xref: /freebsd/bin/sh/parser.c (revision 780fb4a2)
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 <stdlib.h>
44 #include <unistd.h>
45 #include <stdio.h>
46 
47 #include "shell.h"
48 #include "parser.h"
49 #include "nodes.h"
50 #include "expand.h"	/* defines rmescapes() */
51 #include "syntax.h"
52 #include "options.h"
53 #include "input.h"
54 #include "output.h"
55 #include "var.h"
56 #include "error.h"
57 #include "memalloc.h"
58 #include "mystring.h"
59 #include "alias.h"
60 #include "show.h"
61 #include "eval.h"
62 #include "exec.h"	/* to check for special builtins */
63 #ifndef NO_HISTORY
64 #include "myhistedit.h"
65 #endif
66 
67 /*
68  * Shell command parser.
69  */
70 
71 #define	PROMPTLEN	128
72 
73 /* values of checkkwd variable */
74 #define CHKALIAS	0x1
75 #define CHKKWD		0x2
76 #define CHKNL		0x4
77 
78 /* values returned by readtoken */
79 #include "token.h"
80 
81 
82 
83 struct heredoc {
84 	struct heredoc *next;	/* next here document in list */
85 	union node *here;		/* redirection node */
86 	char *eofmark;		/* string indicating end of input */
87 	int striptabs;		/* if set, strip leading tabs */
88 };
89 
90 struct parser_temp {
91 	struct parser_temp *next;
92 	void *data;
93 };
94 
95 
96 static struct heredoc *heredoclist;	/* list of here documents to read */
97 static int doprompt;		/* if set, prompt the user */
98 static int needprompt;		/* true if interactive and at start of line */
99 static int lasttoken;		/* last token read */
100 static int tokpushback;		/* last token pushed back */
101 static char *wordtext;		/* text of last word returned by readtoken */
102 static int checkkwd;
103 static struct nodelist *backquotelist;
104 static union node *redirnode;
105 static struct heredoc *heredoc;
106 static int quoteflag;		/* set if (part of) last token was quoted */
107 static int startlinno;		/* line # where last token started */
108 static int funclinno;		/* line # where the current function started */
109 static struct parser_temp *parser_temp;
110 
111 #define NOEOFMARK ((const char *)&heredoclist)
112 
113 
114 static union node *list(int);
115 static union node *andor(void);
116 static union node *pipeline(void);
117 static union node *command(void);
118 static union node *simplecmd(union node **, union node *);
119 static union node *makename(void);
120 static union node *makebinary(int type, union node *n1, union node *n2);
121 static void parsefname(void);
122 static void parseheredoc(void);
123 static int peektoken(void);
124 static int readtoken(void);
125 static int xxreadtoken(void);
126 static int readtoken1(int, const char *, const char *, int);
127 static int noexpand(char *);
128 static void consumetoken(int);
129 static void synexpect(int) __dead2;
130 static void synerror(const char *) __dead2;
131 static void setprompt(int);
132 static int pgetc_linecont(void);
133 
134 
135 static void *
136 parser_temp_alloc(size_t len)
137 {
138 	struct parser_temp *t;
139 
140 	INTOFF;
141 	t = ckmalloc(sizeof(*t));
142 	t->data = NULL;
143 	t->next = parser_temp;
144 	parser_temp = t;
145 	t->data = ckmalloc(len);
146 	INTON;
147 	return t->data;
148 }
149 
150 
151 static void *
152 parser_temp_realloc(void *ptr, size_t len)
153 {
154 	struct parser_temp *t;
155 
156 	INTOFF;
157 	t = parser_temp;
158 	if (ptr != t->data)
159 		error("bug: parser_temp_realloc misused");
160 	t->data = ckrealloc(t->data, len);
161 	INTON;
162 	return t->data;
163 }
164 
165 
166 static void
167 parser_temp_free_upto(void *ptr)
168 {
169 	struct parser_temp *t;
170 	int done = 0;
171 
172 	INTOFF;
173 	while (parser_temp != NULL && !done) {
174 		t = parser_temp;
175 		parser_temp = t->next;
176 		done = t->data == ptr;
177 		ckfree(t->data);
178 		ckfree(t);
179 	}
180 	INTON;
181 	if (!done)
182 		error("bug: parser_temp_free_upto misused");
183 }
184 
185 
186 static void
187 parser_temp_free_all(void)
188 {
189 	struct parser_temp *t;
190 
191 	INTOFF;
192 	while (parser_temp != NULL) {
193 		t = parser_temp;
194 		parser_temp = t->next;
195 		ckfree(t->data);
196 		ckfree(t);
197 	}
198 	INTON;
199 }
200 
201 
202 /*
203  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
204  * valid parse tree indicating a blank line.)
205  */
206 
207 union node *
208 parsecmd(int interact)
209 {
210 	int t;
211 
212 	/* This assumes the parser is not re-entered,
213 	 * which could happen if we add command substitution on PS1/PS2.
214 	 */
215 	parser_temp_free_all();
216 	heredoclist = NULL;
217 
218 	tokpushback = 0;
219 	checkkwd = 0;
220 	doprompt = interact;
221 	if (doprompt)
222 		setprompt(1);
223 	else
224 		setprompt(0);
225 	needprompt = 0;
226 	t = readtoken();
227 	if (t == TEOF)
228 		return NEOF;
229 	if (t == TNL)
230 		return NULL;
231 	tokpushback++;
232 	return list(1);
233 }
234 
235 
236 /*
237  * Read and parse words for wordexp.
238  * Returns a list of NARG nodes; NULL if there are no words.
239  */
240 union node *
241 parsewordexp(void)
242 {
243 	union node *n, *first = NULL, **pnext;
244 	int t;
245 
246 	/* This assumes the parser is not re-entered,
247 	 * which could happen if we add command substitution on PS1/PS2.
248 	 */
249 	parser_temp_free_all();
250 	heredoclist = NULL;
251 
252 	tokpushback = 0;
253 	checkkwd = 0;
254 	doprompt = 0;
255 	setprompt(0);
256 	needprompt = 0;
257 	pnext = &first;
258 	while ((t = readtoken()) != TEOF) {
259 		if (t != TWORD)
260 			synexpect(TWORD);
261 		n = makename();
262 		*pnext = n;
263 		pnext = &n->narg.next;
264 	}
265 	return first;
266 }
267 
268 
269 static union node *
270 list(int nlflag)
271 {
272 	union node *ntop, *n1, *n2, *n3;
273 	int tok;
274 
275 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
276 	if (!nlflag && tokendlist[peektoken()])
277 		return NULL;
278 	ntop = n1 = NULL;
279 	for (;;) {
280 		n2 = andor();
281 		tok = readtoken();
282 		if (tok == TBACKGND) {
283 			if (n2 != NULL && n2->type == NPIPE) {
284 				n2->npipe.backgnd = 1;
285 			} else if (n2 != NULL && n2->type == NREDIR) {
286 				n2->type = NBACKGND;
287 			} else {
288 				n3 = (union node *)stalloc(sizeof (struct nredir));
289 				n3->type = NBACKGND;
290 				n3->nredir.n = n2;
291 				n3->nredir.redirect = NULL;
292 				n2 = n3;
293 			}
294 		}
295 		if (ntop == NULL)
296 			ntop = n2;
297 		else if (n1 == NULL) {
298 			n1 = makebinary(NSEMI, ntop, n2);
299 			ntop = n1;
300 		}
301 		else {
302 			n3 = makebinary(NSEMI, n1->nbinary.ch2, n2);
303 			n1->nbinary.ch2 = n3;
304 			n1 = n3;
305 		}
306 		switch (tok) {
307 		case TBACKGND:
308 		case TSEMI:
309 			tok = readtoken();
310 			/* FALLTHROUGH */
311 		case TNL:
312 			if (tok == TNL) {
313 				parseheredoc();
314 				if (nlflag)
315 					return ntop;
316 			} else if (tok == TEOF && nlflag) {
317 				parseheredoc();
318 				return ntop;
319 			} else {
320 				tokpushback++;
321 			}
322 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
323 			if (!nlflag && tokendlist[peektoken()])
324 				return ntop;
325 			break;
326 		case TEOF:
327 			if (heredoclist)
328 				parseheredoc();
329 			else
330 				pungetc();		/* push back EOF on input */
331 			return ntop;
332 		default:
333 			if (nlflag)
334 				synexpect(-1);
335 			tokpushback++;
336 			return ntop;
337 		}
338 	}
339 }
340 
341 
342 
343 static union node *
344 andor(void)
345 {
346 	union node *n;
347 	int t;
348 
349 	n = pipeline();
350 	for (;;) {
351 		if ((t = readtoken()) == TAND) {
352 			t = NAND;
353 		} else if (t == TOR) {
354 			t = NOR;
355 		} else {
356 			tokpushback++;
357 			return n;
358 		}
359 		n = makebinary(t, n, pipeline());
360 	}
361 }
362 
363 
364 
365 static union node *
366 pipeline(void)
367 {
368 	union node *n1, *n2, *pipenode;
369 	struct nodelist *lp, *prev;
370 	int negate, t;
371 
372 	negate = 0;
373 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
374 	TRACE(("pipeline: entered\n"));
375 	while (readtoken() == TNOT)
376 		negate = !negate;
377 	tokpushback++;
378 	n1 = command();
379 	if (readtoken() == TPIPE) {
380 		pipenode = (union node *)stalloc(sizeof (struct npipe));
381 		pipenode->type = NPIPE;
382 		pipenode->npipe.backgnd = 0;
383 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
384 		pipenode->npipe.cmdlist = lp;
385 		lp->n = n1;
386 		do {
387 			prev = lp;
388 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
389 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
390 			t = readtoken();
391 			tokpushback++;
392 			if (t == TNOT)
393 				lp->n = pipeline();
394 			else
395 				lp->n = command();
396 			prev->next = lp;
397 		} while (readtoken() == TPIPE);
398 		lp->next = NULL;
399 		n1 = pipenode;
400 	}
401 	tokpushback++;
402 	if (negate) {
403 		n2 = (union node *)stalloc(sizeof (struct nnot));
404 		n2->type = NNOT;
405 		n2->nnot.com = n1;
406 		return n2;
407 	} else
408 		return n1;
409 }
410 
411 
412 
413 static union node *
414 command(void)
415 {
416 	union node *n1, *n2;
417 	union node *ap, **app;
418 	union node *cp, **cpp;
419 	union node *redir, **rpp;
420 	int t;
421 	int is_subshell;
422 
423 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
424 	is_subshell = 0;
425 	redir = NULL;
426 	n1 = NULL;
427 	rpp = &redir;
428 
429 	/* Check for redirection which may precede command */
430 	while (readtoken() == TREDIR) {
431 		*rpp = n2 = redirnode;
432 		rpp = &n2->nfile.next;
433 		parsefname();
434 	}
435 	tokpushback++;
436 
437 	switch (readtoken()) {
438 	case TIF:
439 		n1 = (union node *)stalloc(sizeof (struct nif));
440 		n1->type = NIF;
441 		if ((n1->nif.test = list(0)) == NULL)
442 			synexpect(-1);
443 		consumetoken(TTHEN);
444 		n1->nif.ifpart = list(0);
445 		n2 = n1;
446 		while (readtoken() == TELIF) {
447 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
448 			n2 = n2->nif.elsepart;
449 			n2->type = NIF;
450 			if ((n2->nif.test = list(0)) == NULL)
451 				synexpect(-1);
452 			consumetoken(TTHEN);
453 			n2->nif.ifpart = list(0);
454 		}
455 		if (lasttoken == TELSE)
456 			n2->nif.elsepart = list(0);
457 		else {
458 			n2->nif.elsepart = NULL;
459 			tokpushback++;
460 		}
461 		consumetoken(TFI);
462 		checkkwd = CHKKWD | CHKALIAS;
463 		break;
464 	case TWHILE:
465 	case TUNTIL:
466 		t = lasttoken;
467 		if ((n1 = list(0)) == NULL)
468 			synexpect(-1);
469 		consumetoken(TDO);
470 		n1 = makebinary((t == TWHILE)? NWHILE : NUNTIL, n1, list(0));
471 		consumetoken(TDONE);
472 		checkkwd = CHKKWD | CHKALIAS;
473 		break;
474 	case TFOR:
475 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
476 			synerror("Bad for loop variable");
477 		n1 = (union node *)stalloc(sizeof (struct nfor));
478 		n1->type = NFOR;
479 		n1->nfor.var = wordtext;
480 		while (readtoken() == TNL)
481 			;
482 		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
483 			app = &ap;
484 			while (readtoken() == TWORD) {
485 				n2 = makename();
486 				*app = n2;
487 				app = &n2->narg.next;
488 			}
489 			*app = NULL;
490 			n1->nfor.args = ap;
491 			if (lasttoken != TNL && lasttoken != TSEMI)
492 				synexpect(-1);
493 		} else {
494 			static char argvars[5] = {
495 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
496 			};
497 			n2 = (union node *)stalloc(sizeof (struct narg));
498 			n2->type = NARG;
499 			n2->narg.text = argvars;
500 			n2->narg.backquote = NULL;
501 			n2->narg.next = NULL;
502 			n1->nfor.args = n2;
503 			/*
504 			 * Newline or semicolon here is optional (but note
505 			 * that the original Bourne shell only allowed NL).
506 			 */
507 			if (lasttoken != TNL && lasttoken != TSEMI)
508 				tokpushback++;
509 		}
510 		checkkwd = CHKNL | CHKKWD | CHKALIAS;
511 		if ((t = readtoken()) == TDO)
512 			t = TDONE;
513 		else if (t == TBEGIN)
514 			t = TEND;
515 		else
516 			synexpect(-1);
517 		n1->nfor.body = list(0);
518 		consumetoken(t);
519 		checkkwd = CHKKWD | CHKALIAS;
520 		break;
521 	case TCASE:
522 		n1 = (union node *)stalloc(sizeof (struct ncase));
523 		n1->type = NCASE;
524 		consumetoken(TWORD);
525 		n1->ncase.expr = makename();
526 		while (readtoken() == TNL);
527 		if (lasttoken != TWORD || ! equal(wordtext, "in"))
528 			synerror("expecting \"in\"");
529 		cpp = &n1->ncase.cases;
530 		checkkwd = CHKNL | CHKKWD, readtoken();
531 		while (lasttoken != TESAC) {
532 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
533 			cp->type = NCLIST;
534 			app = &cp->nclist.pattern;
535 			if (lasttoken == TLP)
536 				readtoken();
537 			for (;;) {
538 				*app = ap = makename();
539 				checkkwd = CHKNL | CHKKWD;
540 				if (readtoken() != TPIPE)
541 					break;
542 				app = &ap->narg.next;
543 				readtoken();
544 			}
545 			ap->narg.next = NULL;
546 			if (lasttoken != TRP)
547 				synexpect(TRP);
548 			cp->nclist.body = list(0);
549 
550 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
551 			if ((t = readtoken()) != TESAC) {
552 				if (t == TENDCASE)
553 					;
554 				else if (t == TFALLTHRU)
555 					cp->type = NCLISTFALLTHRU;
556 				else
557 					synexpect(TENDCASE);
558 				checkkwd = CHKNL | CHKKWD, readtoken();
559 			}
560 			cpp = &cp->nclist.next;
561 		}
562 		*cpp = NULL;
563 		checkkwd = CHKKWD | CHKALIAS;
564 		break;
565 	case TLP:
566 		n1 = (union node *)stalloc(sizeof (struct nredir));
567 		n1->type = NSUBSHELL;
568 		n1->nredir.n = list(0);
569 		n1->nredir.redirect = NULL;
570 		consumetoken(TRP);
571 		checkkwd = CHKKWD | CHKALIAS;
572 		is_subshell = 1;
573 		break;
574 	case TBEGIN:
575 		n1 = list(0);
576 		consumetoken(TEND);
577 		checkkwd = CHKKWD | CHKALIAS;
578 		break;
579 	/* A simple command must have at least one redirection or word. */
580 	case TBACKGND:
581 	case TSEMI:
582 	case TAND:
583 	case TOR:
584 	case TPIPE:
585 	case TENDCASE:
586 	case TFALLTHRU:
587 	case TEOF:
588 	case TNL:
589 	case TRP:
590 		if (!redir)
591 			synexpect(-1);
592 	case TWORD:
593 		tokpushback++;
594 		n1 = simplecmd(rpp, redir);
595 		return n1;
596 	default:
597 		synexpect(-1);
598 	}
599 
600 	/* Now check for redirection which may follow command */
601 	while (readtoken() == TREDIR) {
602 		*rpp = n2 = redirnode;
603 		rpp = &n2->nfile.next;
604 		parsefname();
605 	}
606 	tokpushback++;
607 	*rpp = NULL;
608 	if (redir) {
609 		if (!is_subshell) {
610 			n2 = (union node *)stalloc(sizeof (struct nredir));
611 			n2->type = NREDIR;
612 			n2->nredir.n = n1;
613 			n1 = n2;
614 		}
615 		n1->nredir.redirect = redir;
616 	}
617 
618 	return n1;
619 }
620 
621 
622 static union node *
623 simplecmd(union node **rpp, union node *redir)
624 {
625 	union node *args, **app;
626 	union node **orig_rpp = rpp;
627 	union node *n = NULL;
628 	int special;
629 	int savecheckkwd;
630 
631 	/* If we don't have any redirections already, then we must reset */
632 	/* rpp to be the address of the local redir variable.  */
633 	if (redir == NULL)
634 		rpp = &redir;
635 
636 	args = NULL;
637 	app = &args;
638 	/*
639 	 * We save the incoming value, because we need this for shell
640 	 * functions.  There can not be a redirect or an argument between
641 	 * the function name and the open parenthesis.
642 	 */
643 	orig_rpp = rpp;
644 
645 	savecheckkwd = CHKALIAS;
646 
647 	for (;;) {
648 		checkkwd = savecheckkwd;
649 		if (readtoken() == TWORD) {
650 			n = makename();
651 			*app = n;
652 			app = &n->narg.next;
653 			if (savecheckkwd != 0 && !isassignment(wordtext))
654 				savecheckkwd = 0;
655 		} else if (lasttoken == TREDIR) {
656 			*rpp = n = redirnode;
657 			rpp = &n->nfile.next;
658 			parsefname();	/* read name of redirection file */
659 		} else if (lasttoken == TLP && app == &args->narg.next
660 					    && rpp == orig_rpp) {
661 			/* We have a function */
662 			consumetoken(TRP);
663 			funclinno = plinno;
664 			/*
665 			 * - Require plain text.
666 			 * - Functions with '/' cannot be called.
667 			 * - Reject name=().
668 			 * - Reject ksh extended glob patterns.
669 			 */
670 			if (!noexpand(n->narg.text) || quoteflag ||
671 			    strchr(n->narg.text, '/') ||
672 			    strchr("!%*+-=?@}~",
673 				n->narg.text[strlen(n->narg.text) - 1]))
674 				synerror("Bad function name");
675 			rmescapes(n->narg.text);
676 			if (find_builtin(n->narg.text, &special) >= 0 &&
677 			    special)
678 				synerror("Cannot override a special builtin with a function");
679 			n->type = NDEFUN;
680 			n->narg.next = command();
681 			funclinno = 0;
682 			return n;
683 		} else {
684 			tokpushback++;
685 			break;
686 		}
687 	}
688 	*app = NULL;
689 	*rpp = NULL;
690 	n = (union node *)stalloc(sizeof (struct ncmd));
691 	n->type = NCMD;
692 	n->ncmd.args = args;
693 	n->ncmd.redirect = redir;
694 	return n;
695 }
696 
697 static union node *
698 makename(void)
699 {
700 	union node *n;
701 
702 	n = (union node *)stalloc(sizeof (struct narg));
703 	n->type = NARG;
704 	n->narg.next = NULL;
705 	n->narg.text = wordtext;
706 	n->narg.backquote = backquotelist;
707 	return n;
708 }
709 
710 static union node *
711 makebinary(int type, union node *n1, union node *n2)
712 {
713 	union node *n;
714 
715 	n = (union node *)stalloc(sizeof (struct nbinary));
716 	n->type = type;
717 	n->nbinary.ch1 = n1;
718 	n->nbinary.ch2 = n2;
719 	return (n);
720 }
721 
722 void
723 forcealias(void)
724 {
725 	checkkwd |= CHKALIAS;
726 }
727 
728 void
729 fixredir(union node *n, const char *text, int err)
730 {
731 	TRACE(("Fix redir %s %d\n", text, err));
732 	if (!err)
733 		n->ndup.vname = NULL;
734 
735 	if (is_digit(text[0]) && text[1] == '\0')
736 		n->ndup.dupfd = digit_val(text[0]);
737 	else if (text[0] == '-' && text[1] == '\0')
738 		n->ndup.dupfd = -1;
739 	else {
740 
741 		if (err)
742 			synerror("Bad fd number");
743 		else
744 			n->ndup.vname = makename();
745 	}
746 }
747 
748 
749 static void
750 parsefname(void)
751 {
752 	union node *n = redirnode;
753 
754 	consumetoken(TWORD);
755 	if (n->type == NHERE) {
756 		struct heredoc *here = heredoc;
757 		struct heredoc *p;
758 
759 		if (quoteflag == 0)
760 			n->type = NXHERE;
761 		TRACE(("Here document %d\n", n->type));
762 		if (here->striptabs) {
763 			while (*wordtext == '\t')
764 				wordtext++;
765 		}
766 		if (! noexpand(wordtext))
767 			synerror("Illegal eof marker for << redirection");
768 		rmescapes(wordtext);
769 		here->eofmark = wordtext;
770 		here->next = NULL;
771 		if (heredoclist == NULL)
772 			heredoclist = here;
773 		else {
774 			for (p = heredoclist ; p->next ; p = p->next);
775 			p->next = here;
776 		}
777 	} else if (n->type == NTOFD || n->type == NFROMFD) {
778 		fixredir(n, wordtext, 0);
779 	} else {
780 		n->nfile.fname = makename();
781 	}
782 }
783 
784 
785 /*
786  * Input any here documents.
787  */
788 
789 static void
790 parseheredoc(void)
791 {
792 	struct heredoc *here;
793 	union node *n;
794 
795 	while (heredoclist) {
796 		here = heredoclist;
797 		heredoclist = here->next;
798 		if (needprompt) {
799 			setprompt(2);
800 			needprompt = 0;
801 		}
802 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
803 				here->eofmark, here->striptabs);
804 		n = makename();
805 		here->here->nhere.doc = n;
806 	}
807 }
808 
809 static int
810 peektoken(void)
811 {
812 	int t;
813 
814 	t = readtoken();
815 	tokpushback++;
816 	return (t);
817 }
818 
819 static int
820 readtoken(void)
821 {
822 	int t;
823 	struct alias *ap;
824 #ifdef DEBUG
825 	int alreadyseen = tokpushback;
826 #endif
827 
828 	top:
829 	t = xxreadtoken();
830 
831 	/*
832 	 * eat newlines
833 	 */
834 	if (checkkwd & CHKNL) {
835 		while (t == TNL) {
836 			parseheredoc();
837 			t = xxreadtoken();
838 		}
839 	}
840 
841 	/*
842 	 * check for keywords and aliases
843 	 */
844 	if (t == TWORD && !quoteflag)
845 	{
846 		const char * const *pp;
847 
848 		if (checkkwd & CHKKWD)
849 			for (pp = parsekwd; *pp; pp++) {
850 				if (**pp == *wordtext && equal(*pp, wordtext))
851 				{
852 					lasttoken = t = pp - parsekwd + KWDOFFSET;
853 					TRACE(("keyword %s recognized\n", tokname[t]));
854 					goto out;
855 				}
856 			}
857 		if (checkkwd & CHKALIAS &&
858 		    (ap = lookupalias(wordtext, 1)) != NULL) {
859 			pushstring(ap->val, strlen(ap->val), ap);
860 			goto top;
861 		}
862 	}
863 out:
864 	if (t != TNOT)
865 		checkkwd = 0;
866 
867 #ifdef DEBUG
868 	if (!alreadyseen)
869 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
870 	else
871 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
872 #endif
873 	return (t);
874 }
875 
876 
877 /*
878  * Read the next input token.
879  * If the token is a word, we set backquotelist to the list of cmds in
880  *	backquotes.  We set quoteflag to true if any part of the word was
881  *	quoted.
882  * If the token is TREDIR, then we set redirnode to a structure containing
883  *	the redirection.
884  * In all cases, the variable startlinno is set to the number of the line
885  *	on which the token starts.
886  *
887  * [Change comment:  here documents and internal procedures]
888  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
889  *  word parsing code into a separate routine.  In this case, readtoken
890  *  doesn't need to have any internal procedures, but parseword does.
891  *  We could also make parseoperator in essence the main routine, and
892  *  have parseword (readtoken1?) handle both words and redirection.]
893  */
894 
895 #define RETURN(token)	return lasttoken = token
896 
897 static int
898 xxreadtoken(void)
899 {
900 	int c;
901 
902 	if (tokpushback) {
903 		tokpushback = 0;
904 		return lasttoken;
905 	}
906 	if (needprompt) {
907 		setprompt(2);
908 		needprompt = 0;
909 	}
910 	startlinno = plinno;
911 	for (;;) {	/* until token or start of word found */
912 		c = pgetc_macro();
913 		switch (c) {
914 		case ' ': case '\t':
915 			continue;
916 		case '#':
917 			while ((c = pgetc()) != '\n' && c != PEOF);
918 			pungetc();
919 			continue;
920 		case '\\':
921 			if (pgetc() == '\n') {
922 				startlinno = ++plinno;
923 				if (doprompt)
924 					setprompt(2);
925 				else
926 					setprompt(0);
927 				continue;
928 			}
929 			pungetc();
930 			/* FALLTHROUGH */
931 		default:
932 			return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
933 		case '\n':
934 			plinno++;
935 			needprompt = doprompt;
936 			RETURN(TNL);
937 		case PEOF:
938 			RETURN(TEOF);
939 		case '&':
940 			if (pgetc_linecont() == '&')
941 				RETURN(TAND);
942 			pungetc();
943 			RETURN(TBACKGND);
944 		case '|':
945 			if (pgetc_linecont() == '|')
946 				RETURN(TOR);
947 			pungetc();
948 			RETURN(TPIPE);
949 		case ';':
950 			c = pgetc_linecont();
951 			if (c == ';')
952 				RETURN(TENDCASE);
953 			else if (c == '&')
954 				RETURN(TFALLTHRU);
955 			pungetc();
956 			RETURN(TSEMI);
957 		case '(':
958 			RETURN(TLP);
959 		case ')':
960 			RETURN(TRP);
961 		}
962 	}
963 #undef RETURN
964 }
965 
966 
967 #define MAXNEST_static 8
968 struct tokenstate
969 {
970 	const char *syntax; /* *SYNTAX */
971 	int parenlevel; /* levels of parentheses in arithmetic */
972 	enum tokenstate_category
973 	{
974 		TSTATE_TOP,
975 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
976 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
977 		TSTATE_ARITH
978 	} category;
979 };
980 
981 
982 /*
983  * Check to see whether we are at the end of the here document.  When this
984  * is called, c is set to the first character of the next input line.  If
985  * we are at the end of the here document, this routine sets the c to PEOF.
986  * The new value of c is returned.
987  */
988 
989 static int
990 checkend(int c, const char *eofmark, int striptabs)
991 {
992 	if (striptabs) {
993 		while (c == '\t')
994 			c = pgetc();
995 	}
996 	if (c == *eofmark) {
997 		int c2;
998 		const char *q;
999 
1000 		for (q = eofmark + 1; c2 = pgetc(), *q != '\0' && c2 == *q; q++)
1001 			;
1002 		if ((c2 == PEOF || c2 == '\n') && *q == '\0') {
1003 			c = PEOF;
1004 			if (c2 == '\n') {
1005 				plinno++;
1006 				needprompt = doprompt;
1007 			}
1008 		} else {
1009 			pungetc();
1010 			pushstring(eofmark + 1, q - (eofmark + 1), NULL);
1011 		}
1012 	} else if (c == '\n' && *eofmark == '\0') {
1013 		c = PEOF;
1014 		plinno++;
1015 		needprompt = doprompt;
1016 	}
1017 	return (c);
1018 }
1019 
1020 
1021 /*
1022  * Parse a redirection operator.  The variable "out" points to a string
1023  * specifying the fd to be redirected.  The variable "c" contains the
1024  * first character of the redirection operator.
1025  */
1026 
1027 static void
1028 parseredir(char *out, int c)
1029 {
1030 	char fd = *out;
1031 	union node *np;
1032 
1033 	np = (union node *)stalloc(sizeof (struct nfile));
1034 	if (c == '>') {
1035 		np->nfile.fd = 1;
1036 		c = pgetc_linecont();
1037 		if (c == '>')
1038 			np->type = NAPPEND;
1039 		else if (c == '&')
1040 			np->type = NTOFD;
1041 		else if (c == '|')
1042 			np->type = NCLOBBER;
1043 		else {
1044 			np->type = NTO;
1045 			pungetc();
1046 		}
1047 	} else {	/* c == '<' */
1048 		np->nfile.fd = 0;
1049 		c = pgetc_linecont();
1050 		if (c == '<') {
1051 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1052 				np = (union node *)stalloc(sizeof (struct nhere));
1053 				np->nfile.fd = 0;
1054 			}
1055 			np->type = NHERE;
1056 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1057 			heredoc->here = np;
1058 			if ((c = pgetc_linecont()) == '-') {
1059 				heredoc->striptabs = 1;
1060 			} else {
1061 				heredoc->striptabs = 0;
1062 				pungetc();
1063 			}
1064 		} else if (c == '&')
1065 			np->type = NFROMFD;
1066 		else if (c == '>')
1067 			np->type = NFROMTO;
1068 		else {
1069 			np->type = NFROM;
1070 			pungetc();
1071 		}
1072 	}
1073 	if (fd != '\0')
1074 		np->nfile.fd = digit_val(fd);
1075 	redirnode = np;
1076 }
1077 
1078 /*
1079  * Called to parse command substitutions.
1080  */
1081 
1082 static char *
1083 parsebackq(char *out, struct nodelist **pbqlist,
1084 		int oldstyle, int dblquote, int quoted)
1085 {
1086 	struct nodelist **nlpp;
1087 	union node *n;
1088 	char *volatile str;
1089 	struct jmploc jmploc;
1090 	struct jmploc *const savehandler = handler;
1091 	size_t savelen;
1092 	int saveprompt;
1093 	const int bq_startlinno = plinno;
1094 	char *volatile ostr = NULL;
1095 	struct parsefile *const savetopfile = getcurrentfile();
1096 	struct heredoc *const saveheredoclist = heredoclist;
1097 	struct heredoc *here;
1098 
1099 	str = NULL;
1100 	if (setjmp(jmploc.loc)) {
1101 		popfilesupto(savetopfile);
1102 		if (str)
1103 			ckfree(str);
1104 		if (ostr)
1105 			ckfree(ostr);
1106 		heredoclist = saveheredoclist;
1107 		handler = savehandler;
1108 		if (exception == EXERROR) {
1109 			startlinno = bq_startlinno;
1110 			synerror("Error in command substitution");
1111 		}
1112 		longjmp(handler->loc, 1);
1113 	}
1114 	INTOFF;
1115 	savelen = out - stackblock();
1116 	if (savelen > 0) {
1117 		str = ckmalloc(savelen);
1118 		memcpy(str, stackblock(), savelen);
1119 	}
1120 	handler = &jmploc;
1121 	heredoclist = NULL;
1122 	INTON;
1123         if (oldstyle) {
1124                 /* We must read until the closing backquote, giving special
1125                    treatment to some slashes, and then push the string and
1126                    reread it as input, interpreting it normally.  */
1127                 char *oout;
1128                 int c;
1129                 int olen;
1130 
1131 
1132                 STARTSTACKSTR(oout);
1133 		for (;;) {
1134 			if (needprompt) {
1135 				setprompt(2);
1136 				needprompt = 0;
1137 			}
1138 			CHECKSTRSPACE(2, oout);
1139 			c = pgetc_linecont();
1140 			if (c == '`')
1141 				break;
1142 			switch (c) {
1143 			case '\\':
1144 				c = pgetc();
1145                                 if (c != '\\' && c != '`' && c != '$'
1146                                     && (!dblquote || c != '"'))
1147                                         USTPUTC('\\', oout);
1148 				break;
1149 
1150 			case '\n':
1151 				plinno++;
1152 				needprompt = doprompt;
1153 				break;
1154 
1155 			case PEOF:
1156 			        startlinno = plinno;
1157 				synerror("EOF in backquote substitution");
1158  				break;
1159 
1160 			default:
1161 				break;
1162 			}
1163 			USTPUTC(c, oout);
1164                 }
1165                 USTPUTC('\0', oout);
1166                 olen = oout - stackblock();
1167 		INTOFF;
1168 		ostr = ckmalloc(olen);
1169 		memcpy(ostr, stackblock(), olen);
1170 		setinputstring(ostr, 1);
1171 		INTON;
1172         }
1173 	nlpp = pbqlist;
1174 	while (*nlpp)
1175 		nlpp = &(*nlpp)->next;
1176 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1177 	(*nlpp)->next = NULL;
1178 
1179 	if (oldstyle) {
1180 		saveprompt = doprompt;
1181 		doprompt = 0;
1182 	}
1183 
1184 	n = list(0);
1185 
1186 	if (oldstyle) {
1187 		if (peektoken() != TEOF)
1188 			synexpect(-1);
1189 		doprompt = saveprompt;
1190 	} else
1191 		consumetoken(TRP);
1192 
1193 	(*nlpp)->n = n;
1194         if (oldstyle) {
1195 		/*
1196 		 * Start reading from old file again, ignoring any pushed back
1197 		 * tokens left from the backquote parsing
1198 		 */
1199                 popfile();
1200 		tokpushback = 0;
1201 	}
1202 	STARTSTACKSTR(out);
1203 	CHECKSTRSPACE(savelen + 1, out);
1204 	INTOFF;
1205 	if (str) {
1206 		memcpy(out, str, savelen);
1207 		STADJUST(savelen, out);
1208 		ckfree(str);
1209 		str = NULL;
1210 	}
1211 	if (ostr) {
1212 		ckfree(ostr);
1213 		ostr = NULL;
1214 	}
1215 	here = saveheredoclist;
1216 	if (here != NULL) {
1217 		while (here->next != NULL)
1218 			here = here->next;
1219 		here->next = heredoclist;
1220 		heredoclist = saveheredoclist;
1221 	}
1222 	handler = savehandler;
1223 	INTON;
1224 	if (quoted)
1225 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1226 	else
1227 		USTPUTC(CTLBACKQ, out);
1228 	return out;
1229 }
1230 
1231 
1232 /*
1233  * Called to parse a backslash escape sequence inside $'...'.
1234  * The backslash has already been read.
1235  */
1236 static char *
1237 readcstyleesc(char *out)
1238 {
1239 	int c, vc, i, n;
1240 	unsigned int v;
1241 
1242 	c = pgetc();
1243 	switch (c) {
1244 	case '\0':
1245 		synerror("Unterminated quoted string");
1246 	case '\n':
1247 		plinno++;
1248 		if (doprompt)
1249 			setprompt(2);
1250 		else
1251 			setprompt(0);
1252 		return out;
1253 	case '\\':
1254 	case '\'':
1255 	case '"':
1256 		v = c;
1257 		break;
1258 	case 'a': v = '\a'; break;
1259 	case 'b': v = '\b'; break;
1260 	case 'e': v = '\033'; break;
1261 	case 'f': v = '\f'; break;
1262 	case 'n': v = '\n'; break;
1263 	case 'r': v = '\r'; break;
1264 	case 't': v = '\t'; break;
1265 	case 'v': v = '\v'; break;
1266 	case 'x':
1267 		  v = 0;
1268 		  for (;;) {
1269 			  c = pgetc();
1270 			  if (c >= '0' && c <= '9')
1271 				  v = (v << 4) + c - '0';
1272 			  else if (c >= 'A' && c <= 'F')
1273 				  v = (v << 4) + c - 'A' + 10;
1274 			  else if (c >= 'a' && c <= 'f')
1275 				  v = (v << 4) + c - 'a' + 10;
1276 			  else
1277 				  break;
1278 		  }
1279 		  pungetc();
1280 		  break;
1281 	case '0': case '1': case '2': case '3':
1282 	case '4': case '5': case '6': case '7':
1283 		  v = c - '0';
1284 		  c = pgetc();
1285 		  if (c >= '0' && c <= '7') {
1286 			  v <<= 3;
1287 			  v += c - '0';
1288 			  c = pgetc();
1289 			  if (c >= '0' && c <= '7') {
1290 				  v <<= 3;
1291 				  v += c - '0';
1292 			  } else
1293 				  pungetc();
1294 		  } else
1295 			  pungetc();
1296 		  break;
1297 	case 'c':
1298 		  c = pgetc();
1299 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1300 			  synerror("Bad escape sequence");
1301 		  if (c == '\\' && pgetc() != '\\')
1302 			  synerror("Bad escape sequence");
1303 		  if (c == '?')
1304 			  v = 127;
1305 		  else
1306 			  v = c & 0x1f;
1307 		  break;
1308 	case 'u':
1309 	case 'U':
1310 		  n = c == 'U' ? 8 : 4;
1311 		  v = 0;
1312 		  for (i = 0; i < n; i++) {
1313 			  c = pgetc();
1314 			  if (c >= '0' && c <= '9')
1315 				  v = (v << 4) + c - '0';
1316 			  else if (c >= 'A' && c <= 'F')
1317 				  v = (v << 4) + c - 'A' + 10;
1318 			  else if (c >= 'a' && c <= 'f')
1319 				  v = (v << 4) + c - 'a' + 10;
1320 			  else
1321 				  synerror("Bad escape sequence");
1322 		  }
1323 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1324 			  synerror("Bad escape sequence");
1325 		  /* We really need iconv here. */
1326 		  if (initial_localeisutf8 && v > 127) {
1327 			  CHECKSTRSPACE(4, out);
1328 			  /*
1329 			   * We cannot use wctomb() as the locale may have
1330 			   * changed.
1331 			   */
1332 			  if (v <= 0x7ff) {
1333 				  USTPUTC(0xc0 | v >> 6, out);
1334 				  USTPUTC(0x80 | (v & 0x3f), out);
1335 				  return out;
1336 			  } else if (v <= 0xffff) {
1337 				  USTPUTC(0xe0 | v >> 12, out);
1338 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1339 				  USTPUTC(0x80 | (v & 0x3f), out);
1340 				  return out;
1341 			  } else if (v <= 0x10ffff) {
1342 				  USTPUTC(0xf0 | v >> 18, out);
1343 				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1344 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1345 				  USTPUTC(0x80 | (v & 0x3f), out);
1346 				  return out;
1347 			  }
1348 		  }
1349 		  if (v > 127)
1350 			  v = '?';
1351 		  break;
1352 	default:
1353 		  synerror("Bad escape sequence");
1354 	}
1355 	vc = (char)v;
1356 	/*
1357 	 * We can't handle NUL bytes.
1358 	 * POSIX says we should skip till the closing quote.
1359 	 */
1360 	if (vc == '\0') {
1361 		while ((c = pgetc()) != '\'') {
1362 			if (c == '\\')
1363 				c = pgetc();
1364 			if (c == PEOF)
1365 				synerror("Unterminated quoted string");
1366 			if (c == '\n') {
1367 				plinno++;
1368 				if (doprompt)
1369 					setprompt(2);
1370 				else
1371 					setprompt(0);
1372 			}
1373 		}
1374 		pungetc();
1375 		return out;
1376 	}
1377 	if (SQSYNTAX[vc] == CCTL)
1378 		USTPUTC(CTLESC, out);
1379 	USTPUTC(vc, out);
1380 	return out;
1381 }
1382 
1383 
1384 /*
1385  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1386  * is not NULL, read a here document.  In the latter case, eofmark is the
1387  * word which marks the end of the document and striptabs is true if
1388  * leading tabs should be stripped from the document.  The argument firstc
1389  * is the first character of the input token or document.
1390  *
1391  * Because C does not have internal subroutines, I have simulated them
1392  * using goto's to implement the subroutine linkage.  The following macros
1393  * will run code that appears at the end of readtoken1.
1394  */
1395 
1396 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1397 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1398 
1399 static int
1400 readtoken1(int firstc, char const *initialsyntax, const char *eofmark,
1401     int striptabs)
1402 {
1403 	int c = firstc;
1404 	char *out;
1405 	int len;
1406 	struct nodelist *bqlist;
1407 	int quotef;
1408 	int newvarnest;
1409 	int level;
1410 	int synentry;
1411 	struct tokenstate state_static[MAXNEST_static];
1412 	int maxnest = MAXNEST_static;
1413 	struct tokenstate *state = state_static;
1414 	int sqiscstyle = 0;
1415 
1416 	startlinno = plinno;
1417 	quotef = 0;
1418 	bqlist = NULL;
1419 	newvarnest = 0;
1420 	level = 0;
1421 	state[level].syntax = initialsyntax;
1422 	state[level].parenlevel = 0;
1423 	state[level].category = TSTATE_TOP;
1424 
1425 	STARTSTACKSTR(out);
1426 	loop: {	/* for each line, until end of word */
1427 		if (eofmark && eofmark != NOEOFMARK)
1428 			/* set c to PEOF if at end of here document */
1429 			c = checkend(c, eofmark, striptabs);
1430 		for (;;) {	/* until end of line or end of word */
1431 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1432 
1433 			synentry = state[level].syntax[c];
1434 
1435 			switch(synentry) {
1436 			case CNL:	/* '\n' */
1437 				if (level == 0)
1438 					goto endword;	/* exit outer loop */
1439 				/* FALLTHROUGH */
1440 			case CQNL:
1441 				USTPUTC(c, out);
1442 				plinno++;
1443 				if (doprompt)
1444 					setprompt(2);
1445 				else
1446 					setprompt(0);
1447 				c = pgetc();
1448 				goto loop;		/* continue outer loop */
1449 			case CSBACK:
1450 				if (sqiscstyle) {
1451 					out = readcstyleesc(out);
1452 					break;
1453 				}
1454 				/* FALLTHROUGH */
1455 			case CWORD:
1456 				USTPUTC(c, out);
1457 				break;
1458 			case CCTL:
1459 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1460 					USTPUTC(CTLESC, out);
1461 				USTPUTC(c, out);
1462 				break;
1463 			case CBACK:	/* backslash */
1464 				c = pgetc();
1465 				if (c == PEOF) {
1466 					USTPUTC('\\', out);
1467 					pungetc();
1468 				} else if (c == '\n') {
1469 					plinno++;
1470 					if (doprompt)
1471 						setprompt(2);
1472 					else
1473 						setprompt(0);
1474 				} else {
1475 					if (state[level].syntax == DQSYNTAX &&
1476 					    c != '\\' && c != '`' && c != '$' &&
1477 					    (c != '"' || (eofmark != NULL &&
1478 						newvarnest == 0)) &&
1479 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1480 						USTPUTC('\\', out);
1481 					if ((eofmark == NULL ||
1482 					    newvarnest > 0) &&
1483 					    state[level].syntax == BASESYNTAX)
1484 						USTPUTC(CTLQUOTEMARK, out);
1485 					if (SQSYNTAX[c] == CCTL)
1486 						USTPUTC(CTLESC, out);
1487 					USTPUTC(c, out);
1488 					if ((eofmark == NULL ||
1489 					    newvarnest > 0) &&
1490 					    state[level].syntax == BASESYNTAX &&
1491 					    state[level].category == TSTATE_VAR_OLD)
1492 						USTPUTC(CTLQUOTEEND, out);
1493 					quotef++;
1494 				}
1495 				break;
1496 			case CSQUOTE:
1497 				USTPUTC(CTLQUOTEMARK, out);
1498 				state[level].syntax = SQSYNTAX;
1499 				sqiscstyle = 0;
1500 				break;
1501 			case CDQUOTE:
1502 				USTPUTC(CTLQUOTEMARK, out);
1503 				state[level].syntax = DQSYNTAX;
1504 				break;
1505 			case CENDQUOTE:
1506 				if (eofmark != NULL && newvarnest == 0)
1507 					USTPUTC(c, out);
1508 				else {
1509 					if (state[level].category == TSTATE_VAR_OLD)
1510 						USTPUTC(CTLQUOTEEND, out);
1511 					state[level].syntax = BASESYNTAX;
1512 					quotef++;
1513 				}
1514 				break;
1515 			case CVAR:	/* '$' */
1516 				PARSESUB();		/* parse substitution */
1517 				break;
1518 			case CENDVAR:	/* '}' */
1519 				if (level > 0 &&
1520 				    ((state[level].category == TSTATE_VAR_OLD &&
1521 				      state[level].syntax ==
1522 				      state[level - 1].syntax) ||
1523 				    (state[level].category == TSTATE_VAR_NEW &&
1524 				     state[level].syntax == BASESYNTAX))) {
1525 					if (state[level].category == TSTATE_VAR_NEW)
1526 						newvarnest--;
1527 					level--;
1528 					USTPUTC(CTLENDVAR, out);
1529 				} else {
1530 					USTPUTC(c, out);
1531 				}
1532 				break;
1533 			case CLP:	/* '(' in arithmetic */
1534 				state[level].parenlevel++;
1535 				USTPUTC(c, out);
1536 				break;
1537 			case CRP:	/* ')' in arithmetic */
1538 				if (state[level].parenlevel > 0) {
1539 					USTPUTC(c, out);
1540 					--state[level].parenlevel;
1541 				} else {
1542 					if (pgetc_linecont() == ')') {
1543 						if (level > 0 &&
1544 						    state[level].category == TSTATE_ARITH) {
1545 							level--;
1546 							USTPUTC(CTLENDARI, out);
1547 						} else
1548 							USTPUTC(')', out);
1549 					} else {
1550 						/*
1551 						 * unbalanced parens
1552 						 *  (don't 2nd guess - no error)
1553 						 */
1554 						pungetc();
1555 						USTPUTC(')', out);
1556 					}
1557 				}
1558 				break;
1559 			case CBQUOTE:	/* '`' */
1560 				out = parsebackq(out, &bqlist, 1,
1561 				    state[level].syntax == DQSYNTAX &&
1562 				    (eofmark == NULL || newvarnest > 0),
1563 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1564 				break;
1565 			case CEOF:
1566 				goto endword;		/* exit outer loop */
1567 			case CIGN:
1568 				break;
1569 			default:
1570 				if (level == 0)
1571 					goto endword;	/* exit outer loop */
1572 				USTPUTC(c, out);
1573 			}
1574 			c = pgetc_macro();
1575 		}
1576 	}
1577 endword:
1578 	if (state[level].syntax == ARISYNTAX)
1579 		synerror("Missing '))'");
1580 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1581 		synerror("Unterminated quoted string");
1582 	if (state[level].category == TSTATE_VAR_OLD ||
1583 	    state[level].category == TSTATE_VAR_NEW) {
1584 		startlinno = plinno;
1585 		synerror("Missing '}'");
1586 	}
1587 	if (state != state_static)
1588 		parser_temp_free_upto(state);
1589 	USTPUTC('\0', out);
1590 	len = out - stackblock();
1591 	out = stackblock();
1592 	if (eofmark == NULL) {
1593 		if ((c == '>' || c == '<')
1594 		 && quotef == 0
1595 		 && len <= 2
1596 		 && (*out == '\0' || is_digit(*out))) {
1597 			parseredir(out, c);
1598 			return lasttoken = TREDIR;
1599 		} else {
1600 			pungetc();
1601 		}
1602 	}
1603 	quoteflag = quotef;
1604 	backquotelist = bqlist;
1605 	grabstackblock(len);
1606 	wordtext = out;
1607 	return lasttoken = TWORD;
1608 /* end of readtoken routine */
1609 
1610 
1611 /*
1612  * Parse a substitution.  At this point, we have read the dollar sign
1613  * and nothing else.
1614  */
1615 
1616 parsesub: {
1617 	int subtype;
1618 	int typeloc;
1619 	int flags;
1620 	char *p;
1621 	static const char types[] = "}-+?=";
1622 	int linno;
1623 	int length;
1624 	int c1;
1625 
1626 	c = pgetc_linecont();
1627 	if (c == '(') {	/* $(command) or $((arith)) */
1628 		if (pgetc_linecont() == '(') {
1629 			PARSEARITH();
1630 		} else {
1631 			pungetc();
1632 			out = parsebackq(out, &bqlist, 0,
1633 			    state[level].syntax == DQSYNTAX &&
1634 			    (eofmark == NULL || newvarnest > 0),
1635 			    state[level].syntax == DQSYNTAX ||
1636 			    state[level].syntax == ARISYNTAX);
1637 		}
1638 	} else if (c == '{' || is_name(c) || is_special(c)) {
1639 		USTPUTC(CTLVAR, out);
1640 		typeloc = out - stackblock();
1641 		USTPUTC(VSNORMAL, out);
1642 		subtype = VSNORMAL;
1643 		flags = 0;
1644 		if (c == '{') {
1645 			c = pgetc_linecont();
1646 			subtype = 0;
1647 		}
1648 varname:
1649 		if (!is_eof(c) && is_name(c)) {
1650 			length = 0;
1651 			do {
1652 				STPUTC(c, out);
1653 				c = pgetc_linecont();
1654 				length++;
1655 			} while (!is_eof(c) && is_in_name(c));
1656 			if (length == 6 &&
1657 			    strncmp(out - length, "LINENO", length) == 0) {
1658 				/* Replace the variable name with the
1659 				 * current line number. */
1660 				STADJUST(-6, out);
1661 				CHECKSTRSPACE(11, out);
1662 				linno = plinno;
1663 				if (funclinno != 0)
1664 					linno -= funclinno - 1;
1665 				length = snprintf(out, 11, "%d", linno);
1666 				if (length > 10)
1667 					length = 10;
1668 				out += length;
1669 				flags |= VSLINENO;
1670 			}
1671 		} else if (is_digit(c)) {
1672 			if (subtype != VSNORMAL) {
1673 				do {
1674 					STPUTC(c, out);
1675 					c = pgetc_linecont();
1676 				} while (is_digit(c));
1677 			} else {
1678 				USTPUTC(c, out);
1679 				c = pgetc_linecont();
1680 			}
1681 		} else if (is_special(c)) {
1682 			c1 = c;
1683 			c = pgetc_linecont();
1684 			if (subtype == 0 && c1 == '#') {
1685 				subtype = VSLENGTH;
1686 				if (strchr(types, c) == NULL && c != ':' &&
1687 				    c != '#' && c != '%')
1688 					goto varname;
1689 				c1 = c;
1690 				c = pgetc_linecont();
1691 				if (c1 != '}' && c == '}') {
1692 					pungetc();
1693 					c = c1;
1694 					goto varname;
1695 				}
1696 				pungetc();
1697 				c = c1;
1698 				c1 = '#';
1699 				subtype = 0;
1700 			}
1701 			USTPUTC(c1, out);
1702 		} else {
1703 			subtype = VSERROR;
1704 			if (c == '}')
1705 				pungetc();
1706 			else if (c == '\n' || c == PEOF)
1707 				synerror("Unexpected end of line in substitution");
1708 			else if (BASESYNTAX[c] != CCTL)
1709 				USTPUTC(c, out);
1710 		}
1711 		if (subtype == 0) {
1712 			switch (c) {
1713 			case ':':
1714 				flags |= VSNUL;
1715 				c = pgetc_linecont();
1716 				/*FALLTHROUGH*/
1717 			default:
1718 				p = strchr(types, c);
1719 				if (p == NULL) {
1720 					if (c == '\n' || c == PEOF)
1721 						synerror("Unexpected end of line in substitution");
1722 					if (flags == VSNUL)
1723 						STPUTC(':', out);
1724 					if (BASESYNTAX[c] != CCTL)
1725 						STPUTC(c, out);
1726 					subtype = VSERROR;
1727 				} else
1728 					subtype = p - types + VSNORMAL;
1729 				break;
1730 			case '%':
1731 			case '#':
1732 				{
1733 					int cc = c;
1734 					subtype = c == '#' ? VSTRIMLEFT :
1735 							     VSTRIMRIGHT;
1736 					c = pgetc_linecont();
1737 					if (c == cc)
1738 						subtype++;
1739 					else
1740 						pungetc();
1741 					break;
1742 				}
1743 			}
1744 		} else if (subtype != VSERROR) {
1745 			if (subtype == VSLENGTH && c != '}')
1746 				subtype = VSERROR;
1747 			pungetc();
1748 		}
1749 		STPUTC('=', out);
1750 		if (state[level].syntax == DQSYNTAX ||
1751 		    state[level].syntax == ARISYNTAX)
1752 			flags |= VSQUOTE;
1753 		*(stackblock() + typeloc) = subtype | flags;
1754 		if (subtype != VSNORMAL) {
1755 			if (level + 1 >= maxnest) {
1756 				maxnest *= 2;
1757 				if (state == state_static) {
1758 					state = parser_temp_alloc(
1759 					    maxnest * sizeof(*state));
1760 					memcpy(state, state_static,
1761 					    MAXNEST_static * sizeof(*state));
1762 				} else
1763 					state = parser_temp_realloc(state,
1764 					    maxnest * sizeof(*state));
1765 			}
1766 			level++;
1767 			state[level].parenlevel = 0;
1768 			if (subtype == VSMINUS || subtype == VSPLUS ||
1769 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1770 				/*
1771 				 * For operators that were in the Bourne shell,
1772 				 * inherit the double-quote state.
1773 				 */
1774 				state[level].syntax = state[level - 1].syntax;
1775 				state[level].category = TSTATE_VAR_OLD;
1776 			} else {
1777 				/*
1778 				 * The other operators take a pattern,
1779 				 * so go to BASESYNTAX.
1780 				 * Also, ' and " are now special, even
1781 				 * in here documents.
1782 				 */
1783 				state[level].syntax = BASESYNTAX;
1784 				state[level].category = TSTATE_VAR_NEW;
1785 				newvarnest++;
1786 			}
1787 		}
1788 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1789 		/* $'cstylequotes' */
1790 		USTPUTC(CTLQUOTEMARK, out);
1791 		state[level].syntax = SQSYNTAX;
1792 		sqiscstyle = 1;
1793 	} else {
1794 		USTPUTC('$', out);
1795 		pungetc();
1796 	}
1797 	goto parsesub_return;
1798 }
1799 
1800 
1801 /*
1802  * Parse an arithmetic expansion (indicate start of one and set state)
1803  */
1804 parsearith: {
1805 
1806 	if (level + 1 >= maxnest) {
1807 		maxnest *= 2;
1808 		if (state == state_static) {
1809 			state = parser_temp_alloc(
1810 			    maxnest * sizeof(*state));
1811 			memcpy(state, state_static,
1812 			    MAXNEST_static * sizeof(*state));
1813 		} else
1814 			state = parser_temp_realloc(state,
1815 			    maxnest * sizeof(*state));
1816 	}
1817 	level++;
1818 	state[level].syntax = ARISYNTAX;
1819 	state[level].parenlevel = 0;
1820 	state[level].category = TSTATE_ARITH;
1821 	USTPUTC(CTLARI, out);
1822 	if (state[level - 1].syntax == DQSYNTAX)
1823 		USTPUTC('"',out);
1824 	else
1825 		USTPUTC(' ',out);
1826 	goto parsearith_return;
1827 }
1828 
1829 } /* end of readtoken */
1830 
1831 
1832 /*
1833  * Returns true if the text contains nothing to expand (no dollar signs
1834  * or backquotes).
1835  */
1836 
1837 static int
1838 noexpand(char *text)
1839 {
1840 	char *p;
1841 	char c;
1842 
1843 	p = text;
1844 	while ((c = *p++) != '\0') {
1845 		if ( c == CTLQUOTEMARK)
1846 			continue;
1847 		if (c == CTLESC)
1848 			p++;
1849 		else if (BASESYNTAX[(int)c] == CCTL)
1850 			return 0;
1851 	}
1852 	return 1;
1853 }
1854 
1855 
1856 /*
1857  * Return true if the argument is a legal variable name (a letter or
1858  * underscore followed by zero or more letters, underscores, and digits).
1859  */
1860 
1861 int
1862 goodname(const char *name)
1863 {
1864 	const char *p;
1865 
1866 	p = name;
1867 	if (! is_name(*p))
1868 		return 0;
1869 	while (*++p) {
1870 		if (! is_in_name(*p))
1871 			return 0;
1872 	}
1873 	return 1;
1874 }
1875 
1876 
1877 int
1878 isassignment(const char *p)
1879 {
1880 	if (!is_name(*p))
1881 		return 0;
1882 	p++;
1883 	for (;;) {
1884 		if (*p == '=')
1885 			return 1;
1886 		else if (!is_in_name(*p))
1887 			return 0;
1888 		p++;
1889 	}
1890 }
1891 
1892 
1893 static void
1894 consumetoken(int token)
1895 {
1896 	if (readtoken() != token)
1897 		synexpect(token);
1898 }
1899 
1900 
1901 /*
1902  * Called when an unexpected token is read during the parse.  The argument
1903  * is the token that is expected, or -1 if more than one type of token can
1904  * occur at this point.
1905  */
1906 
1907 static void
1908 synexpect(int token)
1909 {
1910 	char msg[64];
1911 
1912 	if (token >= 0) {
1913 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1914 			tokname[lasttoken], tokname[token]);
1915 	} else {
1916 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1917 	}
1918 	synerror(msg);
1919 }
1920 
1921 
1922 static void
1923 synerror(const char *msg)
1924 {
1925 	if (commandname)
1926 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1927 	else if (arg0)
1928 		outfmt(out2, "%s: ", arg0);
1929 	outfmt(out2, "Syntax error: %s\n", msg);
1930 	error((char *)NULL);
1931 }
1932 
1933 static void
1934 setprompt(int which)
1935 {
1936 	whichprompt = which;
1937 	if (which == 0)
1938 		return;
1939 
1940 #ifndef NO_HISTORY
1941 	if (!el)
1942 #endif
1943 	{
1944 		out2str(getprompt(NULL));
1945 		flushout(out2);
1946 	}
1947 }
1948 
1949 static int
1950 pgetc_linecont(void)
1951 {
1952 	int c;
1953 
1954 	while ((c = pgetc_macro()) == '\\') {
1955 		c = pgetc();
1956 		if (c == '\n') {
1957 			plinno++;
1958 			if (doprompt)
1959 				setprompt(2);
1960 			else
1961 				setprompt(0);
1962 		} else {
1963 			pungetc();
1964 			/* Allow the backslash to be pushed back. */
1965 			pushstring("\\", 1, NULL);
1966 			return (pgetc());
1967 		}
1968 	}
1969 	return (c);
1970 }
1971 
1972 /*
1973  * called by editline -- any expansions to the prompt
1974  *    should be added here.
1975  */
1976 char *
1977 getprompt(void *unused __unused)
1978 {
1979 	static char ps[PROMPTLEN];
1980 	const char *fmt;
1981 	const char *pwd;
1982 	int i, trim;
1983 	static char internal_error[] = "??";
1984 
1985 	/*
1986 	 * Select prompt format.
1987 	 */
1988 	switch (whichprompt) {
1989 	case 0:
1990 		fmt = "";
1991 		break;
1992 	case 1:
1993 		fmt = ps1val();
1994 		break;
1995 	case 2:
1996 		fmt = ps2val();
1997 		break;
1998 	default:
1999 		return internal_error;
2000 	}
2001 
2002 	/*
2003 	 * Format prompt string.
2004 	 */
2005 	for (i = 0; (i < PROMPTLEN - 1) && (*fmt != '\0'); i++, fmt++)
2006 		if (*fmt == '\\')
2007 			switch (*++fmt) {
2008 
2009 				/*
2010 				 * Hostname.
2011 				 *
2012 				 * \h specifies just the local hostname,
2013 				 * \H specifies fully-qualified hostname.
2014 				 */
2015 			case 'h':
2016 			case 'H':
2017 				ps[i] = '\0';
2018 				gethostname(&ps[i], PROMPTLEN - i - 1);
2019 				ps[PROMPTLEN - 1] = '\0';
2020 				/* Skip to end of hostname. */
2021 				trim = (*fmt == 'h') ? '.' : '\0';
2022 				while ((ps[i] != '\0') && (ps[i] != trim))
2023 					i++;
2024 				--i;
2025 				break;
2026 
2027 				/*
2028 				 * Working directory.
2029 				 *
2030 				 * \W specifies just the final component,
2031 				 * \w specifies the entire path.
2032 				 */
2033 			case 'W':
2034 			case 'w':
2035 				pwd = lookupvar("PWD");
2036 				if (pwd == NULL || *pwd == '\0')
2037 					pwd = "?";
2038 				if (*fmt == 'W' &&
2039 				    *pwd == '/' && pwd[1] != '\0')
2040 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
2041 					    PROMPTLEN - i);
2042 				else
2043 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
2044 				/* Skip to end of path. */
2045 				while (ps[i + 1] != '\0')
2046 					i++;
2047 				break;
2048 
2049 				/*
2050 				 * Superuser status.
2051 				 *
2052 				 * '$' for normal users, '#' for root.
2053 				 */
2054 			case '$':
2055 				ps[i] = (geteuid() != 0) ? '$' : '#';
2056 				break;
2057 
2058 				/*
2059 				 * A literal \.
2060 				 */
2061 			case '\\':
2062 				ps[i] = '\\';
2063 				break;
2064 
2065 				/*
2066 				 * Emit unrecognized formats verbatim.
2067 				 */
2068 			default:
2069 				ps[i] = '\\';
2070 				if (i < PROMPTLEN - 2)
2071 					ps[++i] = *fmt;
2072 				break;
2073 			}
2074 		else
2075 			ps[i] = *fmt;
2076 	ps[i] = '\0';
2077 	return (ps);
2078 }
2079 
2080 
2081 const char *
2082 expandstr(const char *ps)
2083 {
2084 	union node n;
2085 	struct jmploc jmploc;
2086 	struct jmploc *const savehandler = handler;
2087 	const int saveprompt = doprompt;
2088 	struct parsefile *const savetopfile = getcurrentfile();
2089 	struct parser_temp *const saveparser_temp = parser_temp;
2090 	const char *result = NULL;
2091 
2092 	if (!setjmp(jmploc.loc)) {
2093 		handler = &jmploc;
2094 		parser_temp = NULL;
2095 		setinputstring(ps, 1);
2096 		doprompt = 0;
2097 		readtoken1(pgetc(), DQSYNTAX, NOEOFMARK, 0);
2098 		if (backquotelist != NULL)
2099 			error("Command substitution not allowed here");
2100 
2101 		n.narg.type = NARG;
2102 		n.narg.next = NULL;
2103 		n.narg.text = wordtext;
2104 		n.narg.backquote = backquotelist;
2105 
2106 		expandarg(&n, NULL, 0);
2107 		result = stackblock();
2108 		INTOFF;
2109 	}
2110 	handler = savehandler;
2111 	doprompt = saveprompt;
2112 	popfilesupto(savetopfile);
2113 	if (parser_temp != saveparser_temp) {
2114 		parser_temp_free_all();
2115 		parser_temp = saveparser_temp;
2116 	}
2117 	if (result != NULL) {
2118 		INTON;
2119 	} else if (exception == EXINT)
2120 		raise(SIGINT);
2121 	return result;
2122 }
2123