xref: /dflybsd-src/bin/sh/parser.c (revision 088f2cdc6274ddce9e65245341b5ed3059305a5c)
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  * 3. All advertising materials mentioning features or use of this software
17  *    must display the following acknowledgement:
18  *	This product includes software developed by the University of
19  *	California, Berkeley and its contributors.
20  * 4. Neither the name of the University nor the names of its contributors
21  *    may be used to endorse or promote products derived from this software
22  *    without specific prior written permission.
23  *
24  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34  * SUCH DAMAGE.
35  *
36  * @(#)parser.c	8.7 (Berkeley) 5/16/95
37  * $FreeBSD: src/bin/sh/parser.c,v 1.112 2011/05/21 22:03:06 jilles Exp $
38  */
39 
40 #include <stdio.h>
41 #include <stdlib.h>
42 #include <unistd.h>
43 
44 #include "shell.h"
45 #include "parser.h"
46 #include "nodes.h"
47 #include "expand.h"	/* defines rmescapes() */
48 #include "syntax.h"
49 #include "options.h"
50 #include "input.h"
51 #include "output.h"
52 #include "var.h"
53 #include "error.h"
54 #include "memalloc.h"
55 #include "mystring.h"
56 #include "alias.h"
57 #include "show.h"
58 #include "eval.h"
59 #include "exec.h"	/* to check for special builtins */
60 #ifndef NO_HISTORY
61 #include "myhistedit.h"
62 #endif
63 
64 /*
65  * Shell command parser.
66  */
67 
68 #define	EOFMARKLEN	79
69 #define	PROMPTLEN	128
70 
71 /* values of checkkwd variable */
72 #define CHKALIAS	0x1
73 #define CHKKWD		0x2
74 #define CHKNL		0x4
75 
76 /* values returned by readtoken */
77 #include "token.h"
78 
79 
80 
81 struct heredoc {
82 	struct heredoc *next;	/* next here document in list */
83 	union node *here;		/* redirection node */
84 	char *eofmark;		/* string indicating end of input */
85 	int striptabs;		/* if set, strip leading tabs */
86 };
87 
88 struct parser_temp {
89 	struct parser_temp *next;
90 	void *data;
91 };
92 
93 
94 static struct heredoc *heredoclist;	/* list of here documents to read */
95 static int doprompt;		/* if set, prompt the user */
96 static int needprompt;		/* true if interactive and at start of line */
97 static int lasttoken;		/* last token read */
98 MKINIT int tokpushback;		/* last token pushed back */
99 static char *wordtext;		/* text of last word returned by readtoken */
100 MKINIT int checkkwd;            /* 1 == check for kwds, 2 == also eat newlines */
101 static struct nodelist *backquotelist;
102 static union node *redirnode;
103 static struct heredoc *heredoc;
104 static int quoteflag;		/* set if (part of) last token was quoted */
105 static int startlinno;		/* line # where last token started */
106 static int funclinno;		/* line # where the current function started */
107 static struct parser_temp *parser_temp;
108 
109 
110 static union node *list(int, int);
111 static union node *andor(void);
112 static union node *pipeline(void);
113 static union node *command(void);
114 static union node *simplecmd(union node **, union node *);
115 static union node *makename(void);
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, char const *, char *, int);
122 static int noexpand(char *);
123 static void synexpect(int) __dead2;
124 static void synerror(const char *) __dead2;
125 static void setprompt(int);
126 
127 
128 static void *
129 parser_temp_alloc(size_t len)
130 {
131 	struct parser_temp *t;
132 
133 	INTOFF;
134 	t = ckmalloc(sizeof(*t));
135 	t->data = NULL;
136 	t->next = parser_temp;
137 	parser_temp = t;
138 	t->data = ckmalloc(len);
139 	INTON;
140 	return t->data;
141 }
142 
143 
144 static void *
145 parser_temp_realloc(void *ptr, size_t len)
146 {
147 	struct parser_temp *t;
148 
149 	INTOFF;
150 	t = parser_temp;
151 	if (ptr != t->data)
152 		error("bug: parser_temp_realloc misused");
153 	t->data = ckrealloc(t->data, len);
154 	INTON;
155 	return t->data;
156 }
157 
158 
159 static void
160 parser_temp_free_upto(void *ptr)
161 {
162 	struct parser_temp *t;
163 	int done = 0;
164 
165 	INTOFF;
166 	while (parser_temp != NULL && !done) {
167 		t = parser_temp;
168 		parser_temp = t->next;
169 		done = t->data == ptr;
170 		ckfree(t->data);
171 		ckfree(t);
172 	}
173 	INTON;
174 	if (!done)
175 		error("bug: parser_temp_free_upto misused");
176 }
177 
178 
179 static void
180 parser_temp_free_all(void)
181 {
182 	struct parser_temp *t;
183 
184 	INTOFF;
185 	while (parser_temp != NULL) {
186 		t = parser_temp;
187 		parser_temp = t->next;
188 		ckfree(t->data);
189 		ckfree(t);
190 	}
191 	INTON;
192 }
193 
194 
195 /*
196  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
197  * valid parse tree indicating a blank line.)
198  */
199 
200 union node *
201 parsecmd(int interact)
202 {
203 	int t;
204 
205 	/* This assumes the parser is not re-entered,
206 	 * which could happen if we add command substitution on PS1/PS2.
207 	 */
208 	parser_temp_free_all();
209 	heredoclist = NULL;
210 
211 	tokpushback = 0;
212 	doprompt = interact;
213 	if (doprompt)
214 		setprompt(1);
215 	else
216 		setprompt(0);
217 	needprompt = 0;
218 	t = readtoken();
219 	if (t == TEOF)
220 		return NEOF;
221 	if (t == TNL)
222 		return NULL;
223 	tokpushback++;
224 	return list(1, 1);
225 }
226 
227 
228 static union node *
229 list(int nlflag, int erflag)
230 {
231 	union node *ntop, *n1, *n2, *n3;
232 	int tok;
233 
234 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
235 	if (!nlflag && !erflag && tokendlist[peektoken()])
236 		return NULL;
237 	ntop = n1 = NULL;
238 	for (;;) {
239 		n2 = andor();
240 		tok = readtoken();
241 		if (tok == TBACKGND) {
242 			if (n2->type == NCMD || n2->type == NPIPE) {
243 				n2->ncmd.backgnd = 1;
244 			} else if (n2->type == NREDIR) {
245 				n2->type = NBACKGND;
246 			} else {
247 				n3 = (union node *)stalloc(sizeof (struct nredir));
248 				n3->type = NBACKGND;
249 				n3->nredir.n = n2;
250 				n3->nredir.redirect = NULL;
251 				n2 = n3;
252 			}
253 		}
254 		if (ntop == NULL)
255 			ntop = n2;
256 		else if (n1 == NULL) {
257 			n1 = (union node *)stalloc(sizeof (struct nbinary));
258 			n1->type = NSEMI;
259 			n1->nbinary.ch1 = ntop;
260 			n1->nbinary.ch2 = n2;
261 			ntop = n1;
262 		}
263 		else {
264 			n3 = (union node *)stalloc(sizeof (struct nbinary));
265 			n3->type = NSEMI;
266 			n3->nbinary.ch1 = n1->nbinary.ch2;
267 			n3->nbinary.ch2 = n2;
268 			n1->nbinary.ch2 = n3;
269 			n1 = n3;
270 		}
271 		switch (tok) {
272 		case TBACKGND:
273 		case TSEMI:
274 			tok = readtoken();
275 			/* FALLTHROUGH */
276 		case TNL:
277 			if (tok == TNL) {
278 				parseheredoc();
279 				if (nlflag)
280 					return ntop;
281 			} else if (tok == TEOF && nlflag) {
282 				parseheredoc();
283 				return ntop;
284 			} else {
285 				tokpushback++;
286 			}
287 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
288 			if (!nlflag && !erflag && tokendlist[peektoken()])
289 				return ntop;
290 			break;
291 		case TEOF:
292 			if (heredoclist)
293 				parseheredoc();
294 			else
295 				pungetc();		/* push back EOF on input */
296 			return ntop;
297 		default:
298 			if (nlflag || erflag)
299 				synexpect(-1);
300 			tokpushback++;
301 			return ntop;
302 		}
303 	}
304 }
305 
306 
307 
308 static union node *
309 andor(void)
310 {
311 	union node *n1, *n2, *n3;
312 	int t;
313 
314 	n1 = pipeline();
315 	for (;;) {
316 		if ((t = readtoken()) == TAND) {
317 			t = NAND;
318 		} else if (t == TOR) {
319 			t = NOR;
320 		} else {
321 			tokpushback++;
322 			return n1;
323 		}
324 		n2 = pipeline();
325 		n3 = (union node *)stalloc(sizeof (struct nbinary));
326 		n3->type = t;
327 		n3->nbinary.ch1 = n1;
328 		n3->nbinary.ch2 = n2;
329 		n1 = n3;
330 	}
331 }
332 
333 
334 
335 static union node *
336 pipeline(void)
337 {
338 	union node *n1, *n2, *pipenode;
339 	struct nodelist *lp, *prev;
340 	int negate, t;
341 
342 	negate = 0;
343 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
344 	TRACE(("pipeline: entered\n"));
345 	while (readtoken() == TNOT)
346 		negate = !negate;
347 	tokpushback++;
348 	n1 = command();
349 	if (readtoken() == TPIPE) {
350 		pipenode = (union node *)stalloc(sizeof (struct npipe));
351 		pipenode->type = NPIPE;
352 		pipenode->npipe.backgnd = 0;
353 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
354 		pipenode->npipe.cmdlist = lp;
355 		lp->n = n1;
356 		do {
357 			prev = lp;
358 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
359 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
360 			t = readtoken();
361 			tokpushback++;
362 			if (t == TNOT)
363 				lp->n = pipeline();
364 			else
365 				lp->n = command();
366 			prev->next = lp;
367 		} while (readtoken() == TPIPE);
368 		lp->next = NULL;
369 		n1 = pipenode;
370 	}
371 	tokpushback++;
372 	if (negate) {
373 		n2 = (union node *)stalloc(sizeof (struct nnot));
374 		n2->type = NNOT;
375 		n2->nnot.com = n1;
376 		return n2;
377 	} else
378 		return n1;
379 }
380 
381 
382 
383 static union node *
384 command(void)
385 {
386 	union node *n1, *n2;
387 	union node *ap, **app;
388 	union node *cp, **cpp;
389 	union node *redir, **rpp;
390 	int t;
391 	int is_subshell;
392 
393 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
394 	is_subshell = 0;
395 	redir = NULL;
396 	n1 = NULL;
397 	rpp = &redir;
398 
399 	/* Check for redirection which may precede command */
400 	while (readtoken() == TREDIR) {
401 		*rpp = n2 = redirnode;
402 		rpp = &n2->nfile.next;
403 		parsefname();
404 	}
405 	tokpushback++;
406 
407 	switch (readtoken()) {
408 	case TIF:
409 		n1 = (union node *)stalloc(sizeof (struct nif));
410 		n1->type = NIF;
411 		if ((n1->nif.test = list(0, 0)) == NULL)
412 			synexpect(-1);
413 		if (readtoken() != TTHEN)
414 			synexpect(TTHEN);
415 		n1->nif.ifpart = list(0, 0);
416 		n2 = n1;
417 		while (readtoken() == TELIF) {
418 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
419 			n2 = n2->nif.elsepart;
420 			n2->type = NIF;
421 			if ((n2->nif.test = list(0, 0)) == NULL)
422 				synexpect(-1);
423 			if (readtoken() != TTHEN)
424 				synexpect(TTHEN);
425 			n2->nif.ifpart = list(0, 0);
426 		}
427 		if (lasttoken == TELSE)
428 			n2->nif.elsepart = list(0, 0);
429 		else {
430 			n2->nif.elsepart = NULL;
431 			tokpushback++;
432 		}
433 		if (readtoken() != TFI)
434 			synexpect(TFI);
435 		checkkwd = CHKKWD | CHKALIAS;
436 		break;
437 	case TWHILE:
438 	case TUNTIL: {
439 		int got;
440 		n1 = (union node *)stalloc(sizeof (struct nbinary));
441 		n1->type = (lasttoken == TWHILE)? NWHILE : NUNTIL;
442 		if ((n1->nbinary.ch1 = list(0, 0)) == NULL)
443 			synexpect(-1);
444 		if ((got=readtoken()) != TDO) {
445 TRACE(("expecting DO got %s %s\n", tokname[got], got == TWORD ? wordtext : ""));
446 			synexpect(TDO);
447 		}
448 		n1->nbinary.ch2 = list(0, 0);
449 		if (readtoken() != TDONE)
450 			synexpect(TDONE);
451 		checkkwd = CHKKWD | CHKALIAS;
452 		break;
453 	}
454 	case TFOR:
455 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
456 			synerror("Bad for loop variable");
457 		n1 = (union node *)stalloc(sizeof (struct nfor));
458 		n1->type = NFOR;
459 		n1->nfor.var = wordtext;
460 		while (readtoken() == TNL)
461 			;
462 		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
463 			app = &ap;
464 			while (readtoken() == TWORD) {
465 				n2 = (union node *)stalloc(sizeof (struct narg));
466 				n2->type = NARG;
467 				n2->narg.text = wordtext;
468 				n2->narg.backquote = backquotelist;
469 				*app = n2;
470 				app = &n2->narg.next;
471 			}
472 			*app = NULL;
473 			n1->nfor.args = ap;
474 			if (lasttoken != TNL && lasttoken != TSEMI)
475 				synexpect(-1);
476 		} else {
477 			static char argvars[5] = {
478 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
479 			};
480 			n2 = (union node *)stalloc(sizeof (struct narg));
481 			n2->type = NARG;
482 			n2->narg.text = argvars;
483 			n2->narg.backquote = NULL;
484 			n2->narg.next = NULL;
485 			n1->nfor.args = n2;
486 			/*
487 			 * Newline or semicolon here is optional (but note
488 			 * that the original Bourne shell only allowed NL).
489 			 */
490 			if (lasttoken != TNL && lasttoken != TSEMI)
491 				tokpushback++;
492 		}
493 		checkkwd = CHKNL | CHKKWD | CHKALIAS;
494 		if ((t = readtoken()) == TDO)
495 			t = TDONE;
496 		else if (t == TBEGIN)
497 			t = TEND;
498 		else
499 			synexpect(-1);
500 		n1->nfor.body = list(0, 0);
501 		if (readtoken() != t)
502 			synexpect(t);
503 		checkkwd = CHKKWD | CHKALIAS;
504 		break;
505 	case TCASE:
506 		n1 = (union node *)stalloc(sizeof (struct ncase));
507 		n1->type = NCASE;
508 		if (readtoken() != TWORD)
509 			synexpect(TWORD);
510 		n1->ncase.expr = n2 = (union node *)stalloc(sizeof (struct narg));
511 		n2->type = NARG;
512 		n2->narg.text = wordtext;
513 		n2->narg.backquote = backquotelist;
514 		n2->narg.next = NULL;
515 		while (readtoken() == TNL);
516 		if (lasttoken != TWORD || ! equal(wordtext, "in"))
517 			synerror("expecting \"in\"");
518 		cpp = &n1->ncase.cases;
519 		checkkwd = CHKNL | CHKKWD, readtoken();
520 		while (lasttoken != TESAC) {
521 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
522 			cp->type = NCLIST;
523 			app = &cp->nclist.pattern;
524 			if (lasttoken == TLP)
525 				readtoken();
526 			for (;;) {
527 				*app = ap = (union node *)stalloc(sizeof (struct narg));
528 				ap->type = NARG;
529 				ap->narg.text = wordtext;
530 				ap->narg.backquote = backquotelist;
531 				checkkwd = CHKNL | CHKKWD;
532 				if (readtoken() != TPIPE)
533 					break;
534 				app = &ap->narg.next;
535 				readtoken();
536 			}
537 			ap->narg.next = NULL;
538 			if (lasttoken != TRP)
539 				synexpect(TRP);
540 			cp->nclist.body = list(0, 0);
541 
542 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
543 			if ((t = readtoken()) != TESAC) {
544 				if (t != TENDCASE)
545 					synexpect(TENDCASE);
546 				else
547 					checkkwd = CHKNL | CHKKWD, readtoken();
548 			}
549 			cpp = &cp->nclist.next;
550 		}
551 		*cpp = NULL;
552 		checkkwd = CHKKWD | CHKALIAS;
553 		break;
554 	case TLP:
555 		n1 = (union node *)stalloc(sizeof (struct nredir));
556 		n1->type = NSUBSHELL;
557 		n1->nredir.n = list(0, 0);
558 		n1->nredir.redirect = NULL;
559 		if (readtoken() != TRP)
560 			synexpect(TRP);
561 		checkkwd = CHKKWD | CHKALIAS;
562 		is_subshell = 1;
563 		break;
564 	case TBEGIN:
565 		n1 = list(0, 0);
566 		if (readtoken() != TEND)
567 			synexpect(TEND);
568 		checkkwd = CHKKWD | CHKALIAS;
569 		break;
570 	/* Handle an empty command like other simple commands.  */
571 	case TBACKGND:
572 	case TSEMI:
573 	case TAND:
574 	case TOR:
575 		/*
576 		 * An empty command before a ; doesn't make much sense, and
577 		 * should certainly be disallowed in the case of `if ;'.
578 		 */
579 		if (!redir)
580 			synexpect(-1);
581 	case TNL:
582 	case TEOF:
583 	case TWORD:
584 	case TRP:
585 		tokpushback++;
586 		n1 = simplecmd(rpp, redir);
587 		return n1;
588 	default:
589 		synexpect(-1);
590 	}
591 
592 	/* Now check for redirection which may follow command */
593 	while (readtoken() == TREDIR) {
594 		*rpp = n2 = redirnode;
595 		rpp = &n2->nfile.next;
596 		parsefname();
597 	}
598 	tokpushback++;
599 	*rpp = NULL;
600 	if (redir) {
601 		if (!is_subshell) {
602 			n2 = (union node *)stalloc(sizeof (struct nredir));
603 			n2->type = NREDIR;
604 			n2->nredir.n = n1;
605 			n1 = n2;
606 		}
607 		n1->nredir.redirect = redir;
608 	}
609 
610 	return n1;
611 }
612 
613 
614 static union node *
615 simplecmd(union node **rpp, union node *redir)
616 {
617 	union node *args, **app;
618 	union node **orig_rpp = rpp;
619 	union node *n = NULL;
620 	int special;
621 	int savecheckkwd;
622 
623 	/* If we don't have any redirections already, then we must reset */
624 	/* rpp to be the address of the local redir variable.  */
625 	if (redir == 0)
626 		rpp = &redir;
627 
628 	args = NULL;
629 	app = &args;
630 	/*
631 	 * We save the incoming value, because we need this for shell
632 	 * functions.  There can not be a redirect or an argument between
633 	 * the function name and the open parenthesis.
634 	 */
635 	orig_rpp = rpp;
636 
637 	savecheckkwd = CHKALIAS;
638 
639 	for (;;) {
640 		checkkwd = savecheckkwd;
641 		if (readtoken() == TWORD) {
642 			n = (union node *)stalloc(sizeof (struct narg));
643 			n->type = NARG;
644 			n->narg.text = wordtext;
645 			n->narg.backquote = backquotelist;
646 			*app = n;
647 			app = &n->narg.next;
648 			if (savecheckkwd != 0 && !isassignment(wordtext))
649 				savecheckkwd = 0;
650 		} else if (lasttoken == TREDIR) {
651 			*rpp = n = redirnode;
652 			rpp = &n->nfile.next;
653 			parsefname();	/* read name of redirection file */
654 		} else if (lasttoken == TLP && app == &args->narg.next
655 					    && rpp == orig_rpp) {
656 			/* We have a function */
657 			if (readtoken() != TRP)
658 				synexpect(TRP);
659 			funclinno = plinno;
660 			/*
661 			 * - Require plain text.
662 			 * - Functions with '/' cannot be called.
663 			 * - Reject name=().
664 			 * - Reject ksh extended glob patterns.
665 			 */
666 			if (!noexpand(n->narg.text) || quoteflag ||
667 			    strchr(n->narg.text, '/') ||
668 			    strchr("!%*+-=?@}~",
669 				n->narg.text[strlen(n->narg.text) - 1]))
670 				synerror("Bad function name");
671 			rmescapes(n->narg.text);
672 			if (find_builtin(n->narg.text, &special) >= 0 &&
673 			    special)
674 				synerror("Cannot override a special builtin with a function");
675 			n->type = NDEFUN;
676 			n->narg.next = command();
677 			funclinno = 0;
678 			return n;
679 		} else {
680 			tokpushback++;
681 			break;
682 		}
683 	}
684 	*app = NULL;
685 	*rpp = NULL;
686 	n = (union node *)stalloc(sizeof (struct ncmd));
687 	n->type = NCMD;
688 	n->ncmd.backgnd = 0;
689 	n->ncmd.args = args;
690 	n->ncmd.redirect = redir;
691 	return n;
692 }
693 
694 static union node *
695 makename(void)
696 {
697 	union node *n;
698 
699 	n = (union node *)stalloc(sizeof (struct narg));
700 	n->type = NARG;
701 	n->narg.next = NULL;
702 	n->narg.text = wordtext;
703 	n->narg.backquote = backquotelist;
704 	return n;
705 }
706 
707 void
708 fixredir(union node *n, const char *text, int err)
709 {
710 	TRACE(("Fix redir %s %d\n", text, err));
711 	if (!err)
712 		n->ndup.vname = NULL;
713 
714 	if (is_digit(text[0]) && text[1] == '\0')
715 		n->ndup.dupfd = digit_val(text[0]);
716 	else if (text[0] == '-' && text[1] == '\0')
717 		n->ndup.dupfd = -1;
718 	else {
719 
720 		if (err)
721 			synerror("Bad fd number");
722 		else
723 			n->ndup.vname = makename();
724 	}
725 }
726 
727 
728 static void
729 parsefname(void)
730 {
731 	union node *n = redirnode;
732 
733 	if (readtoken() != TWORD)
734 		synexpect(-1);
735 	if (n->type == NHERE) {
736 		struct heredoc *here = heredoc;
737 		struct heredoc *p;
738 		int i;
739 
740 		if (quoteflag == 0)
741 			n->type = NXHERE;
742 		TRACE(("Here document %d\n", n->type));
743 		if (here->striptabs) {
744 			while (*wordtext == '\t')
745 				wordtext++;
746 		}
747 		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
748 			synerror("Illegal eof marker for << redirection");
749 		rmescapes(wordtext);
750 		here->eofmark = wordtext;
751 		here->next = NULL;
752 		if (heredoclist == NULL)
753 			heredoclist = here;
754 		else {
755 			for (p = heredoclist ; p->next ; p = p->next);
756 			p->next = here;
757 		}
758 	} else if (n->type == NTOFD || n->type == NFROMFD) {
759 		fixredir(n, wordtext, 0);
760 	} else {
761 		n->nfile.fname = makename();
762 	}
763 }
764 
765 
766 /*
767  * Input any here documents.
768  */
769 
770 static void
771 parseheredoc(void)
772 {
773 	struct heredoc *here;
774 	union node *n;
775 
776 	while (heredoclist) {
777 		here = heredoclist;
778 		heredoclist = here->next;
779 		if (needprompt) {
780 			setprompt(2);
781 			needprompt = 0;
782 		}
783 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
784 				here->eofmark, here->striptabs);
785 		n = (union node *)stalloc(sizeof (struct narg));
786 		n->narg.type = NARG;
787 		n->narg.next = NULL;
788 		n->narg.text = wordtext;
789 		n->narg.backquote = backquotelist;
790 		here->here->nhere.doc = n;
791 	}
792 }
793 
794 static int
795 peektoken(void)
796 {
797 	int t;
798 
799 	t = readtoken();
800 	tokpushback++;
801 	return (t);
802 }
803 
804 static int
805 readtoken(void)
806 {
807 	int t;
808 	struct alias *ap;
809 #ifdef DEBUG
810 	int alreadyseen = tokpushback;
811 #endif
812 
813 	top:
814 	t = xxreadtoken();
815 
816 	/*
817 	 * eat newlines
818 	 */
819 	if (checkkwd & CHKNL) {
820 		while (t == TNL) {
821 			parseheredoc();
822 			t = xxreadtoken();
823 		}
824 	}
825 
826 	/*
827 	 * check for keywords and aliases
828 	 */
829 	if (t == TWORD && !quoteflag)
830 	{
831 		const char * const *pp;
832 
833 		if (checkkwd & CHKKWD)
834 			for (pp = parsekwd; *pp; pp++) {
835 				if (**pp == *wordtext && equal(*pp, wordtext))
836 				{
837 					lasttoken = t = pp - parsekwd + KWDOFFSET;
838 					TRACE(("keyword %s recognized\n", tokname[t]));
839 					goto out;
840 				}
841 			}
842 		if (checkkwd & CHKALIAS &&
843 		    (ap = lookupalias(wordtext, 1)) != NULL) {
844 			pushstring(ap->val, strlen(ap->val), ap);
845 			goto top;
846 		}
847 	}
848 out:
849 	if (t != TNOT)
850 		checkkwd = 0;
851 
852 #ifdef DEBUG
853 	if (!alreadyseen)
854 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
855 	else
856 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
857 #endif
858 	return (t);
859 }
860 
861 
862 /*
863  * Read the next input token.
864  * If the token is a word, we set backquotelist to the list of cmds in
865  *	backquotes.  We set quoteflag to true if any part of the word was
866  *	quoted.
867  * If the token is TREDIR, then we set redirnode to a structure containing
868  *	the redirection.
869  * In all cases, the variable startlinno is set to the number of the line
870  *	on which the token starts.
871  *
872  * [Change comment:  here documents and internal procedures]
873  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
874  *  word parsing code into a separate routine.  In this case, readtoken
875  *  doesn't need to have any internal procedures, but parseword does.
876  *  We could also make parseoperator in essence the main routine, and
877  *  have parseword (readtoken1?) handle both words and redirection.]
878  */
879 
880 #define RETURN(token)	return lasttoken = token
881 
882 static int
883 xxreadtoken(void)
884 {
885 	int c;
886 
887 	if (tokpushback) {
888 		tokpushback = 0;
889 		return lasttoken;
890 	}
891 	if (needprompt) {
892 		setprompt(2);
893 		needprompt = 0;
894 	}
895 	startlinno = plinno;
896 	for (;;) {	/* until token or start of word found */
897 		c = pgetc_macro();
898 		switch (c) {
899 		case ' ': case '\t':
900 			continue;
901 		case '#':
902 			while ((c = pgetc()) != '\n' && c != PEOF);
903 			pungetc();
904 			continue;
905 		case '\\':
906 			if (pgetc() == '\n') {
907 				startlinno = ++plinno;
908 				if (doprompt)
909 					setprompt(2);
910 				else
911 					setprompt(0);
912 				continue;
913 			}
914 			pungetc();
915 			goto breakloop;
916 		case '\n':
917 			plinno++;
918 			needprompt = doprompt;
919 			RETURN(TNL);
920 		case PEOF:
921 			RETURN(TEOF);
922 		case '&':
923 			if (pgetc() == '&')
924 				RETURN(TAND);
925 			pungetc();
926 			RETURN(TBACKGND);
927 		case '|':
928 			if (pgetc() == '|')
929 				RETURN(TOR);
930 			pungetc();
931 			RETURN(TPIPE);
932 		case ';':
933 			if (pgetc() == ';')
934 				RETURN(TENDCASE);
935 			pungetc();
936 			RETURN(TSEMI);
937 		case '(':
938 			RETURN(TLP);
939 		case ')':
940 			RETURN(TRP);
941 		default:
942 			goto breakloop;
943 		}
944 	}
945 breakloop:
946 	return readtoken1(c, BASESYNTAX, NULL, 0);
947 #undef RETURN
948 }
949 
950 
951 #define MAXNEST_STATIC 8
952 struct tokenstate
953 {
954 	const char *syntax; /* *SYNTAX */
955 	int parenlevel; /* levels of parentheses in arithmetic */
956 	enum tokenstate_category
957 	{
958 		TSTATE_TOP,
959 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
960 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
961 		TSTATE_ARITH
962 	} category;
963 };
964 
965 
966 /*
967  * Called to parse command substitutions.
968  */
969 
970 static char *
971 parsebackq(char *out, struct nodelist **pbqlist,
972     int oldstyle, int dblquote, int quoted)
973 {
974 	struct nodelist **nlpp;
975 	union node *n;
976 	char *volatile str;
977 	struct jmploc jmploc;
978 	struct jmploc *const savehandler = handler;
979 	int savelen;
980 	int saveprompt;
981 	const int bq_startlinno = plinno;
982 	char *volatile ostr = NULL;
983 	struct parsefile *const savetopfile = getcurrentfile();
984 	struct heredoc *const saveheredoclist = heredoclist;
985 	struct heredoc *here;
986 
987 	str = NULL;
988 	if (setjmp(jmploc.loc)) {
989 		popfilesupto(savetopfile);
990 		if (str)
991 			ckfree(str);
992 		if (ostr)
993 			ckfree(ostr);
994 		heredoclist = saveheredoclist;
995 		handler = savehandler;
996 		if (exception == EXERROR) {
997 			startlinno = bq_startlinno;
998 			synerror("Error in command substitution");
999 		}
1000 		longjmp(handler->loc, 1);
1001 	}
1002 	INTOFF;
1003 	savelen = out - stackblock();
1004 	if (savelen > 0) {
1005 		str = ckmalloc(savelen);
1006 		memcpy(str, stackblock(), savelen);
1007 	}
1008 	handler = &jmploc;
1009 	heredoclist = NULL;
1010 	INTON;
1011 	if (oldstyle) {
1012 		/*
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 		 */
1017 		char *oout;
1018 		int c, olen;
1019 
1020 		STARTSTACKSTR(oout);
1021 		for (;;) {
1022 			if (needprompt) {
1023 				setprompt(2);
1024 				needprompt = 0;
1025 			}
1026 			CHECKSTRSPACE(2, oout);
1027 			switch (c = pgetc()) {
1028 			case '`':
1029 				goto done;
1030 
1031 			case '\\':
1032 				if ((c = pgetc()) == '\n') {
1033 					plinno++;
1034 					if (doprompt)
1035 						setprompt(2);
1036 					else
1037 						setprompt(0);
1038 					/*
1039 					 * If eating a newline, avoid putting
1040 					 * the newline into the new character
1041 					 * stream (via the USTPUTC after the
1042 					 * switch).
1043 					 */
1044 					continue;
1045 				}
1046 				if (c != '\\' && c != '`' && c != '$'
1047 				    && (!dblquote || c != '"'))
1048 					USTPUTC('\\', oout);
1049 				break;
1050 
1051 			case '\n':
1052 				plinno++;
1053 				needprompt = doprompt;
1054 				break;
1055 
1056 			case PEOF:
1057 				startlinno = plinno;
1058 				synerror("EOF in backquote substitution");
1059 				break;
1060 
1061 			default:
1062 				break;
1063 			}
1064 			USTPUTC(c, oout);
1065 		}
1066 done:
1067 		USTPUTC('\0', oout);
1068 		olen = oout - stackblock();
1069 		INTOFF;
1070 		ostr = ckmalloc(olen);
1071 		memcpy(ostr, stackblock(), olen);
1072 		setinputstring(ostr, 1);
1073 		INTON;
1074 	}
1075 	nlpp = pbqlist;
1076 	while (*nlpp)
1077 		nlpp = &(*nlpp)->next;
1078 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1079 	(*nlpp)->next = NULL;
1080 
1081 	if (oldstyle) {
1082 		saveprompt = doprompt;
1083 		doprompt = 0;
1084 	}
1085 
1086 	n = list(0, oldstyle);
1087 
1088 	if (oldstyle)
1089 		doprompt = saveprompt;
1090 	else {
1091 		if (readtoken() != TRP)
1092 			synexpect(TRP);
1093 	}
1094 
1095 	(*nlpp)->n = n;
1096 	if (oldstyle) {
1097 		/*
1098 		 * Start reading from old file again, ignoring any pushed back
1099 		 * tokens left from the backquote parsing
1100 		 */
1101 		popfile();
1102 		tokpushback = 0;
1103 	}
1104 	STARTSTACKSTR(out);
1105 	CHECKSTRSPACE(savelen + 1, out);
1106 	INTOFF;
1107 	if (str) {
1108 		memcpy(out, str, savelen);
1109 		STADJUST(savelen, out);
1110 		ckfree(str);
1111 		str = NULL;
1112 	}
1113 	if (ostr) {
1114 		ckfree(ostr);
1115 		ostr = NULL;
1116 	}
1117 	here = saveheredoclist;
1118 	if (here != NULL) {
1119 		while (here->next != NULL)
1120 			here = here->next;
1121 		here->next = heredoclist;
1122 		heredoclist = saveheredoclist;
1123 	}
1124 	handler = savehandler;
1125 	INTON;
1126 	if (quoted)
1127 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1128 	else
1129 		USTPUTC(CTLBACKQ, out);
1130 	return out;
1131 }
1132 
1133 
1134 /*
1135  * Called to parse a backslash escape sequence inside $'...'.
1136  * The backslash has already been read.
1137  */
1138 static char *
1139 readcstyleesc(char *out)
1140 {
1141 	int c, v, i, n;
1142 
1143 	c = pgetc();
1144 	switch (c) {
1145 	case '\0':
1146 		synerror("Unterminated quoted string");
1147 	case '\n':
1148 		plinno++;
1149 		if (doprompt)
1150 			setprompt(2);
1151 		else
1152 			setprompt(0);
1153 		return out;
1154 	case '\\':
1155 	case '\'':
1156 	case '"':
1157 		v = c;
1158 		break;
1159 	case 'a': v = '\a'; break;
1160 	case 'b': v = '\b'; break;
1161 	case 'e': v = '\033'; break;
1162 	case 'f': v = '\f'; break;
1163 	case 'n': v = '\n'; break;
1164 	case 'r': v = '\r'; break;
1165 	case 't': v = '\t'; break;
1166 	case 'v': v = '\v'; break;
1167 	case 'x':
1168 		  v = 0;
1169 		  for (;;) {
1170 			  c = pgetc();
1171 			  if (c >= '0' && c <= '9')
1172 				  v = (v << 4) + c - '0';
1173 			  else if (c >= 'A' && c <= 'F')
1174 				  v = (v << 4) + c - 'A' + 10;
1175 			  else if (c >= 'a' && c <= 'f')
1176 				  v = (v << 4) + c - 'a' + 10;
1177 			  else
1178 				  break;
1179 		  }
1180 		  pungetc();
1181 		  break;
1182 	case '0': case '1': case '2': case '3':
1183 	case '4': case '5': case '6': case '7':
1184 		  v = c - '0';
1185 		  c = pgetc();
1186 		  if (c >= '0' && c <= '7') {
1187 			  v <<= 3;
1188 			  v += c - '0';
1189 			  c = pgetc();
1190 			  if (c >= '0' && c <= '7') {
1191 				  v <<= 3;
1192 				  v += c - '0';
1193 			  } else
1194 				  pungetc();
1195 		  } else
1196 			  pungetc();
1197 		  break;
1198 	case 'c':
1199 		  c = pgetc();
1200 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1201 			  synerror("Bad escape sequence");
1202 		  if (c == '\\' && pgetc() != '\\')
1203 			  synerror("Bad escape sequence");
1204 		  if (c == '?')
1205 			  v = 127;
1206 		  else
1207 			  v = c & 0x1f;
1208 		  break;
1209 	case 'u':
1210 	case 'U':
1211 		  n = c == 'U' ? 8 : 4;
1212 		  v = 0;
1213 		  for (i = 0; i < n; i++) {
1214 			  c = pgetc();
1215 			  if (c >= '0' && c <= '9')
1216 				  v = (v << 4) + c - '0';
1217 			  else if (c >= 'A' && c <= 'F')
1218 				  v = (v << 4) + c - 'A' + 10;
1219 			  else if (c >= 'a' && c <= 'f')
1220 				  v = (v << 4) + c - 'a' + 10;
1221 			  else
1222 				  synerror("Bad escape sequence");
1223 		  }
1224 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1225 			  synerror("Bad escape sequence");
1226 		  /* We really need iconv here. */
1227 		  if (initial_localeisutf8 && v > 127) {
1228 			  CHECKSTRSPACE(4, out);
1229 			  /*
1230 			   * We cannot use wctomb() as the locale may have
1231 			   * changed.
1232 			   */
1233 			  if (v <= 0x7ff) {
1234 				  USTPUTC(0xc0 | v >> 6, out);
1235 				  USTPUTC(0x80 | (v & 0x3f), out);
1236 				  return out;
1237 			  } else if (v <= 0xffff) {
1238 				  USTPUTC(0xe0 | v >> 12, out);
1239 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1240 				  USTPUTC(0x80 | (v & 0x3f), out);
1241 				  return out;
1242 			  } else if (v <= 0x10ffff) {
1243 				  USTPUTC(0xf0 | v >> 18, out);
1244 				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1245 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1246 				  USTPUTC(0x80 | (v & 0x3f), out);
1247 				  return out;
1248 			  }
1249 		  }
1250 		  if (v > 127)
1251 			  v = '?';
1252 		  break;
1253 	default:
1254 		  synerror("Bad escape sequence");
1255 	}
1256 	v = (char)v;
1257 	/*
1258 	 * We can't handle NUL bytes.
1259 	 * POSIX says we should skip till the closing quote.
1260 	 */
1261 	if (v == '\0') {
1262 		while ((c = pgetc()) != '\'') {
1263 			if (c == '\\')
1264 				c = pgetc();
1265 			if (c == PEOF)
1266 				synerror("Unterminated quoted string");
1267 		}
1268 		pungetc();
1269 		return out;
1270 	}
1271 	if (SQSYNTAX[v] == CCTL)
1272 		USTPUTC(CTLESC, out);
1273 	USTPUTC(v, out);
1274 	return out;
1275 }
1276 
1277 
1278 /*
1279  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1280  * is not NULL, read a here document.  In the latter case, eofmark is the
1281  * word which marks the end of the document and striptabs is true if
1282  * leading tabs should be stripped from the document.  The argument firstc
1283  * is the first character of the input token or document.
1284  *
1285  * Because C does not have internal subroutines, I have simulated them
1286  * using goto's to implement the subroutine linkage.  The following macros
1287  * will run code that appears at the end of readtoken1.
1288  */
1289 
1290 #define CHECKEND()	{goto checkend; checkend_return:;}
1291 #define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1292 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1293 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1294 
1295 static int
1296 readtoken1(int firstc, char const *initialsyntax, char *eofmark, int striptabs)
1297 {
1298 	int c = firstc;
1299 	char * volatile out;
1300 	int len;
1301 	char line[EOFMARKLEN + 1];
1302 	struct nodelist *bqlist;
1303 	volatile int quotef;
1304 	int newvarnest;
1305 	int level;
1306 	int synentry;
1307 	struct tokenstate state_static[MAXNEST_STATIC];
1308 	int maxnest = MAXNEST_STATIC;
1309 	struct tokenstate *state = state_static;
1310 	int sqiscstyle = 0;
1311 
1312 	startlinno = plinno;
1313 	quotef = 0;
1314 	bqlist = NULL;
1315 	newvarnest = 0;
1316 	level = 0;
1317 	state[level].syntax = initialsyntax;
1318 	state[level].parenlevel = 0;
1319 	state[level].category = TSTATE_TOP;
1320 
1321 	STARTSTACKSTR(out);
1322 	loop: {	/* for each line, until end of word */
1323 		CHECKEND();	/* set c to PEOF if at end of here document */
1324 		for (;;) {	/* until end of line or end of word */
1325 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1326 
1327 			synentry = state[level].syntax[c];
1328 
1329 			switch(synentry) {
1330 			case CNL:	/* '\n' */
1331 				if (state[level].syntax == BASESYNTAX)
1332 					goto endword;	/* exit outer loop */
1333 				USTPUTC(c, out);
1334 				plinno++;
1335 				if (doprompt)
1336 					setprompt(2);
1337 				else
1338 					setprompt(0);
1339 				c = pgetc();
1340 				goto loop;		/* continue outer loop */
1341 			case CSBACK:
1342 				if (sqiscstyle) {
1343 					out = readcstyleesc(out);
1344 					break;
1345 				}
1346 				/* FALLTHROUGH */
1347 			case CWORD:
1348 				USTPUTC(c, out);
1349 				break;
1350 			case CCTL:
1351 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1352 					USTPUTC(CTLESC, out);
1353 				USTPUTC(c, out);
1354 				break;
1355 			case CBACK:	/* backslash */
1356 				c = pgetc();
1357 				if (c == PEOF) {
1358 					USTPUTC('\\', out);
1359 					pungetc();
1360 				} else if (c == '\n') {
1361 					plinno++;
1362 					if (doprompt)
1363 						setprompt(2);
1364 					else
1365 						setprompt(0);
1366 				} else {
1367 					if (state[level].syntax == DQSYNTAX &&
1368 					    c != '\\' && c != '`' && c != '$' &&
1369 					    (c != '"' || (eofmark != NULL &&
1370 						newvarnest == 0)) &&
1371 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1372 						USTPUTC('\\', out);
1373 					if ((eofmark == NULL ||
1374 					    newvarnest > 0) &&
1375 					    state[level].syntax == BASESYNTAX)
1376 						USTPUTC(CTLQUOTEMARK, out);
1377 					if (SQSYNTAX[c] == CCTL)
1378 						USTPUTC(CTLESC, out);
1379 					USTPUTC(c, out);
1380 					if ((eofmark == NULL ||
1381 					    newvarnest > 0) &&
1382 					    state[level].syntax == BASESYNTAX &&
1383 					    state[level].category == TSTATE_VAR_OLD)
1384 						USTPUTC(CTLQUOTEEND, out);
1385 					quotef++;
1386 				}
1387 				break;
1388 			case CSQUOTE:
1389 				USTPUTC(CTLQUOTEMARK, out);
1390 				state[level].syntax = SQSYNTAX;
1391 				sqiscstyle = 0;
1392 				break;
1393 			case CDQUOTE:
1394 				USTPUTC(CTLQUOTEMARK, out);
1395 				state[level].syntax = DQSYNTAX;
1396 				break;
1397 			case CENDQUOTE:
1398 				if (eofmark != NULL && newvarnest == 0)
1399 					USTPUTC(c, out);
1400 				else {
1401 					if (state[level].category == TSTATE_VAR_OLD)
1402 						USTPUTC(CTLQUOTEEND, out);
1403 					state[level].syntax = BASESYNTAX;
1404 					quotef++;
1405 				}
1406 				break;
1407 			case CVAR:	/* '$' */
1408 				PARSESUB();		/* parse substitution */
1409 				break;
1410 			case CENDVAR:	/* '}' */
1411 				if (level > 0 &&
1412 				    ((state[level].category == TSTATE_VAR_OLD &&
1413 				      state[level].syntax ==
1414 				      state[level - 1].syntax) ||
1415 				    (state[level].category == TSTATE_VAR_NEW &&
1416 				     state[level].syntax == BASESYNTAX))) {
1417 					if (state[level].category == TSTATE_VAR_NEW)
1418 						newvarnest--;
1419 					level--;
1420 					USTPUTC(CTLENDVAR, out);
1421 				} else {
1422 					USTPUTC(c, out);
1423 				}
1424 				break;
1425 			case CLP:	/* '(' in arithmetic */
1426 				state[level].parenlevel++;
1427 				USTPUTC(c, out);
1428 				break;
1429 			case CRP:	/* ')' in arithmetic */
1430 				if (state[level].parenlevel > 0) {
1431 					USTPUTC(c, out);
1432 					--state[level].parenlevel;
1433 				} else {
1434 					if (pgetc() == ')') {
1435 						if (level > 0 &&
1436 						    state[level].category == TSTATE_ARITH) {
1437 							level--;
1438 							USTPUTC(CTLENDARI, out);
1439 						} else
1440 							USTPUTC(')', out);
1441 					} else {
1442 						/*
1443 						 * unbalanced parens
1444 						 *  (don't 2nd guess - no error)
1445 						 */
1446 						pungetc();
1447 						USTPUTC(')', out);
1448 					}
1449 				}
1450 				break;
1451 			case CBQUOTE:	/* '`' */
1452 				out = parsebackq(out, &bqlist, 1,
1453 				    state[level].syntax == DQSYNTAX &&
1454 				    (eofmark == NULL || newvarnest > 0),
1455 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1456 				break;
1457 			case CEOF:
1458 				goto endword;		/* exit outer loop */
1459 			case CIGN:
1460 				break;
1461 			default:
1462 				if (level == 0)
1463 					goto endword;	/* exit outer loop */
1464 				USTPUTC(c, out);
1465 			}
1466 			c = pgetc_macro();
1467 		}
1468 	}
1469 endword:
1470 	if (state[level].syntax == ARISYNTAX)
1471 		synerror("Missing '))'");
1472 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1473 		synerror("Unterminated quoted string");
1474 	if (state[level].category == TSTATE_VAR_OLD ||
1475 	    state[level].category == TSTATE_VAR_NEW) {
1476 		startlinno = plinno;
1477 		synerror("Missing '}'");
1478 	}
1479 	if (state != state_static)
1480 		parser_temp_free_upto(state);
1481 	USTPUTC('\0', out);
1482 	len = out - stackblock();
1483 	out = stackblock();
1484 	if (eofmark == NULL) {
1485 		if ((c == '>' || c == '<')
1486 		 && quotef == 0
1487 		 && len <= 2
1488 		 && (*out == '\0' || is_digit(*out))) {
1489 			PARSEREDIR();
1490 			return lasttoken = TREDIR;
1491 		} else {
1492 			pungetc();
1493 		}
1494 	}
1495 	quoteflag = quotef;
1496 	backquotelist = bqlist;
1497 	grabstackblock(len);
1498 	wordtext = out;
1499 	return lasttoken = TWORD;
1500 /* end of readtoken routine */
1501 
1502 
1503 /*
1504  * Check to see whether we are at the end of the here document.  When this
1505  * is called, c is set to the first character of the next input line.  If
1506  * we are at the end of the here document, this routine sets the c to PEOF.
1507  */
1508 
1509 checkend: {
1510 	if (eofmark) {
1511 		if (striptabs) {
1512 			while (c == '\t')
1513 				c = pgetc();
1514 		}
1515 		if (c == *eofmark) {
1516 			if (pfgets(line, sizeof line) != NULL) {
1517 				char *p, *q;
1518 
1519 				p = line;
1520 				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1521 				if ((*p == '\0' || *p == '\n') && *q == '\0') {
1522 					c = PEOF;
1523 					if (*p == '\n') {
1524 						plinno++;
1525 						needprompt = doprompt;
1526 					}
1527 				} else {
1528 					pushstring(line, strlen(line), NULL);
1529 				}
1530 			}
1531 		}
1532 	}
1533 	goto checkend_return;
1534 }
1535 
1536 
1537 /*
1538  * Parse a redirection operator.  The variable "out" points to a string
1539  * specifying the fd to be redirected.  The variable "c" contains the
1540  * first character of the redirection operator.
1541  */
1542 
1543 parseredir: {
1544 	char fd = *out;
1545 	union node *np;
1546 
1547 	np = (union node *)stalloc(sizeof (struct nfile));
1548 	if (c == '>') {
1549 		np->nfile.fd = 1;
1550 		c = pgetc();
1551 		if (c == '>')
1552 			np->type = NAPPEND;
1553 		else if (c == '&')
1554 			np->type = NTOFD;
1555 		else if (c == '|')
1556 			np->type = NCLOBBER;
1557 		else {
1558 			np->type = NTO;
1559 			pungetc();
1560 		}
1561 	} else {	/* c == '<' */
1562 		np->nfile.fd = 0;
1563 		c = pgetc();
1564 		if (c == '<') {
1565 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1566 				np = (union node *)stalloc(sizeof (struct nhere));
1567 				np->nfile.fd = 0;
1568 			}
1569 			np->type = NHERE;
1570 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1571 			heredoc->here = np;
1572 			if ((c = pgetc()) == '-') {
1573 				heredoc->striptabs = 1;
1574 			} else {
1575 				heredoc->striptabs = 0;
1576 				pungetc();
1577 			}
1578 		} else if (c == '&')
1579 			np->type = NFROMFD;
1580 		else if (c == '>')
1581 			np->type = NFROMTO;
1582 		else {
1583 			np->type = NFROM;
1584 			pungetc();
1585 		}
1586 	}
1587 	if (fd != '\0')
1588 		np->nfile.fd = digit_val(fd);
1589 	redirnode = np;
1590 	goto parseredir_return;
1591 }
1592 
1593 
1594 /*
1595  * Parse a substitution.  At this point, we have read the dollar sign
1596  * and nothing else.
1597  */
1598 
1599 parsesub: {
1600 	char buf[10];
1601 	int subtype;
1602 	int typeloc;
1603 	int flags;
1604 	char *p;
1605 	static const char types[] = "}-+?=";
1606 	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1607 	int linno;
1608 	int length;
1609 	int c1;
1610 
1611 	c = pgetc();
1612 	if (c == '(') {	/* $(command) or $((arith)) */
1613 		if (pgetc() == '(') {
1614 			PARSEARITH();
1615 		} else {
1616 			pungetc();
1617 			out = parsebackq(out, &bqlist, 0,
1618 			    state[level].syntax == DQSYNTAX &&
1619 			    (eofmark == NULL || newvarnest > 0),
1620 			    state[level].syntax == DQSYNTAX ||
1621 			    state[level].syntax == ARISYNTAX);
1622 		}
1623 	} else if (c == '{' || is_name(c) || is_special(c)) {
1624 		USTPUTC(CTLVAR, out);
1625 		typeloc = out - stackblock();
1626 		USTPUTC(VSNORMAL, out);
1627 		subtype = VSNORMAL;
1628 		flags = 0;
1629 		if (c == '{') {
1630 			bracketed_name = 1;
1631 			c = pgetc();
1632 			subtype = 0;
1633 		}
1634 varname:
1635 		if (!is_eof(c) && is_name(c)) {
1636 			length = 0;
1637 			do {
1638 				STPUTC(c, out);
1639 				c = pgetc();
1640 				length++;
1641 			} while (!is_eof(c) && is_in_name(c));
1642 			if (length == 6 &&
1643 			    strncmp(out - length, "LINENO", length) == 0) {
1644 				/* Replace the variable name with the
1645 				 * current line number. */
1646 				linno = plinno;
1647 				if (funclinno != 0)
1648 					linno -= funclinno - 1;
1649 				snprintf(buf, sizeof(buf), "%d", linno);
1650 				STADJUST(-6, out);
1651 				STPUTS(buf, out);
1652 				flags |= VSLINENO;
1653 			}
1654 		} else if (is_digit(c)) {
1655 			if (bracketed_name) {
1656 				do {
1657 					STPUTC(c, out);
1658 					c = pgetc();
1659 				} while (is_digit(c));
1660 			} else {
1661 				STPUTC(c, out);
1662 				c = pgetc();
1663 			}
1664 		} else if (is_special(c)) {
1665 			c1 = c;
1666 			c = pgetc();
1667 			if (subtype == 0 && c1 == '#') {
1668 				subtype = VSLENGTH;
1669 				if (strchr(types, c) == NULL && c != ':' &&
1670 				    c != '#' && c != '%')
1671 					goto varname;
1672 				c1 = c;
1673 				c = pgetc();
1674 				if (c1 != '}' && c == '}') {
1675 					pungetc();
1676 					c = c1;
1677 					goto varname;
1678 				}
1679 				pungetc();
1680 				c = c1;
1681 				c1 = '#';
1682 				subtype = 0;
1683 			}
1684 			USTPUTC(c1, out);
1685 		} else {
1686 			subtype = VSERROR;
1687 			if (c == '}')
1688 				pungetc();
1689 			else if (c == '\n' || c == PEOF)
1690 				synerror("Unexpected end of line in substitution");
1691 			else
1692 				USTPUTC(c, out);
1693 		}
1694 		if (subtype == 0) {
1695 			switch (c) {
1696 			case ':':
1697 				flags |= VSNUL;
1698 				c = pgetc();
1699 				/*FALLTHROUGH*/
1700 			default:
1701 				p = strchr(types, c);
1702 				if (p == NULL) {
1703 					if (c == '\n' || c == PEOF)
1704 						synerror("Unexpected end of line in substitution");
1705 					if (flags == VSNUL)
1706 						STPUTC(':', out);
1707 					STPUTC(c, out);
1708 					subtype = VSERROR;
1709 				} else
1710 					subtype = p - types + VSNORMAL;
1711 				break;
1712 			case '%':
1713 			case '#':
1714 				{
1715 					int cc = c;
1716 					subtype = c == '#' ? VSTRIMLEFT :
1717 							     VSTRIMRIGHT;
1718 					c = pgetc();
1719 					if (c == cc)
1720 						subtype++;
1721 					else
1722 						pungetc();
1723 					break;
1724 				}
1725 			}
1726 		} else if (subtype != VSERROR) {
1727 			if (subtype == VSLENGTH && c != '}')
1728 				subtype = VSERROR;
1729 			pungetc();
1730 		}
1731 		STPUTC('=', out);
1732 		if (state[level].syntax == DQSYNTAX ||
1733 		    state[level].syntax == ARISYNTAX)
1734 			flags |= VSQUOTE;
1735 		*(stackblock() + typeloc) = subtype | flags;
1736 		if (subtype != VSNORMAL) {
1737 			if (level + 1 >= maxnest) {
1738 				maxnest *= 2;
1739 				if (state == state_static) {
1740 					state = parser_temp_alloc(
1741 					    maxnest * sizeof(*state));
1742 					memcpy(state, state_static,
1743 					    MAXNEST_STATIC * sizeof(*state));
1744 				} else
1745 					state = parser_temp_realloc(state,
1746 					    maxnest * sizeof(*state));
1747 			}
1748 			level++;
1749 			state[level].parenlevel = 0;
1750 			if (subtype == VSMINUS || subtype == VSPLUS ||
1751 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1752 				/*
1753 				 * For operators that were in the Bourne shell,
1754 				 * inherit the double-quote state.
1755 				 */
1756 				state[level].syntax = state[level - 1].syntax;
1757 				state[level].category = TSTATE_VAR_OLD;
1758 			} else {
1759 				/*
1760 				 * The other operators take a pattern,
1761 				 * so go to BASESYNTAX.
1762 				 * Also, ' and " are now special, even
1763 				 * in here documents.
1764 				 */
1765 				state[level].syntax = BASESYNTAX;
1766 				state[level].category = TSTATE_VAR_NEW;
1767 				newvarnest++;
1768 			}
1769 		}
1770 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1771 		/* $'cstylequotes' */
1772 		USTPUTC(CTLQUOTEMARK, out);
1773 		state[level].syntax = SQSYNTAX;
1774 		sqiscstyle = 1;
1775 	} else {
1776 		USTPUTC('$', out);
1777 		pungetc();
1778 	}
1779 	goto parsesub_return;
1780 }
1781 
1782 
1783 /*
1784  * Parse an arithmetic expansion (indicate start of one and set state)
1785  */
1786 parsearith: {
1787 
1788 	if (level + 1 >= maxnest) {
1789 		maxnest *= 2;
1790 		if (state == state_static) {
1791 			state = parser_temp_alloc(
1792 			    maxnest * sizeof(*state));
1793 			memcpy(state, state_static,
1794 			    MAXNEST_STATIC * sizeof(*state));
1795 		} else
1796 			state = parser_temp_realloc(state,
1797 			    maxnest * sizeof(*state));
1798 	}
1799 	level++;
1800 	state[level].syntax = ARISYNTAX;
1801 	state[level].parenlevel = 0;
1802 	state[level].category = TSTATE_ARITH;
1803 	USTPUTC(CTLARI, out);
1804 	if (state[level - 1].syntax == DQSYNTAX)
1805 		USTPUTC('"',out);
1806 	else
1807 		USTPUTC(' ',out);
1808 	goto parsearith_return;
1809 }
1810 
1811 } /* end of readtoken */
1812 
1813 
1814 
1815 #ifdef mkinit
1816 RESET {
1817 	tokpushback = 0;
1818 	checkkwd = 0;
1819 }
1820 #endif
1821 
1822 /*
1823  * Returns true if the text contains nothing to expand (no dollar signs
1824  * or backquotes).
1825  */
1826 
1827 static int
1828 noexpand(char *text)
1829 {
1830 	char *p;
1831 	char c;
1832 
1833 	p = text;
1834 	while ((c = *p++) != '\0') {
1835 		if ( c == CTLQUOTEMARK)
1836 			continue;
1837 		if (c == CTLESC)
1838 			p++;
1839 		else if (BASESYNTAX[(int)c] == CCTL)
1840 			return 0;
1841 	}
1842 	return 1;
1843 }
1844 
1845 
1846 /*
1847  * Return true if the argument is a legal variable name (a letter or
1848  * underscore followed by zero or more letters, underscores, and digits).
1849  */
1850 
1851 int
1852 goodname(const char *name)
1853 {
1854 	const char *p;
1855 
1856 	p = name;
1857 	if (! is_name(*p))
1858 		return 0;
1859 	while (*++p) {
1860 		if (! is_in_name(*p))
1861 			return 0;
1862 	}
1863 	return 1;
1864 }
1865 
1866 
1867 int
1868 isassignment(const char *p)
1869 {
1870 	if (!is_name(*p))
1871 		return 0;
1872 	p++;
1873 	for (;;) {
1874 		if (*p == '=')
1875 			return 1;
1876 		else if (!is_in_name(*p))
1877 			return 0;
1878 		p++;
1879 	}
1880 }
1881 
1882 
1883 /*
1884  * Called when an unexpected token is read during the parse.  The argument
1885  * is the token that is expected, or -1 if more than one type of token can
1886  * occur at this point.
1887  */
1888 
1889 static void
1890 synexpect(int token)
1891 {
1892 	char msg[64];
1893 
1894 	if (token >= 0) {
1895 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1896 			tokname[lasttoken], tokname[token]);
1897 	} else {
1898 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1899 	}
1900 	synerror(msg);
1901 }
1902 
1903 
1904 static void
1905 synerror(const char *msg)
1906 {
1907 	if (commandname)
1908 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1909 	outfmt(out2, "Syntax error: %s\n", msg);
1910 	error(NULL);
1911 }
1912 
1913 static void
1914 setprompt(int which)
1915 {
1916 	whichprompt = which;
1917 
1918 #ifndef NO_HISTORY
1919 	if (!el)
1920 #endif
1921 	{
1922 		out2str(getprompt(NULL));
1923 		flushout(out2);
1924 	}
1925 }
1926 
1927 /*
1928  * called by editline -- any expansions to the prompt
1929  *    should be added here.
1930  */
1931 const char *
1932 getprompt(void *unused __unused)
1933 {
1934 	static char ps[PROMPTLEN];
1935 	const char *fmt;
1936 	const char *pwd;
1937 	int i, trim;
1938 
1939 	/*
1940 	 * Select prompt format.
1941 	 */
1942 	switch (whichprompt) {
1943 	case 0:
1944 		fmt = "";
1945 		break;
1946 	case 1:
1947 		fmt = ps1val();
1948 		break;
1949 	case 2:
1950 		fmt = ps2val();
1951 		break;
1952 	default:
1953 		return "??";
1954 	}
1955 
1956 	/*
1957 	 * Format prompt string.
1958 	 */
1959 	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1960 		if (*fmt == '\\')
1961 			switch (*++fmt) {
1962 
1963 				/*
1964 				 * Hostname.
1965 				 *
1966 				 * \h specifies just the local hostname,
1967 				 * \H specifies fully-qualified hostname.
1968 				 */
1969 			case 'h':
1970 			case 'H':
1971 				ps[i] = '\0';
1972 				gethostname(&ps[i], PROMPTLEN - i);
1973 				/* Skip to end of hostname. */
1974 				trim = (*fmt == 'h') ? '.' : '\0';
1975 				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1976 					i++;
1977 				break;
1978 
1979 				/*
1980 				 * Working directory.
1981 				 *
1982 				 * \W specifies just the final component,
1983 				 * \w specifies the entire path.
1984 				 */
1985 			case 'W':
1986 			case 'w':
1987 				pwd = lookupvar("PWD");
1988 				if (pwd == NULL)
1989 					pwd = "?";
1990 				if (*fmt == 'W' &&
1991 				    *pwd == '/' && pwd[1] != '\0')
1992 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
1993 					    PROMPTLEN - i);
1994 				else
1995 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
1996 				/* Skip to end of path. */
1997 				while (ps[i + 1] != '\0')
1998 					i++;
1999 				break;
2000 
2001 				/*
2002 				 * Superuser status.
2003 				 *
2004 				 * '$' for normal users, '#' for root.
2005 				 */
2006 			case '$':
2007 				ps[i] = (geteuid() != 0) ? '$' : '#';
2008 				break;
2009 
2010 				/*
2011 				 * A literal \.
2012 				 */
2013 			case '\\':
2014 				ps[i] = '\\';
2015 				break;
2016 
2017 				/*
2018 				 * Emit unrecognized formats verbatim.
2019 				 */
2020 			default:
2021 				ps[i++] = '\\';
2022 				ps[i] = *fmt;
2023 				break;
2024 			}
2025 		else
2026 			ps[i] = *fmt;
2027 	ps[i] = '\0';
2028 	return (ps);
2029 }
2030