xref: /dflybsd-src/bin/sh/parser.c (revision e1489450c38a053dad66768fc12706aa2968f646)
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 1.109 2011/05/05 20:55:55 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 
622 	/* If we don't have any redirections already, then we must reset */
623 	/* rpp to be the address of the local redir variable.  */
624 	if (redir == 0)
625 		rpp = &redir;
626 
627 	args = NULL;
628 	app = &args;
629 	/*
630 	 * We save the incoming value, because we need this for shell
631 	 * functions.  There can not be a redirect or an argument between
632 	 * the function name and the open parenthesis.
633 	 */
634 	orig_rpp = rpp;
635 
636 	for (;;) {
637 		if (readtoken() == TWORD) {
638 			n = (union node *)stalloc(sizeof (struct narg));
639 			n->type = NARG;
640 			n->narg.text = wordtext;
641 			n->narg.backquote = backquotelist;
642 			*app = n;
643 			app = &n->narg.next;
644 		} else if (lasttoken == TREDIR) {
645 			*rpp = n = redirnode;
646 			rpp = &n->nfile.next;
647 			parsefname();	/* read name of redirection file */
648 		} else if (lasttoken == TLP && app == &args->narg.next
649 					    && rpp == orig_rpp) {
650 			/* We have a function */
651 			if (readtoken() != TRP)
652 				synexpect(TRP);
653 			funclinno = plinno;
654 			/*
655 			 * - Require plain text.
656 			 * - Functions with '/' cannot be called.
657 			 * - Reject name=().
658 			 * - Reject ksh extended glob patterns.
659 			 */
660 			if (!noexpand(n->narg.text) || quoteflag ||
661 			    strchr(n->narg.text, '/') ||
662 			    strchr("!%*+-=?@}~",
663 				n->narg.text[strlen(n->narg.text) - 1]))
664 				synerror("Bad function name");
665 			rmescapes(n->narg.text);
666 			if (find_builtin(n->narg.text, &special) >= 0 &&
667 			    special)
668 				synerror("Cannot override a special builtin with a function");
669 			n->type = NDEFUN;
670 			n->narg.next = command();
671 			funclinno = 0;
672 			return n;
673 		} else {
674 			tokpushback++;
675 			break;
676 		}
677 	}
678 	*app = NULL;
679 	*rpp = NULL;
680 	n = (union node *)stalloc(sizeof (struct ncmd));
681 	n->type = NCMD;
682 	n->ncmd.backgnd = 0;
683 	n->ncmd.args = args;
684 	n->ncmd.redirect = redir;
685 	return n;
686 }
687 
688 static union node *
689 makename(void)
690 {
691 	union node *n;
692 
693 	n = (union node *)stalloc(sizeof (struct narg));
694 	n->type = NARG;
695 	n->narg.next = NULL;
696 	n->narg.text = wordtext;
697 	n->narg.backquote = backquotelist;
698 	return n;
699 }
700 
701 void
702 fixredir(union node *n, const char *text, int err)
703 {
704 	TRACE(("Fix redir %s %d\n", text, err));
705 	if (!err)
706 		n->ndup.vname = NULL;
707 
708 	if (is_digit(text[0]) && text[1] == '\0')
709 		n->ndup.dupfd = digit_val(text[0]);
710 	else if (text[0] == '-' && text[1] == '\0')
711 		n->ndup.dupfd = -1;
712 	else {
713 
714 		if (err)
715 			synerror("Bad fd number");
716 		else
717 			n->ndup.vname = makename();
718 	}
719 }
720 
721 
722 static void
723 parsefname(void)
724 {
725 	union node *n = redirnode;
726 
727 	if (readtoken() != TWORD)
728 		synexpect(-1);
729 	if (n->type == NHERE) {
730 		struct heredoc *here = heredoc;
731 		struct heredoc *p;
732 		int i;
733 
734 		if (quoteflag == 0)
735 			n->type = NXHERE;
736 		TRACE(("Here document %d\n", n->type));
737 		if (here->striptabs) {
738 			while (*wordtext == '\t')
739 				wordtext++;
740 		}
741 		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
742 			synerror("Illegal eof marker for << redirection");
743 		rmescapes(wordtext);
744 		here->eofmark = wordtext;
745 		here->next = NULL;
746 		if (heredoclist == NULL)
747 			heredoclist = here;
748 		else {
749 			for (p = heredoclist ; p->next ; p = p->next);
750 			p->next = here;
751 		}
752 	} else if (n->type == NTOFD || n->type == NFROMFD) {
753 		fixredir(n, wordtext, 0);
754 	} else {
755 		n->nfile.fname = makename();
756 	}
757 }
758 
759 
760 /*
761  * Input any here documents.
762  */
763 
764 static void
765 parseheredoc(void)
766 {
767 	struct heredoc *here;
768 	union node *n;
769 
770 	while (heredoclist) {
771 		here = heredoclist;
772 		heredoclist = here->next;
773 		if (needprompt) {
774 			setprompt(2);
775 			needprompt = 0;
776 		}
777 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
778 				here->eofmark, here->striptabs);
779 		n = (union node *)stalloc(sizeof (struct narg));
780 		n->narg.type = NARG;
781 		n->narg.next = NULL;
782 		n->narg.text = wordtext;
783 		n->narg.backquote = backquotelist;
784 		here->here->nhere.doc = n;
785 	}
786 }
787 
788 static int
789 peektoken(void)
790 {
791 	int t;
792 
793 	t = readtoken();
794 	tokpushback++;
795 	return (t);
796 }
797 
798 static int
799 readtoken(void)
800 {
801 	int t;
802 	struct alias *ap;
803 #ifdef DEBUG
804 	int alreadyseen = tokpushback;
805 #endif
806 
807 	top:
808 	t = xxreadtoken();
809 
810 	/*
811 	 * eat newlines
812 	 */
813 	if (checkkwd & CHKNL) {
814 		while (t == TNL) {
815 			parseheredoc();
816 			t = xxreadtoken();
817 		}
818 	}
819 
820 	/*
821 	 * check for keywords and aliases
822 	 */
823 	if (t == TWORD && !quoteflag)
824 	{
825 		const char * const *pp;
826 
827 		if (checkkwd & CHKKWD)
828 			for (pp = parsekwd; *pp; pp++) {
829 				if (**pp == *wordtext && equal(*pp, wordtext))
830 				{
831 					lasttoken = t = pp - parsekwd + KWDOFFSET;
832 					TRACE(("keyword %s recognized\n", tokname[t]));
833 					goto out;
834 				}
835 			}
836 		if (checkkwd & CHKALIAS &&
837 		    (ap = lookupalias(wordtext, 1)) != NULL) {
838 			pushstring(ap->val, strlen(ap->val), ap);
839 			goto top;
840 		}
841 	}
842 out:
843 	if (t != TNOT)
844 		checkkwd = 0;
845 
846 #ifdef DEBUG
847 	if (!alreadyseen)
848 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
849 	else
850 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
851 #endif
852 	return (t);
853 }
854 
855 
856 /*
857  * Read the next input token.
858  * If the token is a word, we set backquotelist to the list of cmds in
859  *	backquotes.  We set quoteflag to true if any part of the word was
860  *	quoted.
861  * If the token is TREDIR, then we set redirnode to a structure containing
862  *	the redirection.
863  * In all cases, the variable startlinno is set to the number of the line
864  *	on which the token starts.
865  *
866  * [Change comment:  here documents and internal procedures]
867  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
868  *  word parsing code into a separate routine.  In this case, readtoken
869  *  doesn't need to have any internal procedures, but parseword does.
870  *  We could also make parseoperator in essence the main routine, and
871  *  have parseword (readtoken1?) handle both words and redirection.]
872  */
873 
874 #define RETURN(token)	return lasttoken = token
875 
876 static int
877 xxreadtoken(void)
878 {
879 	int c;
880 
881 	if (tokpushback) {
882 		tokpushback = 0;
883 		return lasttoken;
884 	}
885 	if (needprompt) {
886 		setprompt(2);
887 		needprompt = 0;
888 	}
889 	startlinno = plinno;
890 	for (;;) {	/* until token or start of word found */
891 		c = pgetc_macro();
892 		switch (c) {
893 		case ' ': case '\t':
894 			continue;
895 		case '#':
896 			while ((c = pgetc()) != '\n' && c != PEOF);
897 			pungetc();
898 			continue;
899 		case '\\':
900 			if (pgetc() == '\n') {
901 				startlinno = ++plinno;
902 				if (doprompt)
903 					setprompt(2);
904 				else
905 					setprompt(0);
906 				continue;
907 			}
908 			pungetc();
909 			goto breakloop;
910 		case '\n':
911 			plinno++;
912 			needprompt = doprompt;
913 			RETURN(TNL);
914 		case PEOF:
915 			RETURN(TEOF);
916 		case '&':
917 			if (pgetc() == '&')
918 				RETURN(TAND);
919 			pungetc();
920 			RETURN(TBACKGND);
921 		case '|':
922 			if (pgetc() == '|')
923 				RETURN(TOR);
924 			pungetc();
925 			RETURN(TPIPE);
926 		case ';':
927 			if (pgetc() == ';')
928 				RETURN(TENDCASE);
929 			pungetc();
930 			RETURN(TSEMI);
931 		case '(':
932 			RETURN(TLP);
933 		case ')':
934 			RETURN(TRP);
935 		default:
936 			goto breakloop;
937 		}
938 	}
939 breakloop:
940 	return readtoken1(c, BASESYNTAX, NULL, 0);
941 #undef RETURN
942 }
943 
944 
945 #define MAXNEST_STATIC 8
946 struct tokenstate
947 {
948 	const char *syntax; /* *SYNTAX */
949 	int parenlevel; /* levels of parentheses in arithmetic */
950 	enum tokenstate_category
951 	{
952 		TSTATE_TOP,
953 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
954 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
955 		TSTATE_ARITH
956 	} category;
957 };
958 
959 
960 /*
961  * Called to parse command substitutions.
962  */
963 
964 static char *
965 parsebackq(char *out, struct nodelist **pbqlist,
966     int oldstyle, int dblquote, int quoted)
967 {
968 	struct nodelist **nlpp;
969 	union node *n;
970 	char *volatile str;
971 	struct jmploc jmploc;
972 	struct jmploc *const savehandler = handler;
973 	int savelen;
974 	int saveprompt;
975 	const int bq_startlinno = plinno;
976 	char *volatile ostr = NULL;
977 	struct parsefile *const savetopfile = getcurrentfile();
978 	struct heredoc *const saveheredoclist = heredoclist;
979 	struct heredoc *here;
980 
981 	str = NULL;
982 	if (setjmp(jmploc.loc)) {
983 		popfilesupto(savetopfile);
984 		if (str)
985 			ckfree(str);
986 		if (ostr)
987 			ckfree(ostr);
988 		heredoclist = saveheredoclist;
989 		handler = savehandler;
990 		if (exception == EXERROR) {
991 			startlinno = bq_startlinno;
992 			synerror("Error in command substitution");
993 		}
994 		longjmp(handler->loc, 1);
995 	}
996 	INTOFF;
997 	savelen = out - stackblock();
998 	if (savelen > 0) {
999 		str = ckmalloc(savelen);
1000 		memcpy(str, stackblock(), savelen);
1001 	}
1002 	handler = &jmploc;
1003 	heredoclist = NULL;
1004 	INTON;
1005 	if (oldstyle) {
1006 		/*
1007 		 * We must read until the closing backquote, giving special
1008 		 * treatment to some slashes, and then push the string and
1009 		 * reread it as input, interpreting it normally.
1010 		 */
1011 		char *oout;
1012 		int c, olen;
1013 
1014 		STARTSTACKSTR(oout);
1015 		for (;;) {
1016 			if (needprompt) {
1017 				setprompt(2);
1018 				needprompt = 0;
1019 			}
1020 			CHECKSTRSPACE(2, oout);
1021 			switch (c = pgetc()) {
1022 			case '`':
1023 				goto done;
1024 
1025 			case '\\':
1026 				if ((c = pgetc()) == '\n') {
1027 					plinno++;
1028 					if (doprompt)
1029 						setprompt(2);
1030 					else
1031 						setprompt(0);
1032 					/*
1033 					 * If eating a newline, avoid putting
1034 					 * the newline into the new character
1035 					 * stream (via the USTPUTC after the
1036 					 * switch).
1037 					 */
1038 					continue;
1039 				}
1040 				if (c != '\\' && c != '`' && c != '$'
1041 				    && (!dblquote || c != '"'))
1042 					USTPUTC('\\', oout);
1043 				break;
1044 
1045 			case '\n':
1046 				plinno++;
1047 				needprompt = doprompt;
1048 				break;
1049 
1050 			case PEOF:
1051 				startlinno = plinno;
1052 				synerror("EOF in backquote substitution");
1053 				break;
1054 
1055 			default:
1056 				break;
1057 			}
1058 			USTPUTC(c, oout);
1059 		}
1060 done:
1061 		USTPUTC('\0', oout);
1062 		olen = oout - stackblock();
1063 		INTOFF;
1064 		ostr = ckmalloc(olen);
1065 		memcpy(ostr, stackblock(), olen);
1066 		setinputstring(ostr, 1);
1067 		INTON;
1068 	}
1069 	nlpp = pbqlist;
1070 	while (*nlpp)
1071 		nlpp = &(*nlpp)->next;
1072 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1073 	(*nlpp)->next = NULL;
1074 
1075 	if (oldstyle) {
1076 		saveprompt = doprompt;
1077 		doprompt = 0;
1078 	}
1079 
1080 	n = list(0, oldstyle);
1081 
1082 	if (oldstyle)
1083 		doprompt = saveprompt;
1084 	else {
1085 		if (readtoken() != TRP)
1086 			synexpect(TRP);
1087 	}
1088 
1089 	(*nlpp)->n = n;
1090 	if (oldstyle) {
1091 		/*
1092 		 * Start reading from old file again, ignoring any pushed back
1093 		 * tokens left from the backquote parsing
1094 		 */
1095 		popfile();
1096 		tokpushback = 0;
1097 	}
1098 	STARTSTACKSTR(out);
1099 	CHECKSTRSPACE(savelen + 1, out);
1100 	INTOFF;
1101 	if (str) {
1102 		memcpy(out, str, savelen);
1103 		STADJUST(savelen, out);
1104 		ckfree(str);
1105 		str = NULL;
1106 	}
1107 	if (ostr) {
1108 		ckfree(ostr);
1109 		ostr = NULL;
1110 	}
1111 	here = saveheredoclist;
1112 	if (here != NULL) {
1113 		while (here->next != NULL)
1114 			here = here->next;
1115 		here->next = heredoclist;
1116 		heredoclist = saveheredoclist;
1117 	}
1118 	handler = savehandler;
1119 	INTON;
1120 	if (quoted)
1121 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1122 	else
1123 		USTPUTC(CTLBACKQ, out);
1124 	return out;
1125 }
1126 
1127 
1128 /*
1129  * Called to parse a backslash escape sequence inside $'...'.
1130  * The backslash has already been read.
1131  */
1132 static char *
1133 readcstyleesc(char *out)
1134 {
1135 	int c, v, i, n;
1136 
1137 	c = pgetc();
1138 	switch (c) {
1139 	case '\0':
1140 		synerror("Unterminated quoted string");
1141 	case '\n':
1142 		plinno++;
1143 		if (doprompt)
1144 			setprompt(2);
1145 		else
1146 			setprompt(0);
1147 		return out;
1148 	case '\\':
1149 	case '\'':
1150 	case '"':
1151 		v = c;
1152 		break;
1153 	case 'a': v = '\a'; break;
1154 	case 'b': v = '\b'; break;
1155 	case 'e': v = '\033'; break;
1156 	case 'f': v = '\f'; break;
1157 	case 'n': v = '\n'; break;
1158 	case 'r': v = '\r'; break;
1159 	case 't': v = '\t'; break;
1160 	case 'v': v = '\v'; break;
1161 	case 'x':
1162 		  v = 0;
1163 		  for (;;) {
1164 			  c = pgetc();
1165 			  if (c >= '0' && c <= '9')
1166 				  v = (v << 4) + c - '0';
1167 			  else if (c >= 'A' && c <= 'F')
1168 				  v = (v << 4) + c - 'A' + 10;
1169 			  else if (c >= 'a' && c <= 'f')
1170 				  v = (v << 4) + c - 'a' + 10;
1171 			  else
1172 				  break;
1173 		  }
1174 		  pungetc();
1175 		  break;
1176 	case '0': case '1': case '2': case '3':
1177 	case '4': case '5': case '6': case '7':
1178 		  v = c - '0';
1179 		  c = pgetc();
1180 		  if (c >= '0' && c <= '7') {
1181 			  v <<= 3;
1182 			  v += c - '0';
1183 			  c = pgetc();
1184 			  if (c >= '0' && c <= '7') {
1185 				  v <<= 3;
1186 				  v += c - '0';
1187 			  } else
1188 				  pungetc();
1189 		  } else
1190 			  pungetc();
1191 		  break;
1192 	case 'c':
1193 		  c = pgetc();
1194 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1195 			  synerror("Bad escape sequence");
1196 		  if (c == '\\' && pgetc() != '\\')
1197 			  synerror("Bad escape sequence");
1198 		  if (c == '?')
1199 			  v = 127;
1200 		  else
1201 			  v = c & 0x1f;
1202 		  break;
1203 	case 'u':
1204 	case 'U':
1205 		  n = c == 'U' ? 8 : 4;
1206 		  v = 0;
1207 		  for (i = 0; i < n; i++) {
1208 			  c = pgetc();
1209 			  if (c >= '0' && c <= '9')
1210 				  v = (v << 4) + c - '0';
1211 			  else if (c >= 'A' && c <= 'F')
1212 				  v = (v << 4) + c - 'A' + 10;
1213 			  else if (c >= 'a' && c <= 'f')
1214 				  v = (v << 4) + c - 'a' + 10;
1215 			  else
1216 				  synerror("Bad escape sequence");
1217 		  }
1218 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1219 			  synerror("Bad escape sequence");
1220 		  /* We really need iconv here. */
1221 		  if (v > 127)
1222 			  v = '?';
1223 		  break;
1224 	default:
1225 		  synerror("Bad escape sequence");
1226 	}
1227 	v = (char)v;
1228 	/*
1229 	 * We can't handle NUL bytes.
1230 	 * POSIX says we should skip till the closing quote.
1231 	 */
1232 	if (v == '\0') {
1233 		while ((c = pgetc()) != '\'') {
1234 			if (c == '\\')
1235 				c = pgetc();
1236 			if (c == PEOF)
1237 				synerror("Unterminated quoted string");
1238 		}
1239 		pungetc();
1240 		return out;
1241 	}
1242 	if (SQSYNTAX[v] == CCTL)
1243 		USTPUTC(CTLESC, out);
1244 	USTPUTC(v, out);
1245 	return out;
1246 }
1247 
1248 
1249 /*
1250  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1251  * is not NULL, read a here document.  In the latter case, eofmark is the
1252  * word which marks the end of the document and striptabs is true if
1253  * leading tabs should be stripped from the document.  The argument firstc
1254  * is the first character of the input token or document.
1255  *
1256  * Because C does not have internal subroutines, I have simulated them
1257  * using goto's to implement the subroutine linkage.  The following macros
1258  * will run code that appears at the end of readtoken1.
1259  */
1260 
1261 #define CHECKEND()	{goto checkend; checkend_return:;}
1262 #define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1263 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1264 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1265 
1266 static int
1267 readtoken1(int firstc, char const *initialsyntax, char *eofmark, int striptabs)
1268 {
1269 	int c = firstc;
1270 	char * volatile out;
1271 	int len;
1272 	char line[EOFMARKLEN + 1];
1273 	struct nodelist *bqlist;
1274 	volatile int quotef;
1275 	int newvarnest;
1276 	int level;
1277 	int synentry;
1278 	struct tokenstate state_static[MAXNEST_STATIC];
1279 	int maxnest = MAXNEST_STATIC;
1280 	struct tokenstate *state = state_static;
1281 	int sqiscstyle = 0;
1282 
1283 	startlinno = plinno;
1284 	quotef = 0;
1285 	bqlist = NULL;
1286 	newvarnest = 0;
1287 	level = 0;
1288 	state[level].syntax = initialsyntax;
1289 	state[level].parenlevel = 0;
1290 	state[level].category = TSTATE_TOP;
1291 
1292 	STARTSTACKSTR(out);
1293 	loop: {	/* for each line, until end of word */
1294 		CHECKEND();	/* set c to PEOF if at end of here document */
1295 		for (;;) {	/* until end of line or end of word */
1296 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1297 
1298 			synentry = state[level].syntax[c];
1299 
1300 			switch(synentry) {
1301 			case CNL:	/* '\n' */
1302 				if (state[level].syntax == BASESYNTAX)
1303 					goto endword;	/* exit outer loop */
1304 				USTPUTC(c, out);
1305 				plinno++;
1306 				if (doprompt)
1307 					setprompt(2);
1308 				else
1309 					setprompt(0);
1310 				c = pgetc();
1311 				goto loop;		/* continue outer loop */
1312 			case CSBACK:
1313 				if (sqiscstyle) {
1314 					out = readcstyleesc(out);
1315 					break;
1316 				}
1317 				/* FALLTHROUGH */
1318 			case CWORD:
1319 				USTPUTC(c, out);
1320 				break;
1321 			case CCTL:
1322 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1323 					USTPUTC(CTLESC, out);
1324 				USTPUTC(c, out);
1325 				break;
1326 			case CBACK:	/* backslash */
1327 				c = pgetc();
1328 				if (c == PEOF) {
1329 					USTPUTC('\\', out);
1330 					pungetc();
1331 				} else if (c == '\n') {
1332 					plinno++;
1333 					if (doprompt)
1334 						setprompt(2);
1335 					else
1336 						setprompt(0);
1337 				} else {
1338 					if (state[level].syntax == DQSYNTAX &&
1339 					    c != '\\' && c != '`' && c != '$' &&
1340 					    (c != '"' || (eofmark != NULL &&
1341 						newvarnest == 0)) &&
1342 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1343 						USTPUTC('\\', out);
1344 					if ((eofmark == NULL ||
1345 					    newvarnest > 0) &&
1346 					    state[level].syntax == BASESYNTAX)
1347 						USTPUTC(CTLQUOTEMARK, out);
1348 					if (SQSYNTAX[c] == CCTL)
1349 						USTPUTC(CTLESC, out);
1350 					USTPUTC(c, out);
1351 					if ((eofmark == NULL ||
1352 					    newvarnest > 0) &&
1353 					    state[level].syntax == BASESYNTAX &&
1354 					    state[level].category == TSTATE_VAR_OLD)
1355 						USTPUTC(CTLQUOTEEND, out);
1356 					quotef++;
1357 				}
1358 				break;
1359 			case CSQUOTE:
1360 				USTPUTC(CTLQUOTEMARK, out);
1361 				state[level].syntax = SQSYNTAX;
1362 				sqiscstyle = 0;
1363 				break;
1364 			case CDQUOTE:
1365 				USTPUTC(CTLQUOTEMARK, out);
1366 				state[level].syntax = DQSYNTAX;
1367 				break;
1368 			case CENDQUOTE:
1369 				if (eofmark != NULL && newvarnest == 0)
1370 					USTPUTC(c, out);
1371 				else {
1372 					if (state[level].category == TSTATE_VAR_OLD)
1373 						USTPUTC(CTLQUOTEEND, out);
1374 					state[level].syntax = BASESYNTAX;
1375 					quotef++;
1376 				}
1377 				break;
1378 			case CVAR:	/* '$' */
1379 				PARSESUB();		/* parse substitution */
1380 				break;
1381 			case CENDVAR:	/* '}' */
1382 				if (level > 0 &&
1383 				    ((state[level].category == TSTATE_VAR_OLD &&
1384 				      state[level].syntax ==
1385 				      state[level - 1].syntax) ||
1386 				    (state[level].category == TSTATE_VAR_NEW &&
1387 				     state[level].syntax == BASESYNTAX))) {
1388 					if (state[level].category == TSTATE_VAR_NEW)
1389 						newvarnest--;
1390 					level--;
1391 					USTPUTC(CTLENDVAR, out);
1392 				} else {
1393 					USTPUTC(c, out);
1394 				}
1395 				break;
1396 			case CLP:	/* '(' in arithmetic */
1397 				state[level].parenlevel++;
1398 				USTPUTC(c, out);
1399 				break;
1400 			case CRP:	/* ')' in arithmetic */
1401 				if (state[level].parenlevel > 0) {
1402 					USTPUTC(c, out);
1403 					--state[level].parenlevel;
1404 				} else {
1405 					if (pgetc() == ')') {
1406 						if (level > 0 &&
1407 						    state[level].category == TSTATE_ARITH) {
1408 							level--;
1409 							USTPUTC(CTLENDARI, out);
1410 						} else
1411 							USTPUTC(')', out);
1412 					} else {
1413 						/*
1414 						 * unbalanced parens
1415 						 *  (don't 2nd guess - no error)
1416 						 */
1417 						pungetc();
1418 						USTPUTC(')', out);
1419 					}
1420 				}
1421 				break;
1422 			case CBQUOTE:	/* '`' */
1423 				out = parsebackq(out, &bqlist, 1,
1424 				    state[level].syntax == DQSYNTAX &&
1425 				    (eofmark == NULL || newvarnest > 0),
1426 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1427 				break;
1428 			case CEOF:
1429 				goto endword;		/* exit outer loop */
1430 			case CIGN:
1431 				break;
1432 			default:
1433 				if (level == 0)
1434 					goto endword;	/* exit outer loop */
1435 				USTPUTC(c, out);
1436 			}
1437 			c = pgetc_macro();
1438 		}
1439 	}
1440 endword:
1441 	if (state[level].syntax == ARISYNTAX)
1442 		synerror("Missing '))'");
1443 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1444 		synerror("Unterminated quoted string");
1445 	if (state[level].category == TSTATE_VAR_OLD ||
1446 	    state[level].category == TSTATE_VAR_NEW) {
1447 		startlinno = plinno;
1448 		synerror("Missing '}'");
1449 	}
1450 	if (state != state_static)
1451 		parser_temp_free_upto(state);
1452 	USTPUTC('\0', out);
1453 	len = out - stackblock();
1454 	out = stackblock();
1455 	if (eofmark == NULL) {
1456 		if ((c == '>' || c == '<')
1457 		 && quotef == 0
1458 		 && len <= 2
1459 		 && (*out == '\0' || is_digit(*out))) {
1460 			PARSEREDIR();
1461 			return lasttoken = TREDIR;
1462 		} else {
1463 			pungetc();
1464 		}
1465 	}
1466 	quoteflag = quotef;
1467 	backquotelist = bqlist;
1468 	grabstackblock(len);
1469 	wordtext = out;
1470 	return lasttoken = TWORD;
1471 /* end of readtoken routine */
1472 
1473 
1474 /*
1475  * Check to see whether we are at the end of the here document.  When this
1476  * is called, c is set to the first character of the next input line.  If
1477  * we are at the end of the here document, this routine sets the c to PEOF.
1478  */
1479 
1480 checkend: {
1481 	if (eofmark) {
1482 		if (striptabs) {
1483 			while (c == '\t')
1484 				c = pgetc();
1485 		}
1486 		if (c == *eofmark) {
1487 			if (pfgets(line, sizeof line) != NULL) {
1488 				char *p, *q;
1489 
1490 				p = line;
1491 				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1492 				if (*p == '\n' && *q == '\0') {
1493 					c = PEOF;
1494 					plinno++;
1495 					needprompt = doprompt;
1496 				} else {
1497 					pushstring(line, strlen(line), NULL);
1498 				}
1499 			}
1500 		}
1501 	}
1502 	goto checkend_return;
1503 }
1504 
1505 
1506 /*
1507  * Parse a redirection operator.  The variable "out" points to a string
1508  * specifying the fd to be redirected.  The variable "c" contains the
1509  * first character of the redirection operator.
1510  */
1511 
1512 parseredir: {
1513 	char fd = *out;
1514 	union node *np;
1515 
1516 	np = (union node *)stalloc(sizeof (struct nfile));
1517 	if (c == '>') {
1518 		np->nfile.fd = 1;
1519 		c = pgetc();
1520 		if (c == '>')
1521 			np->type = NAPPEND;
1522 		else if (c == '&')
1523 			np->type = NTOFD;
1524 		else if (c == '|')
1525 			np->type = NCLOBBER;
1526 		else {
1527 			np->type = NTO;
1528 			pungetc();
1529 		}
1530 	} else {	/* c == '<' */
1531 		np->nfile.fd = 0;
1532 		c = pgetc();
1533 		if (c == '<') {
1534 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1535 				np = (union node *)stalloc(sizeof (struct nhere));
1536 				np->nfile.fd = 0;
1537 			}
1538 			np->type = NHERE;
1539 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1540 			heredoc->here = np;
1541 			if ((c = pgetc()) == '-') {
1542 				heredoc->striptabs = 1;
1543 			} else {
1544 				heredoc->striptabs = 0;
1545 				pungetc();
1546 			}
1547 		} else if (c == '&')
1548 			np->type = NFROMFD;
1549 		else if (c == '>')
1550 			np->type = NFROMTO;
1551 		else {
1552 			np->type = NFROM;
1553 			pungetc();
1554 		}
1555 	}
1556 	if (fd != '\0')
1557 		np->nfile.fd = digit_val(fd);
1558 	redirnode = np;
1559 	goto parseredir_return;
1560 }
1561 
1562 
1563 /*
1564  * Parse a substitution.  At this point, we have read the dollar sign
1565  * and nothing else.
1566  */
1567 
1568 parsesub: {
1569 	char buf[10];
1570 	int subtype;
1571 	int typeloc;
1572 	int flags;
1573 	char *p;
1574 	static const char types[] = "}-+?=";
1575 	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1576 	int linno;
1577 	int length;
1578 	int c1;
1579 
1580 	c = pgetc();
1581 	if (c == '(') {	/* $(command) or $((arith)) */
1582 		if (pgetc() == '(') {
1583 			PARSEARITH();
1584 		} else {
1585 			pungetc();
1586 			out = parsebackq(out, &bqlist, 0,
1587 			    state[level].syntax == DQSYNTAX &&
1588 			    (eofmark == NULL || newvarnest > 0),
1589 			    state[level].syntax == DQSYNTAX ||
1590 			    state[level].syntax == ARISYNTAX);
1591 		}
1592 	} else if (c == '{' || is_name(c) || is_special(c)) {
1593 		USTPUTC(CTLVAR, out);
1594 		typeloc = out - stackblock();
1595 		USTPUTC(VSNORMAL, out);
1596 		subtype = VSNORMAL;
1597 		flags = 0;
1598 		if (c == '{') {
1599 			bracketed_name = 1;
1600 			c = pgetc();
1601 			subtype = 0;
1602 		}
1603 varname:
1604 		if (!is_eof(c) && is_name(c)) {
1605 			length = 0;
1606 			do {
1607 				STPUTC(c, out);
1608 				c = pgetc();
1609 				length++;
1610 			} while (!is_eof(c) && is_in_name(c));
1611 			if (length == 6 &&
1612 			    strncmp(out - length, "LINENO", length) == 0) {
1613 				/* Replace the variable name with the
1614 				 * current line number. */
1615 				linno = plinno;
1616 				if (funclinno != 0)
1617 					linno -= funclinno - 1;
1618 				snprintf(buf, sizeof(buf), "%d", linno);
1619 				STADJUST(-6, out);
1620 				STPUTS(buf, out);
1621 				flags |= VSLINENO;
1622 			}
1623 		} else if (is_digit(c)) {
1624 			if (bracketed_name) {
1625 				do {
1626 					STPUTC(c, out);
1627 					c = pgetc();
1628 				} while (is_digit(c));
1629 			} else {
1630 				STPUTC(c, out);
1631 				c = pgetc();
1632 			}
1633 		} else if (is_special(c)) {
1634 			c1 = c;
1635 			c = pgetc();
1636 			if (subtype == 0 && c1 == '#') {
1637 				subtype = VSLENGTH;
1638 				if (strchr(types, c) == NULL && c != ':' &&
1639 				    c != '#' && c != '%')
1640 					goto varname;
1641 				c1 = c;
1642 				c = pgetc();
1643 				if (c1 != '}' && c == '}') {
1644 					pungetc();
1645 					c = c1;
1646 					goto varname;
1647 				}
1648 				pungetc();
1649 				c = c1;
1650 				c1 = '#';
1651 				subtype = 0;
1652 			}
1653 			USTPUTC(c1, out);
1654 		} else {
1655 			subtype = VSERROR;
1656 			if (c == '}')
1657 				pungetc();
1658 			else if (c == '\n' || c == PEOF)
1659 				synerror("Unexpected end of line in substitution");
1660 			else
1661 				USTPUTC(c, out);
1662 		}
1663 		if (subtype == 0) {
1664 			switch (c) {
1665 			case ':':
1666 				flags |= VSNUL;
1667 				c = pgetc();
1668 				/*FALLTHROUGH*/
1669 			default:
1670 				p = strchr(types, c);
1671 				if (p == NULL) {
1672 					if (c == '\n' || c == PEOF)
1673 						synerror("Unexpected end of line in substitution");
1674 					if (flags == VSNUL)
1675 						STPUTC(':', out);
1676 					STPUTC(c, out);
1677 					subtype = VSERROR;
1678 				} else
1679 					subtype = p - types + VSNORMAL;
1680 				break;
1681 			case '%':
1682 			case '#':
1683 				{
1684 					int cc = c;
1685 					subtype = c == '#' ? VSTRIMLEFT :
1686 							     VSTRIMRIGHT;
1687 					c = pgetc();
1688 					if (c == cc)
1689 						subtype++;
1690 					else
1691 						pungetc();
1692 					break;
1693 				}
1694 			}
1695 		} else if (subtype != VSERROR) {
1696 			if (subtype == VSLENGTH && c != '}')
1697 				subtype = VSERROR;
1698 			pungetc();
1699 		}
1700 		STPUTC('=', out);
1701 		if (state[level].syntax == DQSYNTAX ||
1702 		    state[level].syntax == ARISYNTAX)
1703 			flags |= VSQUOTE;
1704 		*(stackblock() + typeloc) = subtype | flags;
1705 		if (subtype != VSNORMAL) {
1706 			if (level + 1 >= maxnest) {
1707 				maxnest *= 2;
1708 				if (state == state_static) {
1709 					state = parser_temp_alloc(
1710 					    maxnest * sizeof(*state));
1711 					memcpy(state, state_static,
1712 					    MAXNEST_STATIC * sizeof(*state));
1713 				} else
1714 					state = parser_temp_realloc(state,
1715 					    maxnest * sizeof(*state));
1716 			}
1717 			level++;
1718 			state[level].parenlevel = 0;
1719 			if (subtype == VSMINUS || subtype == VSPLUS ||
1720 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1721 				/*
1722 				 * For operators that were in the Bourne shell,
1723 				 * inherit the double-quote state.
1724 				 */
1725 				state[level].syntax = state[level - 1].syntax;
1726 				state[level].category = TSTATE_VAR_OLD;
1727 			} else {
1728 				/*
1729 				 * The other operators take a pattern,
1730 				 * so go to BASESYNTAX.
1731 				 * Also, ' and " are now special, even
1732 				 * in here documents.
1733 				 */
1734 				state[level].syntax = BASESYNTAX;
1735 				state[level].category = TSTATE_VAR_NEW;
1736 				newvarnest++;
1737 			}
1738 		}
1739 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1740 		/* $'cstylequotes' */
1741 		USTPUTC(CTLQUOTEMARK, out);
1742 		state[level].syntax = SQSYNTAX;
1743 		sqiscstyle = 1;
1744 	} else {
1745 		USTPUTC('$', out);
1746 		pungetc();
1747 	}
1748 	goto parsesub_return;
1749 }
1750 
1751 
1752 /*
1753  * Parse an arithmetic expansion (indicate start of one and set state)
1754  */
1755 parsearith: {
1756 
1757 	if (level + 1 >= maxnest) {
1758 		maxnest *= 2;
1759 		if (state == state_static) {
1760 			state = parser_temp_alloc(
1761 			    maxnest * sizeof(*state));
1762 			memcpy(state, state_static,
1763 			    MAXNEST_STATIC * sizeof(*state));
1764 		} else
1765 			state = parser_temp_realloc(state,
1766 			    maxnest * sizeof(*state));
1767 	}
1768 	level++;
1769 	state[level].syntax = ARISYNTAX;
1770 	state[level].parenlevel = 0;
1771 	state[level].category = TSTATE_ARITH;
1772 	USTPUTC(CTLARI, out);
1773 	if (state[level - 1].syntax == DQSYNTAX)
1774 		USTPUTC('"',out);
1775 	else
1776 		USTPUTC(' ',out);
1777 	goto parsearith_return;
1778 }
1779 
1780 } /* end of readtoken */
1781 
1782 
1783 
1784 #ifdef mkinit
1785 RESET {
1786 	tokpushback = 0;
1787 	checkkwd = 0;
1788 }
1789 #endif
1790 
1791 /*
1792  * Returns true if the text contains nothing to expand (no dollar signs
1793  * or backquotes).
1794  */
1795 
1796 static int
1797 noexpand(char *text)
1798 {
1799 	char *p;
1800 	char c;
1801 
1802 	p = text;
1803 	while ((c = *p++) != '\0') {
1804 		if ( c == CTLQUOTEMARK)
1805 			continue;
1806 		if (c == CTLESC)
1807 			p++;
1808 		else if (BASESYNTAX[(int)c] == CCTL)
1809 			return 0;
1810 	}
1811 	return 1;
1812 }
1813 
1814 
1815 /*
1816  * Return true if the argument is a legal variable name (a letter or
1817  * underscore followed by zero or more letters, underscores, and digits).
1818  */
1819 
1820 int
1821 goodname(const char *name)
1822 {
1823 	const char *p;
1824 
1825 	p = name;
1826 	if (! is_name(*p))
1827 		return 0;
1828 	while (*++p) {
1829 		if (! is_in_name(*p))
1830 			return 0;
1831 	}
1832 	return 1;
1833 }
1834 
1835 
1836 /*
1837  * Called when an unexpected token is read during the parse.  The argument
1838  * is the token that is expected, or -1 if more than one type of token can
1839  * occur at this point.
1840  */
1841 
1842 static void
1843 synexpect(int token)
1844 {
1845 	char msg[64];
1846 
1847 	if (token >= 0) {
1848 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1849 			tokname[lasttoken], tokname[token]);
1850 	} else {
1851 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1852 	}
1853 	synerror(msg);
1854 }
1855 
1856 
1857 static void
1858 synerror(const char *msg)
1859 {
1860 	if (commandname)
1861 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1862 	outfmt(out2, "Syntax error: %s\n", msg);
1863 	error(NULL);
1864 }
1865 
1866 static void
1867 setprompt(int which)
1868 {
1869 	whichprompt = which;
1870 
1871 #ifndef NO_HISTORY
1872 	if (!el)
1873 #endif
1874 	{
1875 		out2str(getprompt(NULL));
1876 		flushout(out2);
1877 	}
1878 }
1879 
1880 /*
1881  * called by editline -- any expansions to the prompt
1882  *    should be added here.
1883  */
1884 const char *
1885 getprompt(void *unused __unused)
1886 {
1887 	static char ps[PROMPTLEN];
1888 	const char *fmt;
1889 	const char *pwd;
1890 	int i, trim;
1891 
1892 	/*
1893 	 * Select prompt format.
1894 	 */
1895 	switch (whichprompt) {
1896 	case 0:
1897 		fmt = "";
1898 		break;
1899 	case 1:
1900 		fmt = ps1val();
1901 		break;
1902 	case 2:
1903 		fmt = ps2val();
1904 		break;
1905 	default:
1906 		return "??";
1907 	}
1908 
1909 	/*
1910 	 * Format prompt string.
1911 	 */
1912 	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1913 		if (*fmt == '\\')
1914 			switch (*++fmt) {
1915 
1916 				/*
1917 				 * Hostname.
1918 				 *
1919 				 * \h specifies just the local hostname,
1920 				 * \H specifies fully-qualified hostname.
1921 				 */
1922 			case 'h':
1923 			case 'H':
1924 				ps[i] = '\0';
1925 				gethostname(&ps[i], PROMPTLEN - i);
1926 				/* Skip to end of hostname. */
1927 				trim = (*fmt == 'h') ? '.' : '\0';
1928 				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1929 					i++;
1930 				break;
1931 
1932 				/*
1933 				 * Working directory.
1934 				 *
1935 				 * \W specifies just the final component,
1936 				 * \w specifies the entire path.
1937 				 */
1938 			case 'W':
1939 			case 'w':
1940 				pwd = lookupvar("PWD");
1941 				if (pwd == NULL)
1942 					pwd = "?";
1943 				if (*fmt == 'W' &&
1944 				    *pwd == '/' && pwd[1] != '\0')
1945 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
1946 					    PROMPTLEN - i);
1947 				else
1948 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
1949 				/* Skip to end of path. */
1950 				while (ps[i + 1] != '\0')
1951 					i++;
1952 				break;
1953 
1954 				/*
1955 				 * Superuser status.
1956 				 *
1957 				 * '$' for normal users, '#' for root.
1958 				 */
1959 			case '$':
1960 				ps[i] = (geteuid() != 0) ? '$' : '#';
1961 				break;
1962 
1963 				/*
1964 				 * A literal \.
1965 				 */
1966 			case '\\':
1967 				ps[i] = '\\';
1968 				break;
1969 
1970 				/*
1971 				 * Emit unrecognized formats verbatim.
1972 				 */
1973 			default:
1974 				ps[i++] = '\\';
1975 				ps[i] = *fmt;
1976 				break;
1977 			}
1978 		else
1979 			ps[i] = *fmt;
1980 	ps[i] = '\0';
1981 	return (ps);
1982 }
1983