xref: /freebsd/bin/sh/parser.c (revision 884a2a699669ec61e2366e3e358342dbc94be24a)
1 /*-
2  * Copyright (c) 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Kenneth Almquist.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #ifndef lint
34 #if 0
35 static char sccsid[] = "@(#)parser.c	8.7 (Berkeley) 5/16/95";
36 #endif
37 #endif /* not lint */
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include <stdlib.h>
42 #include <unistd.h>
43 #include <stdio.h>
44 
45 #include "shell.h"
46 #include "parser.h"
47 #include "nodes.h"
48 #include "expand.h"	/* defines rmescapes() */
49 #include "syntax.h"
50 #include "options.h"
51 #include "input.h"
52 #include "output.h"
53 #include "var.h"
54 #include "error.h"
55 #include "memalloc.h"
56 #include "mystring.h"
57 #include "alias.h"
58 #include "show.h"
59 #include "eval.h"
60 #include "exec.h"	/* to check for special builtins */
61 #ifndef NO_HISTORY
62 #include "myhistedit.h"
63 #endif
64 
65 /*
66  * Shell command parser.
67  */
68 
69 #define	EOFMARKLEN	79
70 #define	PROMPTLEN	128
71 
72 /* values of checkkwd variable */
73 #define CHKALIAS	0x1
74 #define CHKKWD		0x2
75 #define CHKNL		0x4
76 
77 /* values returned by readtoken */
78 #include "token.h"
79 
80 
81 
82 struct heredoc {
83 	struct heredoc *next;	/* next here document in list */
84 	union node *here;		/* redirection node */
85 	char *eofmark;		/* string indicating end of input */
86 	int striptabs;		/* if set, strip leading tabs */
87 };
88 
89 struct parser_temp {
90 	struct parser_temp *next;
91 	void *data;
92 };
93 
94 
95 static struct heredoc *heredoclist;	/* list of here documents to read */
96 static int doprompt;		/* if set, prompt the user */
97 static int needprompt;		/* true if interactive and at start of line */
98 static int lasttoken;		/* last token read */
99 MKINIT int tokpushback;		/* last token pushed back */
100 static char *wordtext;		/* text of last word returned by readtoken */
101 MKINIT int checkkwd;            /* 1 == check for kwds, 2 == also eat newlines */
102 static struct nodelist *backquotelist;
103 static union node *redirnode;
104 static struct heredoc *heredoc;
105 static int quoteflag;		/* set if (part of) last token was quoted */
106 static int startlinno;		/* line # where last token started */
107 static int funclinno;		/* line # where the current function started */
108 static struct parser_temp *parser_temp;
109 
110 
111 static union node *list(int, int);
112 static union node *andor(void);
113 static union node *pipeline(void);
114 static union node *command(void);
115 static union node *simplecmd(union node **, union node *);
116 static union node *makename(void);
117 static void parsefname(void);
118 static void parseheredoc(void);
119 static int peektoken(void);
120 static int readtoken(void);
121 static int xxreadtoken(void);
122 static int readtoken1(int, char const *, char *, int);
123 static int noexpand(char *);
124 static void synexpect(int) __dead2;
125 static void synerror(const char *) __dead2;
126 static void setprompt(int);
127 
128 
129 static void *
130 parser_temp_alloc(size_t len)
131 {
132 	struct parser_temp *t;
133 
134 	INTOFF;
135 	t = ckmalloc(sizeof(*t));
136 	t->data = NULL;
137 	t->next = parser_temp;
138 	parser_temp = t;
139 	t->data = ckmalloc(len);
140 	INTON;
141 	return t->data;
142 }
143 
144 
145 static void *
146 parser_temp_realloc(void *ptr, size_t len)
147 {
148 	struct parser_temp *t;
149 
150 	INTOFF;
151 	t = parser_temp;
152 	if (ptr != t->data)
153 		error("bug: parser_temp_realloc misused");
154 	t->data = ckrealloc(t->data, len);
155 	INTON;
156 	return t->data;
157 }
158 
159 
160 static void
161 parser_temp_free_upto(void *ptr)
162 {
163 	struct parser_temp *t;
164 	int done = 0;
165 
166 	INTOFF;
167 	while (parser_temp != NULL && !done) {
168 		t = parser_temp;
169 		parser_temp = t->next;
170 		done = t->data == ptr;
171 		ckfree(t->data);
172 		ckfree(t);
173 	}
174 	INTON;
175 	if (!done)
176 		error("bug: parser_temp_free_upto misused");
177 }
178 
179 
180 static void
181 parser_temp_free_all(void)
182 {
183 	struct parser_temp *t;
184 
185 	INTOFF;
186 	while (parser_temp != NULL) {
187 		t = parser_temp;
188 		parser_temp = t->next;
189 		ckfree(t->data);
190 		ckfree(t);
191 	}
192 	INTON;
193 }
194 
195 
196 /*
197  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
198  * valid parse tree indicating a blank line.)
199  */
200 
201 union node *
202 parsecmd(int interact)
203 {
204 	int t;
205 
206 	/* This assumes the parser is not re-entered,
207 	 * which could happen if we add command substitution on PS1/PS2.
208 	 */
209 	parser_temp_free_all();
210 	heredoclist = NULL;
211 
212 	tokpushback = 0;
213 	doprompt = interact;
214 	if (doprompt)
215 		setprompt(1);
216 	else
217 		setprompt(0);
218 	needprompt = 0;
219 	t = readtoken();
220 	if (t == TEOF)
221 		return NEOF;
222 	if (t == TNL)
223 		return NULL;
224 	tokpushback++;
225 	return list(1, 1);
226 }
227 
228 
229 static union node *
230 list(int nlflag, int erflag)
231 {
232 	union node *ntop, *n1, *n2, *n3;
233 	int tok;
234 
235 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
236 	if (!nlflag && !erflag && tokendlist[peektoken()])
237 		return NULL;
238 	ntop = n1 = NULL;
239 	for (;;) {
240 		n2 = andor();
241 		tok = readtoken();
242 		if (tok == TBACKGND) {
243 			if (n2->type == NCMD || n2->type == NPIPE) {
244 				n2->ncmd.backgnd = 1;
245 			} else if (n2->type == NREDIR) {
246 				n2->type = NBACKGND;
247 			} else {
248 				n3 = (union node *)stalloc(sizeof (struct nredir));
249 				n3->type = NBACKGND;
250 				n3->nredir.n = n2;
251 				n3->nredir.redirect = NULL;
252 				n2 = n3;
253 			}
254 		}
255 		if (ntop == NULL)
256 			ntop = n2;
257 		else if (n1 == NULL) {
258 			n1 = (union node *)stalloc(sizeof (struct nbinary));
259 			n1->type = NSEMI;
260 			n1->nbinary.ch1 = ntop;
261 			n1->nbinary.ch2 = n2;
262 			ntop = n1;
263 		}
264 		else {
265 			n3 = (union node *)stalloc(sizeof (struct nbinary));
266 			n3->type = NSEMI;
267 			n3->nbinary.ch1 = n1->nbinary.ch2;
268 			n3->nbinary.ch2 = n2;
269 			n1->nbinary.ch2 = n3;
270 			n1 = n3;
271 		}
272 		switch (tok) {
273 		case TBACKGND:
274 		case TSEMI:
275 			tok = readtoken();
276 			/* FALLTHROUGH */
277 		case TNL:
278 			if (tok == TNL) {
279 				parseheredoc();
280 				if (nlflag)
281 					return ntop;
282 			} else if (tok == TEOF && nlflag) {
283 				parseheredoc();
284 				return ntop;
285 			} else {
286 				tokpushback++;
287 			}
288 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
289 			if (!nlflag && !erflag && tokendlist[peektoken()])
290 				return ntop;
291 			break;
292 		case TEOF:
293 			if (heredoclist)
294 				parseheredoc();
295 			else
296 				pungetc();		/* push back EOF on input */
297 			return ntop;
298 		default:
299 			if (nlflag || erflag)
300 				synexpect(-1);
301 			tokpushback++;
302 			return ntop;
303 		}
304 	}
305 }
306 
307 
308 
309 static union node *
310 andor(void)
311 {
312 	union node *n1, *n2, *n3;
313 	int t;
314 
315 	n1 = pipeline();
316 	for (;;) {
317 		if ((t = readtoken()) == TAND) {
318 			t = NAND;
319 		} else if (t == TOR) {
320 			t = NOR;
321 		} else {
322 			tokpushback++;
323 			return n1;
324 		}
325 		n2 = pipeline();
326 		n3 = (union node *)stalloc(sizeof (struct nbinary));
327 		n3->type = t;
328 		n3->nbinary.ch1 = n1;
329 		n3->nbinary.ch2 = n2;
330 		n1 = n3;
331 	}
332 }
333 
334 
335 
336 static union node *
337 pipeline(void)
338 {
339 	union node *n1, *n2, *pipenode;
340 	struct nodelist *lp, *prev;
341 	int negate, t;
342 
343 	negate = 0;
344 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
345 	TRACE(("pipeline: entered\n"));
346 	while (readtoken() == TNOT)
347 		negate = !negate;
348 	tokpushback++;
349 	n1 = command();
350 	if (readtoken() == TPIPE) {
351 		pipenode = (union node *)stalloc(sizeof (struct npipe));
352 		pipenode->type = NPIPE;
353 		pipenode->npipe.backgnd = 0;
354 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
355 		pipenode->npipe.cmdlist = lp;
356 		lp->n = n1;
357 		do {
358 			prev = lp;
359 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
360 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
361 			t = readtoken();
362 			tokpushback++;
363 			if (t == TNOT)
364 				lp->n = pipeline();
365 			else
366 				lp->n = command();
367 			prev->next = lp;
368 		} while (readtoken() == TPIPE);
369 		lp->next = NULL;
370 		n1 = pipenode;
371 	}
372 	tokpushback++;
373 	if (negate) {
374 		n2 = (union node *)stalloc(sizeof (struct nnot));
375 		n2->type = NNOT;
376 		n2->nnot.com = n1;
377 		return n2;
378 	} else
379 		return n1;
380 }
381 
382 
383 
384 static union node *
385 command(void)
386 {
387 	union node *n1, *n2;
388 	union node *ap, **app;
389 	union node *cp, **cpp;
390 	union node *redir, **rpp;
391 	int t;
392 	int is_subshell;
393 
394 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
395 	is_subshell = 0;
396 	redir = NULL;
397 	n1 = NULL;
398 	rpp = &redir;
399 
400 	/* Check for redirection which may precede command */
401 	while (readtoken() == TREDIR) {
402 		*rpp = n2 = redirnode;
403 		rpp = &n2->nfile.next;
404 		parsefname();
405 	}
406 	tokpushback++;
407 
408 	switch (readtoken()) {
409 	case TIF:
410 		n1 = (union node *)stalloc(sizeof (struct nif));
411 		n1->type = NIF;
412 		if ((n1->nif.test = list(0, 0)) == NULL)
413 			synexpect(-1);
414 		if (readtoken() != TTHEN)
415 			synexpect(TTHEN);
416 		n1->nif.ifpart = list(0, 0);
417 		n2 = n1;
418 		while (readtoken() == TELIF) {
419 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
420 			n2 = n2->nif.elsepart;
421 			n2->type = NIF;
422 			if ((n2->nif.test = list(0, 0)) == NULL)
423 				synexpect(-1);
424 			if (readtoken() != TTHEN)
425 				synexpect(TTHEN);
426 			n2->nif.ifpart = list(0, 0);
427 		}
428 		if (lasttoken == TELSE)
429 			n2->nif.elsepart = list(0, 0);
430 		else {
431 			n2->nif.elsepart = NULL;
432 			tokpushback++;
433 		}
434 		if (readtoken() != TFI)
435 			synexpect(TFI);
436 		checkkwd = CHKKWD | CHKALIAS;
437 		break;
438 	case TWHILE:
439 	case TUNTIL: {
440 		int got;
441 		n1 = (union node *)stalloc(sizeof (struct nbinary));
442 		n1->type = (lasttoken == TWHILE)? NWHILE : NUNTIL;
443 		if ((n1->nbinary.ch1 = list(0, 0)) == NULL)
444 			synexpect(-1);
445 		if ((got=readtoken()) != TDO) {
446 TRACE(("expecting DO got %s %s\n", tokname[got], got == TWORD ? wordtext : ""));
447 			synexpect(TDO);
448 		}
449 		n1->nbinary.ch2 = list(0, 0);
450 		if (readtoken() != TDONE)
451 			synexpect(TDONE);
452 		checkkwd = CHKKWD | CHKALIAS;
453 		break;
454 	}
455 	case TFOR:
456 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
457 			synerror("Bad for loop variable");
458 		n1 = (union node *)stalloc(sizeof (struct nfor));
459 		n1->type = NFOR;
460 		n1->nfor.var = wordtext;
461 		while (readtoken() == TNL)
462 			;
463 		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
464 			app = &ap;
465 			while (readtoken() == TWORD) {
466 				n2 = (union node *)stalloc(sizeof (struct narg));
467 				n2->type = NARG;
468 				n2->narg.text = wordtext;
469 				n2->narg.backquote = backquotelist;
470 				*app = n2;
471 				app = &n2->narg.next;
472 			}
473 			*app = NULL;
474 			n1->nfor.args = ap;
475 			if (lasttoken != TNL && lasttoken != TSEMI)
476 				synexpect(-1);
477 		} else {
478 			static char argvars[5] = {
479 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
480 			};
481 			n2 = (union node *)stalloc(sizeof (struct narg));
482 			n2->type = NARG;
483 			n2->narg.text = argvars;
484 			n2->narg.backquote = NULL;
485 			n2->narg.next = NULL;
486 			n1->nfor.args = n2;
487 			/*
488 			 * Newline or semicolon here is optional (but note
489 			 * that the original Bourne shell only allowed NL).
490 			 */
491 			if (lasttoken != TNL && lasttoken != TSEMI)
492 				tokpushback++;
493 		}
494 		checkkwd = CHKNL | CHKKWD | CHKALIAS;
495 		if ((t = readtoken()) == TDO)
496 			t = TDONE;
497 		else if (t == TBEGIN)
498 			t = TEND;
499 		else
500 			synexpect(-1);
501 		n1->nfor.body = list(0, 0);
502 		if (readtoken() != t)
503 			synexpect(t);
504 		checkkwd = CHKKWD | CHKALIAS;
505 		break;
506 	case TCASE:
507 		n1 = (union node *)stalloc(sizeof (struct ncase));
508 		n1->type = NCASE;
509 		if (readtoken() != TWORD)
510 			synexpect(TWORD);
511 		n1->ncase.expr = n2 = (union node *)stalloc(sizeof (struct narg));
512 		n2->type = NARG;
513 		n2->narg.text = wordtext;
514 		n2->narg.backquote = backquotelist;
515 		n2->narg.next = NULL;
516 		while (readtoken() == TNL);
517 		if (lasttoken != TWORD || ! equal(wordtext, "in"))
518 			synerror("expecting \"in\"");
519 		cpp = &n1->ncase.cases;
520 		checkkwd = CHKNL | CHKKWD, readtoken();
521 		while (lasttoken != TESAC) {
522 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
523 			cp->type = NCLIST;
524 			app = &cp->nclist.pattern;
525 			if (lasttoken == TLP)
526 				readtoken();
527 			for (;;) {
528 				*app = ap = (union node *)stalloc(sizeof (struct narg));
529 				ap->type = NARG;
530 				ap->narg.text = wordtext;
531 				ap->narg.backquote = backquotelist;
532 				checkkwd = CHKNL | CHKKWD;
533 				if (readtoken() != TPIPE)
534 					break;
535 				app = &ap->narg.next;
536 				readtoken();
537 			}
538 			ap->narg.next = NULL;
539 			if (lasttoken != TRP)
540 				synexpect(TRP);
541 			cp->nclist.body = list(0, 0);
542 
543 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
544 			if ((t = readtoken()) != TESAC) {
545 				if (t != TENDCASE)
546 					synexpect(TENDCASE);
547 				else
548 					checkkwd = CHKNL | CHKKWD, readtoken();
549 			}
550 			cpp = &cp->nclist.next;
551 		}
552 		*cpp = NULL;
553 		checkkwd = CHKKWD | CHKALIAS;
554 		break;
555 	case TLP:
556 		n1 = (union node *)stalloc(sizeof (struct nredir));
557 		n1->type = NSUBSHELL;
558 		n1->nredir.n = list(0, 0);
559 		n1->nredir.redirect = NULL;
560 		if (readtoken() != TRP)
561 			synexpect(TRP);
562 		checkkwd = CHKKWD | CHKALIAS;
563 		is_subshell = 1;
564 		break;
565 	case TBEGIN:
566 		n1 = list(0, 0);
567 		if (readtoken() != TEND)
568 			synexpect(TEND);
569 		checkkwd = CHKKWD | CHKALIAS;
570 		break;
571 	/* Handle an empty command like other simple commands.  */
572 	case TBACKGND:
573 	case TSEMI:
574 	case TAND:
575 	case TOR:
576 		/*
577 		 * An empty command before a ; doesn't make much sense, and
578 		 * should certainly be disallowed in the case of `if ;'.
579 		 */
580 		if (!redir)
581 			synexpect(-1);
582 	case TNL:
583 	case TEOF:
584 	case TWORD:
585 	case TRP:
586 		tokpushback++;
587 		n1 = simplecmd(rpp, redir);
588 		return n1;
589 	default:
590 		synexpect(-1);
591 	}
592 
593 	/* Now check for redirection which may follow command */
594 	while (readtoken() == TREDIR) {
595 		*rpp = n2 = redirnode;
596 		rpp = &n2->nfile.next;
597 		parsefname();
598 	}
599 	tokpushback++;
600 	*rpp = NULL;
601 	if (redir) {
602 		if (!is_subshell) {
603 			n2 = (union node *)stalloc(sizeof (struct nredir));
604 			n2->type = NREDIR;
605 			n2->nredir.n = n1;
606 			n1 = n2;
607 		}
608 		n1->nredir.redirect = redir;
609 	}
610 
611 	return n1;
612 }
613 
614 
615 static union node *
616 simplecmd(union node **rpp, union node *redir)
617 {
618 	union node *args, **app;
619 	union node **orig_rpp = rpp;
620 	union node *n = NULL;
621 	int special;
622 	int savecheckkwd;
623 
624 	/* If we don't have any redirections already, then we must reset */
625 	/* rpp to be the address of the local redir variable.  */
626 	if (redir == 0)
627 		rpp = &redir;
628 
629 	args = NULL;
630 	app = &args;
631 	/*
632 	 * We save the incoming value, because we need this for shell
633 	 * functions.  There can not be a redirect or an argument between
634 	 * the function name and the open parenthesis.
635 	 */
636 	orig_rpp = rpp;
637 
638 	savecheckkwd = CHKALIAS;
639 
640 	for (;;) {
641 		checkkwd = savecheckkwd;
642 		if (readtoken() == TWORD) {
643 			n = (union node *)stalloc(sizeof (struct narg));
644 			n->type = NARG;
645 			n->narg.text = wordtext;
646 			n->narg.backquote = backquotelist;
647 			*app = n;
648 			app = &n->narg.next;
649 			if (savecheckkwd != 0 && !isassignment(wordtext))
650 				savecheckkwd = 0;
651 		} else if (lasttoken == TREDIR) {
652 			*rpp = n = redirnode;
653 			rpp = &n->nfile.next;
654 			parsefname();	/* read name of redirection file */
655 		} else if (lasttoken == TLP && app == &args->narg.next
656 					    && rpp == orig_rpp) {
657 			/* We have a function */
658 			if (readtoken() != TRP)
659 				synexpect(TRP);
660 			funclinno = plinno;
661 			/*
662 			 * - Require plain text.
663 			 * - Functions with '/' cannot be called.
664 			 * - Reject name=().
665 			 * - Reject ksh extended glob patterns.
666 			 */
667 			if (!noexpand(n->narg.text) || quoteflag ||
668 			    strchr(n->narg.text, '/') ||
669 			    strchr("!%*+-=?@}~",
670 				n->narg.text[strlen(n->narg.text) - 1]))
671 				synerror("Bad function name");
672 			rmescapes(n->narg.text);
673 			if (find_builtin(n->narg.text, &special) >= 0 &&
674 			    special)
675 				synerror("Cannot override a special builtin with a function");
676 			n->type = NDEFUN;
677 			n->narg.next = command();
678 			funclinno = 0;
679 			return n;
680 		} else {
681 			tokpushback++;
682 			break;
683 		}
684 	}
685 	*app = NULL;
686 	*rpp = NULL;
687 	n = (union node *)stalloc(sizeof (struct ncmd));
688 	n->type = NCMD;
689 	n->ncmd.backgnd = 0;
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 void
709 fixredir(union node *n, const char *text, int err)
710 {
711 	TRACE(("Fix redir %s %d\n", text, err));
712 	if (!err)
713 		n->ndup.vname = NULL;
714 
715 	if (is_digit(text[0]) && text[1] == '\0')
716 		n->ndup.dupfd = digit_val(text[0]);
717 	else if (text[0] == '-' && text[1] == '\0')
718 		n->ndup.dupfd = -1;
719 	else {
720 
721 		if (err)
722 			synerror("Bad fd number");
723 		else
724 			n->ndup.vname = makename();
725 	}
726 }
727 
728 
729 static void
730 parsefname(void)
731 {
732 	union node *n = redirnode;
733 
734 	if (readtoken() != TWORD)
735 		synexpect(-1);
736 	if (n->type == NHERE) {
737 		struct heredoc *here = heredoc;
738 		struct heredoc *p;
739 		int i;
740 
741 		if (quoteflag == 0)
742 			n->type = NXHERE;
743 		TRACE(("Here document %d\n", n->type));
744 		if (here->striptabs) {
745 			while (*wordtext == '\t')
746 				wordtext++;
747 		}
748 		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
749 			synerror("Illegal eof marker for << redirection");
750 		rmescapes(wordtext);
751 		here->eofmark = wordtext;
752 		here->next = NULL;
753 		if (heredoclist == NULL)
754 			heredoclist = here;
755 		else {
756 			for (p = heredoclist ; p->next ; p = p->next);
757 			p->next = here;
758 		}
759 	} else if (n->type == NTOFD || n->type == NFROMFD) {
760 		fixredir(n, wordtext, 0);
761 	} else {
762 		n->nfile.fname = makename();
763 	}
764 }
765 
766 
767 /*
768  * Input any here documents.
769  */
770 
771 static void
772 parseheredoc(void)
773 {
774 	struct heredoc *here;
775 	union node *n;
776 
777 	while (heredoclist) {
778 		here = heredoclist;
779 		heredoclist = here->next;
780 		if (needprompt) {
781 			setprompt(2);
782 			needprompt = 0;
783 		}
784 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
785 				here->eofmark, here->striptabs);
786 		n = (union node *)stalloc(sizeof (struct narg));
787 		n->narg.type = NARG;
788 		n->narg.next = NULL;
789 		n->narg.text = wordtext;
790 		n->narg.backquote = backquotelist;
791 		here->here->nhere.doc = n;
792 	}
793 }
794 
795 static int
796 peektoken(void)
797 {
798 	int t;
799 
800 	t = readtoken();
801 	tokpushback++;
802 	return (t);
803 }
804 
805 static int
806 readtoken(void)
807 {
808 	int t;
809 	struct alias *ap;
810 #ifdef DEBUG
811 	int alreadyseen = tokpushback;
812 #endif
813 
814 	top:
815 	t = xxreadtoken();
816 
817 	/*
818 	 * eat newlines
819 	 */
820 	if (checkkwd & CHKNL) {
821 		while (t == TNL) {
822 			parseheredoc();
823 			t = xxreadtoken();
824 		}
825 	}
826 
827 	/*
828 	 * check for keywords and aliases
829 	 */
830 	if (t == TWORD && !quoteflag)
831 	{
832 		const char * const *pp;
833 
834 		if (checkkwd & CHKKWD)
835 			for (pp = parsekwd; *pp; pp++) {
836 				if (**pp == *wordtext && equal(*pp, wordtext))
837 				{
838 					lasttoken = t = pp - parsekwd + KWDOFFSET;
839 					TRACE(("keyword %s recognized\n", tokname[t]));
840 					goto out;
841 				}
842 			}
843 		if (checkkwd & CHKALIAS &&
844 		    (ap = lookupalias(wordtext, 1)) != NULL) {
845 			pushstring(ap->val, strlen(ap->val), ap);
846 			goto top;
847 		}
848 	}
849 out:
850 	if (t != TNOT)
851 		checkkwd = 0;
852 
853 #ifdef DEBUG
854 	if (!alreadyseen)
855 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
856 	else
857 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
858 #endif
859 	return (t);
860 }
861 
862 
863 /*
864  * Read the next input token.
865  * If the token is a word, we set backquotelist to the list of cmds in
866  *	backquotes.  We set quoteflag to true if any part of the word was
867  *	quoted.
868  * If the token is TREDIR, then we set redirnode to a structure containing
869  *	the redirection.
870  * In all cases, the variable startlinno is set to the number of the line
871  *	on which the token starts.
872  *
873  * [Change comment:  here documents and internal procedures]
874  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
875  *  word parsing code into a separate routine.  In this case, readtoken
876  *  doesn't need to have any internal procedures, but parseword does.
877  *  We could also make parseoperator in essence the main routine, and
878  *  have parseword (readtoken1?) handle both words and redirection.]
879  */
880 
881 #define RETURN(token)	return lasttoken = token
882 
883 static int
884 xxreadtoken(void)
885 {
886 	int c;
887 
888 	if (tokpushback) {
889 		tokpushback = 0;
890 		return lasttoken;
891 	}
892 	if (needprompt) {
893 		setprompt(2);
894 		needprompt = 0;
895 	}
896 	startlinno = plinno;
897 	for (;;) {	/* until token or start of word found */
898 		c = pgetc_macro();
899 		switch (c) {
900 		case ' ': case '\t':
901 			continue;
902 		case '#':
903 			while ((c = pgetc()) != '\n' && c != PEOF);
904 			pungetc();
905 			continue;
906 		case '\\':
907 			if (pgetc() == '\n') {
908 				startlinno = ++plinno;
909 				if (doprompt)
910 					setprompt(2);
911 				else
912 					setprompt(0);
913 				continue;
914 			}
915 			pungetc();
916 			goto breakloop;
917 		case '\n':
918 			plinno++;
919 			needprompt = doprompt;
920 			RETURN(TNL);
921 		case PEOF:
922 			RETURN(TEOF);
923 		case '&':
924 			if (pgetc() == '&')
925 				RETURN(TAND);
926 			pungetc();
927 			RETURN(TBACKGND);
928 		case '|':
929 			if (pgetc() == '|')
930 				RETURN(TOR);
931 			pungetc();
932 			RETURN(TPIPE);
933 		case ';':
934 			if (pgetc() == ';')
935 				RETURN(TENDCASE);
936 			pungetc();
937 			RETURN(TSEMI);
938 		case '(':
939 			RETURN(TLP);
940 		case ')':
941 			RETURN(TRP);
942 		default:
943 			goto breakloop;
944 		}
945 	}
946 breakloop:
947 	return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
948 #undef RETURN
949 }
950 
951 
952 #define MAXNEST_static 8
953 struct tokenstate
954 {
955 	const char *syntax; /* *SYNTAX */
956 	int parenlevel; /* levels of parentheses in arithmetic */
957 	enum tokenstate_category
958 	{
959 		TSTATE_TOP,
960 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
961 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
962 		TSTATE_ARITH
963 	} category;
964 };
965 
966 
967 /*
968  * Called to parse command substitutions.
969  */
970 
971 static char *
972 parsebackq(char *out, struct nodelist **pbqlist,
973 		int oldstyle, int dblquote, int quoted)
974 {
975 	struct nodelist **nlpp;
976 	union node *n;
977 	char *volatile str;
978 	struct jmploc jmploc;
979 	struct jmploc *const savehandler = handler;
980 	int savelen;
981 	int saveprompt;
982 	const int bq_startlinno = plinno;
983 	char *volatile ostr = NULL;
984 	struct parsefile *const savetopfile = getcurrentfile();
985 	struct heredoc *const saveheredoclist = heredoclist;
986 	struct heredoc *here;
987 
988 	str = NULL;
989 	if (setjmp(jmploc.loc)) {
990 		popfilesupto(savetopfile);
991 		if (str)
992 			ckfree(str);
993 		if (ostr)
994 			ckfree(ostr);
995 		heredoclist = saveheredoclist;
996 		handler = savehandler;
997 		if (exception == EXERROR) {
998 			startlinno = bq_startlinno;
999 			synerror("Error in command substitution");
1000 		}
1001 		longjmp(handler->loc, 1);
1002 	}
1003 	INTOFF;
1004 	savelen = out - stackblock();
1005 	if (savelen > 0) {
1006 		str = ckmalloc(savelen);
1007 		memcpy(str, stackblock(), savelen);
1008 	}
1009 	handler = &jmploc;
1010 	heredoclist = NULL;
1011 	INTON;
1012         if (oldstyle) {
1013                 /* We must read until the closing backquote, giving special
1014                    treatment to some slashes, and then push the string and
1015                    reread it as input, interpreting it normally.  */
1016                 char *oout;
1017                 int c;
1018                 int olen;
1019 
1020 
1021                 STARTSTACKSTR(oout);
1022 		for (;;) {
1023 			if (needprompt) {
1024 				setprompt(2);
1025 				needprompt = 0;
1026 			}
1027 			CHECKSTRSPACE(2, oout);
1028 			switch (c = pgetc()) {
1029 			case '`':
1030 				goto done;
1031 
1032 			case '\\':
1033                                 if ((c = pgetc()) == '\n') {
1034 					plinno++;
1035 					if (doprompt)
1036 						setprompt(2);
1037 					else
1038 						setprompt(0);
1039 					/*
1040 					 * If eating a newline, avoid putting
1041 					 * the newline into the new character
1042 					 * stream (via the USTPUTC after the
1043 					 * switch).
1044 					 */
1045 					continue;
1046 				}
1047                                 if (c != '\\' && c != '`' && c != '$'
1048                                     && (!dblquote || c != '"'))
1049                                         USTPUTC('\\', oout);
1050 				break;
1051 
1052 			case '\n':
1053 				plinno++;
1054 				needprompt = doprompt;
1055 				break;
1056 
1057 			case PEOF:
1058 			        startlinno = plinno;
1059 				synerror("EOF in backquote substitution");
1060  				break;
1061 
1062 			default:
1063 				break;
1064 			}
1065 			USTPUTC(c, oout);
1066                 }
1067 done:
1068                 USTPUTC('\0', oout);
1069                 olen = oout - stackblock();
1070 		INTOFF;
1071 		ostr = ckmalloc(olen);
1072 		memcpy(ostr, stackblock(), olen);
1073 		setinputstring(ostr, 1);
1074 		INTON;
1075         }
1076 	nlpp = pbqlist;
1077 	while (*nlpp)
1078 		nlpp = &(*nlpp)->next;
1079 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1080 	(*nlpp)->next = NULL;
1081 
1082 	if (oldstyle) {
1083 		saveprompt = doprompt;
1084 		doprompt = 0;
1085 	}
1086 
1087 	n = list(0, oldstyle);
1088 
1089 	if (oldstyle)
1090 		doprompt = saveprompt;
1091 	else {
1092 		if (readtoken() != TRP)
1093 			synexpect(TRP);
1094 	}
1095 
1096 	(*nlpp)->n = n;
1097         if (oldstyle) {
1098 		/*
1099 		 * Start reading from old file again, ignoring any pushed back
1100 		 * tokens left from the backquote parsing
1101 		 */
1102                 popfile();
1103 		tokpushback = 0;
1104 	}
1105 	STARTSTACKSTR(out);
1106 	CHECKSTRSPACE(savelen + 1, out);
1107 	INTOFF;
1108 	if (str) {
1109 		memcpy(out, str, savelen);
1110 		STADJUST(savelen, out);
1111 		ckfree(str);
1112 		str = NULL;
1113 	}
1114 	if (ostr) {
1115 		ckfree(ostr);
1116 		ostr = NULL;
1117 	}
1118 	here = saveheredoclist;
1119 	if (here != NULL) {
1120 		while (here->next != NULL)
1121 			here = here->next;
1122 		here->next = heredoclist;
1123 		heredoclist = saveheredoclist;
1124 	}
1125 	handler = savehandler;
1126 	INTON;
1127 	if (quoted)
1128 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1129 	else
1130 		USTPUTC(CTLBACKQ, out);
1131 	return out;
1132 }
1133 
1134 
1135 /*
1136  * Called to parse a backslash escape sequence inside $'...'.
1137  * The backslash has already been read.
1138  */
1139 static char *
1140 readcstyleesc(char *out)
1141 {
1142 	int c, v, i, n;
1143 
1144 	c = pgetc();
1145 	switch (c) {
1146 	case '\0':
1147 		synerror("Unterminated quoted string");
1148 	case '\n':
1149 		plinno++;
1150 		if (doprompt)
1151 			setprompt(2);
1152 		else
1153 			setprompt(0);
1154 		return out;
1155 	case '\\':
1156 	case '\'':
1157 	case '"':
1158 		v = c;
1159 		break;
1160 	case 'a': v = '\a'; break;
1161 	case 'b': v = '\b'; break;
1162 	case 'e': v = '\033'; break;
1163 	case 'f': v = '\f'; break;
1164 	case 'n': v = '\n'; break;
1165 	case 'r': v = '\r'; break;
1166 	case 't': v = '\t'; break;
1167 	case 'v': v = '\v'; break;
1168 	case 'x':
1169 		  v = 0;
1170 		  for (;;) {
1171 			  c = pgetc();
1172 			  if (c >= '0' && c <= '9')
1173 				  v = (v << 4) + c - '0';
1174 			  else if (c >= 'A' && c <= 'F')
1175 				  v = (v << 4) + c - 'A' + 10;
1176 			  else if (c >= 'a' && c <= 'f')
1177 				  v = (v << 4) + c - 'a' + 10;
1178 			  else
1179 				  break;
1180 		  }
1181 		  pungetc();
1182 		  break;
1183 	case '0': case '1': case '2': case '3':
1184 	case '4': case '5': case '6': case '7':
1185 		  v = c - '0';
1186 		  c = pgetc();
1187 		  if (c >= '0' && c <= '7') {
1188 			  v <<= 3;
1189 			  v += c - '0';
1190 			  c = pgetc();
1191 			  if (c >= '0' && c <= '7') {
1192 				  v <<= 3;
1193 				  v += c - '0';
1194 			  } else
1195 				  pungetc();
1196 		  } else
1197 			  pungetc();
1198 		  break;
1199 	case 'c':
1200 		  c = pgetc();
1201 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1202 			  synerror("Bad escape sequence");
1203 		  if (c == '\\' && pgetc() != '\\')
1204 			  synerror("Bad escape sequence");
1205 		  if (c == '?')
1206 			  v = 127;
1207 		  else
1208 			  v = c & 0x1f;
1209 		  break;
1210 	case 'u':
1211 	case 'U':
1212 		  n = c == 'U' ? 8 : 4;
1213 		  v = 0;
1214 		  for (i = 0; i < n; i++) {
1215 			  c = pgetc();
1216 			  if (c >= '0' && c <= '9')
1217 				  v = (v << 4) + c - '0';
1218 			  else if (c >= 'A' && c <= 'F')
1219 				  v = (v << 4) + c - 'A' + 10;
1220 			  else if (c >= 'a' && c <= 'f')
1221 				  v = (v << 4) + c - 'a' + 10;
1222 			  else
1223 				  synerror("Bad escape sequence");
1224 		  }
1225 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1226 			  synerror("Bad escape sequence");
1227 		  /* We really need iconv here. */
1228 		  if (initial_localeisutf8 && v > 127) {
1229 			  CHECKSTRSPACE(4, out);
1230 			  /*
1231 			   * We cannot use wctomb() as the locale may have
1232 			   * changed.
1233 			   */
1234 			  if (v <= 0x7ff) {
1235 				  USTPUTC(0xc0 | v >> 6, out);
1236 				  USTPUTC(0x80 | (v & 0x3f), out);
1237 				  return out;
1238 			  } else if (v <= 0xffff) {
1239 				  USTPUTC(0xe0 | v >> 12, out);
1240 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1241 				  USTPUTC(0x80 | (v & 0x3f), out);
1242 				  return out;
1243 			  } else if (v <= 0x10ffff) {
1244 				  USTPUTC(0xf0 | v >> 18, out);
1245 				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1246 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1247 				  USTPUTC(0x80 | (v & 0x3f), out);
1248 				  return out;
1249 			  }
1250 		  }
1251 		  if (v > 127)
1252 			  v = '?';
1253 		  break;
1254 	default:
1255 		  synerror("Bad escape sequence");
1256 	}
1257 	v = (char)v;
1258 	/*
1259 	 * We can't handle NUL bytes.
1260 	 * POSIX says we should skip till the closing quote.
1261 	 */
1262 	if (v == '\0') {
1263 		while ((c = pgetc()) != '\'') {
1264 			if (c == '\\')
1265 				c = pgetc();
1266 			if (c == PEOF)
1267 				synerror("Unterminated quoted string");
1268 		}
1269 		pungetc();
1270 		return out;
1271 	}
1272 	if (SQSYNTAX[v] == CCTL)
1273 		USTPUTC(CTLESC, out);
1274 	USTPUTC(v, out);
1275 	return out;
1276 }
1277 
1278 
1279 /*
1280  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1281  * is not NULL, read a here document.  In the latter case, eofmark is the
1282  * word which marks the end of the document and striptabs is true if
1283  * leading tabs should be stripped from the document.  The argument firstc
1284  * is the first character of the input token or document.
1285  *
1286  * Because C does not have internal subroutines, I have simulated them
1287  * using goto's to implement the subroutine linkage.  The following macros
1288  * will run code that appears at the end of readtoken1.
1289  */
1290 
1291 #define CHECKEND()	{goto checkend; checkend_return:;}
1292 #define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1293 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1294 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1295 
1296 static int
1297 readtoken1(int firstc, char const *initialsyntax, char *eofmark, int striptabs)
1298 {
1299 	int c = firstc;
1300 	char *out;
1301 	int len;
1302 	char line[EOFMARKLEN + 1];
1303 	struct nodelist *bqlist;
1304 	int quotef;
1305 	int newvarnest;
1306 	int level;
1307 	int synentry;
1308 	struct tokenstate state_static[MAXNEST_static];
1309 	int maxnest = MAXNEST_static;
1310 	struct tokenstate *state = state_static;
1311 	int sqiscstyle = 0;
1312 
1313 	startlinno = plinno;
1314 	quotef = 0;
1315 	bqlist = NULL;
1316 	newvarnest = 0;
1317 	level = 0;
1318 	state[level].syntax = initialsyntax;
1319 	state[level].parenlevel = 0;
1320 	state[level].category = TSTATE_TOP;
1321 
1322 	STARTSTACKSTR(out);
1323 	loop: {	/* for each line, until end of word */
1324 		CHECKEND();	/* set c to PEOF if at end of here document */
1325 		for (;;) {	/* until end of line or end of word */
1326 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1327 
1328 			synentry = state[level].syntax[c];
1329 
1330 			switch(synentry) {
1331 			case CNL:	/* '\n' */
1332 				if (state[level].syntax == BASESYNTAX)
1333 					goto endword;	/* exit outer loop */
1334 				USTPUTC(c, out);
1335 				plinno++;
1336 				if (doprompt)
1337 					setprompt(2);
1338 				else
1339 					setprompt(0);
1340 				c = pgetc();
1341 				goto loop;		/* continue outer loop */
1342 			case CSBACK:
1343 				if (sqiscstyle) {
1344 					out = readcstyleesc(out);
1345 					break;
1346 				}
1347 				/* FALLTHROUGH */
1348 			case CWORD:
1349 				USTPUTC(c, out);
1350 				break;
1351 			case CCTL:
1352 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1353 					USTPUTC(CTLESC, out);
1354 				USTPUTC(c, out);
1355 				break;
1356 			case CBACK:	/* backslash */
1357 				c = pgetc();
1358 				if (c == PEOF) {
1359 					USTPUTC('\\', out);
1360 					pungetc();
1361 				} else if (c == '\n') {
1362 					plinno++;
1363 					if (doprompt)
1364 						setprompt(2);
1365 					else
1366 						setprompt(0);
1367 				} else {
1368 					if (state[level].syntax == DQSYNTAX &&
1369 					    c != '\\' && c != '`' && c != '$' &&
1370 					    (c != '"' || (eofmark != NULL &&
1371 						newvarnest == 0)) &&
1372 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1373 						USTPUTC('\\', out);
1374 					if ((eofmark == NULL ||
1375 					    newvarnest > 0) &&
1376 					    state[level].syntax == BASESYNTAX)
1377 						USTPUTC(CTLQUOTEMARK, out);
1378 					if (SQSYNTAX[c] == CCTL)
1379 						USTPUTC(CTLESC, out);
1380 					USTPUTC(c, out);
1381 					if ((eofmark == NULL ||
1382 					    newvarnest > 0) &&
1383 					    state[level].syntax == BASESYNTAX &&
1384 					    state[level].category == TSTATE_VAR_OLD)
1385 						USTPUTC(CTLQUOTEEND, out);
1386 					quotef++;
1387 				}
1388 				break;
1389 			case CSQUOTE:
1390 				USTPUTC(CTLQUOTEMARK, out);
1391 				state[level].syntax = SQSYNTAX;
1392 				sqiscstyle = 0;
1393 				break;
1394 			case CDQUOTE:
1395 				USTPUTC(CTLQUOTEMARK, out);
1396 				state[level].syntax = DQSYNTAX;
1397 				break;
1398 			case CENDQUOTE:
1399 				if (eofmark != NULL && newvarnest == 0)
1400 					USTPUTC(c, out);
1401 				else {
1402 					if (state[level].category == TSTATE_VAR_OLD)
1403 						USTPUTC(CTLQUOTEEND, out);
1404 					state[level].syntax = BASESYNTAX;
1405 					quotef++;
1406 				}
1407 				break;
1408 			case CVAR:	/* '$' */
1409 				PARSESUB();		/* parse substitution */
1410 				break;
1411 			case CENDVAR:	/* '}' */
1412 				if (level > 0 &&
1413 				    ((state[level].category == TSTATE_VAR_OLD &&
1414 				      state[level].syntax ==
1415 				      state[level - 1].syntax) ||
1416 				    (state[level].category == TSTATE_VAR_NEW &&
1417 				     state[level].syntax == BASESYNTAX))) {
1418 					if (state[level].category == TSTATE_VAR_NEW)
1419 						newvarnest--;
1420 					level--;
1421 					USTPUTC(CTLENDVAR, out);
1422 				} else {
1423 					USTPUTC(c, out);
1424 				}
1425 				break;
1426 			case CLP:	/* '(' in arithmetic */
1427 				state[level].parenlevel++;
1428 				USTPUTC(c, out);
1429 				break;
1430 			case CRP:	/* ')' in arithmetic */
1431 				if (state[level].parenlevel > 0) {
1432 					USTPUTC(c, out);
1433 					--state[level].parenlevel;
1434 				} else {
1435 					if (pgetc() == ')') {
1436 						if (level > 0 &&
1437 						    state[level].category == TSTATE_ARITH) {
1438 							level--;
1439 							USTPUTC(CTLENDARI, out);
1440 						} else
1441 							USTPUTC(')', out);
1442 					} else {
1443 						/*
1444 						 * unbalanced parens
1445 						 *  (don't 2nd guess - no error)
1446 						 */
1447 						pungetc();
1448 						USTPUTC(')', out);
1449 					}
1450 				}
1451 				break;
1452 			case CBQUOTE:	/* '`' */
1453 				out = parsebackq(out, &bqlist, 1,
1454 				    state[level].syntax == DQSYNTAX &&
1455 				    (eofmark == NULL || newvarnest > 0),
1456 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1457 				break;
1458 			case CEOF:
1459 				goto endword;		/* exit outer loop */
1460 			case CIGN:
1461 				break;
1462 			default:
1463 				if (level == 0)
1464 					goto endword;	/* exit outer loop */
1465 				USTPUTC(c, out);
1466 			}
1467 			c = pgetc_macro();
1468 		}
1469 	}
1470 endword:
1471 	if (state[level].syntax == ARISYNTAX)
1472 		synerror("Missing '))'");
1473 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1474 		synerror("Unterminated quoted string");
1475 	if (state[level].category == TSTATE_VAR_OLD ||
1476 	    state[level].category == TSTATE_VAR_NEW) {
1477 		startlinno = plinno;
1478 		synerror("Missing '}'");
1479 	}
1480 	if (state != state_static)
1481 		parser_temp_free_upto(state);
1482 	USTPUTC('\0', out);
1483 	len = out - stackblock();
1484 	out = stackblock();
1485 	if (eofmark == NULL) {
1486 		if ((c == '>' || c == '<')
1487 		 && quotef == 0
1488 		 && len <= 2
1489 		 && (*out == '\0' || is_digit(*out))) {
1490 			PARSEREDIR();
1491 			return lasttoken = TREDIR;
1492 		} else {
1493 			pungetc();
1494 		}
1495 	}
1496 	quoteflag = quotef;
1497 	backquotelist = bqlist;
1498 	grabstackblock(len);
1499 	wordtext = out;
1500 	return lasttoken = TWORD;
1501 /* end of readtoken routine */
1502 
1503 
1504 /*
1505  * Check to see whether we are at the end of the here document.  When this
1506  * is called, c is set to the first character of the next input line.  If
1507  * we are at the end of the here document, this routine sets the c to PEOF.
1508  */
1509 
1510 checkend: {
1511 	if (eofmark) {
1512 		if (striptabs) {
1513 			while (c == '\t')
1514 				c = pgetc();
1515 		}
1516 		if (c == *eofmark) {
1517 			if (pfgets(line, sizeof line) != NULL) {
1518 				char *p, *q;
1519 
1520 				p = line;
1521 				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1522 				if ((*p == '\0' || *p == '\n') && *q == '\0') {
1523 					c = PEOF;
1524 					if (*p == '\n') {
1525 						plinno++;
1526 						needprompt = doprompt;
1527 					}
1528 				} else {
1529 					pushstring(line, strlen(line), NULL);
1530 				}
1531 			}
1532 		}
1533 	}
1534 	goto checkend_return;
1535 }
1536 
1537 
1538 /*
1539  * Parse a redirection operator.  The variable "out" points to a string
1540  * specifying the fd to be redirected.  The variable "c" contains the
1541  * first character of the redirection operator.
1542  */
1543 
1544 parseredir: {
1545 	char fd = *out;
1546 	union node *np;
1547 
1548 	np = (union node *)stalloc(sizeof (struct nfile));
1549 	if (c == '>') {
1550 		np->nfile.fd = 1;
1551 		c = pgetc();
1552 		if (c == '>')
1553 			np->type = NAPPEND;
1554 		else if (c == '&')
1555 			np->type = NTOFD;
1556 		else if (c == '|')
1557 			np->type = NCLOBBER;
1558 		else {
1559 			np->type = NTO;
1560 			pungetc();
1561 		}
1562 	} else {	/* c == '<' */
1563 		np->nfile.fd = 0;
1564 		c = pgetc();
1565 		if (c == '<') {
1566 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1567 				np = (union node *)stalloc(sizeof (struct nhere));
1568 				np->nfile.fd = 0;
1569 			}
1570 			np->type = NHERE;
1571 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1572 			heredoc->here = np;
1573 			if ((c = pgetc()) == '-') {
1574 				heredoc->striptabs = 1;
1575 			} else {
1576 				heredoc->striptabs = 0;
1577 				pungetc();
1578 			}
1579 		} else if (c == '&')
1580 			np->type = NFROMFD;
1581 		else if (c == '>')
1582 			np->type = NFROMTO;
1583 		else {
1584 			np->type = NFROM;
1585 			pungetc();
1586 		}
1587 	}
1588 	if (fd != '\0')
1589 		np->nfile.fd = digit_val(fd);
1590 	redirnode = np;
1591 	goto parseredir_return;
1592 }
1593 
1594 
1595 /*
1596  * Parse a substitution.  At this point, we have read the dollar sign
1597  * and nothing else.
1598  */
1599 
1600 parsesub: {
1601 	char buf[10];
1602 	int subtype;
1603 	int typeloc;
1604 	int flags;
1605 	char *p;
1606 	static const char types[] = "}-+?=";
1607 	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1608 	int linno;
1609 	int length;
1610 	int c1;
1611 
1612 	c = pgetc();
1613 	if (c == '(') {	/* $(command) or $((arith)) */
1614 		if (pgetc() == '(') {
1615 			PARSEARITH();
1616 		} else {
1617 			pungetc();
1618 			out = parsebackq(out, &bqlist, 0,
1619 			    state[level].syntax == DQSYNTAX &&
1620 			    (eofmark == NULL || newvarnest > 0),
1621 			    state[level].syntax == DQSYNTAX ||
1622 			    state[level].syntax == ARISYNTAX);
1623 		}
1624 	} else if (c == '{' || is_name(c) || is_special(c)) {
1625 		USTPUTC(CTLVAR, out);
1626 		typeloc = out - stackblock();
1627 		USTPUTC(VSNORMAL, out);
1628 		subtype = VSNORMAL;
1629 		flags = 0;
1630 		if (c == '{') {
1631 			bracketed_name = 1;
1632 			c = pgetc();
1633 			subtype = 0;
1634 		}
1635 varname:
1636 		if (!is_eof(c) && is_name(c)) {
1637 			length = 0;
1638 			do {
1639 				STPUTC(c, out);
1640 				c = pgetc();
1641 				length++;
1642 			} while (!is_eof(c) && is_in_name(c));
1643 			if (length == 6 &&
1644 			    strncmp(out - length, "LINENO", length) == 0) {
1645 				/* Replace the variable name with the
1646 				 * current line number. */
1647 				linno = plinno;
1648 				if (funclinno != 0)
1649 					linno -= funclinno - 1;
1650 				snprintf(buf, sizeof(buf), "%d", linno);
1651 				STADJUST(-6, out);
1652 				STPUTS(buf, out);
1653 				flags |= VSLINENO;
1654 			}
1655 		} else if (is_digit(c)) {
1656 			if (bracketed_name) {
1657 				do {
1658 					STPUTC(c, out);
1659 					c = pgetc();
1660 				} while (is_digit(c));
1661 			} else {
1662 				STPUTC(c, out);
1663 				c = pgetc();
1664 			}
1665 		} else if (is_special(c)) {
1666 			c1 = c;
1667 			c = pgetc();
1668 			if (subtype == 0 && c1 == '#') {
1669 				subtype = VSLENGTH;
1670 				if (strchr(types, c) == NULL && c != ':' &&
1671 				    c != '#' && c != '%')
1672 					goto varname;
1673 				c1 = c;
1674 				c = pgetc();
1675 				if (c1 != '}' && c == '}') {
1676 					pungetc();
1677 					c = c1;
1678 					goto varname;
1679 				}
1680 				pungetc();
1681 				c = c1;
1682 				c1 = '#';
1683 				subtype = 0;
1684 			}
1685 			USTPUTC(c1, out);
1686 		} else {
1687 			subtype = VSERROR;
1688 			if (c == '}')
1689 				pungetc();
1690 			else if (c == '\n' || c == PEOF)
1691 				synerror("Unexpected end of line in substitution");
1692 			else
1693 				USTPUTC(c, out);
1694 		}
1695 		if (subtype == 0) {
1696 			switch (c) {
1697 			case ':':
1698 				flags |= VSNUL;
1699 				c = pgetc();
1700 				/*FALLTHROUGH*/
1701 			default:
1702 				p = strchr(types, c);
1703 				if (p == NULL) {
1704 					if (c == '\n' || c == PEOF)
1705 						synerror("Unexpected end of line in substitution");
1706 					if (flags == VSNUL)
1707 						STPUTC(':', out);
1708 					STPUTC(c, out);
1709 					subtype = VSERROR;
1710 				} else
1711 					subtype = p - types + VSNORMAL;
1712 				break;
1713 			case '%':
1714 			case '#':
1715 				{
1716 					int cc = c;
1717 					subtype = c == '#' ? VSTRIMLEFT :
1718 							     VSTRIMRIGHT;
1719 					c = pgetc();
1720 					if (c == cc)
1721 						subtype++;
1722 					else
1723 						pungetc();
1724 					break;
1725 				}
1726 			}
1727 		} else if (subtype != VSERROR) {
1728 			if (subtype == VSLENGTH && c != '}')
1729 				subtype = VSERROR;
1730 			pungetc();
1731 		}
1732 		STPUTC('=', out);
1733 		if (state[level].syntax == DQSYNTAX ||
1734 		    state[level].syntax == ARISYNTAX)
1735 			flags |= VSQUOTE;
1736 		*(stackblock() + typeloc) = subtype | flags;
1737 		if (subtype != VSNORMAL) {
1738 			if (level + 1 >= maxnest) {
1739 				maxnest *= 2;
1740 				if (state == state_static) {
1741 					state = parser_temp_alloc(
1742 					    maxnest * sizeof(*state));
1743 					memcpy(state, state_static,
1744 					    MAXNEST_static * sizeof(*state));
1745 				} else
1746 					state = parser_temp_realloc(state,
1747 					    maxnest * sizeof(*state));
1748 			}
1749 			level++;
1750 			state[level].parenlevel = 0;
1751 			if (subtype == VSMINUS || subtype == VSPLUS ||
1752 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1753 				/*
1754 				 * For operators that were in the Bourne shell,
1755 				 * inherit the double-quote state.
1756 				 */
1757 				state[level].syntax = state[level - 1].syntax;
1758 				state[level].category = TSTATE_VAR_OLD;
1759 			} else {
1760 				/*
1761 				 * The other operators take a pattern,
1762 				 * so go to BASESYNTAX.
1763 				 * Also, ' and " are now special, even
1764 				 * in here documents.
1765 				 */
1766 				state[level].syntax = BASESYNTAX;
1767 				state[level].category = TSTATE_VAR_NEW;
1768 				newvarnest++;
1769 			}
1770 		}
1771 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1772 		/* $'cstylequotes' */
1773 		USTPUTC(CTLQUOTEMARK, out);
1774 		state[level].syntax = SQSYNTAX;
1775 		sqiscstyle = 1;
1776 	} else {
1777 		USTPUTC('$', out);
1778 		pungetc();
1779 	}
1780 	goto parsesub_return;
1781 }
1782 
1783 
1784 /*
1785  * Parse an arithmetic expansion (indicate start of one and set state)
1786  */
1787 parsearith: {
1788 
1789 	if (level + 1 >= maxnest) {
1790 		maxnest *= 2;
1791 		if (state == state_static) {
1792 			state = parser_temp_alloc(
1793 			    maxnest * sizeof(*state));
1794 			memcpy(state, state_static,
1795 			    MAXNEST_static * sizeof(*state));
1796 		} else
1797 			state = parser_temp_realloc(state,
1798 			    maxnest * sizeof(*state));
1799 	}
1800 	level++;
1801 	state[level].syntax = ARISYNTAX;
1802 	state[level].parenlevel = 0;
1803 	state[level].category = TSTATE_ARITH;
1804 	USTPUTC(CTLARI, out);
1805 	if (state[level - 1].syntax == DQSYNTAX)
1806 		USTPUTC('"',out);
1807 	else
1808 		USTPUTC(' ',out);
1809 	goto parsearith_return;
1810 }
1811 
1812 } /* end of readtoken */
1813 
1814 
1815 
1816 #ifdef mkinit
1817 RESET {
1818 	tokpushback = 0;
1819 	checkkwd = 0;
1820 }
1821 #endif
1822 
1823 /*
1824  * Returns true if the text contains nothing to expand (no dollar signs
1825  * or backquotes).
1826  */
1827 
1828 static int
1829 noexpand(char *text)
1830 {
1831 	char *p;
1832 	char c;
1833 
1834 	p = text;
1835 	while ((c = *p++) != '\0') {
1836 		if ( c == CTLQUOTEMARK)
1837 			continue;
1838 		if (c == CTLESC)
1839 			p++;
1840 		else if (BASESYNTAX[(int)c] == CCTL)
1841 			return 0;
1842 	}
1843 	return 1;
1844 }
1845 
1846 
1847 /*
1848  * Return true if the argument is a legal variable name (a letter or
1849  * underscore followed by zero or more letters, underscores, and digits).
1850  */
1851 
1852 int
1853 goodname(const char *name)
1854 {
1855 	const char *p;
1856 
1857 	p = name;
1858 	if (! is_name(*p))
1859 		return 0;
1860 	while (*++p) {
1861 		if (! is_in_name(*p))
1862 			return 0;
1863 	}
1864 	return 1;
1865 }
1866 
1867 
1868 int
1869 isassignment(const char *p)
1870 {
1871 	if (!is_name(*p))
1872 		return 0;
1873 	p++;
1874 	for (;;) {
1875 		if (*p == '=')
1876 			return 1;
1877 		else if (!is_in_name(*p))
1878 			return 0;
1879 		p++;
1880 	}
1881 }
1882 
1883 
1884 /*
1885  * Called when an unexpected token is read during the parse.  The argument
1886  * is the token that is expected, or -1 if more than one type of token can
1887  * occur at this point.
1888  */
1889 
1890 static void
1891 synexpect(int token)
1892 {
1893 	char msg[64];
1894 
1895 	if (token >= 0) {
1896 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1897 			tokname[lasttoken], tokname[token]);
1898 	} else {
1899 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1900 	}
1901 	synerror(msg);
1902 }
1903 
1904 
1905 static void
1906 synerror(const char *msg)
1907 {
1908 	if (commandname)
1909 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1910 	outfmt(out2, "Syntax error: %s\n", msg);
1911 	error((char *)NULL);
1912 }
1913 
1914 static void
1915 setprompt(int which)
1916 {
1917 	whichprompt = which;
1918 
1919 #ifndef NO_HISTORY
1920 	if (!el)
1921 #endif
1922 	{
1923 		out2str(getprompt(NULL));
1924 		flushout(out2);
1925 	}
1926 }
1927 
1928 /*
1929  * called by editline -- any expansions to the prompt
1930  *    should be added here.
1931  */
1932 char *
1933 getprompt(void *unused __unused)
1934 {
1935 	static char ps[PROMPTLEN];
1936 	char *fmt;
1937 	const char *pwd;
1938 	int i, trim;
1939 	static char internal_error[] = "??";
1940 
1941 	/*
1942 	 * Select prompt format.
1943 	 */
1944 	switch (whichprompt) {
1945 	case 0:
1946 		fmt = nullstr;
1947 		break;
1948 	case 1:
1949 		fmt = ps1val();
1950 		break;
1951 	case 2:
1952 		fmt = ps2val();
1953 		break;
1954 	default:
1955 		return internal_error;
1956 	}
1957 
1958 	/*
1959 	 * Format prompt string.
1960 	 */
1961 	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1962 		if (*fmt == '\\')
1963 			switch (*++fmt) {
1964 
1965 				/*
1966 				 * Hostname.
1967 				 *
1968 				 * \h specifies just the local hostname,
1969 				 * \H specifies fully-qualified hostname.
1970 				 */
1971 			case 'h':
1972 			case 'H':
1973 				ps[i] = '\0';
1974 				gethostname(&ps[i], PROMPTLEN - i);
1975 				/* Skip to end of hostname. */
1976 				trim = (*fmt == 'h') ? '.' : '\0';
1977 				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1978 					i++;
1979 				break;
1980 
1981 				/*
1982 				 * Working directory.
1983 				 *
1984 				 * \W specifies just the final component,
1985 				 * \w specifies the entire path.
1986 				 */
1987 			case 'W':
1988 			case 'w':
1989 				pwd = lookupvar("PWD");
1990 				if (pwd == NULL)
1991 					pwd = "?";
1992 				if (*fmt == 'W' &&
1993 				    *pwd == '/' && pwd[1] != '\0')
1994 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
1995 					    PROMPTLEN - i);
1996 				else
1997 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
1998 				/* Skip to end of path. */
1999 				while (ps[i + 1] != '\0')
2000 					i++;
2001 				break;
2002 
2003 				/*
2004 				 * Superuser status.
2005 				 *
2006 				 * '$' for normal users, '#' for root.
2007 				 */
2008 			case '$':
2009 				ps[i] = (geteuid() != 0) ? '$' : '#';
2010 				break;
2011 
2012 				/*
2013 				 * A literal \.
2014 				 */
2015 			case '\\':
2016 				ps[i] = '\\';
2017 				break;
2018 
2019 				/*
2020 				 * Emit unrecognized formats verbatim.
2021 				 */
2022 			default:
2023 				ps[i++] = '\\';
2024 				ps[i] = *fmt;
2025 				break;
2026 			}
2027 		else
2028 			ps[i] = *fmt;
2029 	ps[i] = '\0';
2030 	return (ps);
2031 }
2032