xref: /freebsd/bin/sh/parser.c (revision 9a14aa017b21c292740c00ee098195cd46642730)
1 /*-
2  * Copyright (c) 1991, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Kenneth Almquist.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 4. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #ifndef lint
34 #if 0
35 static char sccsid[] = "@(#)parser.c	8.7 (Berkeley) 5/16/95";
36 #endif
37 #endif /* not lint */
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
40 
41 #include <stdlib.h>
42 #include <unistd.h>
43 #include <stdio.h>
44 
45 #include "shell.h"
46 #include "parser.h"
47 #include "nodes.h"
48 #include "expand.h"	/* defines rmescapes() */
49 #include "syntax.h"
50 #include "options.h"
51 #include "input.h"
52 #include "output.h"
53 #include "var.h"
54 #include "error.h"
55 #include "memalloc.h"
56 #include "mystring.h"
57 #include "alias.h"
58 #include "show.h"
59 #include "eval.h"
60 #include "exec.h"	/* to check for special builtins */
61 #ifndef NO_HISTORY
62 #include "myhistedit.h"
63 #endif
64 
65 /*
66  * Shell command parser.
67  */
68 
69 #define	EOFMARKLEN	79
70 #define	PROMPTLEN	128
71 
72 /* values of checkkwd variable */
73 #define CHKALIAS	0x1
74 #define CHKKWD		0x2
75 #define CHKNL		0x4
76 
77 /* values returned by readtoken */
78 #include "token.h"
79 
80 
81 
82 struct heredoc {
83 	struct heredoc *next;	/* next here document in list */
84 	union node *here;		/* redirection node */
85 	char *eofmark;		/* string indicating end of input */
86 	int striptabs;		/* if set, strip leading tabs */
87 };
88 
89 struct parser_temp {
90 	struct parser_temp *next;
91 	void *data;
92 };
93 
94 
95 static struct heredoc *heredoclist;	/* list of here documents to read */
96 static int doprompt;		/* if set, prompt the user */
97 static int needprompt;		/* true if interactive and at start of line */
98 static int lasttoken;		/* last token read */
99 MKINIT int tokpushback;		/* last token pushed back */
100 static char *wordtext;		/* text of last word returned by readtoken */
101 MKINIT int checkkwd;            /* 1 == check for kwds, 2 == also eat newlines */
102 static struct nodelist *backquotelist;
103 static union node *redirnode;
104 static struct heredoc *heredoc;
105 static int quoteflag;		/* set if (part of) last token was quoted */
106 static int startlinno;		/* line # where last token started */
107 static int funclinno;		/* line # where the current function started */
108 static struct parser_temp *parser_temp;
109 
110 
111 static union node *list(int, int);
112 static union node *andor(void);
113 static union node *pipeline(void);
114 static union node *command(void);
115 static union node *simplecmd(union node **, union node *);
116 static union node *makename(void);
117 static void parsefname(void);
118 static void parseheredoc(void);
119 static int peektoken(void);
120 static int readtoken(void);
121 static int xxreadtoken(void);
122 static int readtoken1(int, char const *, char *, int);
123 static int noexpand(char *);
124 static void synexpect(int) __dead2;
125 static void synerror(const char *) __dead2;
126 static void setprompt(int);
127 
128 
129 static void *
130 parser_temp_alloc(size_t len)
131 {
132 	struct parser_temp *t;
133 
134 	INTOFF;
135 	t = ckmalloc(sizeof(*t));
136 	t->data = NULL;
137 	t->next = parser_temp;
138 	parser_temp = t;
139 	t->data = ckmalloc(len);
140 	INTON;
141 	return t->data;
142 }
143 
144 
145 static void *
146 parser_temp_realloc(void *ptr, size_t len)
147 {
148 	struct parser_temp *t;
149 
150 	INTOFF;
151 	t = parser_temp;
152 	if (ptr != t->data)
153 		error("bug: parser_temp_realloc misused");
154 	t->data = ckrealloc(t->data, len);
155 	INTON;
156 	return t->data;
157 }
158 
159 
160 static void
161 parser_temp_free_upto(void *ptr)
162 {
163 	struct parser_temp *t;
164 	int done = 0;
165 
166 	INTOFF;
167 	while (parser_temp != NULL && !done) {
168 		t = parser_temp;
169 		parser_temp = t->next;
170 		done = t->data == ptr;
171 		ckfree(t->data);
172 		ckfree(t);
173 	}
174 	INTON;
175 	if (!done)
176 		error("bug: parser_temp_free_upto misused");
177 }
178 
179 
180 static void
181 parser_temp_free_all(void)
182 {
183 	struct parser_temp *t;
184 
185 	INTOFF;
186 	while (parser_temp != NULL) {
187 		t = parser_temp;
188 		parser_temp = t->next;
189 		ckfree(t->data);
190 		ckfree(t);
191 	}
192 	INTON;
193 }
194 
195 
196 /*
197  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
198  * valid parse tree indicating a blank line.)
199  */
200 
201 union node *
202 parsecmd(int interact)
203 {
204 	int t;
205 
206 	/* This assumes the parser is not re-entered,
207 	 * which could happen if we add command substitution on PS1/PS2.
208 	 */
209 	parser_temp_free_all();
210 	heredoclist = NULL;
211 
212 	tokpushback = 0;
213 	doprompt = interact;
214 	if (doprompt)
215 		setprompt(1);
216 	else
217 		setprompt(0);
218 	needprompt = 0;
219 	t = readtoken();
220 	if (t == TEOF)
221 		return NEOF;
222 	if (t == TNL)
223 		return NULL;
224 	tokpushback++;
225 	return list(1, 1);
226 }
227 
228 
229 static union node *
230 list(int nlflag, int erflag)
231 {
232 	union node *ntop, *n1, *n2, *n3;
233 	int tok;
234 
235 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
236 	if (!nlflag && !erflag && tokendlist[peektoken()])
237 		return NULL;
238 	ntop = n1 = NULL;
239 	for (;;) {
240 		n2 = andor();
241 		tok = readtoken();
242 		if (tok == TBACKGND) {
243 			if (n2->type == NPIPE) {
244 				n2->npipe.backgnd = 1;
245 			} else if (n2->type == NREDIR) {
246 				n2->type = NBACKGND;
247 			} else {
248 				n3 = (union node *)stalloc(sizeof (struct nredir));
249 				n3->type = NBACKGND;
250 				n3->nredir.n = n2;
251 				n3->nredir.redirect = NULL;
252 				n2 = n3;
253 			}
254 		}
255 		if (ntop == NULL)
256 			ntop = n2;
257 		else if (n1 == NULL) {
258 			n1 = (union node *)stalloc(sizeof (struct nbinary));
259 			n1->type = NSEMI;
260 			n1->nbinary.ch1 = ntop;
261 			n1->nbinary.ch2 = n2;
262 			ntop = n1;
263 		}
264 		else {
265 			n3 = (union node *)stalloc(sizeof (struct nbinary));
266 			n3->type = NSEMI;
267 			n3->nbinary.ch1 = n1->nbinary.ch2;
268 			n3->nbinary.ch2 = n2;
269 			n1->nbinary.ch2 = n3;
270 			n1 = n3;
271 		}
272 		switch (tok) {
273 		case TBACKGND:
274 		case TSEMI:
275 			tok = readtoken();
276 			/* FALLTHROUGH */
277 		case TNL:
278 			if (tok == TNL) {
279 				parseheredoc();
280 				if (nlflag)
281 					return ntop;
282 			} else if (tok == TEOF && nlflag) {
283 				parseheredoc();
284 				return ntop;
285 			} else {
286 				tokpushback++;
287 			}
288 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
289 			if (!nlflag && !erflag && tokendlist[peektoken()])
290 				return ntop;
291 			break;
292 		case TEOF:
293 			if (heredoclist)
294 				parseheredoc();
295 			else
296 				pungetc();		/* push back EOF on input */
297 			return ntop;
298 		default:
299 			if (nlflag || erflag)
300 				synexpect(-1);
301 			tokpushback++;
302 			return ntop;
303 		}
304 	}
305 }
306 
307 
308 
309 static union node *
310 andor(void)
311 {
312 	union node *n1, *n2, *n3;
313 	int t;
314 
315 	n1 = pipeline();
316 	for (;;) {
317 		if ((t = readtoken()) == TAND) {
318 			t = NAND;
319 		} else if (t == TOR) {
320 			t = NOR;
321 		} else {
322 			tokpushback++;
323 			return n1;
324 		}
325 		n2 = pipeline();
326 		n3 = (union node *)stalloc(sizeof (struct nbinary));
327 		n3->type = t;
328 		n3->nbinary.ch1 = n1;
329 		n3->nbinary.ch2 = n2;
330 		n1 = n3;
331 	}
332 }
333 
334 
335 
336 static union node *
337 pipeline(void)
338 {
339 	union node *n1, *n2, *pipenode;
340 	struct nodelist *lp, *prev;
341 	int negate, t;
342 
343 	negate = 0;
344 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
345 	TRACE(("pipeline: entered\n"));
346 	while (readtoken() == TNOT)
347 		negate = !negate;
348 	tokpushback++;
349 	n1 = command();
350 	if (readtoken() == TPIPE) {
351 		pipenode = (union node *)stalloc(sizeof (struct npipe));
352 		pipenode->type = NPIPE;
353 		pipenode->npipe.backgnd = 0;
354 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
355 		pipenode->npipe.cmdlist = lp;
356 		lp->n = n1;
357 		do {
358 			prev = lp;
359 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
360 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
361 			t = readtoken();
362 			tokpushback++;
363 			if (t == TNOT)
364 				lp->n = pipeline();
365 			else
366 				lp->n = command();
367 			prev->next = lp;
368 		} while (readtoken() == TPIPE);
369 		lp->next = NULL;
370 		n1 = pipenode;
371 	}
372 	tokpushback++;
373 	if (negate) {
374 		n2 = (union node *)stalloc(sizeof (struct nnot));
375 		n2->type = NNOT;
376 		n2->nnot.com = n1;
377 		return n2;
378 	} else
379 		return n1;
380 }
381 
382 
383 
384 static union node *
385 command(void)
386 {
387 	union node *n1, *n2;
388 	union node *ap, **app;
389 	union node *cp, **cpp;
390 	union node *redir, **rpp;
391 	int t;
392 	int is_subshell;
393 
394 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
395 	is_subshell = 0;
396 	redir = NULL;
397 	n1 = NULL;
398 	rpp = &redir;
399 
400 	/* Check for redirection which may precede command */
401 	while (readtoken() == TREDIR) {
402 		*rpp = n2 = redirnode;
403 		rpp = &n2->nfile.next;
404 		parsefname();
405 	}
406 	tokpushback++;
407 
408 	switch (readtoken()) {
409 	case TIF:
410 		n1 = (union node *)stalloc(sizeof (struct nif));
411 		n1->type = NIF;
412 		if ((n1->nif.test = list(0, 0)) == NULL)
413 			synexpect(-1);
414 		if (readtoken() != TTHEN)
415 			synexpect(TTHEN);
416 		n1->nif.ifpart = list(0, 0);
417 		n2 = n1;
418 		while (readtoken() == TELIF) {
419 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
420 			n2 = n2->nif.elsepart;
421 			n2->type = NIF;
422 			if ((n2->nif.test = list(0, 0)) == NULL)
423 				synexpect(-1);
424 			if (readtoken() != TTHEN)
425 				synexpect(TTHEN);
426 			n2->nif.ifpart = list(0, 0);
427 		}
428 		if (lasttoken == TELSE)
429 			n2->nif.elsepart = list(0, 0);
430 		else {
431 			n2->nif.elsepart = NULL;
432 			tokpushback++;
433 		}
434 		if (readtoken() != TFI)
435 			synexpect(TFI);
436 		checkkwd = CHKKWD | CHKALIAS;
437 		break;
438 	case TWHILE:
439 	case TUNTIL: {
440 		int got;
441 		n1 = (union node *)stalloc(sizeof (struct nbinary));
442 		n1->type = (lasttoken == TWHILE)? NWHILE : NUNTIL;
443 		if ((n1->nbinary.ch1 = list(0, 0)) == NULL)
444 			synexpect(-1);
445 		if ((got=readtoken()) != TDO) {
446 TRACE(("expecting DO got %s %s\n", tokname[got], got == TWORD ? wordtext : ""));
447 			synexpect(TDO);
448 		}
449 		n1->nbinary.ch2 = list(0, 0);
450 		if (readtoken() != TDONE)
451 			synexpect(TDONE);
452 		checkkwd = CHKKWD | CHKALIAS;
453 		break;
454 	}
455 	case TFOR:
456 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
457 			synerror("Bad for loop variable");
458 		n1 = (union node *)stalloc(sizeof (struct nfor));
459 		n1->type = NFOR;
460 		n1->nfor.var = wordtext;
461 		while (readtoken() == TNL)
462 			;
463 		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
464 			app = &ap;
465 			while (readtoken() == TWORD) {
466 				n2 = (union node *)stalloc(sizeof (struct narg));
467 				n2->type = NARG;
468 				n2->narg.text = wordtext;
469 				n2->narg.backquote = backquotelist;
470 				*app = n2;
471 				app = &n2->narg.next;
472 			}
473 			*app = NULL;
474 			n1->nfor.args = ap;
475 			if (lasttoken != TNL && lasttoken != TSEMI)
476 				synexpect(-1);
477 		} else {
478 			static char argvars[5] = {
479 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
480 			};
481 			n2 = (union node *)stalloc(sizeof (struct narg));
482 			n2->type = NARG;
483 			n2->narg.text = argvars;
484 			n2->narg.backquote = NULL;
485 			n2->narg.next = NULL;
486 			n1->nfor.args = n2;
487 			/*
488 			 * Newline or semicolon here is optional (but note
489 			 * that the original Bourne shell only allowed NL).
490 			 */
491 			if (lasttoken != TNL && lasttoken != TSEMI)
492 				tokpushback++;
493 		}
494 		checkkwd = CHKNL | CHKKWD | CHKALIAS;
495 		if ((t = readtoken()) == TDO)
496 			t = TDONE;
497 		else if (t == TBEGIN)
498 			t = TEND;
499 		else
500 			synexpect(-1);
501 		n1->nfor.body = list(0, 0);
502 		if (readtoken() != t)
503 			synexpect(t);
504 		checkkwd = CHKKWD | CHKALIAS;
505 		break;
506 	case TCASE:
507 		n1 = (union node *)stalloc(sizeof (struct ncase));
508 		n1->type = NCASE;
509 		if (readtoken() != TWORD)
510 			synexpect(TWORD);
511 		n1->ncase.expr = n2 = (union node *)stalloc(sizeof (struct narg));
512 		n2->type = NARG;
513 		n2->narg.text = wordtext;
514 		n2->narg.backquote = backquotelist;
515 		n2->narg.next = NULL;
516 		while (readtoken() == TNL);
517 		if (lasttoken != TWORD || ! equal(wordtext, "in"))
518 			synerror("expecting \"in\"");
519 		cpp = &n1->ncase.cases;
520 		checkkwd = CHKNL | CHKKWD, readtoken();
521 		while (lasttoken != TESAC) {
522 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
523 			cp->type = NCLIST;
524 			app = &cp->nclist.pattern;
525 			if (lasttoken == TLP)
526 				readtoken();
527 			for (;;) {
528 				*app = ap = (union node *)stalloc(sizeof (struct narg));
529 				ap->type = NARG;
530 				ap->narg.text = wordtext;
531 				ap->narg.backquote = backquotelist;
532 				checkkwd = CHKNL | CHKKWD;
533 				if (readtoken() != TPIPE)
534 					break;
535 				app = &ap->narg.next;
536 				readtoken();
537 			}
538 			ap->narg.next = NULL;
539 			if (lasttoken != TRP)
540 				synexpect(TRP);
541 			cp->nclist.body = list(0, 0);
542 
543 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
544 			if ((t = readtoken()) != TESAC) {
545 				if (t == TENDCASE)
546 					;
547 				else if (t == TFALLTHRU)
548 					cp->type = NCLISTFALLTHRU;
549 				else
550 					synexpect(TENDCASE);
551 				checkkwd = CHKNL | CHKKWD, readtoken();
552 			}
553 			cpp = &cp->nclist.next;
554 		}
555 		*cpp = NULL;
556 		checkkwd = CHKKWD | CHKALIAS;
557 		break;
558 	case TLP:
559 		n1 = (union node *)stalloc(sizeof (struct nredir));
560 		n1->type = NSUBSHELL;
561 		n1->nredir.n = list(0, 0);
562 		n1->nredir.redirect = NULL;
563 		if (readtoken() != TRP)
564 			synexpect(TRP);
565 		checkkwd = CHKKWD | CHKALIAS;
566 		is_subshell = 1;
567 		break;
568 	case TBEGIN:
569 		n1 = list(0, 0);
570 		if (readtoken() != TEND)
571 			synexpect(TEND);
572 		checkkwd = CHKKWD | CHKALIAS;
573 		break;
574 	/* Handle an empty command like other simple commands.  */
575 	case TBACKGND:
576 	case TSEMI:
577 	case TAND:
578 	case TOR:
579 		/*
580 		 * An empty command before a ; doesn't make much sense, and
581 		 * should certainly be disallowed in the case of `if ;'.
582 		 */
583 		if (!redir)
584 			synexpect(-1);
585 	case TNL:
586 	case TEOF:
587 	case TWORD:
588 	case TRP:
589 		tokpushback++;
590 		n1 = simplecmd(rpp, redir);
591 		return n1;
592 	default:
593 		synexpect(-1);
594 	}
595 
596 	/* Now check for redirection which may follow command */
597 	while (readtoken() == TREDIR) {
598 		*rpp = n2 = redirnode;
599 		rpp = &n2->nfile.next;
600 		parsefname();
601 	}
602 	tokpushback++;
603 	*rpp = NULL;
604 	if (redir) {
605 		if (!is_subshell) {
606 			n2 = (union node *)stalloc(sizeof (struct nredir));
607 			n2->type = NREDIR;
608 			n2->nredir.n = n1;
609 			n1 = n2;
610 		}
611 		n1->nredir.redirect = redir;
612 	}
613 
614 	return n1;
615 }
616 
617 
618 static union node *
619 simplecmd(union node **rpp, union node *redir)
620 {
621 	union node *args, **app;
622 	union node **orig_rpp = rpp;
623 	union node *n = NULL;
624 	int special;
625 	int savecheckkwd;
626 
627 	/* If we don't have any redirections already, then we must reset */
628 	/* rpp to be the address of the local redir variable.  */
629 	if (redir == 0)
630 		rpp = &redir;
631 
632 	args = NULL;
633 	app = &args;
634 	/*
635 	 * We save the incoming value, because we need this for shell
636 	 * functions.  There can not be a redirect or an argument between
637 	 * the function name and the open parenthesis.
638 	 */
639 	orig_rpp = rpp;
640 
641 	savecheckkwd = CHKALIAS;
642 
643 	for (;;) {
644 		checkkwd = savecheckkwd;
645 		if (readtoken() == TWORD) {
646 			n = (union node *)stalloc(sizeof (struct narg));
647 			n->type = NARG;
648 			n->narg.text = wordtext;
649 			n->narg.backquote = backquotelist;
650 			*app = n;
651 			app = &n->narg.next;
652 			if (savecheckkwd != 0 && !isassignment(wordtext))
653 				savecheckkwd = 0;
654 		} else if (lasttoken == TREDIR) {
655 			*rpp = n = redirnode;
656 			rpp = &n->nfile.next;
657 			parsefname();	/* read name of redirection file */
658 		} else if (lasttoken == TLP && app == &args->narg.next
659 					    && rpp == orig_rpp) {
660 			/* We have a function */
661 			if (readtoken() != TRP)
662 				synexpect(TRP);
663 			funclinno = plinno;
664 			/*
665 			 * - Require plain text.
666 			 * - Functions with '/' cannot be called.
667 			 * - Reject name=().
668 			 * - Reject ksh extended glob patterns.
669 			 */
670 			if (!noexpand(n->narg.text) || quoteflag ||
671 			    strchr(n->narg.text, '/') ||
672 			    strchr("!%*+-=?@}~",
673 				n->narg.text[strlen(n->narg.text) - 1]))
674 				synerror("Bad function name");
675 			rmescapes(n->narg.text);
676 			if (find_builtin(n->narg.text, &special) >= 0 &&
677 			    special)
678 				synerror("Cannot override a special builtin with a function");
679 			n->type = NDEFUN;
680 			n->narg.next = command();
681 			funclinno = 0;
682 			return n;
683 		} else {
684 			tokpushback++;
685 			break;
686 		}
687 	}
688 	*app = NULL;
689 	*rpp = NULL;
690 	n = (union node *)stalloc(sizeof (struct ncmd));
691 	n->type = NCMD;
692 	n->ncmd.args = args;
693 	n->ncmd.redirect = redir;
694 	return n;
695 }
696 
697 static union node *
698 makename(void)
699 {
700 	union node *n;
701 
702 	n = (union node *)stalloc(sizeof (struct narg));
703 	n->type = NARG;
704 	n->narg.next = NULL;
705 	n->narg.text = wordtext;
706 	n->narg.backquote = backquotelist;
707 	return n;
708 }
709 
710 void
711 fixredir(union node *n, const char *text, int err)
712 {
713 	TRACE(("Fix redir %s %d\n", text, err));
714 	if (!err)
715 		n->ndup.vname = NULL;
716 
717 	if (is_digit(text[0]) && text[1] == '\0')
718 		n->ndup.dupfd = digit_val(text[0]);
719 	else if (text[0] == '-' && text[1] == '\0')
720 		n->ndup.dupfd = -1;
721 	else {
722 
723 		if (err)
724 			synerror("Bad fd number");
725 		else
726 			n->ndup.vname = makename();
727 	}
728 }
729 
730 
731 static void
732 parsefname(void)
733 {
734 	union node *n = redirnode;
735 
736 	if (readtoken() != TWORD)
737 		synexpect(-1);
738 	if (n->type == NHERE) {
739 		struct heredoc *here = heredoc;
740 		struct heredoc *p;
741 		int i;
742 
743 		if (quoteflag == 0)
744 			n->type = NXHERE;
745 		TRACE(("Here document %d\n", n->type));
746 		if (here->striptabs) {
747 			while (*wordtext == '\t')
748 				wordtext++;
749 		}
750 		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
751 			synerror("Illegal eof marker for << redirection");
752 		rmescapes(wordtext);
753 		here->eofmark = wordtext;
754 		here->next = NULL;
755 		if (heredoclist == NULL)
756 			heredoclist = here;
757 		else {
758 			for (p = heredoclist ; p->next ; p = p->next);
759 			p->next = here;
760 		}
761 	} else if (n->type == NTOFD || n->type == NFROMFD) {
762 		fixredir(n, wordtext, 0);
763 	} else {
764 		n->nfile.fname = makename();
765 	}
766 }
767 
768 
769 /*
770  * Input any here documents.
771  */
772 
773 static void
774 parseheredoc(void)
775 {
776 	struct heredoc *here;
777 	union node *n;
778 
779 	while (heredoclist) {
780 		here = heredoclist;
781 		heredoclist = here->next;
782 		if (needprompt) {
783 			setprompt(2);
784 			needprompt = 0;
785 		}
786 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
787 				here->eofmark, here->striptabs);
788 		n = (union node *)stalloc(sizeof (struct narg));
789 		n->narg.type = NARG;
790 		n->narg.next = NULL;
791 		n->narg.text = wordtext;
792 		n->narg.backquote = backquotelist;
793 		here->here->nhere.doc = n;
794 	}
795 }
796 
797 static int
798 peektoken(void)
799 {
800 	int t;
801 
802 	t = readtoken();
803 	tokpushback++;
804 	return (t);
805 }
806 
807 static int
808 readtoken(void)
809 {
810 	int t;
811 	struct alias *ap;
812 #ifdef DEBUG
813 	int alreadyseen = tokpushback;
814 #endif
815 
816 	top:
817 	t = xxreadtoken();
818 
819 	/*
820 	 * eat newlines
821 	 */
822 	if (checkkwd & CHKNL) {
823 		while (t == TNL) {
824 			parseheredoc();
825 			t = xxreadtoken();
826 		}
827 	}
828 
829 	/*
830 	 * check for keywords and aliases
831 	 */
832 	if (t == TWORD && !quoteflag)
833 	{
834 		const char * const *pp;
835 
836 		if (checkkwd & CHKKWD)
837 			for (pp = parsekwd; *pp; pp++) {
838 				if (**pp == *wordtext && equal(*pp, wordtext))
839 				{
840 					lasttoken = t = pp - parsekwd + KWDOFFSET;
841 					TRACE(("keyword %s recognized\n", tokname[t]));
842 					goto out;
843 				}
844 			}
845 		if (checkkwd & CHKALIAS &&
846 		    (ap = lookupalias(wordtext, 1)) != NULL) {
847 			pushstring(ap->val, strlen(ap->val), ap);
848 			goto top;
849 		}
850 	}
851 out:
852 	if (t != TNOT)
853 		checkkwd = 0;
854 
855 #ifdef DEBUG
856 	if (!alreadyseen)
857 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
858 	else
859 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
860 #endif
861 	return (t);
862 }
863 
864 
865 /*
866  * Read the next input token.
867  * If the token is a word, we set backquotelist to the list of cmds in
868  *	backquotes.  We set quoteflag to true if any part of the word was
869  *	quoted.
870  * If the token is TREDIR, then we set redirnode to a structure containing
871  *	the redirection.
872  * In all cases, the variable startlinno is set to the number of the line
873  *	on which the token starts.
874  *
875  * [Change comment:  here documents and internal procedures]
876  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
877  *  word parsing code into a separate routine.  In this case, readtoken
878  *  doesn't need to have any internal procedures, but parseword does.
879  *  We could also make parseoperator in essence the main routine, and
880  *  have parseword (readtoken1?) handle both words and redirection.]
881  */
882 
883 #define RETURN(token)	return lasttoken = token
884 
885 static int
886 xxreadtoken(void)
887 {
888 	int c;
889 
890 	if (tokpushback) {
891 		tokpushback = 0;
892 		return lasttoken;
893 	}
894 	if (needprompt) {
895 		setprompt(2);
896 		needprompt = 0;
897 	}
898 	startlinno = plinno;
899 	for (;;) {	/* until token or start of word found */
900 		c = pgetc_macro();
901 		switch (c) {
902 		case ' ': case '\t':
903 			continue;
904 		case '#':
905 			while ((c = pgetc()) != '\n' && c != PEOF);
906 			pungetc();
907 			continue;
908 		case '\\':
909 			if (pgetc() == '\n') {
910 				startlinno = ++plinno;
911 				if (doprompt)
912 					setprompt(2);
913 				else
914 					setprompt(0);
915 				continue;
916 			}
917 			pungetc();
918 			goto breakloop;
919 		case '\n':
920 			plinno++;
921 			needprompt = doprompt;
922 			RETURN(TNL);
923 		case PEOF:
924 			RETURN(TEOF);
925 		case '&':
926 			if (pgetc() == '&')
927 				RETURN(TAND);
928 			pungetc();
929 			RETURN(TBACKGND);
930 		case '|':
931 			if (pgetc() == '|')
932 				RETURN(TOR);
933 			pungetc();
934 			RETURN(TPIPE);
935 		case ';':
936 			c = pgetc();
937 			if (c == ';')
938 				RETURN(TENDCASE);
939 			else if (c == '&')
940 				RETURN(TFALLTHRU);
941 			pungetc();
942 			RETURN(TSEMI);
943 		case '(':
944 			RETURN(TLP);
945 		case ')':
946 			RETURN(TRP);
947 		default:
948 			goto breakloop;
949 		}
950 	}
951 breakloop:
952 	return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
953 #undef RETURN
954 }
955 
956 
957 #define MAXNEST_static 8
958 struct tokenstate
959 {
960 	const char *syntax; /* *SYNTAX */
961 	int parenlevel; /* levels of parentheses in arithmetic */
962 	enum tokenstate_category
963 	{
964 		TSTATE_TOP,
965 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
966 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
967 		TSTATE_ARITH
968 	} category;
969 };
970 
971 
972 /*
973  * Called to parse command substitutions.
974  */
975 
976 static char *
977 parsebackq(char *out, struct nodelist **pbqlist,
978 		int oldstyle, int dblquote, int quoted)
979 {
980 	struct nodelist **nlpp;
981 	union node *n;
982 	char *volatile str;
983 	struct jmploc jmploc;
984 	struct jmploc *const savehandler = handler;
985 	int savelen;
986 	int saveprompt;
987 	const int bq_startlinno = plinno;
988 	char *volatile ostr = NULL;
989 	struct parsefile *const savetopfile = getcurrentfile();
990 	struct heredoc *const saveheredoclist = heredoclist;
991 	struct heredoc *here;
992 
993 	str = NULL;
994 	if (setjmp(jmploc.loc)) {
995 		popfilesupto(savetopfile);
996 		if (str)
997 			ckfree(str);
998 		if (ostr)
999 			ckfree(ostr);
1000 		heredoclist = saveheredoclist;
1001 		handler = savehandler;
1002 		if (exception == EXERROR) {
1003 			startlinno = bq_startlinno;
1004 			synerror("Error in command substitution");
1005 		}
1006 		longjmp(handler->loc, 1);
1007 	}
1008 	INTOFF;
1009 	savelen = out - stackblock();
1010 	if (savelen > 0) {
1011 		str = ckmalloc(savelen);
1012 		memcpy(str, stackblock(), savelen);
1013 	}
1014 	handler = &jmploc;
1015 	heredoclist = NULL;
1016 	INTON;
1017         if (oldstyle) {
1018                 /* We must read until the closing backquote, giving special
1019                    treatment to some slashes, and then push the string and
1020                    reread it as input, interpreting it normally.  */
1021                 char *oout;
1022                 int c;
1023                 int olen;
1024 
1025 
1026                 STARTSTACKSTR(oout);
1027 		for (;;) {
1028 			if (needprompt) {
1029 				setprompt(2);
1030 				needprompt = 0;
1031 			}
1032 			CHECKSTRSPACE(2, oout);
1033 			switch (c = pgetc()) {
1034 			case '`':
1035 				goto done;
1036 
1037 			case '\\':
1038                                 if ((c = pgetc()) == '\n') {
1039 					plinno++;
1040 					if (doprompt)
1041 						setprompt(2);
1042 					else
1043 						setprompt(0);
1044 					/*
1045 					 * If eating a newline, avoid putting
1046 					 * the newline into the new character
1047 					 * stream (via the USTPUTC after the
1048 					 * switch).
1049 					 */
1050 					continue;
1051 				}
1052                                 if (c != '\\' && c != '`' && c != '$'
1053                                     && (!dblquote || c != '"'))
1054                                         USTPUTC('\\', oout);
1055 				break;
1056 
1057 			case '\n':
1058 				plinno++;
1059 				needprompt = doprompt;
1060 				break;
1061 
1062 			case PEOF:
1063 			        startlinno = plinno;
1064 				synerror("EOF in backquote substitution");
1065  				break;
1066 
1067 			default:
1068 				break;
1069 			}
1070 			USTPUTC(c, oout);
1071                 }
1072 done:
1073                 USTPUTC('\0', oout);
1074                 olen = oout - stackblock();
1075 		INTOFF;
1076 		ostr = ckmalloc(olen);
1077 		memcpy(ostr, stackblock(), olen);
1078 		setinputstring(ostr, 1);
1079 		INTON;
1080         }
1081 	nlpp = pbqlist;
1082 	while (*nlpp)
1083 		nlpp = &(*nlpp)->next;
1084 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1085 	(*nlpp)->next = NULL;
1086 
1087 	if (oldstyle) {
1088 		saveprompt = doprompt;
1089 		doprompt = 0;
1090 	}
1091 
1092 	n = list(0, oldstyle);
1093 
1094 	if (oldstyle)
1095 		doprompt = saveprompt;
1096 	else {
1097 		if (readtoken() != TRP)
1098 			synexpect(TRP);
1099 	}
1100 
1101 	(*nlpp)->n = n;
1102         if (oldstyle) {
1103 		/*
1104 		 * Start reading from old file again, ignoring any pushed back
1105 		 * tokens left from the backquote parsing
1106 		 */
1107                 popfile();
1108 		tokpushback = 0;
1109 	}
1110 	STARTSTACKSTR(out);
1111 	CHECKSTRSPACE(savelen + 1, out);
1112 	INTOFF;
1113 	if (str) {
1114 		memcpy(out, str, savelen);
1115 		STADJUST(savelen, out);
1116 		ckfree(str);
1117 		str = NULL;
1118 	}
1119 	if (ostr) {
1120 		ckfree(ostr);
1121 		ostr = NULL;
1122 	}
1123 	here = saveheredoclist;
1124 	if (here != NULL) {
1125 		while (here->next != NULL)
1126 			here = here->next;
1127 		here->next = heredoclist;
1128 		heredoclist = saveheredoclist;
1129 	}
1130 	handler = savehandler;
1131 	INTON;
1132 	if (quoted)
1133 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1134 	else
1135 		USTPUTC(CTLBACKQ, out);
1136 	return out;
1137 }
1138 
1139 
1140 /*
1141  * Called to parse a backslash escape sequence inside $'...'.
1142  * The backslash has already been read.
1143  */
1144 static char *
1145 readcstyleesc(char *out)
1146 {
1147 	int c, v, i, n;
1148 
1149 	c = pgetc();
1150 	switch (c) {
1151 	case '\0':
1152 		synerror("Unterminated quoted string");
1153 	case '\n':
1154 		plinno++;
1155 		if (doprompt)
1156 			setprompt(2);
1157 		else
1158 			setprompt(0);
1159 		return out;
1160 	case '\\':
1161 	case '\'':
1162 	case '"':
1163 		v = c;
1164 		break;
1165 	case 'a': v = '\a'; break;
1166 	case 'b': v = '\b'; break;
1167 	case 'e': v = '\033'; break;
1168 	case 'f': v = '\f'; break;
1169 	case 'n': v = '\n'; break;
1170 	case 'r': v = '\r'; break;
1171 	case 't': v = '\t'; break;
1172 	case 'v': v = '\v'; break;
1173 	case 'x':
1174 		  v = 0;
1175 		  for (;;) {
1176 			  c = pgetc();
1177 			  if (c >= '0' && c <= '9')
1178 				  v = (v << 4) + c - '0';
1179 			  else if (c >= 'A' && c <= 'F')
1180 				  v = (v << 4) + c - 'A' + 10;
1181 			  else if (c >= 'a' && c <= 'f')
1182 				  v = (v << 4) + c - 'a' + 10;
1183 			  else
1184 				  break;
1185 		  }
1186 		  pungetc();
1187 		  break;
1188 	case '0': case '1': case '2': case '3':
1189 	case '4': case '5': case '6': case '7':
1190 		  v = c - '0';
1191 		  c = pgetc();
1192 		  if (c >= '0' && c <= '7') {
1193 			  v <<= 3;
1194 			  v += c - '0';
1195 			  c = pgetc();
1196 			  if (c >= '0' && c <= '7') {
1197 				  v <<= 3;
1198 				  v += c - '0';
1199 			  } else
1200 				  pungetc();
1201 		  } else
1202 			  pungetc();
1203 		  break;
1204 	case 'c':
1205 		  c = pgetc();
1206 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1207 			  synerror("Bad escape sequence");
1208 		  if (c == '\\' && pgetc() != '\\')
1209 			  synerror("Bad escape sequence");
1210 		  if (c == '?')
1211 			  v = 127;
1212 		  else
1213 			  v = c & 0x1f;
1214 		  break;
1215 	case 'u':
1216 	case 'U':
1217 		  n = c == 'U' ? 8 : 4;
1218 		  v = 0;
1219 		  for (i = 0; i < n; i++) {
1220 			  c = pgetc();
1221 			  if (c >= '0' && c <= '9')
1222 				  v = (v << 4) + c - '0';
1223 			  else if (c >= 'A' && c <= 'F')
1224 				  v = (v << 4) + c - 'A' + 10;
1225 			  else if (c >= 'a' && c <= 'f')
1226 				  v = (v << 4) + c - 'a' + 10;
1227 			  else
1228 				  synerror("Bad escape sequence");
1229 		  }
1230 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1231 			  synerror("Bad escape sequence");
1232 		  /* We really need iconv here. */
1233 		  if (initial_localeisutf8 && v > 127) {
1234 			  CHECKSTRSPACE(4, out);
1235 			  /*
1236 			   * We cannot use wctomb() as the locale may have
1237 			   * changed.
1238 			   */
1239 			  if (v <= 0x7ff) {
1240 				  USTPUTC(0xc0 | v >> 6, out);
1241 				  USTPUTC(0x80 | (v & 0x3f), out);
1242 				  return out;
1243 			  } else if (v <= 0xffff) {
1244 				  USTPUTC(0xe0 | v >> 12, out);
1245 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1246 				  USTPUTC(0x80 | (v & 0x3f), out);
1247 				  return out;
1248 			  } else if (v <= 0x10ffff) {
1249 				  USTPUTC(0xf0 | v >> 18, out);
1250 				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1251 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1252 				  USTPUTC(0x80 | (v & 0x3f), out);
1253 				  return out;
1254 			  }
1255 		  }
1256 		  if (v > 127)
1257 			  v = '?';
1258 		  break;
1259 	default:
1260 		  synerror("Bad escape sequence");
1261 	}
1262 	v = (char)v;
1263 	/*
1264 	 * We can't handle NUL bytes.
1265 	 * POSIX says we should skip till the closing quote.
1266 	 */
1267 	if (v == '\0') {
1268 		while ((c = pgetc()) != '\'') {
1269 			if (c == '\\')
1270 				c = pgetc();
1271 			if (c == PEOF)
1272 				synerror("Unterminated quoted string");
1273 		}
1274 		pungetc();
1275 		return out;
1276 	}
1277 	if (SQSYNTAX[v] == CCTL)
1278 		USTPUTC(CTLESC, out);
1279 	USTPUTC(v, out);
1280 	return out;
1281 }
1282 
1283 
1284 /*
1285  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1286  * is not NULL, read a here document.  In the latter case, eofmark is the
1287  * word which marks the end of the document and striptabs is true if
1288  * leading tabs should be stripped from the document.  The argument firstc
1289  * is the first character of the input token or document.
1290  *
1291  * Because C does not have internal subroutines, I have simulated them
1292  * using goto's to implement the subroutine linkage.  The following macros
1293  * will run code that appears at the end of readtoken1.
1294  */
1295 
1296 #define CHECKEND()	{goto checkend; checkend_return:;}
1297 #define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1298 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1299 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1300 
1301 static int
1302 readtoken1(int firstc, char const *initialsyntax, char *eofmark, int striptabs)
1303 {
1304 	int c = firstc;
1305 	char *out;
1306 	int len;
1307 	char line[EOFMARKLEN + 1];
1308 	struct nodelist *bqlist;
1309 	int quotef;
1310 	int newvarnest;
1311 	int level;
1312 	int synentry;
1313 	struct tokenstate state_static[MAXNEST_static];
1314 	int maxnest = MAXNEST_static;
1315 	struct tokenstate *state = state_static;
1316 	int sqiscstyle = 0;
1317 
1318 	startlinno = plinno;
1319 	quotef = 0;
1320 	bqlist = NULL;
1321 	newvarnest = 0;
1322 	level = 0;
1323 	state[level].syntax = initialsyntax;
1324 	state[level].parenlevel = 0;
1325 	state[level].category = TSTATE_TOP;
1326 
1327 	STARTSTACKSTR(out);
1328 	loop: {	/* for each line, until end of word */
1329 		CHECKEND();	/* set c to PEOF if at end of here document */
1330 		for (;;) {	/* until end of line or end of word */
1331 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1332 
1333 			synentry = state[level].syntax[c];
1334 
1335 			switch(synentry) {
1336 			case CNL:	/* '\n' */
1337 				if (state[level].syntax == BASESYNTAX)
1338 					goto endword;	/* exit outer loop */
1339 				USTPUTC(c, out);
1340 				plinno++;
1341 				if (doprompt)
1342 					setprompt(2);
1343 				else
1344 					setprompt(0);
1345 				c = pgetc();
1346 				goto loop;		/* continue outer loop */
1347 			case CSBACK:
1348 				if (sqiscstyle) {
1349 					out = readcstyleesc(out);
1350 					break;
1351 				}
1352 				/* FALLTHROUGH */
1353 			case CWORD:
1354 				USTPUTC(c, out);
1355 				break;
1356 			case CCTL:
1357 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1358 					USTPUTC(CTLESC, out);
1359 				USTPUTC(c, out);
1360 				break;
1361 			case CBACK:	/* backslash */
1362 				c = pgetc();
1363 				if (c == PEOF) {
1364 					USTPUTC('\\', out);
1365 					pungetc();
1366 				} else if (c == '\n') {
1367 					plinno++;
1368 					if (doprompt)
1369 						setprompt(2);
1370 					else
1371 						setprompt(0);
1372 				} else {
1373 					if (state[level].syntax == DQSYNTAX &&
1374 					    c != '\\' && c != '`' && c != '$' &&
1375 					    (c != '"' || (eofmark != NULL &&
1376 						newvarnest == 0)) &&
1377 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1378 						USTPUTC('\\', out);
1379 					if ((eofmark == NULL ||
1380 					    newvarnest > 0) &&
1381 					    state[level].syntax == BASESYNTAX)
1382 						USTPUTC(CTLQUOTEMARK, out);
1383 					if (SQSYNTAX[c] == CCTL)
1384 						USTPUTC(CTLESC, out);
1385 					USTPUTC(c, out);
1386 					if ((eofmark == NULL ||
1387 					    newvarnest > 0) &&
1388 					    state[level].syntax == BASESYNTAX &&
1389 					    state[level].category == TSTATE_VAR_OLD)
1390 						USTPUTC(CTLQUOTEEND, out);
1391 					quotef++;
1392 				}
1393 				break;
1394 			case CSQUOTE:
1395 				USTPUTC(CTLQUOTEMARK, out);
1396 				state[level].syntax = SQSYNTAX;
1397 				sqiscstyle = 0;
1398 				break;
1399 			case CDQUOTE:
1400 				USTPUTC(CTLQUOTEMARK, out);
1401 				state[level].syntax = DQSYNTAX;
1402 				break;
1403 			case CENDQUOTE:
1404 				if (eofmark != NULL && newvarnest == 0)
1405 					USTPUTC(c, out);
1406 				else {
1407 					if (state[level].category == TSTATE_VAR_OLD)
1408 						USTPUTC(CTLQUOTEEND, out);
1409 					state[level].syntax = BASESYNTAX;
1410 					quotef++;
1411 				}
1412 				break;
1413 			case CVAR:	/* '$' */
1414 				PARSESUB();		/* parse substitution */
1415 				break;
1416 			case CENDVAR:	/* '}' */
1417 				if (level > 0 &&
1418 				    ((state[level].category == TSTATE_VAR_OLD &&
1419 				      state[level].syntax ==
1420 				      state[level - 1].syntax) ||
1421 				    (state[level].category == TSTATE_VAR_NEW &&
1422 				     state[level].syntax == BASESYNTAX))) {
1423 					if (state[level].category == TSTATE_VAR_NEW)
1424 						newvarnest--;
1425 					level--;
1426 					USTPUTC(CTLENDVAR, out);
1427 				} else {
1428 					USTPUTC(c, out);
1429 				}
1430 				break;
1431 			case CLP:	/* '(' in arithmetic */
1432 				state[level].parenlevel++;
1433 				USTPUTC(c, out);
1434 				break;
1435 			case CRP:	/* ')' in arithmetic */
1436 				if (state[level].parenlevel > 0) {
1437 					USTPUTC(c, out);
1438 					--state[level].parenlevel;
1439 				} else {
1440 					if (pgetc() == ')') {
1441 						if (level > 0 &&
1442 						    state[level].category == TSTATE_ARITH) {
1443 							level--;
1444 							USTPUTC(CTLENDARI, out);
1445 						} else
1446 							USTPUTC(')', out);
1447 					} else {
1448 						/*
1449 						 * unbalanced parens
1450 						 *  (don't 2nd guess - no error)
1451 						 */
1452 						pungetc();
1453 						USTPUTC(')', out);
1454 					}
1455 				}
1456 				break;
1457 			case CBQUOTE:	/* '`' */
1458 				out = parsebackq(out, &bqlist, 1,
1459 				    state[level].syntax == DQSYNTAX &&
1460 				    (eofmark == NULL || newvarnest > 0),
1461 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1462 				break;
1463 			case CEOF:
1464 				goto endword;		/* exit outer loop */
1465 			case CIGN:
1466 				break;
1467 			default:
1468 				if (level == 0)
1469 					goto endword;	/* exit outer loop */
1470 				USTPUTC(c, out);
1471 			}
1472 			c = pgetc_macro();
1473 		}
1474 	}
1475 endword:
1476 	if (state[level].syntax == ARISYNTAX)
1477 		synerror("Missing '))'");
1478 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1479 		synerror("Unterminated quoted string");
1480 	if (state[level].category == TSTATE_VAR_OLD ||
1481 	    state[level].category == TSTATE_VAR_NEW) {
1482 		startlinno = plinno;
1483 		synerror("Missing '}'");
1484 	}
1485 	if (state != state_static)
1486 		parser_temp_free_upto(state);
1487 	USTPUTC('\0', out);
1488 	len = out - stackblock();
1489 	out = stackblock();
1490 	if (eofmark == NULL) {
1491 		if ((c == '>' || c == '<')
1492 		 && quotef == 0
1493 		 && len <= 2
1494 		 && (*out == '\0' || is_digit(*out))) {
1495 			PARSEREDIR();
1496 			return lasttoken = TREDIR;
1497 		} else {
1498 			pungetc();
1499 		}
1500 	}
1501 	quoteflag = quotef;
1502 	backquotelist = bqlist;
1503 	grabstackblock(len);
1504 	wordtext = out;
1505 	return lasttoken = TWORD;
1506 /* end of readtoken routine */
1507 
1508 
1509 /*
1510  * Check to see whether we are at the end of the here document.  When this
1511  * is called, c is set to the first character of the next input line.  If
1512  * we are at the end of the here document, this routine sets the c to PEOF.
1513  */
1514 
1515 checkend: {
1516 	if (eofmark) {
1517 		if (striptabs) {
1518 			while (c == '\t')
1519 				c = pgetc();
1520 		}
1521 		if (c == *eofmark) {
1522 			if (pfgets(line, sizeof line) != NULL) {
1523 				char *p, *q;
1524 
1525 				p = line;
1526 				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1527 				if ((*p == '\0' || *p == '\n') && *q == '\0') {
1528 					c = PEOF;
1529 					if (*p == '\n') {
1530 						plinno++;
1531 						needprompt = doprompt;
1532 					}
1533 				} else {
1534 					pushstring(line, strlen(line), NULL);
1535 				}
1536 			}
1537 		}
1538 	}
1539 	goto checkend_return;
1540 }
1541 
1542 
1543 /*
1544  * Parse a redirection operator.  The variable "out" points to a string
1545  * specifying the fd to be redirected.  The variable "c" contains the
1546  * first character of the redirection operator.
1547  */
1548 
1549 parseredir: {
1550 	char fd = *out;
1551 	union node *np;
1552 
1553 	np = (union node *)stalloc(sizeof (struct nfile));
1554 	if (c == '>') {
1555 		np->nfile.fd = 1;
1556 		c = pgetc();
1557 		if (c == '>')
1558 			np->type = NAPPEND;
1559 		else if (c == '&')
1560 			np->type = NTOFD;
1561 		else if (c == '|')
1562 			np->type = NCLOBBER;
1563 		else {
1564 			np->type = NTO;
1565 			pungetc();
1566 		}
1567 	} else {	/* c == '<' */
1568 		np->nfile.fd = 0;
1569 		c = pgetc();
1570 		if (c == '<') {
1571 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1572 				np = (union node *)stalloc(sizeof (struct nhere));
1573 				np->nfile.fd = 0;
1574 			}
1575 			np->type = NHERE;
1576 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1577 			heredoc->here = np;
1578 			if ((c = pgetc()) == '-') {
1579 				heredoc->striptabs = 1;
1580 			} else {
1581 				heredoc->striptabs = 0;
1582 				pungetc();
1583 			}
1584 		} else if (c == '&')
1585 			np->type = NFROMFD;
1586 		else if (c == '>')
1587 			np->type = NFROMTO;
1588 		else {
1589 			np->type = NFROM;
1590 			pungetc();
1591 		}
1592 	}
1593 	if (fd != '\0')
1594 		np->nfile.fd = digit_val(fd);
1595 	redirnode = np;
1596 	goto parseredir_return;
1597 }
1598 
1599 
1600 /*
1601  * Parse a substitution.  At this point, we have read the dollar sign
1602  * and nothing else.
1603  */
1604 
1605 parsesub: {
1606 	char buf[10];
1607 	int subtype;
1608 	int typeloc;
1609 	int flags;
1610 	char *p;
1611 	static const char types[] = "}-+?=";
1612 	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1613 	int linno;
1614 	int length;
1615 	int c1;
1616 
1617 	c = pgetc();
1618 	if (c == '(') {	/* $(command) or $((arith)) */
1619 		if (pgetc() == '(') {
1620 			PARSEARITH();
1621 		} else {
1622 			pungetc();
1623 			out = parsebackq(out, &bqlist, 0,
1624 			    state[level].syntax == DQSYNTAX &&
1625 			    (eofmark == NULL || newvarnest > 0),
1626 			    state[level].syntax == DQSYNTAX ||
1627 			    state[level].syntax == ARISYNTAX);
1628 		}
1629 	} else if (c == '{' || is_name(c) || is_special(c)) {
1630 		USTPUTC(CTLVAR, out);
1631 		typeloc = out - stackblock();
1632 		USTPUTC(VSNORMAL, out);
1633 		subtype = VSNORMAL;
1634 		flags = 0;
1635 		if (c == '{') {
1636 			bracketed_name = 1;
1637 			c = pgetc();
1638 			subtype = 0;
1639 		}
1640 varname:
1641 		if (!is_eof(c) && is_name(c)) {
1642 			length = 0;
1643 			do {
1644 				STPUTC(c, out);
1645 				c = pgetc();
1646 				length++;
1647 			} while (!is_eof(c) && is_in_name(c));
1648 			if (length == 6 &&
1649 			    strncmp(out - length, "LINENO", length) == 0) {
1650 				/* Replace the variable name with the
1651 				 * current line number. */
1652 				linno = plinno;
1653 				if (funclinno != 0)
1654 					linno -= funclinno - 1;
1655 				snprintf(buf, sizeof(buf), "%d", linno);
1656 				STADJUST(-6, out);
1657 				STPUTS(buf, out);
1658 				flags |= VSLINENO;
1659 			}
1660 		} else if (is_digit(c)) {
1661 			if (bracketed_name) {
1662 				do {
1663 					STPUTC(c, out);
1664 					c = pgetc();
1665 				} while (is_digit(c));
1666 			} else {
1667 				STPUTC(c, out);
1668 				c = pgetc();
1669 			}
1670 		} else if (is_special(c)) {
1671 			c1 = c;
1672 			c = pgetc();
1673 			if (subtype == 0 && c1 == '#') {
1674 				subtype = VSLENGTH;
1675 				if (strchr(types, c) == NULL && c != ':' &&
1676 				    c != '#' && c != '%')
1677 					goto varname;
1678 				c1 = c;
1679 				c = pgetc();
1680 				if (c1 != '}' && c == '}') {
1681 					pungetc();
1682 					c = c1;
1683 					goto varname;
1684 				}
1685 				pungetc();
1686 				c = c1;
1687 				c1 = '#';
1688 				subtype = 0;
1689 			}
1690 			USTPUTC(c1, out);
1691 		} else {
1692 			subtype = VSERROR;
1693 			if (c == '}')
1694 				pungetc();
1695 			else if (c == '\n' || c == PEOF)
1696 				synerror("Unexpected end of line in substitution");
1697 			else
1698 				USTPUTC(c, out);
1699 		}
1700 		if (subtype == 0) {
1701 			switch (c) {
1702 			case ':':
1703 				flags |= VSNUL;
1704 				c = pgetc();
1705 				/*FALLTHROUGH*/
1706 			default:
1707 				p = strchr(types, c);
1708 				if (p == NULL) {
1709 					if (c == '\n' || c == PEOF)
1710 						synerror("Unexpected end of line in substitution");
1711 					if (flags == VSNUL)
1712 						STPUTC(':', out);
1713 					STPUTC(c, out);
1714 					subtype = VSERROR;
1715 				} else
1716 					subtype = p - types + VSNORMAL;
1717 				break;
1718 			case '%':
1719 			case '#':
1720 				{
1721 					int cc = c;
1722 					subtype = c == '#' ? VSTRIMLEFT :
1723 							     VSTRIMRIGHT;
1724 					c = pgetc();
1725 					if (c == cc)
1726 						subtype++;
1727 					else
1728 						pungetc();
1729 					break;
1730 				}
1731 			}
1732 		} else if (subtype != VSERROR) {
1733 			if (subtype == VSLENGTH && c != '}')
1734 				subtype = VSERROR;
1735 			pungetc();
1736 		}
1737 		STPUTC('=', out);
1738 		if (state[level].syntax == DQSYNTAX ||
1739 		    state[level].syntax == ARISYNTAX)
1740 			flags |= VSQUOTE;
1741 		*(stackblock() + typeloc) = subtype | flags;
1742 		if (subtype != VSNORMAL) {
1743 			if (level + 1 >= maxnest) {
1744 				maxnest *= 2;
1745 				if (state == state_static) {
1746 					state = parser_temp_alloc(
1747 					    maxnest * sizeof(*state));
1748 					memcpy(state, state_static,
1749 					    MAXNEST_static * sizeof(*state));
1750 				} else
1751 					state = parser_temp_realloc(state,
1752 					    maxnest * sizeof(*state));
1753 			}
1754 			level++;
1755 			state[level].parenlevel = 0;
1756 			if (subtype == VSMINUS || subtype == VSPLUS ||
1757 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1758 				/*
1759 				 * For operators that were in the Bourne shell,
1760 				 * inherit the double-quote state.
1761 				 */
1762 				state[level].syntax = state[level - 1].syntax;
1763 				state[level].category = TSTATE_VAR_OLD;
1764 			} else {
1765 				/*
1766 				 * The other operators take a pattern,
1767 				 * so go to BASESYNTAX.
1768 				 * Also, ' and " are now special, even
1769 				 * in here documents.
1770 				 */
1771 				state[level].syntax = BASESYNTAX;
1772 				state[level].category = TSTATE_VAR_NEW;
1773 				newvarnest++;
1774 			}
1775 		}
1776 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1777 		/* $'cstylequotes' */
1778 		USTPUTC(CTLQUOTEMARK, out);
1779 		state[level].syntax = SQSYNTAX;
1780 		sqiscstyle = 1;
1781 	} else {
1782 		USTPUTC('$', out);
1783 		pungetc();
1784 	}
1785 	goto parsesub_return;
1786 }
1787 
1788 
1789 /*
1790  * Parse an arithmetic expansion (indicate start of one and set state)
1791  */
1792 parsearith: {
1793 
1794 	if (level + 1 >= maxnest) {
1795 		maxnest *= 2;
1796 		if (state == state_static) {
1797 			state = parser_temp_alloc(
1798 			    maxnest * sizeof(*state));
1799 			memcpy(state, state_static,
1800 			    MAXNEST_static * sizeof(*state));
1801 		} else
1802 			state = parser_temp_realloc(state,
1803 			    maxnest * sizeof(*state));
1804 	}
1805 	level++;
1806 	state[level].syntax = ARISYNTAX;
1807 	state[level].parenlevel = 0;
1808 	state[level].category = TSTATE_ARITH;
1809 	USTPUTC(CTLARI, out);
1810 	if (state[level - 1].syntax == DQSYNTAX)
1811 		USTPUTC('"',out);
1812 	else
1813 		USTPUTC(' ',out);
1814 	goto parsearith_return;
1815 }
1816 
1817 } /* end of readtoken */
1818 
1819 
1820 
1821 #ifdef mkinit
1822 RESET {
1823 	tokpushback = 0;
1824 	checkkwd = 0;
1825 }
1826 #endif
1827 
1828 /*
1829  * Returns true if the text contains nothing to expand (no dollar signs
1830  * or backquotes).
1831  */
1832 
1833 static int
1834 noexpand(char *text)
1835 {
1836 	char *p;
1837 	char c;
1838 
1839 	p = text;
1840 	while ((c = *p++) != '\0') {
1841 		if ( c == CTLQUOTEMARK)
1842 			continue;
1843 		if (c == CTLESC)
1844 			p++;
1845 		else if (BASESYNTAX[(int)c] == CCTL)
1846 			return 0;
1847 	}
1848 	return 1;
1849 }
1850 
1851 
1852 /*
1853  * Return true if the argument is a legal variable name (a letter or
1854  * underscore followed by zero or more letters, underscores, and digits).
1855  */
1856 
1857 int
1858 goodname(const char *name)
1859 {
1860 	const char *p;
1861 
1862 	p = name;
1863 	if (! is_name(*p))
1864 		return 0;
1865 	while (*++p) {
1866 		if (! is_in_name(*p))
1867 			return 0;
1868 	}
1869 	return 1;
1870 }
1871 
1872 
1873 int
1874 isassignment(const char *p)
1875 {
1876 	if (!is_name(*p))
1877 		return 0;
1878 	p++;
1879 	for (;;) {
1880 		if (*p == '=')
1881 			return 1;
1882 		else if (!is_in_name(*p))
1883 			return 0;
1884 		p++;
1885 	}
1886 }
1887 
1888 
1889 /*
1890  * Called when an unexpected token is read during the parse.  The argument
1891  * is the token that is expected, or -1 if more than one type of token can
1892  * occur at this point.
1893  */
1894 
1895 static void
1896 synexpect(int token)
1897 {
1898 	char msg[64];
1899 
1900 	if (token >= 0) {
1901 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1902 			tokname[lasttoken], tokname[token]);
1903 	} else {
1904 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1905 	}
1906 	synerror(msg);
1907 }
1908 
1909 
1910 static void
1911 synerror(const char *msg)
1912 {
1913 	if (commandname)
1914 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1915 	outfmt(out2, "Syntax error: %s\n", msg);
1916 	error((char *)NULL);
1917 }
1918 
1919 static void
1920 setprompt(int which)
1921 {
1922 	whichprompt = which;
1923 
1924 #ifndef NO_HISTORY
1925 	if (!el)
1926 #endif
1927 	{
1928 		out2str(getprompt(NULL));
1929 		flushout(out2);
1930 	}
1931 }
1932 
1933 /*
1934  * called by editline -- any expansions to the prompt
1935  *    should be added here.
1936  */
1937 char *
1938 getprompt(void *unused __unused)
1939 {
1940 	static char ps[PROMPTLEN];
1941 	char *fmt;
1942 	const char *pwd;
1943 	int i, trim;
1944 	static char internal_error[] = "??";
1945 
1946 	/*
1947 	 * Select prompt format.
1948 	 */
1949 	switch (whichprompt) {
1950 	case 0:
1951 		fmt = nullstr;
1952 		break;
1953 	case 1:
1954 		fmt = ps1val();
1955 		break;
1956 	case 2:
1957 		fmt = ps2val();
1958 		break;
1959 	default:
1960 		return internal_error;
1961 	}
1962 
1963 	/*
1964 	 * Format prompt string.
1965 	 */
1966 	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1967 		if (*fmt == '\\')
1968 			switch (*++fmt) {
1969 
1970 				/*
1971 				 * Hostname.
1972 				 *
1973 				 * \h specifies just the local hostname,
1974 				 * \H specifies fully-qualified hostname.
1975 				 */
1976 			case 'h':
1977 			case 'H':
1978 				ps[i] = '\0';
1979 				gethostname(&ps[i], PROMPTLEN - i);
1980 				/* Skip to end of hostname. */
1981 				trim = (*fmt == 'h') ? '.' : '\0';
1982 				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1983 					i++;
1984 				break;
1985 
1986 				/*
1987 				 * Working directory.
1988 				 *
1989 				 * \W specifies just the final component,
1990 				 * \w specifies the entire path.
1991 				 */
1992 			case 'W':
1993 			case 'w':
1994 				pwd = lookupvar("PWD");
1995 				if (pwd == NULL)
1996 					pwd = "?";
1997 				if (*fmt == 'W' &&
1998 				    *pwd == '/' && pwd[1] != '\0')
1999 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
2000 					    PROMPTLEN - i);
2001 				else
2002 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
2003 				/* Skip to end of path. */
2004 				while (ps[i + 1] != '\0')
2005 					i++;
2006 				break;
2007 
2008 				/*
2009 				 * Superuser status.
2010 				 *
2011 				 * '$' for normal users, '#' for root.
2012 				 */
2013 			case '$':
2014 				ps[i] = (geteuid() != 0) ? '$' : '#';
2015 				break;
2016 
2017 				/*
2018 				 * A literal \.
2019 				 */
2020 			case '\\':
2021 				ps[i] = '\\';
2022 				break;
2023 
2024 				/*
2025 				 * Emit unrecognized formats verbatim.
2026 				 */
2027 			default:
2028 				ps[i++] = '\\';
2029 				ps[i] = *fmt;
2030 				break;
2031 			}
2032 		else
2033 			ps[i] = *fmt;
2034 	ps[i] = '\0';
2035 	return (ps);
2036 }
2037 
2038 
2039 const char *
2040 expandstr(char *ps)
2041 {
2042 	union node n;
2043 	struct jmploc jmploc;
2044 	struct jmploc *const savehandler = handler;
2045 	const int saveprompt = doprompt;
2046 	struct parsefile *const savetopfile = getcurrentfile();
2047 	struct parser_temp *const saveparser_temp = parser_temp;
2048 	const char *result = NULL;
2049 
2050 	if (!setjmp(jmploc.loc)) {
2051 		handler = &jmploc;
2052 		parser_temp = NULL;
2053 		setinputstring(ps, 1);
2054 		doprompt = 0;
2055 		readtoken1(pgetc(), DQSYNTAX, "\n\n", 0);
2056 		if (backquotelist != NULL)
2057 			error("Command substitution not allowed here");
2058 
2059 		n.narg.type = NARG;
2060 		n.narg.next = NULL;
2061 		n.narg.text = wordtext;
2062 		n.narg.backquote = backquotelist;
2063 
2064 		expandarg(&n, NULL, 0);
2065 		result = stackblock();
2066 		INTOFF;
2067 	}
2068 	handler = savehandler;
2069 	doprompt = saveprompt;
2070 	popfilesupto(savetopfile);
2071 	if (parser_temp != saveparser_temp) {
2072 		parser_temp_free_all();
2073 		parser_temp = saveparser_temp;
2074 	}
2075 	if (result != NULL) {
2076 		INTON;
2077 	} else if (exception == EXINT)
2078 		raise(SIGINT);
2079 	return result;
2080 }
2081