xref: /dflybsd-src/bin/sh/parser.c (revision 7f8c68295613ce24cc71827cf210cb3d1e3bc69b)
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. 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  * @(#)parser.c	8.7 (Berkeley) 5/16/95
33  * $FreeBSD: head/bin/sh/parser.c 245382 2013-01-13 19:26:33Z jilles $
34  */
35 
36 #include <stdio.h>
37 #include <stdlib.h>
38 #include <unistd.h>
39 
40 #include "shell.h"
41 #include "parser.h"
42 #include "nodes.h"
43 #include "expand.h"	/* defines rmescapes() */
44 #include "syntax.h"
45 #include "options.h"
46 #include "input.h"
47 #include "output.h"
48 #include "var.h"
49 #include "error.h"
50 #include "memalloc.h"
51 #include "mystring.h"
52 #include "alias.h"
53 #include "show.h"
54 #include "eval.h"
55 #include "exec.h"	/* to check for special builtins */
56 #ifndef NO_HISTORY
57 #include "myhistedit.h"
58 #endif
59 
60 /*
61  * Shell command parser.
62  */
63 
64 #define	EOFMARKLEN	79
65 #define	PROMPTLEN	128
66 
67 /* values of checkkwd variable */
68 #define CHKALIAS	0x1
69 #define CHKKWD		0x2
70 #define CHKNL		0x4
71 
72 /* values returned by readtoken */
73 #include "token.h"
74 
75 
76 
77 struct heredoc {
78 	struct heredoc *next;	/* next here document in list */
79 	union node *here;		/* redirection node */
80 	char *eofmark;		/* string indicating end of input */
81 	int striptabs;		/* if set, strip leading tabs */
82 };
83 
84 struct parser_temp {
85 	struct parser_temp *next;
86 	void *data;
87 };
88 
89 
90 static struct heredoc *heredoclist;	/* list of here documents to read */
91 static int doprompt;		/* if set, prompt the user */
92 static int needprompt;		/* true if interactive and at start of line */
93 static int lasttoken;		/* last token read */
94 MKINIT int tokpushback;		/* last token pushed back */
95 static char *wordtext;		/* text of last word returned by readtoken */
96 MKINIT int checkkwd;            /* 1 == check for kwds, 2 == also eat newlines */
97 static struct nodelist *backquotelist;
98 static union node *redirnode;
99 static struct heredoc *heredoc;
100 static int quoteflag;		/* set if (part of) last token was quoted */
101 static int startlinno;		/* line # where last token started */
102 static int funclinno;		/* line # where the current function started */
103 static struct parser_temp *parser_temp;
104 
105 
106 static union node *list(int, int);
107 static union node *andor(void);
108 static union node *pipeline(void);
109 static union node *command(void);
110 static union node *simplecmd(union node **, union node *);
111 static union node *makename(void);
112 static void parsefname(void);
113 static void parseheredoc(void);
114 static int peektoken(void);
115 static int readtoken(void);
116 static int xxreadtoken(void);
117 static int readtoken1(int, char const *, char *, int);
118 static int noexpand(char *);
119 static void synexpect(int) __dead2;
120 static void synerror(const char *) __dead2;
121 static void setprompt(int);
122 
123 
124 static void *
125 parser_temp_alloc(size_t len)
126 {
127 	struct parser_temp *t;
128 
129 	INTOFF;
130 	t = ckmalloc(sizeof(*t));
131 	t->data = NULL;
132 	t->next = parser_temp;
133 	parser_temp = t;
134 	t->data = ckmalloc(len);
135 	INTON;
136 	return t->data;
137 }
138 
139 
140 static void *
141 parser_temp_realloc(void *ptr, size_t len)
142 {
143 	struct parser_temp *t;
144 
145 	INTOFF;
146 	t = parser_temp;
147 	if (ptr != t->data)
148 		error("bug: parser_temp_realloc misused");
149 	t->data = ckrealloc(t->data, len);
150 	INTON;
151 	return t->data;
152 }
153 
154 
155 static void
156 parser_temp_free_upto(void *ptr)
157 {
158 	struct parser_temp *t;
159 	int done = 0;
160 
161 	INTOFF;
162 	while (parser_temp != NULL && !done) {
163 		t = parser_temp;
164 		parser_temp = t->next;
165 		done = t->data == ptr;
166 		ckfree(t->data);
167 		ckfree(t);
168 	}
169 	INTON;
170 	if (!done)
171 		error("bug: parser_temp_free_upto misused");
172 }
173 
174 
175 static void
176 parser_temp_free_all(void)
177 {
178 	struct parser_temp *t;
179 
180 	INTOFF;
181 	while (parser_temp != NULL) {
182 		t = parser_temp;
183 		parser_temp = t->next;
184 		ckfree(t->data);
185 		ckfree(t);
186 	}
187 	INTON;
188 }
189 
190 
191 /*
192  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
193  * valid parse tree indicating a blank line.)
194  */
195 
196 union node *
197 parsecmd(int interact)
198 {
199 	int t;
200 
201 	/* This assumes the parser is not re-entered,
202 	 * which could happen if we add command substitution on PS1/PS2.
203 	 */
204 	parser_temp_free_all();
205 	heredoclist = NULL;
206 
207 	tokpushback = 0;
208 	doprompt = interact;
209 	if (doprompt)
210 		setprompt(1);
211 	else
212 		setprompt(0);
213 	needprompt = 0;
214 	t = readtoken();
215 	if (t == TEOF)
216 		return NEOF;
217 	if (t == TNL)
218 		return NULL;
219 	tokpushback++;
220 	return list(1, 1);
221 }
222 
223 
224 static union node *
225 list(int nlflag, int erflag)
226 {
227 	union node *ntop, *n1, *n2, *n3;
228 	int tok;
229 
230 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
231 	if (!nlflag && !erflag && tokendlist[peektoken()])
232 		return NULL;
233 	ntop = n1 = NULL;
234 	for (;;) {
235 		n2 = andor();
236 		tok = readtoken();
237 		if (tok == TBACKGND) {
238 			if (n2 != NULL && n2->type == NPIPE) {
239 				n2->npipe.backgnd = 1;
240 			} else if (n2 != NULL && n2->type == NREDIR) {
241 				n2->type = NBACKGND;
242 			} else {
243 				n3 = (union node *)stalloc(sizeof (struct nredir));
244 				n3->type = NBACKGND;
245 				n3->nredir.n = n2;
246 				n3->nredir.redirect = NULL;
247 				n2 = n3;
248 			}
249 		}
250 		if (ntop == NULL)
251 			ntop = n2;
252 		else if (n1 == NULL) {
253 			n1 = (union node *)stalloc(sizeof (struct nbinary));
254 			n1->type = NSEMI;
255 			n1->nbinary.ch1 = ntop;
256 			n1->nbinary.ch2 = n2;
257 			ntop = n1;
258 		}
259 		else {
260 			n3 = (union node *)stalloc(sizeof (struct nbinary));
261 			n3->type = NSEMI;
262 			n3->nbinary.ch1 = n1->nbinary.ch2;
263 			n3->nbinary.ch2 = n2;
264 			n1->nbinary.ch2 = n3;
265 			n1 = n3;
266 		}
267 		switch (tok) {
268 		case TBACKGND:
269 		case TSEMI:
270 			tok = readtoken();
271 			/* FALLTHROUGH */
272 		case TNL:
273 			if (tok == TNL) {
274 				parseheredoc();
275 				if (nlflag)
276 					return ntop;
277 			} else if (tok == TEOF && nlflag) {
278 				parseheredoc();
279 				return ntop;
280 			} else {
281 				tokpushback++;
282 			}
283 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
284 			if (!nlflag && (erflag ? peektoken() == TEOF :
285 			    tokendlist[peektoken()]))
286 				return ntop;
287 			break;
288 		case TEOF:
289 			if (heredoclist)
290 				parseheredoc();
291 			else
292 				pungetc();		/* push back EOF on input */
293 			return ntop;
294 		default:
295 			if (nlflag || erflag)
296 				synexpect(-1);
297 			tokpushback++;
298 			return ntop;
299 		}
300 	}
301 }
302 
303 
304 
305 static union node *
306 andor(void)
307 {
308 	union node *n1, *n2, *n3;
309 	int t;
310 
311 	n1 = pipeline();
312 	for (;;) {
313 		if ((t = readtoken()) == TAND) {
314 			t = NAND;
315 		} else if (t == TOR) {
316 			t = NOR;
317 		} else {
318 			tokpushback++;
319 			return n1;
320 		}
321 		n2 = pipeline();
322 		n3 = (union node *)stalloc(sizeof (struct nbinary));
323 		n3->type = t;
324 		n3->nbinary.ch1 = n1;
325 		n3->nbinary.ch2 = n2;
326 		n1 = n3;
327 	}
328 }
329 
330 
331 
332 static union node *
333 pipeline(void)
334 {
335 	union node *n1, *n2, *pipenode;
336 	struct nodelist *lp, *prev;
337 	int negate, t;
338 
339 	negate = 0;
340 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
341 	TRACE(("pipeline: entered\n"));
342 	while (readtoken() == TNOT)
343 		negate = !negate;
344 	tokpushback++;
345 	n1 = command();
346 	if (readtoken() == TPIPE) {
347 		pipenode = (union node *)stalloc(sizeof (struct npipe));
348 		pipenode->type = NPIPE;
349 		pipenode->npipe.backgnd = 0;
350 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
351 		pipenode->npipe.cmdlist = lp;
352 		lp->n = n1;
353 		do {
354 			prev = lp;
355 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
356 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
357 			t = readtoken();
358 			tokpushback++;
359 			if (t == TNOT)
360 				lp->n = pipeline();
361 			else
362 				lp->n = command();
363 			prev->next = lp;
364 		} while (readtoken() == TPIPE);
365 		lp->next = NULL;
366 		n1 = pipenode;
367 	}
368 	tokpushback++;
369 	if (negate) {
370 		n2 = (union node *)stalloc(sizeof (struct nnot));
371 		n2->type = NNOT;
372 		n2->nnot.com = n1;
373 		return n2;
374 	} else
375 		return n1;
376 }
377 
378 
379 
380 static union node *
381 command(void)
382 {
383 	union node *n1, *n2;
384 	union node *ap, **app;
385 	union node *cp, **cpp;
386 	union node *redir, **rpp;
387 	int t;
388 	int is_subshell;
389 
390 	checkkwd = CHKNL | CHKKWD | CHKALIAS;
391 	is_subshell = 0;
392 	redir = NULL;
393 	n1 = NULL;
394 	rpp = &redir;
395 
396 	/* Check for redirection which may precede command */
397 	while (readtoken() == TREDIR) {
398 		*rpp = n2 = redirnode;
399 		rpp = &n2->nfile.next;
400 		parsefname();
401 	}
402 	tokpushback++;
403 
404 	switch (readtoken()) {
405 	case TIF:
406 		n1 = (union node *)stalloc(sizeof (struct nif));
407 		n1->type = NIF;
408 		if ((n1->nif.test = list(0, 0)) == NULL)
409 			synexpect(-1);
410 		if (readtoken() != TTHEN)
411 			synexpect(TTHEN);
412 		n1->nif.ifpart = list(0, 0);
413 		n2 = n1;
414 		while (readtoken() == TELIF) {
415 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
416 			n2 = n2->nif.elsepart;
417 			n2->type = NIF;
418 			if ((n2->nif.test = list(0, 0)) == NULL)
419 				synexpect(-1);
420 			if (readtoken() != TTHEN)
421 				synexpect(TTHEN);
422 			n2->nif.ifpart = list(0, 0);
423 		}
424 		if (lasttoken == TELSE)
425 			n2->nif.elsepart = list(0, 0);
426 		else {
427 			n2->nif.elsepart = NULL;
428 			tokpushback++;
429 		}
430 		if (readtoken() != TFI)
431 			synexpect(TFI);
432 		checkkwd = CHKKWD | CHKALIAS;
433 		break;
434 	case TWHILE:
435 	case TUNTIL: {
436 		int got;
437 		n1 = (union node *)stalloc(sizeof (struct nbinary));
438 		n1->type = (lasttoken == TWHILE)? NWHILE : NUNTIL;
439 		if ((n1->nbinary.ch1 = list(0, 0)) == NULL)
440 			synexpect(-1);
441 		if ((got=readtoken()) != TDO) {
442 TRACE(("expecting DO got %s %s\n", tokname[got], got == TWORD ? wordtext : ""));
443 			synexpect(TDO);
444 		}
445 		n1->nbinary.ch2 = list(0, 0);
446 		if (readtoken() != TDONE)
447 			synexpect(TDONE);
448 		checkkwd = CHKKWD | CHKALIAS;
449 		break;
450 	}
451 	case TFOR:
452 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
453 			synerror("Bad for loop variable");
454 		n1 = (union node *)stalloc(sizeof (struct nfor));
455 		n1->type = NFOR;
456 		n1->nfor.var = wordtext;
457 		while (readtoken() == TNL)
458 			;
459 		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
460 			app = &ap;
461 			while (readtoken() == TWORD) {
462 				n2 = (union node *)stalloc(sizeof (struct narg));
463 				n2->type = NARG;
464 				n2->narg.text = wordtext;
465 				n2->narg.backquote = backquotelist;
466 				*app = n2;
467 				app = &n2->narg.next;
468 			}
469 			*app = NULL;
470 			n1->nfor.args = ap;
471 			if (lasttoken != TNL && lasttoken != TSEMI)
472 				synexpect(-1);
473 		} else {
474 			static char argvars[5] = {
475 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
476 			};
477 			n2 = (union node *)stalloc(sizeof (struct narg));
478 			n2->type = NARG;
479 			n2->narg.text = argvars;
480 			n2->narg.backquote = NULL;
481 			n2->narg.next = NULL;
482 			n1->nfor.args = n2;
483 			/*
484 			 * Newline or semicolon here is optional (but note
485 			 * that the original Bourne shell only allowed NL).
486 			 */
487 			if (lasttoken != TNL && lasttoken != TSEMI)
488 				tokpushback++;
489 		}
490 		checkkwd = CHKNL | CHKKWD | CHKALIAS;
491 		if ((t = readtoken()) == TDO)
492 			t = TDONE;
493 		else if (t == TBEGIN)
494 			t = TEND;
495 		else
496 			synexpect(-1);
497 		n1->nfor.body = list(0, 0);
498 		if (readtoken() != t)
499 			synexpect(t);
500 		checkkwd = CHKKWD | CHKALIAS;
501 		break;
502 	case TCASE:
503 		n1 = (union node *)stalloc(sizeof (struct ncase));
504 		n1->type = NCASE;
505 		if (readtoken() != TWORD)
506 			synexpect(TWORD);
507 		n1->ncase.expr = n2 = (union node *)stalloc(sizeof (struct narg));
508 		n2->type = NARG;
509 		n2->narg.text = wordtext;
510 		n2->narg.backquote = backquotelist;
511 		n2->narg.next = NULL;
512 		while (readtoken() == TNL);
513 		if (lasttoken != TWORD || ! equal(wordtext, "in"))
514 			synerror("expecting \"in\"");
515 		cpp = &n1->ncase.cases;
516 		checkkwd = CHKNL | CHKKWD, readtoken();
517 		while (lasttoken != TESAC) {
518 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
519 			cp->type = NCLIST;
520 			app = &cp->nclist.pattern;
521 			if (lasttoken == TLP)
522 				readtoken();
523 			for (;;) {
524 				*app = ap = (union node *)stalloc(sizeof (struct narg));
525 				ap->type = NARG;
526 				ap->narg.text = wordtext;
527 				ap->narg.backquote = backquotelist;
528 				checkkwd = CHKNL | CHKKWD;
529 				if (readtoken() != TPIPE)
530 					break;
531 				app = &ap->narg.next;
532 				readtoken();
533 			}
534 			ap->narg.next = NULL;
535 			if (lasttoken != TRP)
536 				synexpect(TRP);
537 			cp->nclist.body = list(0, 0);
538 
539 			checkkwd = CHKNL | CHKKWD | CHKALIAS;
540 			if ((t = readtoken()) != TESAC) {
541 				if (t == TENDCASE)
542 					;
543 				else if (t == TFALLTHRU)
544 					cp->type = NCLISTFALLTHRU;
545 				else
546 					synexpect(TENDCASE);
547 				checkkwd = CHKNL | CHKKWD, readtoken();
548 			}
549 			cpp = &cp->nclist.next;
550 		}
551 		*cpp = NULL;
552 		checkkwd = CHKKWD | CHKALIAS;
553 		break;
554 	case TLP:
555 		n1 = (union node *)stalloc(sizeof (struct nredir));
556 		n1->type = NSUBSHELL;
557 		n1->nredir.n = list(0, 0);
558 		n1->nredir.redirect = NULL;
559 		if (readtoken() != TRP)
560 			synexpect(TRP);
561 		checkkwd = CHKKWD | CHKALIAS;
562 		is_subshell = 1;
563 		break;
564 	case TBEGIN:
565 		n1 = list(0, 0);
566 		if (readtoken() != TEND)
567 			synexpect(TEND);
568 		checkkwd = CHKKWD | CHKALIAS;
569 		break;
570 	/* Handle an empty command like other simple commands.  */
571 	case TBACKGND:
572 	case TSEMI:
573 	case TAND:
574 	case TOR:
575 		/*
576 		 * An empty command before a ; doesn't make much sense, and
577 		 * should certainly be disallowed in the case of `if ;'.
578 		 */
579 		if (!redir)
580 			synexpect(-1);
581 	case TNL:
582 	case TEOF:
583 	case TWORD:
584 	case TRP:
585 		tokpushback++;
586 		n1 = simplecmd(rpp, redir);
587 		return n1;
588 	default:
589 		synexpect(-1);
590 	}
591 
592 	/* Now check for redirection which may follow command */
593 	while (readtoken() == TREDIR) {
594 		*rpp = n2 = redirnode;
595 		rpp = &n2->nfile.next;
596 		parsefname();
597 	}
598 	tokpushback++;
599 	*rpp = NULL;
600 	if (redir) {
601 		if (!is_subshell) {
602 			n2 = (union node *)stalloc(sizeof (struct nredir));
603 			n2->type = NREDIR;
604 			n2->nredir.n = n1;
605 			n1 = n2;
606 		}
607 		n1->nredir.redirect = redir;
608 	}
609 
610 	return n1;
611 }
612 
613 
614 static union node *
615 simplecmd(union node **rpp, union node *redir)
616 {
617 	union node *args, **app;
618 	union node **orig_rpp = rpp;
619 	union node *n = NULL;
620 	int special;
621 	int savecheckkwd;
622 
623 	/* If we don't have any redirections already, then we must reset */
624 	/* rpp to be the address of the local redir variable.  */
625 	if (redir == NULL)
626 		rpp = &redir;
627 
628 	args = NULL;
629 	app = &args;
630 	/*
631 	 * We save the incoming value, because we need this for shell
632 	 * functions.  There can not be a redirect or an argument between
633 	 * the function name and the open parenthesis.
634 	 */
635 	orig_rpp = rpp;
636 
637 	savecheckkwd = CHKALIAS;
638 
639 	for (;;) {
640 		checkkwd = savecheckkwd;
641 		if (readtoken() == TWORD) {
642 			n = (union node *)stalloc(sizeof (struct narg));
643 			n->type = NARG;
644 			n->narg.text = wordtext;
645 			n->narg.backquote = backquotelist;
646 			*app = n;
647 			app = &n->narg.next;
648 			if (savecheckkwd != 0 && !isassignment(wordtext))
649 				savecheckkwd = 0;
650 		} else if (lasttoken == TREDIR) {
651 			*rpp = n = redirnode;
652 			rpp = &n->nfile.next;
653 			parsefname();	/* read name of redirection file */
654 		} else if (lasttoken == TLP && app == &args->narg.next
655 					    && rpp == orig_rpp) {
656 			/* We have a function */
657 			if (readtoken() != TRP)
658 				synexpect(TRP);
659 			funclinno = plinno;
660 			/*
661 			 * - Require plain text.
662 			 * - Functions with '/' cannot be called.
663 			 * - Reject name=().
664 			 * - Reject ksh extended glob patterns.
665 			 */
666 			if (!noexpand(n->narg.text) || quoteflag ||
667 			    strchr(n->narg.text, '/') ||
668 			    strchr("!%*+-=?@}~",
669 				n->narg.text[strlen(n->narg.text) - 1]))
670 				synerror("Bad function name");
671 			rmescapes(n->narg.text);
672 			if (find_builtin(n->narg.text, &special) >= 0 &&
673 			    special)
674 				synerror("Cannot override a special builtin with a function");
675 			n->type = NDEFUN;
676 			n->narg.next = command();
677 			funclinno = 0;
678 			return n;
679 		} else {
680 			tokpushback++;
681 			break;
682 		}
683 	}
684 	*app = NULL;
685 	*rpp = NULL;
686 	n = (union node *)stalloc(sizeof (struct ncmd));
687 	n->type = NCMD;
688 	n->ncmd.args = args;
689 	n->ncmd.redirect = redir;
690 	return n;
691 }
692 
693 static union node *
694 makename(void)
695 {
696 	union node *n;
697 
698 	n = (union node *)stalloc(sizeof (struct narg));
699 	n->type = NARG;
700 	n->narg.next = NULL;
701 	n->narg.text = wordtext;
702 	n->narg.backquote = backquotelist;
703 	return n;
704 }
705 
706 void
707 fixredir(union node *n, const char *text, int err)
708 {
709 	TRACE(("Fix redir %s %d\n", text, err));
710 	if (!err)
711 		n->ndup.vname = NULL;
712 
713 	if (is_digit(text[0]) && text[1] == '\0')
714 		n->ndup.dupfd = digit_val(text[0]);
715 	else if (text[0] == '-' && text[1] == '\0')
716 		n->ndup.dupfd = -1;
717 	else {
718 
719 		if (err)
720 			synerror("Bad fd number");
721 		else
722 			n->ndup.vname = makename();
723 	}
724 }
725 
726 
727 static void
728 parsefname(void)
729 {
730 	union node *n = redirnode;
731 
732 	if (readtoken() != TWORD)
733 		synexpect(-1);
734 	if (n->type == NHERE) {
735 		struct heredoc *here = heredoc;
736 		struct heredoc *p;
737 		int i;
738 
739 		if (quoteflag == 0)
740 			n->type = NXHERE;
741 		TRACE(("Here document %d\n", n->type));
742 		if (here->striptabs) {
743 			while (*wordtext == '\t')
744 				wordtext++;
745 		}
746 		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
747 			synerror("Illegal eof marker for << redirection");
748 		rmescapes(wordtext);
749 		here->eofmark = wordtext;
750 		here->next = NULL;
751 		if (heredoclist == NULL)
752 			heredoclist = here;
753 		else {
754 			for (p = heredoclist ; p->next ; p = p->next);
755 			p->next = here;
756 		}
757 	} else if (n->type == NTOFD || n->type == NFROMFD) {
758 		fixredir(n, wordtext, 0);
759 	} else {
760 		n->nfile.fname = makename();
761 	}
762 }
763 
764 
765 /*
766  * Input any here documents.
767  */
768 
769 static void
770 parseheredoc(void)
771 {
772 	struct heredoc *here;
773 	union node *n;
774 
775 	while (heredoclist) {
776 		here = heredoclist;
777 		heredoclist = here->next;
778 		if (needprompt) {
779 			setprompt(2);
780 			needprompt = 0;
781 		}
782 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
783 				here->eofmark, here->striptabs);
784 		n = (union node *)stalloc(sizeof (struct narg));
785 		n->narg.type = NARG;
786 		n->narg.next = NULL;
787 		n->narg.text = wordtext;
788 		n->narg.backquote = backquotelist;
789 		here->here->nhere.doc = n;
790 	}
791 }
792 
793 static int
794 peektoken(void)
795 {
796 	int t;
797 
798 	t = readtoken();
799 	tokpushback++;
800 	return (t);
801 }
802 
803 static int
804 readtoken(void)
805 {
806 	int t;
807 	struct alias *ap;
808 #ifdef DEBUG
809 	int alreadyseen = tokpushback;
810 #endif
811 
812 	top:
813 	t = xxreadtoken();
814 
815 	/*
816 	 * eat newlines
817 	 */
818 	if (checkkwd & CHKNL) {
819 		while (t == TNL) {
820 			parseheredoc();
821 			t = xxreadtoken();
822 		}
823 	}
824 
825 	/*
826 	 * check for keywords and aliases
827 	 */
828 	if (t == TWORD && !quoteflag)
829 	{
830 		const char * const *pp;
831 
832 		if (checkkwd & CHKKWD)
833 			for (pp = parsekwd; *pp; pp++) {
834 				if (**pp == *wordtext && equal(*pp, wordtext))
835 				{
836 					lasttoken = t = pp - parsekwd + KWDOFFSET;
837 					TRACE(("keyword %s recognized\n", tokname[t]));
838 					goto out;
839 				}
840 			}
841 		if (checkkwd & CHKALIAS &&
842 		    (ap = lookupalias(wordtext, 1)) != NULL) {
843 			pushstring(ap->val, strlen(ap->val), ap);
844 			goto top;
845 		}
846 	}
847 out:
848 	if (t != TNOT)
849 		checkkwd = 0;
850 
851 #ifdef DEBUG
852 	if (!alreadyseen)
853 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
854 	else
855 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
856 #endif
857 	return (t);
858 }
859 
860 
861 /*
862  * Read the next input token.
863  * If the token is a word, we set backquotelist to the list of cmds in
864  *	backquotes.  We set quoteflag to true if any part of the word was
865  *	quoted.
866  * If the token is TREDIR, then we set redirnode to a structure containing
867  *	the redirection.
868  * In all cases, the variable startlinno is set to the number of the line
869  *	on which the token starts.
870  *
871  * [Change comment:  here documents and internal procedures]
872  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
873  *  word parsing code into a separate routine.  In this case, readtoken
874  *  doesn't need to have any internal procedures, but parseword does.
875  *  We could also make parseoperator in essence the main routine, and
876  *  have parseword (readtoken1?) handle both words and redirection.]
877  */
878 
879 #define RETURN(token)	return lasttoken = token
880 
881 static int
882 xxreadtoken(void)
883 {
884 	int c;
885 
886 	if (tokpushback) {
887 		tokpushback = 0;
888 		return lasttoken;
889 	}
890 	if (needprompt) {
891 		setprompt(2);
892 		needprompt = 0;
893 	}
894 	startlinno = plinno;
895 	for (;;) {	/* until token or start of word found */
896 		c = pgetc_macro();
897 		switch (c) {
898 		case ' ': case '\t':
899 			continue;
900 		case '#':
901 			while ((c = pgetc()) != '\n' && c != PEOF);
902 			pungetc();
903 			continue;
904 		case '\\':
905 			if (pgetc() == '\n') {
906 				startlinno = ++plinno;
907 				if (doprompt)
908 					setprompt(2);
909 				else
910 					setprompt(0);
911 				continue;
912 			}
913 			pungetc();
914 			goto breakloop;
915 		case '\n':
916 			plinno++;
917 			needprompt = doprompt;
918 			RETURN(TNL);
919 		case PEOF:
920 			RETURN(TEOF);
921 		case '&':
922 			if (pgetc() == '&')
923 				RETURN(TAND);
924 			pungetc();
925 			RETURN(TBACKGND);
926 		case '|':
927 			if (pgetc() == '|')
928 				RETURN(TOR);
929 			pungetc();
930 			RETURN(TPIPE);
931 		case ';':
932 			c = pgetc();
933 			if (c == ';')
934 				RETURN(TENDCASE);
935 			else if (c == '&')
936 				RETURN(TFALLTHRU);
937 			pungetc();
938 			RETURN(TSEMI);
939 		case '(':
940 			RETURN(TLP);
941 		case ')':
942 			RETURN(TRP);
943 		default:
944 			goto breakloop;
945 		}
946 	}
947 breakloop:
948 	return readtoken1(c, BASESYNTAX, NULL, 0);
949 #undef RETURN
950 }
951 
952 
953 #define MAXNEST_STATIC 8
954 struct tokenstate
955 {
956 	const char *syntax; /* *SYNTAX */
957 	int parenlevel; /* levels of parentheses in arithmetic */
958 	enum tokenstate_category
959 	{
960 		TSTATE_TOP,
961 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
962 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
963 		TSTATE_ARITH
964 	} category;
965 };
966 
967 
968 /*
969  * Called to parse command substitutions.
970  */
971 
972 static char *
973 parsebackq(char *out, struct nodelist **pbqlist,
974     int oldstyle, int dblquote, int quoted)
975 {
976 	struct nodelist **nlpp;
977 	union node *n;
978 	char *volatile str;
979 	struct jmploc jmploc;
980 	struct jmploc *const savehandler = handler;
981 	int savelen;
982 	int saveprompt;
983 	const int bq_startlinno = plinno;
984 	char *volatile ostr = NULL;
985 	struct parsefile *const savetopfile = getcurrentfile();
986 	struct heredoc *const saveheredoclist = heredoclist;
987 	struct heredoc *here;
988 
989 	str = NULL;
990 	if (setjmp(jmploc.loc)) {
991 		popfilesupto(savetopfile);
992 		if (str)
993 			ckfree(str);
994 		if (ostr)
995 			ckfree(ostr);
996 		heredoclist = saveheredoclist;
997 		handler = savehandler;
998 		if (exception == EXERROR) {
999 			startlinno = bq_startlinno;
1000 			synerror("Error in command substitution");
1001 		}
1002 		longjmp(handler->loc, 1);
1003 	}
1004 	INTOFF;
1005 	savelen = out - stackblock();
1006 	if (savelen > 0) {
1007 		str = ckmalloc(savelen);
1008 		memcpy(str, stackblock(), savelen);
1009 	}
1010 	handler = &jmploc;
1011 	heredoclist = NULL;
1012 	INTON;
1013 	if (oldstyle) {
1014 		/*
1015 		 * We must read until the closing backquote, giving special
1016 		 * treatment to some slashes, and then push the string and
1017 		 * reread it as input, interpreting it normally.
1018 		 */
1019 		char *oout;
1020 		int c, olen;
1021 
1022 		STARTSTACKSTR(oout);
1023 		for (;;) {
1024 			if (needprompt) {
1025 				setprompt(2);
1026 				needprompt = 0;
1027 			}
1028 			CHECKSTRSPACE(2, oout);
1029 			switch (c = pgetc()) {
1030 			case '`':
1031 				goto done;
1032 
1033 			case '\\':
1034 				if ((c = pgetc()) == '\n') {
1035 					plinno++;
1036 					if (doprompt)
1037 						setprompt(2);
1038 					else
1039 						setprompt(0);
1040 					/*
1041 					 * If eating a newline, avoid putting
1042 					 * the newline into the new character
1043 					 * stream (via the USTPUTC after the
1044 					 * switch).
1045 					 */
1046 					continue;
1047 				}
1048 				if (c != '\\' && c != '`' && c != '$'
1049 				    && (!dblquote || c != '"'))
1050 					USTPUTC('\\', oout);
1051 				break;
1052 
1053 			case '\n':
1054 				plinno++;
1055 				needprompt = doprompt;
1056 				break;
1057 
1058 			case PEOF:
1059 				startlinno = plinno;
1060 				synerror("EOF in backquote substitution");
1061 				break;
1062 
1063 			default:
1064 				break;
1065 			}
1066 			USTPUTC(c, oout);
1067 		}
1068 done:
1069 		USTPUTC('\0', oout);
1070 		olen = oout - stackblock();
1071 		INTOFF;
1072 		ostr = ckmalloc(olen);
1073 		memcpy(ostr, stackblock(), olen);
1074 		setinputstring(ostr, 1);
1075 		INTON;
1076 	}
1077 	nlpp = pbqlist;
1078 	while (*nlpp)
1079 		nlpp = &(*nlpp)->next;
1080 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1081 	(*nlpp)->next = NULL;
1082 
1083 	if (oldstyle) {
1084 		saveprompt = doprompt;
1085 		doprompt = 0;
1086 	}
1087 
1088 	n = list(0, oldstyle);
1089 
1090 	if (oldstyle)
1091 		doprompt = saveprompt;
1092 	else {
1093 		if (readtoken() != TRP)
1094 			synexpect(TRP);
1095 	}
1096 
1097 	(*nlpp)->n = n;
1098 	if (oldstyle) {
1099 		/*
1100 		 * Start reading from old file again, ignoring any pushed back
1101 		 * tokens left from the backquote parsing
1102 		 */
1103 		popfile();
1104 		tokpushback = 0;
1105 	}
1106 	STARTSTACKSTR(out);
1107 	CHECKSTRSPACE(savelen + 1, out);
1108 	INTOFF;
1109 	if (str) {
1110 		memcpy(out, str, savelen);
1111 		STADJUST(savelen, out);
1112 		ckfree(str);
1113 		str = NULL;
1114 	}
1115 	if (ostr) {
1116 		ckfree(ostr);
1117 		ostr = NULL;
1118 	}
1119 	here = saveheredoclist;
1120 	if (here != NULL) {
1121 		while (here->next != NULL)
1122 			here = here->next;
1123 		here->next = heredoclist;
1124 		heredoclist = saveheredoclist;
1125 	}
1126 	handler = savehandler;
1127 	INTON;
1128 	if (quoted)
1129 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1130 	else
1131 		USTPUTC(CTLBACKQ, out);
1132 	return out;
1133 }
1134 
1135 
1136 /*
1137  * Called to parse a backslash escape sequence inside $'...'.
1138  * The backslash has already been read.
1139  */
1140 static char *
1141 readcstyleesc(char *out)
1142 {
1143 	int c, v, i, n;
1144 
1145 	c = pgetc();
1146 	switch (c) {
1147 	case '\0':
1148 		synerror("Unterminated quoted string");
1149 	case '\n':
1150 		plinno++;
1151 		if (doprompt)
1152 			setprompt(2);
1153 		else
1154 			setprompt(0);
1155 		return out;
1156 	case '\\':
1157 	case '\'':
1158 	case '"':
1159 		v = c;
1160 		break;
1161 	case 'a': v = '\a'; break;
1162 	case 'b': v = '\b'; break;
1163 	case 'e': v = '\033'; break;
1164 	case 'f': v = '\f'; break;
1165 	case 'n': v = '\n'; break;
1166 	case 'r': v = '\r'; break;
1167 	case 't': v = '\t'; break;
1168 	case 'v': v = '\v'; break;
1169 	case 'x':
1170 		  v = 0;
1171 		  for (;;) {
1172 			  c = pgetc();
1173 			  if (c >= '0' && c <= '9')
1174 				  v = (v << 4) + c - '0';
1175 			  else if (c >= 'A' && c <= 'F')
1176 				  v = (v << 4) + c - 'A' + 10;
1177 			  else if (c >= 'a' && c <= 'f')
1178 				  v = (v << 4) + c - 'a' + 10;
1179 			  else
1180 				  break;
1181 		  }
1182 		  pungetc();
1183 		  break;
1184 	case '0': case '1': case '2': case '3':
1185 	case '4': case '5': case '6': case '7':
1186 		  v = c - '0';
1187 		  c = pgetc();
1188 		  if (c >= '0' && c <= '7') {
1189 			  v <<= 3;
1190 			  v += c - '0';
1191 			  c = pgetc();
1192 			  if (c >= '0' && c <= '7') {
1193 				  v <<= 3;
1194 				  v += c - '0';
1195 			  } else
1196 				  pungetc();
1197 		  } else
1198 			  pungetc();
1199 		  break;
1200 	case 'c':
1201 		  c = pgetc();
1202 		  if (c < 0x3f || c > 0x7a || c == 0x60)
1203 			  synerror("Bad escape sequence");
1204 		  if (c == '\\' && pgetc() != '\\')
1205 			  synerror("Bad escape sequence");
1206 		  if (c == '?')
1207 			  v = 127;
1208 		  else
1209 			  v = c & 0x1f;
1210 		  break;
1211 	case 'u':
1212 	case 'U':
1213 		  n = c == 'U' ? 8 : 4;
1214 		  v = 0;
1215 		  for (i = 0; i < n; i++) {
1216 			  c = pgetc();
1217 			  if (c >= '0' && c <= '9')
1218 				  v = (v << 4) + c - '0';
1219 			  else if (c >= 'A' && c <= 'F')
1220 				  v = (v << 4) + c - 'A' + 10;
1221 			  else if (c >= 'a' && c <= 'f')
1222 				  v = (v << 4) + c - 'a' + 10;
1223 			  else
1224 				  synerror("Bad escape sequence");
1225 		  }
1226 		  if (v == 0 || (v >= 0xd800 && v <= 0xdfff))
1227 			  synerror("Bad escape sequence");
1228 		  /* We really need iconv here. */
1229 		  if (initial_localeisutf8 && v > 127) {
1230 			  CHECKSTRSPACE(4, out);
1231 			  /*
1232 			   * We cannot use wctomb() as the locale may have
1233 			   * changed.
1234 			   */
1235 			  if (v <= 0x7ff) {
1236 				  USTPUTC(0xc0 | v >> 6, out);
1237 				  USTPUTC(0x80 | (v & 0x3f), out);
1238 				  return out;
1239 			  } else if (v <= 0xffff) {
1240 				  USTPUTC(0xe0 | v >> 12, out);
1241 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1242 				  USTPUTC(0x80 | (v & 0x3f), out);
1243 				  return out;
1244 			  } else if (v <= 0x10ffff) {
1245 				  USTPUTC(0xf0 | v >> 18, out);
1246 				  USTPUTC(0x80 | ((v >> 12) & 0x3f), out);
1247 				  USTPUTC(0x80 | ((v >> 6) & 0x3f), out);
1248 				  USTPUTC(0x80 | (v & 0x3f), out);
1249 				  return out;
1250 			  }
1251 		  }
1252 		  if (v > 127)
1253 			  v = '?';
1254 		  break;
1255 	default:
1256 		  synerror("Bad escape sequence");
1257 	}
1258 	v = (char)v;
1259 	/*
1260 	 * We can't handle NUL bytes.
1261 	 * POSIX says we should skip till the closing quote.
1262 	 */
1263 	if (v == '\0') {
1264 		while ((c = pgetc()) != '\'') {
1265 			if (c == '\\')
1266 				c = pgetc();
1267 			if (c == PEOF)
1268 				synerror("Unterminated quoted string");
1269 		}
1270 		pungetc();
1271 		return out;
1272 	}
1273 	if (SQSYNTAX[v] == CCTL)
1274 		USTPUTC(CTLESC, out);
1275 	USTPUTC(v, out);
1276 	return out;
1277 }
1278 
1279 
1280 /*
1281  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1282  * is not NULL, read a here document.  In the latter case, eofmark is the
1283  * word which marks the end of the document and striptabs is true if
1284  * leading tabs should be stripped from the document.  The argument firstc
1285  * is the first character of the input token or document.
1286  *
1287  * Because C does not have internal subroutines, I have simulated them
1288  * using goto's to implement the subroutine linkage.  The following macros
1289  * will run code that appears at the end of readtoken1.
1290  */
1291 
1292 #define CHECKEND()	{goto checkend; checkend_return:;}
1293 #define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1294 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1295 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1296 
1297 static int
1298 readtoken1(int firstc, char const *initialsyntax, char *eofmark, int striptabs)
1299 {
1300 	int c = firstc;
1301 	char * volatile out;
1302 	int len;
1303 	char line[EOFMARKLEN + 1];
1304 	struct nodelist *bqlist;
1305 	volatile int quotef;
1306 	int newvarnest;
1307 	int level;
1308 	int synentry;
1309 	struct tokenstate state_static[MAXNEST_STATIC];
1310 	int maxnest = MAXNEST_STATIC;
1311 	struct tokenstate *state = state_static;
1312 	int sqiscstyle = 0;
1313 
1314 	startlinno = plinno;
1315 	quotef = 0;
1316 	bqlist = NULL;
1317 	newvarnest = 0;
1318 	level = 0;
1319 	state[level].syntax = initialsyntax;
1320 	state[level].parenlevel = 0;
1321 	state[level].category = TSTATE_TOP;
1322 
1323 	STARTSTACKSTR(out);
1324 	loop: {	/* for each line, until end of word */
1325 		CHECKEND();	/* set c to PEOF if at end of here document */
1326 		for (;;) {	/* until end of line or end of word */
1327 			CHECKSTRSPACE(4, out);	/* permit 4 calls to USTPUTC */
1328 
1329 			synentry = state[level].syntax[c];
1330 
1331 			switch(synentry) {
1332 			case CNL:	/* '\n' */
1333 				if (state[level].syntax == BASESYNTAX)
1334 					goto endword;	/* exit outer loop */
1335 				USTPUTC(c, out);
1336 				plinno++;
1337 				if (doprompt)
1338 					setprompt(2);
1339 				else
1340 					setprompt(0);
1341 				c = pgetc();
1342 				goto loop;		/* continue outer loop */
1343 			case CSBACK:
1344 				if (sqiscstyle) {
1345 					out = readcstyleesc(out);
1346 					break;
1347 				}
1348 				/* FALLTHROUGH */
1349 			case CWORD:
1350 				USTPUTC(c, out);
1351 				break;
1352 			case CCTL:
1353 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1354 					USTPUTC(CTLESC, out);
1355 				USTPUTC(c, out);
1356 				break;
1357 			case CBACK:	/* backslash */
1358 				c = pgetc();
1359 				if (c == PEOF) {
1360 					USTPUTC('\\', out);
1361 					pungetc();
1362 				} else if (c == '\n') {
1363 					plinno++;
1364 					if (doprompt)
1365 						setprompt(2);
1366 					else
1367 						setprompt(0);
1368 				} else {
1369 					if (state[level].syntax == DQSYNTAX &&
1370 					    c != '\\' && c != '`' && c != '$' &&
1371 					    (c != '"' || (eofmark != NULL &&
1372 						newvarnest == 0)) &&
1373 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1374 						USTPUTC('\\', out);
1375 					if ((eofmark == NULL ||
1376 					    newvarnest > 0) &&
1377 					    state[level].syntax == BASESYNTAX)
1378 						USTPUTC(CTLQUOTEMARK, out);
1379 					if (SQSYNTAX[c] == CCTL)
1380 						USTPUTC(CTLESC, out);
1381 					USTPUTC(c, out);
1382 					if ((eofmark == NULL ||
1383 					    newvarnest > 0) &&
1384 					    state[level].syntax == BASESYNTAX &&
1385 					    state[level].category == TSTATE_VAR_OLD)
1386 						USTPUTC(CTLQUOTEEND, out);
1387 					quotef++;
1388 				}
1389 				break;
1390 			case CSQUOTE:
1391 				USTPUTC(CTLQUOTEMARK, out);
1392 				state[level].syntax = SQSYNTAX;
1393 				sqiscstyle = 0;
1394 				break;
1395 			case CDQUOTE:
1396 				USTPUTC(CTLQUOTEMARK, out);
1397 				state[level].syntax = DQSYNTAX;
1398 				break;
1399 			case CENDQUOTE:
1400 				if (eofmark != NULL && newvarnest == 0)
1401 					USTPUTC(c, out);
1402 				else {
1403 					if (state[level].category == TSTATE_VAR_OLD)
1404 						USTPUTC(CTLQUOTEEND, out);
1405 					state[level].syntax = BASESYNTAX;
1406 					quotef++;
1407 				}
1408 				break;
1409 			case CVAR:	/* '$' */
1410 				PARSESUB();		/* parse substitution */
1411 				break;
1412 			case CENDVAR:	/* '}' */
1413 				if (level > 0 &&
1414 				    ((state[level].category == TSTATE_VAR_OLD &&
1415 				      state[level].syntax ==
1416 				      state[level - 1].syntax) ||
1417 				    (state[level].category == TSTATE_VAR_NEW &&
1418 				     state[level].syntax == BASESYNTAX))) {
1419 					if (state[level].category == TSTATE_VAR_NEW)
1420 						newvarnest--;
1421 					level--;
1422 					USTPUTC(CTLENDVAR, out);
1423 				} else {
1424 					USTPUTC(c, out);
1425 				}
1426 				break;
1427 			case CLP:	/* '(' in arithmetic */
1428 				state[level].parenlevel++;
1429 				USTPUTC(c, out);
1430 				break;
1431 			case CRP:	/* ')' in arithmetic */
1432 				if (state[level].parenlevel > 0) {
1433 					USTPUTC(c, out);
1434 					--state[level].parenlevel;
1435 				} else {
1436 					if (pgetc() == ')') {
1437 						if (level > 0 &&
1438 						    state[level].category == TSTATE_ARITH) {
1439 							level--;
1440 							USTPUTC(CTLENDARI, out);
1441 						} else
1442 							USTPUTC(')', out);
1443 					} else {
1444 						/*
1445 						 * unbalanced parens
1446 						 *  (don't 2nd guess - no error)
1447 						 */
1448 						pungetc();
1449 						USTPUTC(')', out);
1450 					}
1451 				}
1452 				break;
1453 			case CBQUOTE:	/* '`' */
1454 				out = parsebackq(out, &bqlist, 1,
1455 				    state[level].syntax == DQSYNTAX &&
1456 				    (eofmark == NULL || newvarnest > 0),
1457 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1458 				break;
1459 			case CEOF:
1460 				goto endword;		/* exit outer loop */
1461 			case CIGN:
1462 				break;
1463 			default:
1464 				if (level == 0)
1465 					goto endword;	/* exit outer loop */
1466 				USTPUTC(c, out);
1467 			}
1468 			c = pgetc_macro();
1469 		}
1470 	}
1471 endword:
1472 	if (state[level].syntax == ARISYNTAX)
1473 		synerror("Missing '))'");
1474 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1475 		synerror("Unterminated quoted string");
1476 	if (state[level].category == TSTATE_VAR_OLD ||
1477 	    state[level].category == TSTATE_VAR_NEW) {
1478 		startlinno = plinno;
1479 		synerror("Missing '}'");
1480 	}
1481 	if (state != state_static)
1482 		parser_temp_free_upto(state);
1483 	USTPUTC('\0', out);
1484 	len = out - stackblock();
1485 	out = stackblock();
1486 	if (eofmark == NULL) {
1487 		if ((c == '>' || c == '<')
1488 		 && quotef == 0
1489 		 && len <= 2
1490 		 && (*out == '\0' || is_digit(*out))) {
1491 			PARSEREDIR();
1492 			return lasttoken = TREDIR;
1493 		} else {
1494 			pungetc();
1495 		}
1496 	}
1497 	quoteflag = quotef;
1498 	backquotelist = bqlist;
1499 	grabstackblock(len);
1500 	wordtext = out;
1501 	return lasttoken = TWORD;
1502 /* end of readtoken routine */
1503 
1504 
1505 /*
1506  * Check to see whether we are at the end of the here document.  When this
1507  * is called, c is set to the first character of the next input line.  If
1508  * we are at the end of the here document, this routine sets the c to PEOF.
1509  */
1510 
1511 checkend: {
1512 	if (eofmark) {
1513 		if (striptabs) {
1514 			while (c == '\t')
1515 				c = pgetc();
1516 		}
1517 		if (c == *eofmark) {
1518 			if (pfgets(line, sizeof line) != NULL) {
1519 				char *p, *q;
1520 
1521 				p = line;
1522 				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1523 				if ((*p == '\0' || *p == '\n') && *q == '\0') {
1524 					c = PEOF;
1525 					if (*p == '\n') {
1526 						plinno++;
1527 						needprompt = doprompt;
1528 					}
1529 				} else {
1530 					pushstring(line, strlen(line), NULL);
1531 				}
1532 			}
1533 		}
1534 	}
1535 	goto checkend_return;
1536 }
1537 
1538 
1539 /*
1540  * Parse a redirection operator.  The variable "out" points to a string
1541  * specifying the fd to be redirected.  The variable "c" contains the
1542  * first character of the redirection operator.
1543  */
1544 
1545 parseredir: {
1546 	char fd = *out;
1547 	union node *np;
1548 
1549 	np = (union node *)stalloc(sizeof (struct nfile));
1550 	if (c == '>') {
1551 		np->nfile.fd = 1;
1552 		c = pgetc();
1553 		if (c == '>')
1554 			np->type = NAPPEND;
1555 		else if (c == '&')
1556 			np->type = NTOFD;
1557 		else if (c == '|')
1558 			np->type = NCLOBBER;
1559 		else {
1560 			np->type = NTO;
1561 			pungetc();
1562 		}
1563 	} else {	/* c == '<' */
1564 		np->nfile.fd = 0;
1565 		c = pgetc();
1566 		if (c == '<') {
1567 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1568 				np = (union node *)stalloc(sizeof (struct nhere));
1569 				np->nfile.fd = 0;
1570 			}
1571 			np->type = NHERE;
1572 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1573 			heredoc->here = np;
1574 			if ((c = pgetc()) == '-') {
1575 				heredoc->striptabs = 1;
1576 			} else {
1577 				heredoc->striptabs = 0;
1578 				pungetc();
1579 			}
1580 		} else if (c == '&')
1581 			np->type = NFROMFD;
1582 		else if (c == '>')
1583 			np->type = NFROMTO;
1584 		else {
1585 			np->type = NFROM;
1586 			pungetc();
1587 		}
1588 	}
1589 	if (fd != '\0')
1590 		np->nfile.fd = digit_val(fd);
1591 	redirnode = np;
1592 	goto parseredir_return;
1593 }
1594 
1595 
1596 /*
1597  * Parse a substitution.  At this point, we have read the dollar sign
1598  * and nothing else.
1599  */
1600 
1601 parsesub: {
1602 	char buf[10];
1603 	int subtype;
1604 	int typeloc;
1605 	int flags;
1606 	char *p;
1607 	static const char types[] = "}-+?=";
1608 	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1609 	int linno;
1610 	int length;
1611 	int c1;
1612 
1613 	c = pgetc();
1614 	if (c == '(') {	/* $(command) or $((arith)) */
1615 		if (pgetc() == '(') {
1616 			PARSEARITH();
1617 		} else {
1618 			pungetc();
1619 			out = parsebackq(out, &bqlist, 0,
1620 			    state[level].syntax == DQSYNTAX &&
1621 			    (eofmark == NULL || newvarnest > 0),
1622 			    state[level].syntax == DQSYNTAX ||
1623 			    state[level].syntax == ARISYNTAX);
1624 		}
1625 	} else if (c == '{' || is_name(c) || is_special(c)) {
1626 		USTPUTC(CTLVAR, out);
1627 		typeloc = out - stackblock();
1628 		USTPUTC(VSNORMAL, out);
1629 		subtype = VSNORMAL;
1630 		flags = 0;
1631 		if (c == '{') {
1632 			bracketed_name = 1;
1633 			c = pgetc();
1634 			subtype = 0;
1635 		}
1636 varname:
1637 		if (!is_eof(c) && is_name(c)) {
1638 			length = 0;
1639 			do {
1640 				STPUTC(c, out);
1641 				c = pgetc();
1642 				length++;
1643 			} while (!is_eof(c) && is_in_name(c));
1644 			if (length == 6 &&
1645 			    strncmp(out - length, "LINENO", length) == 0) {
1646 				/* Replace the variable name with the
1647 				 * current line number. */
1648 				linno = plinno;
1649 				if (funclinno != 0)
1650 					linno -= funclinno - 1;
1651 				snprintf(buf, sizeof(buf), "%d", linno);
1652 				STADJUST(-6, out);
1653 				STPUTS(buf, out);
1654 				flags |= VSLINENO;
1655 			}
1656 		} else if (is_digit(c)) {
1657 			if (bracketed_name) {
1658 				do {
1659 					STPUTC(c, out);
1660 					c = pgetc();
1661 				} while (is_digit(c));
1662 			} else {
1663 				STPUTC(c, out);
1664 				c = pgetc();
1665 			}
1666 		} else if (is_special(c)) {
1667 			c1 = c;
1668 			c = pgetc();
1669 			if (subtype == 0 && c1 == '#') {
1670 				subtype = VSLENGTH;
1671 				if (strchr(types, c) == NULL && c != ':' &&
1672 				    c != '#' && c != '%')
1673 					goto varname;
1674 				c1 = c;
1675 				c = pgetc();
1676 				if (c1 != '}' && c == '}') {
1677 					pungetc();
1678 					c = c1;
1679 					goto varname;
1680 				}
1681 				pungetc();
1682 				c = c1;
1683 				c1 = '#';
1684 				subtype = 0;
1685 			}
1686 			USTPUTC(c1, out);
1687 		} else {
1688 			subtype = VSERROR;
1689 			if (c == '}')
1690 				pungetc();
1691 			else if (c == '\n' || c == PEOF)
1692 				synerror("Unexpected end of line in substitution");
1693 			else
1694 				USTPUTC(c, out);
1695 		}
1696 		if (subtype == 0) {
1697 			switch (c) {
1698 			case ':':
1699 				flags |= VSNUL;
1700 				c = pgetc();
1701 				/*FALLTHROUGH*/
1702 			default:
1703 				p = strchr(types, c);
1704 				if (p == NULL) {
1705 					if (c == '\n' || c == PEOF)
1706 						synerror("Unexpected end of line in substitution");
1707 					if (flags == VSNUL)
1708 						STPUTC(':', out);
1709 					STPUTC(c, out);
1710 					subtype = VSERROR;
1711 				} else
1712 					subtype = p - types + VSNORMAL;
1713 				break;
1714 			case '%':
1715 			case '#':
1716 				{
1717 					int cc = c;
1718 					subtype = c == '#' ? VSTRIMLEFT :
1719 							     VSTRIMRIGHT;
1720 					c = pgetc();
1721 					if (c == cc)
1722 						subtype++;
1723 					else
1724 						pungetc();
1725 					break;
1726 				}
1727 			}
1728 		} else if (subtype != VSERROR) {
1729 			if (subtype == VSLENGTH && c != '}')
1730 				subtype = VSERROR;
1731 			pungetc();
1732 		}
1733 		STPUTC('=', out);
1734 		if (state[level].syntax == DQSYNTAX ||
1735 		    state[level].syntax == ARISYNTAX)
1736 			flags |= VSQUOTE;
1737 		*(stackblock() + typeloc) = subtype | flags;
1738 		if (subtype != VSNORMAL) {
1739 			if (level + 1 >= maxnest) {
1740 				maxnest *= 2;
1741 				if (state == state_static) {
1742 					state = parser_temp_alloc(
1743 					    maxnest * sizeof(*state));
1744 					memcpy(state, state_static,
1745 					    MAXNEST_STATIC * sizeof(*state));
1746 				} else
1747 					state = parser_temp_realloc(state,
1748 					    maxnest * sizeof(*state));
1749 			}
1750 			level++;
1751 			state[level].parenlevel = 0;
1752 			if (subtype == VSMINUS || subtype == VSPLUS ||
1753 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1754 				/*
1755 				 * For operators that were in the Bourne shell,
1756 				 * inherit the double-quote state.
1757 				 */
1758 				state[level].syntax = state[level - 1].syntax;
1759 				state[level].category = TSTATE_VAR_OLD;
1760 			} else {
1761 				/*
1762 				 * The other operators take a pattern,
1763 				 * so go to BASESYNTAX.
1764 				 * Also, ' and " are now special, even
1765 				 * in here documents.
1766 				 */
1767 				state[level].syntax = BASESYNTAX;
1768 				state[level].category = TSTATE_VAR_NEW;
1769 				newvarnest++;
1770 			}
1771 		}
1772 	} else if (c == '\'' && state[level].syntax == BASESYNTAX) {
1773 		/* $'cstylequotes' */
1774 		USTPUTC(CTLQUOTEMARK, out);
1775 		state[level].syntax = SQSYNTAX;
1776 		sqiscstyle = 1;
1777 	} else {
1778 		USTPUTC('$', out);
1779 		pungetc();
1780 	}
1781 	goto parsesub_return;
1782 }
1783 
1784 
1785 /*
1786  * Parse an arithmetic expansion (indicate start of one and set state)
1787  */
1788 parsearith: {
1789 
1790 	if (level + 1 >= maxnest) {
1791 		maxnest *= 2;
1792 		if (state == state_static) {
1793 			state = parser_temp_alloc(
1794 			    maxnest * sizeof(*state));
1795 			memcpy(state, state_static,
1796 			    MAXNEST_STATIC * sizeof(*state));
1797 		} else
1798 			state = parser_temp_realloc(state,
1799 			    maxnest * sizeof(*state));
1800 	}
1801 	level++;
1802 	state[level].syntax = ARISYNTAX;
1803 	state[level].parenlevel = 0;
1804 	state[level].category = TSTATE_ARITH;
1805 	USTPUTC(CTLARI, out);
1806 	if (state[level - 1].syntax == DQSYNTAX)
1807 		USTPUTC('"',out);
1808 	else
1809 		USTPUTC(' ',out);
1810 	goto parsearith_return;
1811 }
1812 
1813 } /* end of readtoken */
1814 
1815 
1816 
1817 #ifdef mkinit
1818 RESET {
1819 	tokpushback = 0;
1820 	checkkwd = 0;
1821 }
1822 #endif
1823 
1824 /*
1825  * Returns true if the text contains nothing to expand (no dollar signs
1826  * or backquotes).
1827  */
1828 
1829 static int
1830 noexpand(char *text)
1831 {
1832 	char *p;
1833 	char c;
1834 
1835 	p = text;
1836 	while ((c = *p++) != '\0') {
1837 		if ( c == CTLQUOTEMARK)
1838 			continue;
1839 		if (c == CTLESC)
1840 			p++;
1841 		else if (BASESYNTAX[(int)c] == CCTL)
1842 			return 0;
1843 	}
1844 	return 1;
1845 }
1846 
1847 
1848 /*
1849  * Return true if the argument is a legal variable name (a letter or
1850  * underscore followed by zero or more letters, underscores, and digits).
1851  */
1852 
1853 int
1854 goodname(const char *name)
1855 {
1856 	const char *p;
1857 
1858 	p = name;
1859 	if (! is_name(*p))
1860 		return 0;
1861 	while (*++p) {
1862 		if (! is_in_name(*p))
1863 			return 0;
1864 	}
1865 	return 1;
1866 }
1867 
1868 
1869 int
1870 isassignment(const char *p)
1871 {
1872 	if (!is_name(*p))
1873 		return 0;
1874 	p++;
1875 	for (;;) {
1876 		if (*p == '=')
1877 			return 1;
1878 		else if (!is_in_name(*p))
1879 			return 0;
1880 		p++;
1881 	}
1882 }
1883 
1884 
1885 /*
1886  * Called when an unexpected token is read during the parse.  The argument
1887  * is the token that is expected, or -1 if more than one type of token can
1888  * occur at this point.
1889  */
1890 
1891 static void
1892 synexpect(int token)
1893 {
1894 	char msg[64];
1895 
1896 	if (token >= 0) {
1897 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1898 			tokname[lasttoken], tokname[token]);
1899 	} else {
1900 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1901 	}
1902 	synerror(msg);
1903 }
1904 
1905 
1906 static void
1907 synerror(const char *msg)
1908 {
1909 	if (commandname)
1910 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1911 	outfmt(out2, "Syntax error: %s\n", msg);
1912 	error(NULL);
1913 }
1914 
1915 static void
1916 setprompt(int which)
1917 {
1918 	whichprompt = which;
1919 
1920 #ifndef NO_HISTORY
1921 	if (!el)
1922 #endif
1923 	{
1924 		out2str(getprompt(NULL));
1925 		flushout(out2);
1926 	}
1927 }
1928 
1929 /*
1930  * called by editline -- any expansions to the prompt
1931  *    should be added here.
1932  */
1933 const char *
1934 getprompt(void *unused __unused)
1935 {
1936 	static char ps[PROMPTLEN];
1937 	const char *fmt;
1938 	const char *pwd;
1939 	int i, trim;
1940 
1941 	/*
1942 	 * Select prompt format.
1943 	 */
1944 	switch (whichprompt) {
1945 	case 0:
1946 		fmt = "";
1947 		break;
1948 	case 1:
1949 		fmt = ps1val();
1950 		break;
1951 	case 2:
1952 		fmt = ps2val();
1953 		break;
1954 	default:
1955 		return "??";
1956 	}
1957 
1958 	/*
1959 	 * Format prompt string.
1960 	 */
1961 	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1962 		if (*fmt == '\\')
1963 			switch (*++fmt) {
1964 
1965 				/*
1966 				 * Hostname.
1967 				 *
1968 				 * \h specifies just the local hostname,
1969 				 * \H specifies fully-qualified hostname.
1970 				 */
1971 			case 'h':
1972 			case 'H':
1973 				ps[i] = '\0';
1974 				gethostname(&ps[i], PROMPTLEN - i);
1975 				/* Skip to end of hostname. */
1976 				trim = (*fmt == 'h') ? '.' : '\0';
1977 				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1978 					i++;
1979 				break;
1980 
1981 				/*
1982 				 * Working directory.
1983 				 *
1984 				 * \W specifies just the final component,
1985 				 * \w specifies the entire path.
1986 				 */
1987 			case 'W':
1988 			case 'w':
1989 				pwd = lookupvar("PWD");
1990 				if (pwd == NULL)
1991 					pwd = "?";
1992 				if (*fmt == 'W' &&
1993 				    *pwd == '/' && pwd[1] != '\0')
1994 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
1995 					    PROMPTLEN - i);
1996 				else
1997 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
1998 				/* Skip to end of path. */
1999 				while (ps[i + 1] != '\0')
2000 					i++;
2001 				break;
2002 
2003 				/*
2004 				 * Superuser status.
2005 				 *
2006 				 * '$' for normal users, '#' for root.
2007 				 */
2008 			case '$':
2009 				ps[i] = (geteuid() != 0) ? '$' : '#';
2010 				break;
2011 
2012 				/*
2013 				 * A literal \.
2014 				 */
2015 			case '\\':
2016 				ps[i] = '\\';
2017 				break;
2018 
2019 				/*
2020 				 * Emit unrecognized formats verbatim.
2021 				 */
2022 			default:
2023 				ps[i++] = '\\';
2024 				ps[i] = *fmt;
2025 				break;
2026 			}
2027 		else
2028 			ps[i] = *fmt;
2029 	ps[i] = '\0';
2030 	return (ps);
2031 }
2032 
2033 
2034 const char *
2035 expandstr(char *ps)
2036 {
2037 	union node n;
2038 	struct jmploc jmploc;
2039 	struct jmploc *const savehandler = handler;
2040 	const int saveprompt = doprompt;
2041 	struct parsefile *const savetopfile = getcurrentfile();
2042 	struct parser_temp *const saveparser_temp = parser_temp;
2043 	const char *result = NULL;
2044 
2045 	if (!setjmp(jmploc.loc)) {
2046 		handler = &jmploc;
2047 		parser_temp = NULL;
2048 		setinputstring(ps, 1);
2049 		doprompt = 0;
2050 		readtoken1(pgetc(), DQSYNTAX, __DECONST(char *, "\n\n"), 0);
2051 		if (backquotelist != NULL)
2052 			error("Command substitution not allowed here");
2053 
2054 		n.narg.type = NARG;
2055 		n.narg.next = NULL;
2056 		n.narg.text = wordtext;
2057 		n.narg.backquote = backquotelist;
2058 
2059 		expandarg(&n, NULL, 0);
2060 		result = stackblock();
2061 		INTOFF;
2062 	}
2063 	handler = savehandler;
2064 	doprompt = saveprompt;
2065 	popfilesupto(savetopfile);
2066 	if (parser_temp != saveparser_temp) {
2067 		parser_temp_free_all();
2068 		parser_temp = saveparser_temp;
2069 	}
2070 	if (result != NULL) {
2071 		INTON;
2072 	} else if (exception == EXINT)
2073 		raise(SIGINT);
2074 	return result;
2075 }
2076