2017 		}
2018 	}
2019 
2020 	strlcpy(name, cached_name, namelen);
2021 }
2022 
2023 
2024 /*
2025  * called by editline -- any expansions to the prompt
2026  *    should be added here.
2027  */
2028 char *
2029 getprompt(void *unused __unused)
2030 {
2031 	static char ps[PROMPTLEN];
2032 	const char *fmt;
2033 	const char *home;
2034 	const char *pwd;
2035 	size_t homelen;
2036 	int i, trim;
2037 	static char internal_error[] = "??";
2038 
2039 	/*
2040 	 * Select prompt format.
2041 	 */
2042 	switch (whichprompt) {
2043 	case 0:
2044 		fmt = "";
2045 		break;
2046 	case 1:
2047 		fmt = ps1val();
2048 		break;
2049 	case 2:
2050 		fmt = ps2val();
2051 		break;
2052 	default:
2053 		return internal_error;
2054 	}
2055 
2056 	/*
2057 	 * Format prompt string.
2058 	 */
2059 	for (i = 0; (i < PROMPTLEN - 1) && (*fmt != '\0'); i++, fmt++)
2060 		if (*fmt == '\\')
2061 			switch (*++fmt) {
2062 
2063 				/*
2064 				 * Hostname.
2065 				 *
2066 				 * \h specifies just the local hostname,
2067 				 * \H specifies fully-qualified hostname.
2068 				 */
2069 			case 'h':
2070 			case 'H':
2071 				ps[i] = '\0';
2072 				gethostname(&ps[i], PROMPTLEN - i - 1);
2073 				ps[PROMPTLEN - 1] = '\0';
2074 				/* Skip to end of hostname. */
2075 				trim = (*fmt == 'h') ? '.' : '\0';
2076 				while ((ps[i] != '\0') && (ps[i] != trim))
2077 					i++;
2078 				--i;
2079 				break;
2080 
2081 				/*
2082 				 * User name.
2083 				 */
2084 			case 'u':
2085 				ps[i] = '\0';
2086 				getusername(&ps[i], PROMPTLEN - i);
2087 				/* Skip to end of username. */
2088 				while (ps[i + 1] != '\0')
2089 					i++;
2090 				break;
2091 
2092 				/*
2093 				 * Working directory.
2094 				 *
2095 				 * \W specifies just the final component,
2096 				 * \w specifies the entire path.
2097 				 */
2098 			case 'W':
2099 			case 'w':
2100 				pwd = lookupvar("PWD");
2101 				if (pwd == NULL || *pwd == '\0')
2102 					pwd = "?";
2103 				if (*fmt == 'W' &&
2104 				    *pwd == '/' && pwd[1] != '\0')
2105 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
2106 					    PROMPTLEN - i);
2107 				else {
2108 					home = lookupvar("HOME");
2109 					if (home != NULL)
2110 						homelen = strlen(home);
2111 					if (home != NULL &&
2112 					    strcmp(home, "/") != 0 &&
2113 					    strncmp(pwd, home, homelen) == 0 &&
2114 					    (pwd[homelen] == '/' ||
2115 					    pwd[homelen] == '\0')) {
2116 						strlcpy(&ps[i], "~",
2117 						    PROMPTLEN - i);
2118 						strlcpy(&ps[i + 1],
2119 						    pwd + homelen,
2120 						    PROMPTLEN - i - 1);
2121 					} else {
2122 						strlcpy(&ps[i], pwd, PROMPTLEN - i);
2123 					}
2124 				}
2125 				/* Skip to end of path. */
2126 				while (ps[i + 1] != '\0')
2127 					i++;
2128 				break;
2129 
2130 				/*
2131 				 * Superuser status.
2132 				 *
2133 				 * '$' for normal users, '#' for root.
2134 				 */
2135 			case '$':
2136 				ps[i] = (geteuid() != 0) ? '$' : '#';
2137 				break;
2138 
2139 				/*
2140 				 * A literal \.
2141 				 */
2142 			case '\\':
2143 				ps[i] = '\\';
2144 				break;
2145 
2146 				/*
2147 				 * Emit unrecognized formats verbatim.
2148 				 */
2149 			default:
2150 				ps[i] = '\\';
2151 				if (i < PROMPTLEN - 2)
2152 					ps[++i] = *fmt;
2153 				break;
2154 			}
2155 		else
2156 			ps[i] = *fmt;
2157 	ps[i] = '\0';
2158 	return (ps);
2159 }
2160 
2161 
2162 const char *
2163 expandstr(const char *ps)
2164 {
2165 	union node n;
2166 	struct jmploc jmploc;
2167 	struct jmploc *const savehandler = handler;
2168 	const int saveprompt = doprompt;
2169 	struct parsefile *const savetopfile = getcurrentfile();
2170 	struct parser_temp *const saveparser_temp = parser_temp;
2171 	const char *result = NULL;
2172 
2173 	if (!setjmp(jmploc.loc)) {
2174 		handler = &jmploc;
2175 		parser_temp = NULL;
2176 		setinputstring(ps, 1);
2177 		doprompt = 0;
2178 		readtoken1(pgetc(), DQSYNTAX, NOEOFMARK, 0);
2179 		if (backquotelist != NULL)
2180 			error("Command substitution not allowed here");
2181 
2182 		n.narg.type = NARG;
2183 		n.narg.next = NULL;
2184 		n.narg.text = wordtext;
2185 		n.narg.backquote = backquotelist;
2186 
2187 		expandarg(&n, NULL, 0);
2188 		result = stackblock();
2189 		INTOFF;
2190 	}
2191 	handler = savehandler;
2192 	doprompt = saveprompt;
2193 	popfilesupto(savetopfile);
2194 	if (parser_temp != saveparser_temp) {
2195 		parser_temp_free_all();
2196 		parser_temp = saveparser_temp;
2197 	}
2198 	if (result != NULL) {
2199 		INTON;
2200 	} else if (exception == EXINT)
2201 		raise(SIGINT);
2202 	return result;
2203 }
2204