xref: /freebsd/bin/sh/parser.c (revision f5463265955b829775bbb32e1fd0bc11dafc36ce)
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 #include <sys/param.h>
36 #include <pwd.h>
37 #include <stdlib.h>
38 #include <unistd.h>
39 #include <stdio.h>
40 #include <time.h>
41 
42 #include "shell.h"
43 #include "parser.h"
44 #include "nodes.h"
45 #include "expand.h"	/* defines rmescapes() */
46 #include "syntax.h"
47 #include "options.h"
48 #include "input.h"
49 #include "output.h"
50 #include "var.h"
51 #include "error.h"
52 #include "memalloc.h"
53 #include "mystring.h"
54 #include "alias.h"
55 #include "show.h"
56 #include "eval.h"
57 #include "exec.h"	/* to check for special builtins */
58 #ifndef NO_HISTORY
59 #include "myhistedit.h"
60 #endif
61 
62 /*
63  * Shell command parser.
64  */
65 
66 #define	PROMPTLEN	192
67 
68 /* values of checkkwd variable */
69 #define CHKALIAS	0x1
70 #define CHKKWD		0x2
71 #define CHKNL		0x4
72 
73 /* values returned by readtoken */
74 #include "token.h"
75 
76 
77 
78 struct heredoc {
79 	struct heredoc *next;	/* next here document in list */
80 	union node *here;		/* redirection node */
81 	char *eofmark;		/* string indicating end of input */
82 	int striptabs;		/* if set, strip leading tabs */
83 };
84 
85 struct parser_temp {
86 	struct parser_temp *next;
87 	void *data;
88 };
89 
90 
91 static struct heredoc *heredoclist;	/* list of here documents to read */
92 static int doprompt;		/* if set, prompt the user */
93 static int needprompt;		/* true if interactive and at start of line */
94 static int lasttoken;		/* last token read */
95 static int tokpushback;		/* last token pushed back */
96 static char *wordtext;		/* text of last word returned by readtoken */
97 static int checkkwd;
98 static struct nodelist *backquotelist;
99 static union node *redirnode;
100 static struct heredoc *heredoc;
101 static int quoteflag;		/* set if (part of) last token was quoted */
102 static int startlinno;		/* line # where last token started */
103 static int funclinno;		/* line # where the current function started */
104 static struct parser_temp *parser_temp;
105 
106 #define NOEOFMARK ((const char *)&heredoclist)
107 
108 
109 static union node *list(int);
110 static union node *andor(void);
111 static union node *pipeline(void);
112 static union node *command(void);
113 static union node *simplecmd(union node **, union node *);
114 static union node *makename(void);
115 static union node *makebinary(int type, union node *n1, union node *n2);
116 static void parsefname(void);
117 static void parseheredoc(void);
118 static int peektoken(void);
119 static int readtoken(void);
120 static int xxreadtoken(void);
121 static int readtoken1(int, const char *, const char *, int);
122 static int noexpand(char *);
123 static void consumetoken(int);
124 static void synexpect(int) __dead2;
125 static void synerror(const char *) __dead2;
126 static void setprompt(int);
127 static int pgetc_linecont(void);
128 static void getusername(char *, size_t);
129 
130 
131 static void *
132 parser_temp_alloc(size_t len)
133 {
134 	struct parser_temp *t;
135 
136 	INTOFF;
137 	t = ckmalloc(sizeof(*t));
138 	t->data = NULL;
139 	t->next = parser_temp;
140 	parser_temp = t;
141 	t->data = ckmalloc(len);
142 	INTON;
143 	return t->data;
144 }
145 
146 
147 static void *
148 parser_temp_realloc(void *ptr, size_t len)
149 {
150 	struct parser_temp *t;
151 
152 	INTOFF;
153 	t = parser_temp;
154 	if (ptr != t->data)
155 		error("bug: parser_temp_realloc misused");
156 	t->data = ckrealloc(t->data, len);
157 	INTON;
158 	return t->data;
159 }
160 
161 
162 static void
163 parser_temp_free_upto(void *ptr)
164 {
165 	struct parser_temp *t;
166 	int done = 0;
167 
168 	INTOFF;
169 	while (parser_temp != NULL && !done) {
170 		t = parser_temp;
171 		parser_temp = t->next;
172 		done = t->data == ptr;
173 		ckfree(t->data);
174 		ckfree(t);
175 	}
176 	INTON;
177 	if (!done)
178 		error("bug: parser_temp_free_upto misused");
179 }
180 
181 
182 static void
183 parser_temp_free_all(void)
184 {
185 	struct parser_temp *t;
186 
187 	INTOFF;
188 	while (parser_temp != NULL) {
189 		t = parser_temp;
190 		parser_temp = t->next;
191 		ckfree(t->data);
192 		ckfree(t);
193 	}
194 	INTON;
195 }
196 
197 
198 /*
199  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
200  * valid parse tree indicating a blank line.)
201  */
202 
203 union node *
204 parsecmd(int interact)
205 {
206 	int t;
207 
208 	/* This assumes the parser is not re-entered,
209 	 * which could happen if we add command substitution on PS1/PS2.
210 	 */
211 	parser_temp_free_all();
212 	heredoclist = NULL;
213 
214 	tokpushback = 0;
215 	checkkwd = 0;
216 	doprompt = interact;
217 	if (doprompt)
218 		setprompt(1);
219 	else
220 		setprompt(0);
221 	needprompt = 0;
222 	t = readtoken();
223 	if (t == TEOF)
224 		return NEOF;
225 	if (t == TNL)
226 		return NULL;
227 	tokpushback++;
228 	return list(1);
229 }
230 
231 
232 /*
233  * Read and parse words for wordexp.
234  * Returns a list of NARG nodes; NULL if there are no words.
235  */
236 union node *
237 parsewordexp(void)
238 {
239 	union node *n, *first = NULL, **pnext;
240 	int t;
241 
242 	/* This assumes the parser is not re-entered,
243 	 * which could happen if we add command substitution on PS1/PS2.
244 	 */
245 	parser_temp_free_all();
246 	heredoclist = NULL;
247 
248 	tokpushback = 0;
249 	checkkwd = 0;
250 	doprompt = 0;
251 	setprompt(0);
252 	needprompt = 0;
253 	pnext = &first;
254 	while ((t = readtoken()) != TEOF) {
255 		if (t != TWORD)
256 			synexpect(TWORD);
257 		n = makename();
258 		*pnext = n;
259 		pnext = &n->narg.next;
260 	}
261 	return first;
262 }
263 
264 
265 static union node *
266 list(int nlflag)
267 {
268 	union node *ntop, *n1, *n2, *n3;
269 	int tok;
270 
271 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
272 	if (!nlflag && tokendlist[peektoken()])
273 		return NULL;
274 	ntop = n1 = NULL;
275 	for (;;) {
276 		n2 = andor();
277 		tok = readtoken();
278 		if (tok == TBACKGND) {
279 			if (n2 != NULL && n2->type == NPIPE) {
280 				n2->npipe.backgnd = 1;
281 			} else if (n2 != NULL && n2->type == NREDIR) {
282 				n2->type = NBACKGND;
283 			} else {
284 				n3 = (union node *)stalloc(sizeof (struct nredir));
285 				n3->type = NBACKGND;
286 				n3->nredir.n = n2;
287 				n3->nredir.redirect = NULL;
288 				n2 = n3;
289 			}
290 		}
291 		if (ntop == NULL)
292 			ntop = n2;
293 		else if (n1 == NULL) {
294 			n1 = makebinary(NSEMI, ntop, n2);
295 			ntop = n1;
296 		}
297 		else {
298 			n3 = makebinary(NSEMI, n1->nbinary.ch2, n2);
299 			n1->nbinary.ch2 = n3;
300 			n1 = n3;
301 		}
302 		switch (tok) {
303 		case TBACKGND:
304 		case TSEMI:
305 			tok = readtoken();
306 			/* FALLTHROUGH */
307 		case TNL:
308 			if (tok == TNL) {
309 				parseheredoc();
310 				if (nlflag)
311 					return ntop;
312 			} else if (tok == TEOF && nlflag) {
313 				parseheredoc();
314 				return ntop;
315 			} else {
316 				tokpushback++;
317 			}
318 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
319 			if (!nlflag && tokendlist[peektoken()])
320 				return ntop;
321 			break;
322 		case TEOF:
323 			if (heredoclist)
324 				parseheredoc();
325 			else
326 				pungetc();		/* push back EOF on input */
327 			return ntop;
328 		default:
329 			if (nlflag)
330 				synexpect(-1);
331 			tokpushback++;
332 			return ntop;
333 		}
334 	}
335 }
336 
337 
338 
339 static union node *
340 andor(void)
341 {
342 	union node *n;
343 	int t;
344 
345 	n = pipeline();
346 	for (;;) {
347 		if ((t = readtoken()) == TAND) {
348 			t = NAND;
349 		} else if (t == TOR) {
350 			t = NOR;
351 		} else {
352 			tokpushback++;
353 			return n;
354 		}
355 		n = makebinary(t, n, pipeline());
356 	}
357 }
358 
359 
360 
361 static union node *
362 pipeline(void)
363 {
364 	union node *n1, *n2, *pipenode;
365 	struct nodelist *lp, *prev;
366 	int negate, t;
367 
368 	negate = 0;
369 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
370 	TRACE(("pipeline: entered\n"));
371 	while (readtoken() == TNOT)
372 		negate = !negate;
373 	tokpushback++;
374 	n1 = command();
375 	if (readtoken() == TPIPE) {
376 		pipenode = (union node *)stalloc(sizeof (struct npipe));
377 		pipenode->type = NPIPE;
378 		pipenode->npipe.backgnd = 0;
379 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
380 		pipenode->npipe.cmdlist = lp;
381 		lp->n = n1;
382 		do {
383 			prev = lp;
384 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
385 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
386 			t = readtoken();
387 			tokpushback++;
388 			if (t == TNOT)
389 				lp->n = pipeline();
390 			else
391 				lp->n = command();
392 			prev->next = lp;
393 		} while (readtoken() == TPIPE);
394 		lp->next = NULL;
395 		n1 = pipenode;
396 	}
397 	tokpushback++;
398 	if (negate) {
399 		n2 = (union node *)stalloc(sizeof (struct nnot));
400 		n2->type = NNOT;
401 		n2->nnot.com = n1;
402 		return n2;
403 	} else
404 		return n1;
405 }
406 
407 
408 
409 static union node *
410 command(void)
411 {
412 	union node *n1, *n2;
413 	union node *ap, **app;
414 	union node *cp, **cpp;
415 	union node *redir, **rpp;
416 	int t;
417 	int is_subshell;
418 
419 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
420 	is_subshell = 0;
421 	redir = NULL;
422 	n1 = NULL;
423 	rpp = &redir;
424 
425 	/* Check for redirection which may precede command */
426 	while (readtoken() == TREDIR) {
427 		*rpp = n2 = redirnode;
428 		rpp = &n2->nfile.next;
429 		parsefname();
430 	}
431 	tokpushback++;
432 
433 	switch (readtoken()) {
434 	case TIF:
435 		n1 = (union node *)stalloc(sizeof (struct nif));
436 		n1->type = NIF;
437 		if ((n1->nif.test = list(0)) == NULL)
438 			synexpect(-1);
439 		consumetoken(TTHEN);
440 		n1->nif.ifpart = list(0);
441 		n2 = n1;
442 		while (readtoken() == TELIF) {
443 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
444 			n2 = n2->nif.elsepart;
445 			n2->type = NIF;
446 			if ((n2->nif.test = list(0)) == NULL)
447 				synexpect(-1);
448 			consumetoken(TTHEN);
449 			n2->nif.ifpart = list(0);
450 		}
451 		if (lasttoken == TELSE)
452 			n2->nif.elsepart = list(0);
453 		else {
454 			n2->nif.elsepart = NULL;
455 			tokpushback++;
456 		}
457 		consumetoken(TFI);
458 		checkkwd = CHKKWD | CHKALIAS;
459 		break;
460 	case TWHILE:
461 	case TUNTIL:
462 		t = lasttoken;
463 		if ((n1 = list(0)) == NULL)
464 			synexpect(-1);
465 		consumetoken(TDO);
466 		n1 = makebinary((t == TWHILE)? NWHILE : NUNTIL, n1, list(0));
467 		consumetoken(TDONE);
468 		checkkwd = CHKKWD | CHKALIAS;
469 		break;
470 	case TFOR:
471 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
472 			synerror("Bad for loop variable");
473 		n1 = (union node *)stalloc(sizeof (struct nfor));
474 		n1->type = NFOR;
475 		n1->nfor.var = wordtext;
476 		checkkwd = CHKNL;
477 		if (readtoken() == TWORD && !quoteflag &&
478 		    equal(wordtext, "in")) {
479 			app = &ap;
480 			while (readtoken() == TWORD) {
481 				n2 = makename();
482 				*app = n2;
483 				app = &n2->narg.next;
484 			}
485 			*app = NULL;
486 			n1->nfor.args = ap;
487 			if (lasttoken == TNL)
488 				tokpushback++;
489 			else if (lasttoken != TSEMI)
490 				synexpect(-1);
491 		} else {
492 			static char argvars[5] = {
493 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
494 			};
495 			n2 = (union node *)stalloc(sizeof (struct narg));
496 			n2->type = NARG;
497 			n2->narg.text = argvars;
498 			n2->narg.backquote = NULL;
499 			n2->narg.next = NULL;
500 			n1->nfor.args = n2;
501 			/*
502 			 * Newline or semicolon here is optional (but note
503 			 * that the original Bourne shell only allowed NL).
504 			 */
505 			if (lasttoken != TSEMI)
506 				tokpushback++;
507 		}
508 		checkkwd = CHKNL | CHKKWD | CHKALIAS;
509 		if ((t = readtoken()) == TDO)
510 			t = TDONE;
511 		else if (t == TBEGIN)
512 			t = TEND;
513 		else
514 			synexpect(-1);
515 		n1->nfor.body = list(0);
516 		consumetoken(t);
517 		checkkwd = CHKKWD | CHKALIAS;
518 		break;
519 	case TCASE:
520 		n1 = (union node *)stalloc(sizeof (struct ncase));
521 		n1->type = NCASE;
522 		consumetoken(TWORD);
523 		n1->ncase.expr = makename();
524 		checkkwd = CHKNL;
525 		if (readtoken() != TWORD || ! equal(wordtext, "in"))
526 			synerror("expecting \"in\"");
527 		cpp = &n1->ncase.cases;
528 		checkkwd = CHKNL | CHKKWD, readtoken();
529 		while (lasttoken != TESAC) {
530 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
531 			cp->type = NCLIST;
532 			app = &cp->nclist.pattern;
533 			if (lasttoken == TLP)
534 				readtoken();
535 			for (;;) {
536 				*app = ap = makename();
537 				checkkwd = CHKNL | CHKKWD;
538 				if (readtoken() != TPIPE)
539 					break;
540 				app = &ap->narg.next;
541 				readtoken();
542 			}
543 			ap->narg.next = NULL;
544 			if (lasttoken != TRP)
545 				synexpect(TRP);
546 			cp->nclist.body = list(0);
547 
548 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
549 			if ((t = readtoken()) != TESAC) {
550 				if (t == TENDCASE)
551 					;
552 				else if (t == TFALLTHRU)
553 					cp->type = NCLISTFALLTHRU;
554 				else
555 					synexpect(TENDCASE);
556 				checkkwd = CHKNL | CHKKWD, readtoken();
557 			}
558 			cpp = &cp->nclist.next;
559 		}
560 		*cpp = NULL;
561 		checkkwd = CHKKWD | CHKALIAS;
562 		break;
563 	case TLP:
564 		n1 = (union node *)stalloc(sizeof (struct nredir));
565 		n1->type = NSUBSHELL;
566 		n1->nredir.n = list(0);
567 		n1->nredir.redirect = NULL;
568 		consumetoken(TRP);
569 		checkkwd = CHKKWD | CHKALIAS;
570 		is_subshell = 1;
571 		break;
572 	case TBEGIN:
573 		n1 = list(0);
574 		consumetoken(TEND);
575 		checkkwd = CHKKWD | CHKALIAS;
576 		break;
577 	/* A simple command must have at least one redirection or word. */
578 	case TBACKGND:
579 	case TSEMI:
580 	case TAND:
581 	case TOR:
582 	case TPIPE:
583 	case TENDCASE:
584 	case TFALLTHRU:
585 	case TEOF:
586 	case TNL:
587 	case TRP:
588 		if (!redir)
589 			synexpect(-1);
590 	case TWORD:
591 		tokpushback++;
592 		n1 = simplecmd(rpp, redir);
593 		return n1;
594 	default:
595 		synexpect(-1);
596 	}
597 
598 	/* Now check for redirection which may follow command */
599 	while (readtoken() == TREDIR) {
600 		*rpp = n2 = redirnode;
601 		rpp = &n2->nfile.next;
602 		parsefname();
603 	}
604 	tokpushback++;
605 	*rpp = NULL;
606 	if (redir) {
607 		if (!is_subshell) {
608 			n2 = (union node *)stalloc(sizeof (struct nredir));
609 			n2->type = NREDIR;
610 			n2->nredir.n = n1;
611 			n1 = n2;
612 		}
613 		n1->nredir.redirect = redir;
614 	}
615 
616 	return n1;
617 }
618 
619 
620 static union node *
621 simplecmd(union node **rpp, union node *redir)
622 {
623 	union node *args, **app;
624 	union node **orig_rpp = rpp;
625 	union node *n = NULL;
626 	int special;
627 	int savecheckkwd;
628 
629 	/* If we don't have any redirections already, then we must reset */
630 	/* rpp to be the address of the local redir variable.  */
631 	if (redir == NULL)
632 		rpp = &redir;
633 
634 	args = NULL;
635 	app = &args;
636 	/*
637 	 * We save the incoming value, because we need this for shell
638 	 * functions.  There can not be a redirect or an argument between
639 	 * the function name and the open parenthesis.
640 	 */
641 	orig_rpp = rpp;
642 
643 	savecheckkwd = CHKALIAS;
644 
645 	for (;;) {
646 		checkkwd = savecheckkwd;
647 		if (readtoken() == TWORD) {
648 			n = makename();
649 			*app = n;
650 			app = &n->narg.next;
651 			if (savecheckkwd != 0 && !isassignment(wordtext))
652 				savecheckkwd = 0;
653 		} else if (lasttoken == TREDIR) {
654 			*rpp = n = redirnode;
655 			rpp = &n->nfile.next;
656 			parsefname();	/* read name of redirection file */
657 		} else if (lasttoken == TLP && app == &args->narg.next
658 					    && rpp == orig_rpp) {
659 			/* We have a function */
660 			consumetoken(TRP);
661 			funclinno = plinno;
662 			/*
663 			 * - Require plain text.
664 			 * - Functions with '/' cannot be called.
665 			 * - Reject name=().
666 			 * - Reject ksh extended glob patterns.
667 			 */
668 			if (!noexpand(n->narg.text) || quoteflag ||
669 			    strchr(n->narg.text, '/') ||
670 			    strchr("!%*+-=?@}~",
671 				n->narg.text[strlen(n->narg.text) - 1]))
672 				synerror("Bad function name");
673 			rmescapes(n->narg.text);
674 			if (find_builtin(n->narg.text, &special) >= 0 &&
675 			    special)
676 				synerror("Cannot override a special builtin with a function");
677 			n->type = NDEFUN;
678 			n->narg.next = command();
679 			funclinno = 0;
680 			return n;
681 		} else {
682 			tokpushback++;
683 			break;
684 		}
685 	}
686 	*app = NULL;
687 	*rpp = NULL;
688 	n = (union node *)stalloc(sizeof (struct ncmd));
689 	n->type = NCMD;
690 	n->ncmd.args = args;
691 	n->ncmd.redirect = redir;
692 	return n;
693 }
694 
695 static union node *
696 makename(void)
697 {
698 	union node *n;
699 
700 	n = (union node *)stalloc(sizeof (struct narg));
701 	n->type = NARG;
702 	n->narg.next = NULL;
703 	n->narg.text = wordtext;
704 	n->narg.backquote = backquotelist;
705 	return n;
706 }
707 
708 static union node *
709 makebinary(int type, union node *n1, union node *n2)
710 {
711 	union node *n;
712 
713 	n = (union node *)stalloc(sizeof (struct nbinary));
714 	n->type = type;
715 	n->nbinary.ch1 = n1;
716 	n->nbinary.ch2 = n2;
717 	return (n);
718 }
719 
720 void
721 forcealias(void)
722 {
723 	checkkwd |= CHKALIAS;
724 }
725 
726 void
727 fixredir(union node *n, const char *text, int err)
728 {
729 	TRACE(("Fix redir %s %d\n", text, err));
730 	if (!err)
731 		n->ndup.vname = NULL;
732 
733 	if (is_digit(text[0]) && text[1] == '\0')
734 		n->ndup.dupfd = digit_val(text[0]);
735 	else if (text[0] == '-' && text[1] == '\0')
736 		n->ndup.dupfd = -1;
737 	else {
738 
739 		if (err)
740 			synerror("Bad fd number");
741 		else
742 			n->ndup.vname = makename();
743 	}
744 }
745 
746 
747 static void
748 parsefname(void)
749 {
750 	union node *n = redirnode;
751 
752 	consumetoken(TWORD);
753 	if (n->type == NHERE) {
754 		struct heredoc *here = heredoc;
755 		struct heredoc *p;
756 
757 		if (quoteflag == 0)
758 			n->type = NXHERE;
759 		TRACE(("Here document %d\n", n->type));
760 		if (here->striptabs) {
761 			while (*wordtext == '\t')
762 				wordtext++;
763 		}
764 		if (! noexpand(wordtext))
765 			synerror("Illegal eof marker for << redirection");
766 		rmescapes(wordtext);
767 		here->eofmark = wordtext;
768 		here->next = NULL;
769 		if (heredoclist == NULL)
770 			heredoclist = here;
771 		else {
772 			for (p = heredoclist ; p->next ; p = p->next);
773 			p->next = here;
774 		}
775 	} else if (n->type == NTOFD || n->type == NFROMFD) {
776 		fixredir(n, wordtext, 0);
777 	} else {
778 		n->nfile.fname = makename();
779 	}
780 }
781 
782 
783 /*
784  * Input any here documents.
785  */
786 
787 static void
788 parseheredoc(void)
789 {
790 	struct heredoc *here;
791 	union node *n;
792 
793 	while (heredoclist) {
794 		here = heredoclist;
795 		heredoclist = here->next;
796 		if (needprompt) {
797 			setprompt(2);
798 			needprompt = 0;
799 		}
800 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
801 				here->eofmark, here->striptabs);
802 		n = makename();
803 		here->here->nhere.doc = n;
804 	}
805 }
806 
807 static int
808 peektoken(void)
809 {
810 	int t;
811 
812 	t = readtoken();
813 	tokpushback++;
814 	return (t);
815 }
816 
817 static int
818 readtoken(void)
819 {
820 	int t;
821 	struct alias *ap;
822 #ifdef DEBUG
823 	int alreadyseen = tokpushback;
824 #endif
825 
826 	top:
827 	t = xxreadtoken();
828 
829 	/*
830 	 * eat newlines
831 	 */
832 	if (checkkwd & CHKNL) {
833 		while (t == TNL) {
834 			parseheredoc();
835 			t = xxreadtoken();
836 		}
837 	}
838 
839 	/*
840 	 * check for keywords and aliases
841 	 */
842 	if (t == TWORD && !quoteflag)
843 	{
844 		const char * const *pp;
845 
846 		if (checkkwd & CHKKWD)
847 			for (pp = parsekwd; *pp; pp++) {
848 				if (**pp == *wordtext && equal(*pp, wordtext))
849 				{
850 					lasttoken = t = pp - parsekwd + KWDOFFSET;
851 					TRACE(("keyword %s recognized\n", tokname[t]));
852 					goto out;
853 				}
854 			}
855 		if (checkkwd & CHKALIAS &&
856 		    (ap = lookupalias(wordtext, 1)) != NULL) {
857 			pushstring(ap->val, strlen(ap->val), ap);
858 			goto top;
859 		}
860 	}
861 out:
862 	if (t != TNOT)
863 		checkkwd = 0;
864 
865 #ifdef DEBUG
866 	if (!alreadyseen)
867 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
868 	else
869 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
870 #endif
871 	return (t);
872 }
873 
874 
875 /*
876  * Read the next input token.
877  * If the token is a word, we set backquotelist to the list of cmds in
878  *	backquotes.  We set quoteflag to true if any part of the word was
879  *	quoted.
880  * If the token is TREDIR, then we set redirnode to a structure containing
881  *	the redirection.
882  * In all cases, the variable startlinno is set to the number of the line
883  *	on which the token starts.
884  *
885  * [Change comment:  here documents and internal procedures]
886  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
887  *  word parsing code into a separate routine.  In this case, readtoken
888  *  doesn't need to have any internal procedures, but parseword does.
889  *  We could also make parseoperator in essence the main routine, and
890  *  have parseword (readtoken1?) handle both words and redirection.]
891  */
892 
893 #define RETURN(token)	return lasttoken = token
894 
895 static int
896 xxreadtoken(void)
897 {
898 	int c;
899 
900 	if (tokpushback) {
901 		tokpushback = 0;
902 		return lasttoken;
903 	}
904 	if (needprompt) {
905 		setprompt(2);
906 		needprompt = 0;
907 	}
908 	startlinno = plinno;
909 	for (;;) {	/* until token or start of word found */
910 		c = pgetc_macro();
911 		switch (c) {
912 		case ' ': case '\t':
913 			continue;
914 		case '#':
915 			while ((c = pgetc()) != '\n' && c != PEOF);
916 			pungetc();
917 			continue;
918 		case '\\':
919 			if (pgetc() == '\n') {
920 				startlinno = ++plinno;
921 				if (doprompt)
922 					setprompt(2);
923 				else
924 					setprompt(0);
925 				continue;
926 			}
927 			pungetc();
928 			/* FALLTHROUGH */
929 		default:
930 			return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
931 		case '\n':
932 			plinno++;
933 			needprompt = doprompt;
934 			RETURN(TNL);
935 		case PEOF:
936 			RETURN(TEOF);
937 		case '&':
938 			if (pgetc_linecont() == '&')
939 				RETURN(TAND);
940 			pungetc();
941 			RETURN(TBACKGND);
942 		case '|':
943 			if (pgetc_linecont() == '|')
944 				RETURN(TOR);
945 			pungetc();
946 			RETURN(TPIPE);
947 		case ';':
948 			c = pgetc_linecont();
949 			if (c == ';')
950 				RETURN(TENDCASE);
951 			else if (c == '&')
952 				RETURN(TFALLTHRU);
953 			pungetc();
954 			RETURN(TSEMI);
955 		case '(':
956 			RETURN(TLP);
957 		case ')':
958 			RETURN(TRP);
959 		}
960 	}
961 #undef RETURN
962 }
963 
964 
965 #define MAXNEST_static 8
966 struct tokenstate
967 {
968 	const char *syntax; /* *SYNTAX */
969 	int parenlevel; /* levels of parentheses in arithmetic */
970 	enum tokenstate_category
971 	{
972 		TSTATE_TOP,
973 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
974 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
975 		TSTATE_ARITH
976 	} category;
977 };
978 
979 
980 /*
981  * Check to see whether we are at the end of the here document.  When this
982  * is called, c is set to the first character of the next input line.  If
983  * we are at the end of the here document, this routine sets the c to PEOF.
984  * The new value of c is returned.
985  */
986 
987 static int
988 checkend(int c, const char *eofmark, int striptabs)
989 {
990 	if (striptabs) {
991 		while (c == '\t')
992 			c = pgetc();
993 	}
994 	if (c == *eofmark) {
995 		int c2;
996 		const char *q;
997 
998 		for (q = eofmark + 1; c2 = pgetc(), *q != '\0' && c2 == *q; q++)
999 			;
1000 		if ((c2 == PEOF || c2 == '\n') && *q == '\0') {
1001 			c = PEOF;
1002 			if (c2 == '\n') {
1003 				plinno++;
1004 				needprompt = doprompt;
1005 			}
1006 		} else {
1007 			pungetc();
1008 			pushstring(eofmark + 1, q - (eofmark + 1), NULL);
1009 		}
1010 	} else if (c == '\n' && *eofmark == '\0') {
1011 		c = PEOF;
1012 		plinno++;
1013 		needprompt = doprompt;
1014 	}
1015 	return (c);
1016 }
1017 
1018 
1019 /*
1020  * Parse a redirection operator.  The variable "out" points to a string
1021  * specifying the fd to be redirected.  The variable "c" contains the
1022  * first character of the redirection operator.
1023  */
1024 
1025 static void
1026 parseredir(char *out, int c)
1027 {
1028 	char fd = *out;
1029 	union node *np;
1030 
1031 	np = (union node *)stalloc(sizeof (struct nfile));
1032 	if (c == '>') {
1033 		np->nfile.fd = 1;
1034 		c = pgetc_linecont();
1035 		if (c == '>')
1036 			np->type = NAPPEND;
1037 		else if (c == '&')
1038 			np->type = NTOFD;
1039 		else if (c == '|')
1040 			np->type = NCLOBBER;
1041 		else {
1042 			np->type = NTO;
1043 			pungetc();
1044 		}
1045 	} else {	/* c == '<' */
1046 		np->nfile.fd = 0;
1047 		c = pgetc_linecont();
1048 		if (c == '<') {
1049 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1050 				np = (union node *)stalloc(sizeof (struct nhere));
1051 				np->nfile.fd = 0;
1052 			}
1053 			np->type = NHERE;
1054 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1055 			heredoc->here = np;
1056 			if ((c = pgetc_linecont()) == '-') {
1057 				heredoc->striptabs = 1;
1058 			} else {
1059 				heredoc->striptabs = 0;
1060 				pungetc();
1061 			}
1062 		} else if (c == '&')
1063 			np->type = NFROMFD;
1064 		else if (c == '>')
1065 			np->type = NFROMTO;
1066 		else {
1067 			np->type = NFROM;
1068 			pungetc();
1069 		}
1070 	}
1071 	if (fd != '\0')
1072 		np->nfile.fd = digit_val(fd);
1073 	redirnode = np;
1074 }
1075 
1076 /*
1077  * Called to parse command substitutions.
1078  */
1079 
1080 static char *
1081 parsebackq(char *out, struct nodelist **pbqlist,
1082 		int oldstyle, int dblquote, int quoted)
1083 {
1084 	struct nodelist **nlpp;
1085 	union node *n;
1086 	char *volatile str;
1087 	struct jmploc jmploc;
1088 	struct jmploc *const savehandler = handler;
1089 	size_t savelen;
1090 	int saveprompt;
1091 	const int bq_startlinno = plinno;
1092 	char *volatile ostr = NULL;
1093 	struct parsefile *const savetopfile = getcurrentfile();
1094 	struct heredoc *const saveheredoclist = heredoclist;
1095 	struct heredoc *here;
1096 
1097 	str = NULL;
1098 	if (setjmp(jmploc.loc)) {
1099 		popfilesupto(savetopfile);
1100 		if (str)
1101 			ckfree(str);
1102 		if (ostr)
1103 			ckfree(ostr);
1104 		heredoclist = saveheredoclist;
1105 		handler = savehandler;
1106 		if (exception == EXERROR) {
1107 			startlinno = bq_startlinno;
1108 			synerror("Error in command substitution");
1109 		}
1110 		longjmp(handler->loc, 1);
1111 	}
1112 	INTOFF;
1113 	savelen = out - stackblock();
1114 	if (savelen > 0) {
1115 		str = ckmalloc(savelen);
1116 		memcpy(str, stackblock(), savelen);
1117 	}
1118 	handler = &jmploc;
1119 	heredoclist = NULL;
1120 	INTON;
1121         if (oldstyle) {
1122                 /* We must read until the closing backquote, giving special
1123                    treatment to some slashes, and then push the string and
1124                    reread it as input, interpreting it normally.  */
1125                 char *oout;
1126                 int c;
1127                 int olen;
1128 
1129 
1130                 STARTSTACKSTR(oout);
1131 		for (;;) {
1132 			if (needprompt) {
1133 				setprompt(2);
1134 				needprompt = 0;
1135 			}
1136 			CHECKSTRSPACE(2, oout);
1137 			c = pgetc_linecont();
1138 			if (c == '`')
1139 				break;
1140 			switch (c) {
1141 			case '\\':
1142 				c = pgetc();
1143                                 if (c != '\\' && c != '`' && c != '$'
1144                                     && (!dblquote || c != '"'))
1145                                         USTPUTC('\\', oout);
1146 				break;
1147 
1148 			case '\n':
1149 				plinno++;
1150 				needprompt = doprompt;
1151 				break;
1152 
1153 			case PEOF:
1154 			        startlinno = plinno;
1155 				synerror("EOF in backquote substitution");
1156  				break;
1157 
1158 			default:
1159 				break;
1160 			}
1161 			USTPUTC(c, oout);
1162                 }
1163                 USTPUTC('\0', oout);
1164                 olen = oout - stackblock();
1165 		INTOFF;
1166 		ostr = ckmalloc(olen);
1167 		memcpy(ostr, stackblock(), olen);
1168 		setinputstring(ostr, 1);
1169 		INTON;
1170         }
1171 	nlpp = pbqlist;
1172 	while (*nlpp)
1173 		nlpp = &(*nlpp)->next;
1174 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1175 	(*nlpp)->next = NULL;
1176 
1177 	if (oldstyle) {
1178 		saveprompt = doprompt;
1179 		doprompt = 0;
1180 	}
1181 
1182 	n = list(0);
1183 
1184 	if (oldstyle) {
1185 		if (peektoken() != TEOF)
1186 			synexpect(-1);
1187 		doprompt = saveprompt;
1188 	} else
1189 		consumetoken(TRP);
1190 
1191 	(*nlpp)->n = n;
1192         if (oldstyle) {
1193 		/*
1194 		 * Start reading from old file again, ignoring any pushed back
1195 		 * tokens left from the backquote parsing
1196 		 */
1197                 popfile();
1198 		tokpushback = 0;
1199 	}
1200 	STARTSTACKSTR(out);
1201 	CHECKSTRSPACE(savelen + 1, out);
1202 	INTOFF;
1203 	if (str) {
1204 		memcpy(out, str, savelen);
1205 		STADJUST(savelen, out);
1206 		ckfree(str);
1207 		str = NULL;
1208 	}
1209 	if (ostr) {
1210 		ckfree(ostr);
1211 		ostr = NULL;
1212 	}
1213 	here = saveheredoclist;
1214 	if (here != NULL) {
1215 		while (here->next != NULL)
1216 			here = here->next;
1217 		here->next = heredoclist;
1218 		heredoclist = saveheredoclist;
1219 	}
1220 	handler = savehandler;
1221 	INTON;
1222 	if (quoted)
1223 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1224 	else
1225 		USTPUTC(CTLBACKQ, out);
1226 	return out;
1227 }
1228 
1229 
1230 /*
1231  * Called to parse a backslash escape sequence inside $'...'.
1232  * The backslash has already been read.
1233  */
1234 static char *
1235 readcstyleesc(char *out)
1236 {
1237 	int c, vc, i, n;
1238 	unsigned int v;
1239 
1240 	c = pgetc();
1241 	switch (c) {
1242 	case '\0':
1243 		synerror("Unterminated quoted string");
1244 	case '\n':
1245 		plinno++;
1246 		if (doprompt)
1247 			setprompt(2);
1248 		else
1249 			setprompt(0);
1250 		return out;
1251 	case '\\':
1252 	case '\'':
1253 	case '"':
1254 		v = c;
1255 		break;
1256 	case 'a': v = '\a'; break;
1257 	case 'b': v = '\b'; break;
1258 	case 'e': v = '\033'; break;
1259 	case 'f': v = '\f'; break;
1260 	case 'n': v = '\n'; break;
1261 	case 'r': v = '\r'; break;
1262 	case 't': v = '\t'; break;
1263 	case 'v': v = '\v'; break;
1264 	case 'x':
1265 		  v = 0;
1266 		  for (;;) {
1267 			  c = pgetc();
1268 			  if (c >= '0' && c <= '9')
1269 				  v = (v << 4) + c - '0';
1270 			  else if (c >= 'A' && c <= 'F')
1271 				  v = (v << 4) + c - 'A' + 10;
1272 			  else if (c >= 'a' && c <= 'f')
1273 				  v = (v << 4) + c - 'a' + 10;
1274 			  else
1275 				  break;
1276 		  }
1277 		  pungetc();
1278 		  break;
1279 	case '0': case '1': case '2': case '3':
1280 	case '4': case '5': case '6': case '7':
1281 		  v = c - '0';
1282 		  c = pgetc();
1283 		  if (c >= '0' && c <= '7') {
1284 			  v <<= 3;
1285 			  v += c - '0';
1286 			  c = pgetc();
1287 			  if (c >= '0' && c <= '7') {
1288 				  v <<= 3;
1289 				  v += c - '0';
1290 			  } else
1291 				  pungetc();
1292 		  } else
1293 			  pungetc();
1294 		  break;
1295 	case 'c':
1296 		  c = pgetc();
1297 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1298 			  synerror("Bad escape sequence");
1299 		  if (c == '\\' && pgetc() != '\\')
1300 			  synerror("Bad escape sequence");
1301 		  if (c == '?')
1302 			  v = 127;
1303 		  else
1304 			  v = c & 0x1f;
1305 		  break;
1306 	case 'u':
1307 	case 'U':
1308 		  n = c == 'U' ? 8 : 4;
1309 		  v = 0;
1310 		  for (i = 0; i < n; i++) {
1311 			  c = pgetc();
1312 			  if (c >= '0' && c <= '9')
1313 				  v = (v << 4) + c - '0';
1314 			  else if (c >= 'A' && c <= 'F')
1315 				  v = (v << 4) + c - 'A' + 10;
1316 			  else if (c >= 'a' && c <= 'f')
1317 				  v = (v << 4) + c - 'a' + 10;
1318 			  else
1319 				  synerror("Bad escape sequence");
1320 		  }
1321 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1322 			  synerror("Bad escape sequence");
1323 		  /* We really need iconv here. */
1324 		  if (initial_localeisutf8 && v > 127) {
1325 			  CHECKSTRSPACE(4, out);
1326 			  /*
1327 			   * We cannot use wctomb() as the locale may have
1328 			   * changed.
1329 			   */
1330 			  if (v <= 0x7ff) {
1331 				  USTPUTC(0xc0 | v >> 6, out);
1332 				  USTPUTC(0x80 | (v & 0x3f), out);
1333 				  return out;
1334 			  } else if (v <= 0xffff) {
1335 				  USTPUTC(0xe0 | v >> 12, out);
1336 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1337 				  USTPUTC(0x80 | (v & 0x3f), out);
1338 				  return out;
1339 			  } else if (v <= 0x10ffff) {
1340 				  USTPUTC(0xf0 | v >> 18, out);
1341 				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1342 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1343 				  USTPUTC(0x80 | (v & 0x3f), out);
1344 				  return out;
1345 			  }
1346 		  }
1347 		  if (v > 127)
1348 			  v = '?';
1349 		  break;
1350 	default:
1351 		  synerror("Bad escape sequence");
1352 	}
1353 	vc = (char)v;
1354 	/*
1355 	 * We can't handle NUL bytes.
1356 	 * POSIX says we should skip till the closing quote.
1357 	 */
1358 	if (vc == '\0') {
1359 		while ((c = pgetc()) != '\'') {
1360 			if (c == '\\')
1361 				c = pgetc();
1362 			if (c == PEOF)
1363 				synerror("Unterminated quoted string");
1364 			if (c == '\n') {
1365 				plinno++;
1366 				if (doprompt)
1367 					setprompt(2);
1368 				else
1369 					setprompt(0);
1370 			}
1371 		}
1372 		pungetc();
1373 		return out;
1374 	}
1375 	if (SQSYNTAX[vc] == CCTL)
1376 		USTPUTC(CTLESC, out);
1377 	USTPUTC(vc, out);
1378 	return out;
1379 }
1380 
1381 
1382 /*
1383  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1384  * is not NULL, read a here document.  In the latter case, eofmark is the
1385  * word which marks the end of the document and striptabs is true if
1386  * leading tabs should be stripped from the document.  The argument firstc
1387  * is the first character of the input token or document.
1388  *
1389  * Because C does not have internal subroutines, I have simulated them
1390  * using goto's to implement the subroutine linkage.  The following macros
1391  * will run code that appears at the end of readtoken1.
1392  */
1393 
1394 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1395 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1396 
1397 static int
1398 readtoken1(int firstc, char const *initialsyntax, const char *eofmark,
1399     int striptabs)
1400 {
1401 	int c = firstc;
1402 	char *out;
1403 	int len;
1404 	struct nodelist *bqlist;
1405 	int quotef;
1406 	int newvarnest;
1407 	int level;
1408 	int synentry;
1409 	struct tokenstate state_static[MAXNEST_static];
1410 	int maxnest = MAXNEST_static;
1411 	struct tokenstate *state = state_static;
1412 	int sqiscstyle = 0;
1413 
1414 	startlinno = plinno;
1415 	quotef = 0;
1416 	bqlist = NULL;
1417 	newvarnest = 0;
1418 	level = 0;
1419 	state[level].syntax = initialsyntax;
1420 	state[level].parenlevel = 0;
1421 	state[level].category = TSTATE_TOP;
1422 
1423 	STARTSTACKSTR(out);
1424 	loop: {	/* for each line, until end of word */
1425 		if (eofmark && eofmark != NOEOFMARK)
1426 			/* set c to PEOF if at end of here document */
1427 			c = checkend(c, eofmark, striptabs);
1428 		for (;;) {	/* until end of line or end of word */
1429 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1430 
1431 			synentry = state[level].syntax[c];
1432 
1433 			switch(synentry) {
1434 			case CNL:	/* '\n' */
1435 				if (level == 0)
1436 					goto endword;	/* exit outer loop */
1437 				/* FALLTHROUGH */
1438 			case CQNL:
1439 				USTPUTC(c, out);
1440 				plinno++;
1441 				if (doprompt)
1442 					setprompt(2);
1443 				else
1444 					setprompt(0);
1445 				c = pgetc();
1446 				goto loop;		/* continue outer loop */
1447 			case CSBACK:
1448 				if (sqiscstyle) {
1449 					out = readcstyleesc(out);
1450 					break;
1451 				}
1452 				/* FALLTHROUGH */
1453 			case CWORD:
1454 				USTPUTC(c, out);
1455 				break;
1456 			case CCTL:
1457 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1458 					USTPUTC(CTLESC, out);
1459 				USTPUTC(c, out);
1460 				break;
1461 			case CBACK:	/* backslash */
1462 				c = pgetc();
1463 				if (c == PEOF) {
1464 					USTPUTC('\\', out);
1465 					pungetc();
1466 				} else if (c == '\n') {
1467 					plinno++;
1468 					if (doprompt)
1469 						setprompt(2);
1470 					else
1471 						setprompt(0);
1472 				} else {
1473 					if (state[level].syntax == DQSYNTAX &&
1474 					    c != '\\' && c != '`' && c != '$' &&
1475 					    (c != '"' || (eofmark != NULL &&
1476 						newvarnest == 0)) &&
1477 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1478 						USTPUTC('\\', out);
1479 					if ((eofmark == NULL ||
1480 					    newvarnest > 0) &&
1481 					    state[level].syntax == BASESYNTAX)
1482 						USTPUTC(CTLQUOTEMARK, out);
1483 					if (SQSYNTAX[c] == CCTL)
1484 						USTPUTC(CTLESC, out);
1485 					USTPUTC(c, out);
1486 					if ((eofmark == NULL ||
1487 					    newvarnest > 0) &&
1488 					    state[level].syntax == BASESYNTAX &&
1489 					    state[level].category == TSTATE_VAR_OLD)
1490 						USTPUTC(CTLQUOTEEND, out);
1491 					quotef++;
1492 				}
1493 				break;
1494 			case CSQUOTE:
1495 				USTPUTC(CTLQUOTEMARK, out);
1496 				state[level].syntax = SQSYNTAX;
1497 				sqiscstyle = 0;
1498 				break;
1499 			case CDQUOTE:
1500 				USTPUTC(CTLQUOTEMARK, out);
1501 				state[level].syntax = DQSYNTAX;
1502 				break;
1503 			case CENDQUOTE:
1504 				if (eofmark != NULL && newvarnest == 0)
1505 					USTPUTC(c, out);
1506 				else {
1507 					if (state[level].category == TSTATE_VAR_OLD)
1508 						USTPUTC(CTLQUOTEEND, out);
1509 					state[level].syntax = BASESYNTAX;
1510 					quotef++;
1511 				}
1512 				break;
1513 			case CVAR:	/* '$' */
1514 				PARSESUB();		/* parse substitution */
1515 				break;
1516 			case CENDVAR:	/* '}' */
1517 				if (level > 0 &&
1518 				    ((state[level].category == TSTATE_VAR_OLD &&
1519 				      state[level].syntax ==
1520 				      state[level - 1].syntax) ||
1521 				    (state[level].category == TSTATE_VAR_NEW &&
1522 				     state[level].syntax == BASESYNTAX))) {
1523 					if (state[level].category == TSTATE_VAR_NEW)
1524 						newvarnest--;
1525 					level--;
1526 					USTPUTC(CTLENDVAR, out);
1527 				} else {
1528 					USTPUTC(c, out);
1529 				}
1530 				break;
1531 			case CLP:	/* '(' in arithmetic */
1532 				state[level].parenlevel++;
1533 				USTPUTC(c, out);
1534 				break;
1535 			case CRP:	/* ')' in arithmetic */
1536 				if (state[level].parenlevel > 0) {
1537 					USTPUTC(c, out);
1538 					--state[level].parenlevel;
1539 				} else {
1540 					if (pgetc_linecont() == ')') {
1541 						if (level > 0 &&
1542 						    state[level].category == TSTATE_ARITH) {
1543 							level--;
1544 							USTPUTC(CTLENDARI, out);
1545 						} else
1546 							USTPUTC(')', out);
1547 					} else {
1548 						/*
1549 						 * unbalanced parens
1550 						 *  (don't 2nd guess - no error)
1551 						 */
1552 						pungetc();
1553 						USTPUTC(')', out);
1554 					}
1555 				}
1556 				break;
1557 			case CBQUOTE:	/* '`' */
1558 				out = parsebackq(out, &bqlist, 1,
1559 				    state[level].syntax == DQSYNTAX &&
1560 				    (eofmark == NULL || newvarnest > 0),
1561 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1562 				break;
1563 			case CEOF:
1564 				goto endword;		/* exit outer loop */
1565 			case CIGN:
1566 				break;
1567 			default:
1568 				if (level == 0)
1569 					goto endword;	/* exit outer loop */
1570 				USTPUTC(c, out);
1571 			}
1572 			c = pgetc_macro();
1573 		}
1574 	}
1575 endword:
1576 	if (state[level].syntax == ARISYNTAX)
1577 		synerror("Missing '))'");
1578 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1579 		synerror("Unterminated quoted string");
1580 	if (state[level].category == TSTATE_VAR_OLD ||
1581 	    state[level].category == TSTATE_VAR_NEW) {
1582 		startlinno = plinno;
1583 		synerror("Missing '}'");
1584 	}
1585 	if (state != state_static)
1586 		parser_temp_free_upto(state);
1587 	USTPUTC('\0', out);
1588 	len = out - stackblock();
1589 	out = stackblock();
1590 	if (eofmark == NULL) {
1591 		if ((c == '>' || c == '<')
1592 		 && quotef == 0
1593 		 && len <= 2
1594 		 && (*out == '\0' || is_digit(*out))) {
1595 			parseredir(out, c);
1596 			return lasttoken = TREDIR;
1597 		} else {
1598 			pungetc();
1599 		}
1600 	}
1601 	quoteflag = quotef;
1602 	backquotelist = bqlist;
1603 	grabstackblock(len);
1604 	wordtext = out;
1605 	return lasttoken = TWORD;
1606 /* end of readtoken routine */
1607 
1608 
1609 /*
1610  * Parse a substitution.  At this point, we have read the dollar sign
1611  * and nothing else.
1612  */
1613 
1614 parsesub: {
1615 	int subtype;
1616 	int typeloc;
1617 	int flags;
1618 	char *p;
1619 	static const char types[] = "}-+?=";
1620 	int linno;
1621 	int length;
1622 	int c1;
1623 
1624 	c = pgetc_linecont();
1625 	if (c == '(') {	/* $(command) or $((arith)) */
1626 		if (pgetc_linecont() == '(') {
1627 			PARSEARITH();
1628 		} else {
1629 			pungetc();
1630 			out = parsebackq(out, &bqlist, 0,
1631 			    state[level].syntax == DQSYNTAX &&
1632 			    (eofmark == NULL || newvarnest > 0),
1633 			    state[level].syntax == DQSYNTAX ||
1634 			    state[level].syntax == ARISYNTAX);
1635 		}
1636 	} else if (c == '{' || is_name(c) || is_special(c)) {
1637 		USTPUTC(CTLVAR, out);
1638 		typeloc = out - stackblock();
1639 		USTPUTC(VSNORMAL, out);
1640 		subtype = VSNORMAL;
1641 		flags = 0;
1642 		if (c == '{') {
1643 			c = pgetc_linecont();
1644 			subtype = 0;
1645 		}
1646 varname:
1647 		if (!is_eof(c) && is_name(c)) {
1648 			length = 0;
1649 			do {
1650 				STPUTC(c, out);
1651 				c = pgetc_linecont();
1652 				length++;
1653 			} while (!is_eof(c) && is_in_name(c));
1654 			if (length == 6 &&
1655 			    strncmp(out - length, "LINENO", length) == 0) {
1656 				/* Replace the variable name with the
1657 				 * current line number. */
1658 				STADJUST(-6, out);
1659 				CHECKSTRSPACE(11, out);
1660 				linno = plinno;
1661 				if (funclinno != 0)
1662 					linno -= funclinno - 1;
1663 				length = snprintf(out, 11, "%d", linno);
1664 				if (length > 10)
1665 					length = 10;
1666 				out += length;
1667 				flags |= VSLINENO;
1668 			}
1669 		} else if (is_digit(c)) {
1670 			if (subtype != VSNORMAL) {
1671 				do {
1672 					STPUTC(c, out);
1673 					c = pgetc_linecont();
1674 				} while (is_digit(c));
1675 			} else {
1676 				USTPUTC(c, out);
1677 				c = pgetc_linecont();
1678 			}
1679 		} else if (is_special(c)) {
1680 			c1 = c;
1681 			c = pgetc_linecont();
1682 			if (subtype == 0 && c1 == '#') {
1683 				subtype = VSLENGTH;
1684 				if (strchr(types, c) == NULL && c != ':' &&
1685 				    c != '#' && c != '%')
1686 					goto varname;
1687 				c1 = c;
1688 				c = pgetc_linecont();
1689 				if (c1 != '}' && c == '}') {
1690 					pungetc();
1691 					c = c1;
1692 					goto varname;
1693 				}
1694 				pungetc();
1695 				c = c1;
1696 				c1 = '#';
1697 				subtype = 0;
1698 			}
1699 			USTPUTC(c1, out);
1700 		} else {
1701 			subtype = VSERROR;
1702 			if (c == '}')
1703 				pungetc();
1704 			else if (c == '\n' || c == PEOF)
1705 				synerror("Unexpected end of line in substitution");
1706 			else if (BASESYNTAX[c] != CCTL)
1707 				USTPUTC(c, out);
1708 		}
1709 		if (subtype == 0) {
1710 			switch (c) {
1711 			case ':':
1712 				flags |= VSNUL;
1713 				c = pgetc_linecont();
1714 				/*FALLTHROUGH*/
1715 			default:
1716 				p = strchr(types, c);
1717 				if (p == NULL) {
1718 					if (c == '\n' || c == PEOF)
1719 						synerror("Unexpected end of line in substitution");
1720 					if (flags == VSNUL)
1721 						STPUTC(':', out);
1722 					if (BASESYNTAX[c] != CCTL)
1723 						STPUTC(c, out);
1724 					subtype = VSERROR;
1725 				} else
1726 					subtype = p - types + VSNORMAL;
1727 				break;
1728 			case '%':
1729 			case '#':
1730 				{
1731 					int cc = c;
1732 					subtype = c == '#' ? VSTRIMLEFT :
1733 							     VSTRIMRIGHT;
1734 					c = pgetc_linecont();
1735 					if (c == cc)
1736 						subtype++;
1737 					else
1738 						pungetc();
1739 					break;
1740 				}
1741 			}
1742 		} else if (subtype != VSERROR) {
1743 			if (subtype == VSLENGTH && c != '}')
1744 				subtype = VSERROR;
1745 			pungetc();
1746 		}
1747 		STPUTC('=', out);
1748 		if (state[level].syntax == DQSYNTAX ||
1749 		    state[level].syntax == ARISYNTAX)
1750 			flags |= VSQUOTE;
1751 		*(stackblock() + typeloc) = subtype | flags;
1752 		if (subtype != VSNORMAL) {
1753 			if (level + 1 >= maxnest) {
1754 				maxnest *= 2;
1755 				if (state == state_static) {
1756 					state = parser_temp_alloc(
1757 					    maxnest * sizeof(*state));
1758 					memcpy(state, state_static,
1759 					    MAXNEST_static * sizeof(*state));
1760 				} else
1761 					state = parser_temp_realloc(state,
1762 					    maxnest * sizeof(*state));
1763 			}
1764 			level++;
1765 			state[level].parenlevel = 0;
1766 			if (subtype == VSMINUS || subtype == VSPLUS ||
1767 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1768 				/*
1769 				 * For operators that were in the Bourne shell,
1770 				 * inherit the double-quote state.
1771 				 */
1772 				state[level].syntax = state[level - 1].syntax;
1773 				state[level].category = TSTATE_VAR_OLD;
1774 			} else {
1775 				/*
1776 				 * The other operators take a pattern,
1777 				 * so go to BASESYNTAX.
1778 				 * Also, ' and " are now special, even
1779 				 * in here documents.
1780 				 */
1781 				state[level].syntax = BASESYNTAX;
1782 				state[level].category = TSTATE_VAR_NEW;
1783 				newvarnest++;
1784 			}
1785 		}
1786 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1787 		/* $'cstylequotes' */
1788 		USTPUTC(CTLQUOTEMARK, out);
1789 		state[level].syntax = SQSYNTAX;
1790 		sqiscstyle = 1;
1791 	} else {
1792 		USTPUTC('$', out);
1793 		pungetc();
1794 	}
1795 	goto parsesub_return;
1796 }
1797 
1798 
1799 /*
1800  * Parse an arithmetic expansion (indicate start of one and set state)
1801  */
1802 parsearith: {
1803 
1804 	if (level + 1 >= maxnest) {
1805 		maxnest *= 2;
1806 		if (state == state_static) {
1807 			state = parser_temp_alloc(
1808 			    maxnest * sizeof(*state));
1809 			memcpy(state, state_static,
1810 			    MAXNEST_static * sizeof(*state));
1811 		} else
1812 			state = parser_temp_realloc(state,
1813 			    maxnest * sizeof(*state));
1814 	}
1815 	level++;
1816 	state[level].syntax = ARISYNTAX;
1817 	state[level].parenlevel = 0;
1818 	state[level].category = TSTATE_ARITH;
1819 	USTPUTC(CTLARI, out);
1820 	if (state[level - 1].syntax == DQSYNTAX)
1821 		USTPUTC('"',out);
1822 	else
1823 		USTPUTC(' ',out);
1824 	goto parsearith_return;
1825 }
1826 
1827 } /* end of readtoken */
1828 
1829 
1830 /*
1831  * Returns true if the text contains nothing to expand (no dollar signs
1832  * or backquotes).
1833  */
1834 
1835 static int
1836 noexpand(char *text)
1837 {
1838 	char *p;
1839 	char c;
1840 
1841 	p = text;
1842 	while ((c = *p++) != '\0') {
1843 		if ( c == CTLQUOTEMARK)
1844 			continue;
1845 		if (c == CTLESC)
1846 			p++;
1847 		else if (BASESYNTAX[(int)c] == CCTL)
1848 			return 0;
1849 	}
1850 	return 1;
1851 }
1852 
1853 
1854 /*
1855  * Return true if the argument is a legal variable name (a letter or
1856  * underscore followed by zero or more letters, underscores, and digits).
1857  */
1858 
1859 int
1860 goodname(const char *name)
1861 {
1862 	const char *p;
1863 
1864 	p = name;
1865 	if (! is_name(*p))
1866 		return 0;
1867 	while (*++p) {
1868 		if (! is_in_name(*p))
1869 			return 0;
1870 	}
1871 	return 1;
1872 }
1873 
1874 
1875 int
1876 isassignment(const char *p)
1877 {
1878 	if (!is_name(*p))
1879 		return 0;
1880 	p++;
1881 	for (;;) {
1882 		if (*p == '=')
1883 			return 1;
1884 		else if (!is_in_name(*p))
1885 			return 0;
1886 		p++;
1887 	}
1888 }
1889 
1890 
1891 static void
1892 consumetoken(int token)
1893 {
1894 	if (readtoken() != token)
1895 		synexpect(token);
1896 }
1897 
1898 
1899 /*
1900  * Called when an unexpected token is read during the parse.  The argument
1901  * is the token that is expected, or -1 if more than one type of token can
1902  * occur at this point.
1903  */
1904 
1905 static void
1906 synexpect(int token)
1907 {
1908 	char msg[64];
1909 
1910 	if (token >= 0) {
1911 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1912 			tokname[lasttoken], tokname[token]);
1913 	} else {
1914 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1915 	}
1916 	synerror(msg);
1917 }
1918 
1919 
1920 static void
1921 synerror(const char *msg)
1922 {
1923 	if (commandname)
1924 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1925 	else if (arg0)
1926 		outfmt(out2, "%s: ", arg0);
1927 	outfmt(out2, "Syntax error: %s\n", msg);
1928 	error((char *)NULL);
1929 }
1930 
1931 static void
1932 setprompt(int which)
1933 {
1934 	whichprompt = which;
1935 	if (which == 0)
1936 		return;
1937 
1938 #ifndef NO_HISTORY
1939 	if (!el)
1940 #endif
1941 	{
1942 		out2str(getprompt(NULL));
1943 		flushout(out2);
1944 	}
1945 }
1946 
1947 static int
1948 pgetc_linecont(void)
1949 {
1950 	int c;
1951 
1952 	while ((c = pgetc_macro()) == '\\') {
1953 		c = pgetc();
1954 		if (c == '\n') {
1955 			plinno++;
1956 			if (doprompt)
1957 				setprompt(2);
1958 			else
1959 				setprompt(0);
1960 		} else {
1961 			pungetc();
1962 			/* Allow the backslash to be pushed back. */
1963 			pushstring("\\", 1, NULL);
1964 			return (pgetc());
1965 		}
1966 	}
1967 	return (c);
1968 }
1969 
1970 
1971 static struct passwd *
1972 getpwlogin(void)
1973 {
1974 	const char *login;
1975 
1976 	login = getlogin();
1977 	if (login == NULL)
1978 		return (NULL);
1979 
1980 	return (getpwnam(login));
1981 }
1982 
1983 
1984 static void
1985 getusername(char *name, size_t namelen)
1986 {
1987 	static char cached_name[MAXLOGNAME];
1988 	struct passwd *pw;
1989 	uid_t euid;
1990 
1991 	if (cached_name[0] == '\0') {
1992 		euid = geteuid();
1993 
1994 		/*
1995 		 * Handle the case when there is more than one
1996 		 * login with the same UID, or when the login
1997 		 * returned by getlogin(2) does no longer match
1998 		 * the current UID.
1999 		 */
2000 		pw = getpwlogin();
2001 		if (pw == NULL || pw->pw_uid != euid)
2002 			pw = getpwuid(euid);
2003 
2004 		if (pw != NULL) {
2005 			strlcpy(cached_name, pw->pw_name,
2006 			    sizeof(cached_name));
2007 		} else {
2008 			snprintf(cached_name, sizeof(cached_name),
2009 			    "%u", euid);
2010 		}
2011 	}
2012 
2013 	strlcpy(name, cached_name, namelen);
2014 }
2015 
2016 
2017 /*
2018  * called by editline -- any expansions to the prompt
2019  *    should be added here.
2020  */
2021 char *
2022 getprompt(void *unused __unused)
2023 {
2024 	static char ps[PROMPTLEN];
2025 	const char *fmt;
2026 	const char *home;
2027 	const char *pwd;
2028 	size_t homelen;
2029 	int i, trim;
2030 	static char internal_error[] = "??";
2031 
2032 	/*
2033 	 * Select prompt format.
2034 	 */
2035 	switch (whichprompt) {
2036 	case 0:
2037 		fmt = "";
2038 		break;
2039 	case 1:
2040 		fmt = ps1val();
2041 		break;
2042 	case 2:
2043 		fmt = ps2val();
2044 		break;
2045 	default:
2046 		return internal_error;
2047 	}
2048 
2049 	/*
2050 	 * Format prompt string.
2051 	 */
2052 	for (i = 0; (i < PROMPTLEN - 1) && (*fmt != '\0'); i++, fmt++) {
2053 		if (*fmt != '\\') {
2054 			ps[i] = *fmt;
2055 			continue;
2056 		}
2057 
2058 		switch (*++fmt) {
2059 
2060 		/*
2061 		 * Non-printing sequence begin and end.
2062 		 */
2063 		case '[':
2064 		case ']':
2065 			ps[i] = '\001';
2066 			break;
2067 
2068 		/*
2069 		 * Literal \ and some ASCII characters:
2070 		 * \a	BEL
2071 		 * \e	ESC
2072 		 * \r	CR
2073 		 */
2074 		case '\\':
2075 		case 'a':
2076 		case 'e':
2077 		case 'r':
2078 			if (*fmt == 'a')
2079 				ps[i] = '\007';
2080 			else if (*fmt == 'e')
2081 				ps[i] = '\033';
2082 			else if (*fmt == 'r')
2083 				ps[i] = '\r';
2084 			else
2085 				ps[i] = '\\';
2086 			break;
2087 
2088 		/*
2089 		 * CRLF sequence
2090 		 */
2091 		case 'n':
2092 			if (i < PROMPTLEN - 3) {
2093 				ps[i++] = '\r';
2094 				ps[i] = '\n';
2095 			}
2096 			break;
2097 
2098 		/*
2099 		 * Print the current time as per provided strftime format.
2100 		 */
2101 		case 'D': {
2102 			char tfmt[128] = "%X"; /* \D{} means %X. */
2103 			struct tm *now;
2104 
2105 			if (fmt[1] != '{') {
2106 				/*
2107 				 * "\D" but not "\D{", so treat the '\'
2108 				 * literally and rewind fmt to treat 'D'
2109 				 * literally next iteration.
2110 				 */
2111 				ps[i] = '\\';
2112 				fmt--;
2113 				break;
2114 			}
2115 			fmt += 2; /* Consume "D{". */
2116 			if (fmt[0] != '}') {
2117 				char *end;
2118 
2119 				end = memccpy(tfmt, fmt, '}', sizeof(tfmt));
2120 				if (end == NULL) {
2121 					/*
2122 					 * Format too long or no '}', so
2123 					 * ignore "\D{" altogether.
2124 					 * The loop will do i++, but nothing
2125 					 * was written to ps, so do i-- here.
2126 					 * Rewind fmt for similar reason.
2127 					 */
2128 					i--;
2129 					fmt--;
2130 					break;
2131 				}
2132 				*--end = '\0'; /* Ignore the copy of '}'. */
2133 				fmt += end - tfmt;
2134 			}
2135 			now = localtime(&(time_t){time(NULL)});
2136 			i += strftime(&ps[i], PROMPTLEN - i - 1, tfmt, now);
2137 			i--; /* The loop will do i++. */
2138 			break;
2139 		}
2140 
2141 		/*
2142 		 * Hostname.
2143 		 *
2144 		 * \h specifies just the local hostname,
2145 		 * \H specifies fully-qualified hostname.
2146 		 */
2147 		case 'h':
2148 		case 'H':
2149 			ps[i] = '\0';
2150 			gethostname(&ps[i], PROMPTLEN - i - 1);
2151 			ps[PROMPTLEN - 1] = '\0';
2152 			/* Skip to end of hostname. */
2153 			trim = (*fmt == 'h') ? '.' : '\0';
2154 			while ((ps[i] != '\0') && (ps[i] != trim))
2155 				i++;
2156 			--i;
2157 			break;
2158 
2159 		/*
2160 		 * User name.
2161 		 */
2162 		case 'u':
2163 			ps[i] = '\0';
2164 			getusername(&ps[i], PROMPTLEN - i);
2165 			/* Skip to end of username. */
2166 			while (ps[i + 1] != '\0')
2167 				i++;
2168 			break;
2169 
2170 		/*
2171 		 * Working directory.
2172 		 *
2173 		 * \W specifies just the final component,
2174 		 * \w specifies the entire path.
2175 		 */
2176 		case 'W':
2177 		case 'w':
2178 			pwd = lookupvar("PWD");
2179 			if (pwd == NULL || *pwd == '\0')
2180 				pwd = "?";
2181 			if (*fmt == 'W' &&
2182 			    *pwd == '/' && pwd[1] != '\0')
2183 				strlcpy(&ps[i], strrchr(pwd, '/') + 1,
2184 				    PROMPTLEN - i);
2185 			else {
2186 				home = lookupvar("HOME");
2187 				if (home != NULL)
2188 					homelen = strlen(home);
2189 				if (home != NULL &&
2190 				    strcmp(home, "/") != 0 &&
2191 				    strncmp(pwd, home, homelen) == 0 &&
2192 				    (pwd[homelen] == '/' ||
2193 				    pwd[homelen] == '\0')) {
2194 					strlcpy(&ps[i], "~",
2195 					    PROMPTLEN - i);
2196 					strlcpy(&ps[i + 1],
2197 					    pwd + homelen,
2198 					    PROMPTLEN - i - 1);
2199 				} else {
2200 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
2201 				}
2202 			}
2203 			/* Skip to end of path. */
2204 			while (ps[i + 1] != '\0')
2205 				i++;
2206 			break;
2207 
2208 		/*
2209 		 * Superuser status.
2210 		 *
2211 		 * '$' for normal users, '#' for root.
2212 		 */
2213 		case '$':
2214 			ps[i] = (geteuid() != 0) ? '$' : '#';
2215 			break;
2216 
2217 		/*
2218 		 * Emit unrecognized formats verbatim.
2219 		 */
2220 		default:
2221 			ps[i] = '\\';
2222 			if (i < PROMPTLEN - 2)
2223 				ps[++i] = *fmt;
2224 			break;
2225 		}
2226 
2227 	}
2228 	ps[i] = '\0';
2229 	return (ps);
2230 }
2231 
2232 
2233 const char *
2234 expandstr(const char *ps)
2235 {
2236 	union node n;
2237 	struct jmploc jmploc;
2238 	struct jmploc *const savehandler = handler;
2239 	const int saveprompt = doprompt;
2240 	struct parsefile *const savetopfile = getcurrentfile();
2241 	struct parser_temp *const saveparser_temp = parser_temp;
2242 	const char *result = NULL;
2243 
2244 	if (!setjmp(jmploc.loc)) {
2245 		handler = &jmploc;
2246 		parser_temp = NULL;
2247 		setinputstring(ps, 1);
2248 		doprompt = 0;
2249 		readtoken1(pgetc(), DQSYNTAX, NOEOFMARK, 0);
2250 		if (backquotelist != NULL)
2251 			error("Command substitution not allowed here");
2252 
2253 		n.narg.type = NARG;
2254 		n.narg.next = NULL;
2255 		n.narg.text = wordtext;
2256 		n.narg.backquote = backquotelist;
2257 
2258 		expandarg(&n, NULL, 0);
2259 		result = stackblock();
2260 		INTOFF;
2261 	}
2262 	handler = savehandler;
2263 	doprompt = saveprompt;
2264 	popfilesupto(savetopfile);
2265 	if (parser_temp != saveparser_temp) {
2266 		parser_temp_free_all();
2267 		parser_temp = saveparser_temp;
2268 	}
2269 	if (result != NULL) {
2270 		INTON;
2271 	} else if (exception == EXINT)
2272 		raise(SIGINT);
2273 	return result;
2274 }
2275