xref: /freebsd/bin/sh/parser.c (revision b2db760808f74bb53c232900091c9da801ebbfcc)
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 #ifndef NO_HISTORY
61 #include "myhistedit.h"
62 #endif
63 
64 /*
65  * Shell command parser.
66  */
67 
68 #define	EOFMARKLEN	79
69 #define	PROMPTLEN	128
70 
71 /* values returned by readtoken */
72 #include "token.h"
73 
74 
75 
76 struct heredoc {
77 	struct heredoc *next;	/* next here document in list */
78 	union node *here;		/* redirection node */
79 	char *eofmark;		/* string indicating end of input */
80 	int striptabs;		/* if set, strip leading tabs */
81 };
82 
83 struct parser_temp {
84 	struct parser_temp *next;
85 	void *data;
86 };
87 
88 
89 STATIC struct heredoc *heredoclist;	/* list of here documents to read */
90 STATIC int doprompt;		/* if set, prompt the user */
91 STATIC int needprompt;		/* true if interactive and at start of line */
92 STATIC int lasttoken;		/* last token read */
93 MKINIT int tokpushback;		/* last token pushed back */
94 STATIC char *wordtext;		/* text of last word returned by readtoken */
95 MKINIT int checkkwd;            /* 1 == check for kwds, 2 == also eat newlines */
96 STATIC struct nodelist *backquotelist;
97 STATIC union node *redirnode;
98 STATIC struct heredoc *heredoc;
99 STATIC int quoteflag;		/* set if (part of) last token was quoted */
100 STATIC int startlinno;		/* line # where last token started */
101 STATIC int funclinno;		/* line # where the current function started */
102 STATIC struct parser_temp *parser_temp;
103 
104 /* XXX When 'noaliases' is set to one, no alias expansion takes place. */
105 static int noaliases = 0;
106 
107 
108 STATIC union node *list(int);
109 STATIC union node *andor(void);
110 STATIC union node *pipeline(void);
111 STATIC union node *command(void);
112 STATIC union node *simplecmd(union node **, union node *);
113 STATIC union node *makename(void);
114 STATIC void parsefname(void);
115 STATIC void parseheredoc(void);
116 STATIC int peektoken(void);
117 STATIC int readtoken(void);
118 STATIC int xxreadtoken(void);
119 STATIC int readtoken1(int, char const *, char *, int);
120 STATIC int noexpand(char *);
121 STATIC void synexpect(int);
122 STATIC void synerror(const char *);
123 STATIC void setprompt(int);
124 
125 
126 STATIC void *
127 parser_temp_alloc(size_t len)
128 {
129 	struct parser_temp *t;
130 
131 	INTOFF;
132 	t = ckmalloc(sizeof(*t));
133 	t->data = NULL;
134 	t->next = parser_temp;
135 	parser_temp = t;
136 	t->data = ckmalloc(len);
137 	INTON;
138 	return t->data;
139 }
140 
141 
142 STATIC void *
143 parser_temp_realloc(void *ptr, size_t len)
144 {
145 	struct parser_temp *t;
146 
147 	INTOFF;
148 	t = parser_temp;
149 	if (ptr != t->data)
150 		error("bug: parser_temp_realloc misused");
151 	t->data = ckrealloc(t->data, len);
152 	INTON;
153 	return t->data;
154 }
155 
156 
157 STATIC void
158 parser_temp_free_upto(void *ptr)
159 {
160 	struct parser_temp *t;
161 	int done = 0;
162 
163 	INTOFF;
164 	while (parser_temp != NULL && !done) {
165 		t = parser_temp;
166 		parser_temp = t->next;
167 		done = t->data == ptr;
168 		ckfree(t->data);
169 		ckfree(t);
170 	}
171 	INTON;
172 	if (!done)
173 		error("bug: parser_temp_free_upto misused");
174 }
175 
176 
177 STATIC void
178 parser_temp_free_all(void)
179 {
180 	struct parser_temp *t;
181 
182 	INTOFF;
183 	while (parser_temp != NULL) {
184 		t = parser_temp;
185 		parser_temp = t->next;
186 		ckfree(t->data);
187 		ckfree(t);
188 	}
189 	INTON;
190 }
191 
192 
193 /*
194  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
195  * valid parse tree indicating a blank line.)
196  */
197 
198 union node *
199 parsecmd(int interact)
200 {
201 	int t;
202 
203 	/* This assumes the parser is not re-entered,
204 	 * which could happen if we add command substitution on PS1/PS2.
205 	 */
206 	parser_temp_free_all();
207 	heredoclist = NULL;
208 
209 	tokpushback = 0;
210 	doprompt = interact;
211 	if (doprompt)
212 		setprompt(1);
213 	else
214 		setprompt(0);
215 	needprompt = 0;
216 	t = readtoken();
217 	if (t == TEOF)
218 		return NEOF;
219 	if (t == TNL)
220 		return NULL;
221 	tokpushback++;
222 	return list(1);
223 }
224 
225 
226 STATIC union node *
227 list(int nlflag)
228 {
229 	union node *n1, *n2, *n3;
230 	int tok;
231 
232 	checkkwd = 2;
233 	if (nlflag == 0 && tokendlist[peektoken()])
234 		return NULL;
235 	n1 = NULL;
236 	for (;;) {
237 		n2 = andor();
238 		tok = readtoken();
239 		if (tok == TBACKGND) {
240 			if (n2->type == NCMD || n2->type == NPIPE) {
241 				n2->ncmd.backgnd = 1;
242 			} else if (n2->type == NREDIR) {
243 				n2->type = NBACKGND;
244 			} else {
245 				n3 = (union node *)stalloc(sizeof (struct nredir));
246 				n3->type = NBACKGND;
247 				n3->nredir.n = n2;
248 				n3->nredir.redirect = NULL;
249 				n2 = n3;
250 			}
251 		}
252 		if (n1 == NULL) {
253 			n1 = n2;
254 		}
255 		else {
256 			n3 = (union node *)stalloc(sizeof (struct nbinary));
257 			n3->type = NSEMI;
258 			n3->nbinary.ch1 = n1;
259 			n3->nbinary.ch2 = n2;
260 			n1 = n3;
261 		}
262 		switch (tok) {
263 		case TBACKGND:
264 		case TSEMI:
265 			tok = readtoken();
266 			/* FALLTHROUGH */
267 		case TNL:
268 			if (tok == TNL) {
269 				parseheredoc();
270 				if (nlflag)
271 					return n1;
272 			} else if (tok == TEOF && nlflag) {
273 				parseheredoc();
274 				return n1;
275 			} else {
276 				tokpushback++;
277 			}
278 			checkkwd = 2;
279 			if (tokendlist[peektoken()])
280 				return n1;
281 			break;
282 		case TEOF:
283 			if (heredoclist)
284 				parseheredoc();
285 			else
286 				pungetc();		/* push back EOF on input */
287 			return n1;
288 		default:
289 			if (nlflag)
290 				synexpect(-1);
291 			tokpushback++;
292 			return n1;
293 		}
294 	}
295 }
296 
297 
298 
299 STATIC union node *
300 andor(void)
301 {
302 	union node *n1, *n2, *n3;
303 	int t;
304 
305 	n1 = pipeline();
306 	for (;;) {
307 		if ((t = readtoken()) == TAND) {
308 			t = NAND;
309 		} else if (t == TOR) {
310 			t = NOR;
311 		} else {
312 			tokpushback++;
313 			return n1;
314 		}
315 		n2 = pipeline();
316 		n3 = (union node *)stalloc(sizeof (struct nbinary));
317 		n3->type = t;
318 		n3->nbinary.ch1 = n1;
319 		n3->nbinary.ch2 = n2;
320 		n1 = n3;
321 	}
322 }
323 
324 
325 
326 STATIC union node *
327 pipeline(void)
328 {
329 	union node *n1, *n2, *pipenode;
330 	struct nodelist *lp, *prev;
331 	int negate;
332 
333 	negate = 0;
334 	checkkwd = 2;
335 	TRACE(("pipeline: entered\n"));
336 	while (readtoken() == TNOT)
337 		negate = !negate;
338 	tokpushback++;
339 	n1 = command();
340 	if (readtoken() == TPIPE) {
341 		pipenode = (union node *)stalloc(sizeof (struct npipe));
342 		pipenode->type = NPIPE;
343 		pipenode->npipe.backgnd = 0;
344 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
345 		pipenode->npipe.cmdlist = lp;
346 		lp->n = n1;
347 		do {
348 			prev = lp;
349 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
350 			lp->n = command();
351 			prev->next = lp;
352 		} while (readtoken() == TPIPE);
353 		lp->next = NULL;
354 		n1 = pipenode;
355 	}
356 	tokpushback++;
357 	if (negate) {
358 		n2 = (union node *)stalloc(sizeof (struct nnot));
359 		n2->type = NNOT;
360 		n2->nnot.com = n1;
361 		return n2;
362 	} else
363 		return n1;
364 }
365 
366 
367 
368 STATIC union node *
369 command(void)
370 {
371 	union node *n1, *n2;
372 	union node *ap, **app;
373 	union node *cp, **cpp;
374 	union node *redir, **rpp;
375 	int t, negate = 0;
376 
377 	checkkwd = 2;
378 	redir = NULL;
379 	n1 = NULL;
380 	rpp = &redir;
381 
382 	/* Check for redirection which may precede command */
383 	while (readtoken() == TREDIR) {
384 		*rpp = n2 = redirnode;
385 		rpp = &n2->nfile.next;
386 		parsefname();
387 	}
388 	tokpushback++;
389 
390 	while (readtoken() == TNOT) {
391 		TRACE(("command: TNOT recognized\n"));
392 		negate = !negate;
393 	}
394 	tokpushback++;
395 
396 	switch (readtoken()) {
397 	case TIF:
398 		n1 = (union node *)stalloc(sizeof (struct nif));
399 		n1->type = NIF;
400 		if ((n1->nif.test = list(0)) == NULL)
401 			synexpect(-1);
402 		if (readtoken() != TTHEN)
403 			synexpect(TTHEN);
404 		n1->nif.ifpart = list(0);
405 		n2 = n1;
406 		while (readtoken() == TELIF) {
407 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
408 			n2 = n2->nif.elsepart;
409 			n2->type = NIF;
410 			if ((n2->nif.test = list(0)) == NULL)
411 				synexpect(-1);
412 			if (readtoken() != TTHEN)
413 				synexpect(TTHEN);
414 			n2->nif.ifpart = list(0);
415 		}
416 		if (lasttoken == TELSE)
417 			n2->nif.elsepart = list(0);
418 		else {
419 			n2->nif.elsepart = NULL;
420 			tokpushback++;
421 		}
422 		if (readtoken() != TFI)
423 			synexpect(TFI);
424 		checkkwd = 1;
425 		break;
426 	case TWHILE:
427 	case TUNTIL: {
428 		int got;
429 		n1 = (union node *)stalloc(sizeof (struct nbinary));
430 		n1->type = (lasttoken == TWHILE)? NWHILE : NUNTIL;
431 		if ((n1->nbinary.ch1 = list(0)) == NULL)
432 			synexpect(-1);
433 		if ((got=readtoken()) != TDO) {
434 TRACE(("expecting DO got %s %s\n", tokname[got], got == TWORD ? wordtext : ""));
435 			synexpect(TDO);
436 		}
437 		n1->nbinary.ch2 = list(0);
438 		if (readtoken() != TDONE)
439 			synexpect(TDONE);
440 		checkkwd = 1;
441 		break;
442 	}
443 	case TFOR:
444 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
445 			synerror("Bad for loop variable");
446 		n1 = (union node *)stalloc(sizeof (struct nfor));
447 		n1->type = NFOR;
448 		n1->nfor.var = wordtext;
449 		while (readtoken() == TNL)
450 			;
451 		if (lasttoken == TWORD && ! quoteflag && equal(wordtext, "in")) {
452 			app = &ap;
453 			while (readtoken() == TWORD) {
454 				n2 = (union node *)stalloc(sizeof (struct narg));
455 				n2->type = NARG;
456 				n2->narg.text = wordtext;
457 				n2->narg.backquote = backquotelist;
458 				*app = n2;
459 				app = &n2->narg.next;
460 			}
461 			*app = NULL;
462 			n1->nfor.args = ap;
463 			if (lasttoken != TNL && lasttoken != TSEMI)
464 				synexpect(-1);
465 		} else {
466 			static char argvars[5] = {
467 				CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'
468 			};
469 			n2 = (union node *)stalloc(sizeof (struct narg));
470 			n2->type = NARG;
471 			n2->narg.text = argvars;
472 			n2->narg.backquote = NULL;
473 			n2->narg.next = NULL;
474 			n1->nfor.args = n2;
475 			/*
476 			 * Newline or semicolon here is optional (but note
477 			 * that the original Bourne shell only allowed NL).
478 			 */
479 			if (lasttoken != TNL && lasttoken != TSEMI)
480 				tokpushback++;
481 		}
482 		checkkwd = 2;
483 		if ((t = readtoken()) == TDO)
484 			t = TDONE;
485 		else if (t == TBEGIN)
486 			t = TEND;
487 		else
488 			synexpect(-1);
489 		n1->nfor.body = list(0);
490 		if (readtoken() != t)
491 			synexpect(t);
492 		checkkwd = 1;
493 		break;
494 	case TCASE:
495 		n1 = (union node *)stalloc(sizeof (struct ncase));
496 		n1->type = NCASE;
497 		if (readtoken() != TWORD)
498 			synexpect(TWORD);
499 		n1->ncase.expr = n2 = (union node *)stalloc(sizeof (struct narg));
500 		n2->type = NARG;
501 		n2->narg.text = wordtext;
502 		n2->narg.backquote = backquotelist;
503 		n2->narg.next = NULL;
504 		while (readtoken() == TNL);
505 		if (lasttoken != TWORD || ! equal(wordtext, "in"))
506 			synerror("expecting \"in\"");
507 		cpp = &n1->ncase.cases;
508 		noaliases = 1;	/* turn off alias expansion */
509 		checkkwd = 2, readtoken();
510 		while (lasttoken != TESAC) {
511 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
512 			cp->type = NCLIST;
513 			app = &cp->nclist.pattern;
514 			if (lasttoken == TLP)
515 				readtoken();
516 			for (;;) {
517 				*app = ap = (union node *)stalloc(sizeof (struct narg));
518 				ap->type = NARG;
519 				ap->narg.text = wordtext;
520 				ap->narg.backquote = backquotelist;
521 				if (checkkwd = 2, readtoken() != TPIPE)
522 					break;
523 				app = &ap->narg.next;
524 				readtoken();
525 			}
526 			ap->narg.next = NULL;
527 			if (lasttoken != TRP)
528 				noaliases = 0, synexpect(TRP);
529 			cp->nclist.body = list(0);
530 
531 			checkkwd = 2;
532 			if ((t = readtoken()) != TESAC) {
533 				if (t != TENDCASE)
534 					noaliases = 0, synexpect(TENDCASE);
535 				else
536 					checkkwd = 2, readtoken();
537 			}
538 			cpp = &cp->nclist.next;
539 		}
540 		noaliases = 0;	/* reset alias expansion */
541 		*cpp = NULL;
542 		checkkwd = 1;
543 		break;
544 	case TLP:
545 		n1 = (union node *)stalloc(sizeof (struct nredir));
546 		n1->type = NSUBSHELL;
547 		n1->nredir.n = list(0);
548 		n1->nredir.redirect = NULL;
549 		if (readtoken() != TRP)
550 			synexpect(TRP);
551 		checkkwd = 1;
552 		break;
553 	case TBEGIN:
554 		n1 = list(0);
555 		if (readtoken() != TEND)
556 			synexpect(TEND);
557 		checkkwd = 1;
558 		break;
559 	/* Handle an empty command like other simple commands.  */
560 	case TBACKGND:
561 	case TSEMI:
562 	case TAND:
563 	case TOR:
564 		/*
565 		 * An empty command before a ; doesn't make much sense, and
566 		 * should certainly be disallowed in the case of `if ;'.
567 		 */
568 		if (!redir)
569 			synexpect(-1);
570 	case TNL:
571 	case TEOF:
572 	case TWORD:
573 	case TRP:
574 		tokpushback++;
575 		n1 = simplecmd(rpp, redir);
576 		goto checkneg;
577 	default:
578 		synexpect(-1);
579 	}
580 
581 	/* Now check for redirection which may follow command */
582 	while (readtoken() == TREDIR) {
583 		*rpp = n2 = redirnode;
584 		rpp = &n2->nfile.next;
585 		parsefname();
586 	}
587 	tokpushback++;
588 	*rpp = NULL;
589 	if (redir) {
590 		if (n1->type != NSUBSHELL) {
591 			n2 = (union node *)stalloc(sizeof (struct nredir));
592 			n2->type = NREDIR;
593 			n2->nredir.n = n1;
594 			n1 = n2;
595 		}
596 		n1->nredir.redirect = redir;
597 	}
598 
599 checkneg:
600 	if (negate) {
601 		n2 = (union node *)stalloc(sizeof (struct nnot));
602 		n2->type = NNOT;
603 		n2->nnot.com = n1;
604 		return n2;
605 	}
606 	else
607 		return n1;
608 }
609 
610 
611 STATIC union node *
612 simplecmd(union node **rpp, union node *redir)
613 {
614 	union node *args, **app;
615 	union node **orig_rpp = rpp;
616 	union node *n = NULL;
617 
618 	/* If we don't have any redirections already, then we must reset */
619 	/* rpp to be the address of the local redir variable.  */
620 	if (redir == 0)
621 		rpp = &redir;
622 
623 	args = NULL;
624 	app = &args;
625 	/*
626 	 * We save the incoming value, because we need this for shell
627 	 * functions.  There can not be a redirect or an argument between
628 	 * the function name and the open parenthesis.
629 	 */
630 	orig_rpp = rpp;
631 
632 	for (;;) {
633 		if (readtoken() == TWORD) {
634 			n = (union node *)stalloc(sizeof (struct narg));
635 			n->type = NARG;
636 			n->narg.text = wordtext;
637 			n->narg.backquote = backquotelist;
638 			*app = n;
639 			app = &n->narg.next;
640 		} else if (lasttoken == TREDIR) {
641 			*rpp = n = redirnode;
642 			rpp = &n->nfile.next;
643 			parsefname();	/* read name of redirection file */
644 		} else if (lasttoken == TLP && app == &args->narg.next
645 					    && rpp == orig_rpp) {
646 			/* We have a function */
647 			if (readtoken() != TRP)
648 				synexpect(TRP);
649 			funclinno = plinno;
650 #ifdef notdef
651 			if (! goodname(n->narg.text))
652 				synerror("Bad function name");
653 #endif
654 			n->type = NDEFUN;
655 			n->narg.next = command();
656 			funclinno = 0;
657 			return n;
658 		} else {
659 			tokpushback++;
660 			break;
661 		}
662 	}
663 	*app = NULL;
664 	*rpp = NULL;
665 	n = (union node *)stalloc(sizeof (struct ncmd));
666 	n->type = NCMD;
667 	n->ncmd.backgnd = 0;
668 	n->ncmd.args = args;
669 	n->ncmd.redirect = redir;
670 	return n;
671 }
672 
673 STATIC union node *
674 makename(void)
675 {
676 	union node *n;
677 
678 	n = (union node *)stalloc(sizeof (struct narg));
679 	n->type = NARG;
680 	n->narg.next = NULL;
681 	n->narg.text = wordtext;
682 	n->narg.backquote = backquotelist;
683 	return n;
684 }
685 
686 void fixredir(union node *n, const char *text, int err)
687 {
688 	TRACE(("Fix redir %s %d\n", text, err));
689 	if (!err)
690 		n->ndup.vname = NULL;
691 
692 	if (is_digit(text[0]) && text[1] == '\0')
693 		n->ndup.dupfd = digit_val(text[0]);
694 	else if (text[0] == '-' && text[1] == '\0')
695 		n->ndup.dupfd = -1;
696 	else {
697 
698 		if (err)
699 			synerror("Bad fd number");
700 		else
701 			n->ndup.vname = makename();
702 	}
703 }
704 
705 
706 STATIC void
707 parsefname(void)
708 {
709 	union node *n = redirnode;
710 
711 	if (readtoken() != TWORD)
712 		synexpect(-1);
713 	if (n->type == NHERE) {
714 		struct heredoc *here = heredoc;
715 		struct heredoc *p;
716 		int i;
717 
718 		if (quoteflag == 0)
719 			n->type = NXHERE;
720 		TRACE(("Here document %d\n", n->type));
721 		if (here->striptabs) {
722 			while (*wordtext == '\t')
723 				wordtext++;
724 		}
725 		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
726 			synerror("Illegal eof marker for << redirection");
727 		rmescapes(wordtext);
728 		here->eofmark = wordtext;
729 		here->next = NULL;
730 		if (heredoclist == NULL)
731 			heredoclist = here;
732 		else {
733 			for (p = heredoclist ; p->next ; p = p->next);
734 			p->next = here;
735 		}
736 	} else if (n->type == NTOFD || n->type == NFROMFD) {
737 		fixredir(n, wordtext, 0);
738 	} else {
739 		n->nfile.fname = makename();
740 	}
741 }
742 
743 
744 /*
745  * Input any here documents.
746  */
747 
748 STATIC void
749 parseheredoc(void)
750 {
751 	struct heredoc *here;
752 	union node *n;
753 
754 	while (heredoclist) {
755 		here = heredoclist;
756 		heredoclist = here->next;
757 		if (needprompt) {
758 			setprompt(2);
759 			needprompt = 0;
760 		}
761 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
762 				here->eofmark, here->striptabs);
763 		n = (union node *)stalloc(sizeof (struct narg));
764 		n->narg.type = NARG;
765 		n->narg.next = NULL;
766 		n->narg.text = wordtext;
767 		n->narg.backquote = backquotelist;
768 		here->here->nhere.doc = n;
769 	}
770 }
771 
772 STATIC int
773 peektoken(void)
774 {
775 	int t;
776 
777 	t = readtoken();
778 	tokpushback++;
779 	return (t);
780 }
781 
782 STATIC int
783 readtoken(void)
784 {
785 	int t;
786 	int savecheckkwd = checkkwd;
787 	struct alias *ap;
788 #ifdef DEBUG
789 	int alreadyseen = tokpushback;
790 #endif
791 
792 	top:
793 	t = xxreadtoken();
794 
795 	if (checkkwd) {
796 		/*
797 		 * eat newlines
798 		 */
799 		if (checkkwd == 2) {
800 			checkkwd = 0;
801 			while (t == TNL) {
802 				parseheredoc();
803 				t = xxreadtoken();
804 			}
805 		} else
806 			checkkwd = 0;
807 		/*
808 		 * check for keywords and aliases
809 		 */
810 		if (t == TWORD && !quoteflag)
811 		{
812 			const char * const *pp;
813 
814 			for (pp = parsekwd; *pp; pp++) {
815 				if (**pp == *wordtext && equal(*pp, wordtext))
816 				{
817 					lasttoken = t = pp - parsekwd + KWDOFFSET;
818 					TRACE(("keyword %s recognized\n", tokname[t]));
819 					goto out;
820 				}
821 			}
822 			if (noaliases == 0 &&
823 			    (ap = lookupalias(wordtext, 1)) != NULL) {
824 				pushstring(ap->val, strlen(ap->val), ap);
825 				checkkwd = savecheckkwd;
826 				goto top;
827 			}
828 		}
829 out:
830 		checkkwd = (t == TNOT) ? savecheckkwd : 0;
831 	}
832 #ifdef DEBUG
833 	if (!alreadyseen)
834 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
835 	else
836 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
837 #endif
838 	return (t);
839 }
840 
841 
842 /*
843  * Read the next input token.
844  * If the token is a word, we set backquotelist to the list of cmds in
845  *	backquotes.  We set quoteflag to true if any part of the word was
846  *	quoted.
847  * If the token is TREDIR, then we set redirnode to a structure containing
848  *	the redirection.
849  * In all cases, the variable startlinno is set to the number of the line
850  *	on which the token starts.
851  *
852  * [Change comment:  here documents and internal procedures]
853  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
854  *  word parsing code into a separate routine.  In this case, readtoken
855  *  doesn't need to have any internal procedures, but parseword does.
856  *  We could also make parseoperator in essence the main routine, and
857  *  have parseword (readtoken1?) handle both words and redirection.]
858  */
859 
860 #define RETURN(token)	return lasttoken = token
861 
862 STATIC int
863 xxreadtoken(void)
864 {
865 	int c;
866 
867 	if (tokpushback) {
868 		tokpushback = 0;
869 		return lasttoken;
870 	}
871 	if (needprompt) {
872 		setprompt(2);
873 		needprompt = 0;
874 	}
875 	startlinno = plinno;
876 	for (;;) {	/* until token or start of word found */
877 		c = pgetc_macro();
878 		if (c == ' ' || c == '\t')
879 			continue;		/* quick check for white space first */
880 		switch (c) {
881 		case ' ': case '\t':
882 			continue;
883 		case '#':
884 			while ((c = pgetc()) != '\n' && c != PEOF);
885 			pungetc();
886 			continue;
887 		case '\\':
888 			if (pgetc() == '\n') {
889 				startlinno = ++plinno;
890 				if (doprompt)
891 					setprompt(2);
892 				else
893 					setprompt(0);
894 				continue;
895 			}
896 			pungetc();
897 			goto breakloop;
898 		case '\n':
899 			plinno++;
900 			needprompt = doprompt;
901 			RETURN(TNL);
902 		case PEOF:
903 			RETURN(TEOF);
904 		case '&':
905 			if (pgetc() == '&')
906 				RETURN(TAND);
907 			pungetc();
908 			RETURN(TBACKGND);
909 		case '|':
910 			if (pgetc() == '|')
911 				RETURN(TOR);
912 			pungetc();
913 			RETURN(TPIPE);
914 		case ';':
915 			if (pgetc() == ';')
916 				RETURN(TENDCASE);
917 			pungetc();
918 			RETURN(TSEMI);
919 		case '(':
920 			RETURN(TLP);
921 		case ')':
922 			RETURN(TRP);
923 		default:
924 			goto breakloop;
925 		}
926 	}
927 breakloop:
928 	return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
929 #undef RETURN
930 }
931 
932 
933 #define MAXNEST_STATIC 8
934 struct tokenstate
935 {
936 	const char *syntax; /* *SYNTAX */
937 	int parenlevel; /* levels of parentheses in arithmetic */
938 	enum tokenstate_category
939 	{
940 		TSTATE_TOP,
941 		TSTATE_VAR_OLD, /* ${var+-=?}, inherits dquotes */
942 		TSTATE_VAR_NEW, /* other ${var...}, own dquote state */
943 		TSTATE_ARITH
944 	} category;
945 };
946 
947 
948 /*
949  * Called to parse command substitutions.
950  */
951 
952 STATIC char *
953 parsebackq(char *out, struct nodelist **pbqlist,
954 		int oldstyle, int dblquote, int quoted)
955 {
956 	struct nodelist **nlpp;
957 	union node *n;
958 	char *volatile str;
959 	struct jmploc jmploc;
960 	struct jmploc *const savehandler = handler;
961 	int savelen;
962 	int saveprompt;
963 	const int bq_startlinno = plinno;
964 	char *volatile ostr = NULL;
965 	struct parsefile *const savetopfile = getcurrentfile();
966 	struct heredoc *const saveheredoclist = heredoclist;
967 	struct heredoc *here;
968 
969 	str = NULL;
970 	if (setjmp(jmploc.loc)) {
971 		popfilesupto(savetopfile);
972 		if (str)
973 			ckfree(str);
974 		if (ostr)
975 			ckfree(ostr);
976 		heredoclist = saveheredoclist;
977 		handler = savehandler;
978 		if (exception == EXERROR) {
979 			startlinno = bq_startlinno;
980 			synerror("Error in command substitution");
981 		}
982 		longjmp(handler->loc, 1);
983 	}
984 	INTOFF;
985 	savelen = out - stackblock();
986 	if (savelen > 0) {
987 		str = ckmalloc(savelen);
988 		memcpy(str, stackblock(), savelen);
989 	}
990 	handler = &jmploc;
991 	heredoclist = NULL;
992 	INTON;
993         if (oldstyle) {
994                 /* We must read until the closing backquote, giving special
995                    treatment to some slashes, and then push the string and
996                    reread it as input, interpreting it normally.  */
997                 char *oout;
998                 int c;
999                 int olen;
1000 
1001 
1002                 STARTSTACKSTR(oout);
1003 		for (;;) {
1004 			if (needprompt) {
1005 				setprompt(2);
1006 				needprompt = 0;
1007 			}
1008 			switch (c = pgetc()) {
1009 			case '`':
1010 				goto done;
1011 
1012 			case '\\':
1013                                 if ((c = pgetc()) == '\n') {
1014 					plinno++;
1015 					if (doprompt)
1016 						setprompt(2);
1017 					else
1018 						setprompt(0);
1019 					/*
1020 					 * If eating a newline, avoid putting
1021 					 * the newline into the new character
1022 					 * stream (via the STPUTC after the
1023 					 * switch).
1024 					 */
1025 					continue;
1026 				}
1027                                 if (c != '\\' && c != '`' && c != '$'
1028                                     && (!dblquote || c != '"'))
1029                                         STPUTC('\\', oout);
1030 				break;
1031 
1032 			case '\n':
1033 				plinno++;
1034 				needprompt = doprompt;
1035 				break;
1036 
1037 			case PEOF:
1038 			        startlinno = plinno;
1039 				synerror("EOF in backquote substitution");
1040  				break;
1041 
1042 			default:
1043 				break;
1044 			}
1045 			STPUTC(c, oout);
1046                 }
1047 done:
1048                 STPUTC('\0', oout);
1049                 olen = oout - stackblock();
1050 		INTOFF;
1051 		ostr = ckmalloc(olen);
1052 		memcpy(ostr, stackblock(), olen);
1053 		setinputstring(ostr, 1);
1054 		INTON;
1055         }
1056 	nlpp = pbqlist;
1057 	while (*nlpp)
1058 		nlpp = &(*nlpp)->next;
1059 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1060 	(*nlpp)->next = NULL;
1061 
1062 	if (oldstyle) {
1063 		saveprompt = doprompt;
1064 		doprompt = 0;
1065 	}
1066 
1067 	n = list(0);
1068 
1069 	if (oldstyle)
1070 		doprompt = saveprompt;
1071 	else {
1072 		if (readtoken() != TRP)
1073 			synexpect(TRP);
1074 	}
1075 
1076 	(*nlpp)->n = n;
1077         if (oldstyle) {
1078 		/*
1079 		 * Start reading from old file again, ignoring any pushed back
1080 		 * tokens left from the backquote parsing
1081 		 */
1082                 popfile();
1083 		tokpushback = 0;
1084 	}
1085 	while (stackblocksize() <= savelen)
1086 		growstackblock();
1087 	STARTSTACKSTR(out);
1088 	INTOFF;
1089 	if (str) {
1090 		memcpy(out, str, savelen);
1091 		STADJUST(savelen, out);
1092 		ckfree(str);
1093 		str = NULL;
1094 	}
1095 	if (ostr) {
1096 		ckfree(ostr);
1097 		ostr = NULL;
1098 	}
1099 	here = saveheredoclist;
1100 	if (here != NULL) {
1101 		while (here->next != NULL)
1102 			here = here->next;
1103 		here->next = heredoclist;
1104 		heredoclist = saveheredoclist;
1105 	}
1106 	handler = savehandler;
1107 	INTON;
1108 	if (quoted)
1109 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1110 	else
1111 		USTPUTC(CTLBACKQ, out);
1112 	return out;
1113 }
1114 
1115 
1116 /*
1117  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
1118  * is not NULL, read a here document.  In the latter case, eofmark is the
1119  * word which marks the end of the document and striptabs is true if
1120  * leading tabs should be stripped from the document.  The argument firstc
1121  * is the first character of the input token or document.
1122  *
1123  * Because C does not have internal subroutines, I have simulated them
1124  * using goto's to implement the subroutine linkage.  The following macros
1125  * will run code that appears at the end of readtoken1.
1126  */
1127 
1128 #define CHECKEND()	{goto checkend; checkend_return:;}
1129 #define PARSEREDIR()	{goto parseredir; parseredir_return:;}
1130 #define PARSESUB()	{goto parsesub; parsesub_return:;}
1131 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
1132 
1133 STATIC int
1134 readtoken1(int firstc, char const *initialsyntax, char *eofmark, int striptabs)
1135 {
1136 	int c = firstc;
1137 	char *out;
1138 	int len;
1139 	char line[EOFMARKLEN + 1];
1140 	struct nodelist *bqlist;
1141 	int quotef;
1142 	int newvarnest;
1143 	int level;
1144 	int synentry;
1145 	struct tokenstate state_static[MAXNEST_STATIC];
1146 	int maxnest = MAXNEST_STATIC;
1147 	struct tokenstate *state = state_static;
1148 
1149 	startlinno = plinno;
1150 	quotef = 0;
1151 	bqlist = NULL;
1152 	newvarnest = 0;
1153 	level = 0;
1154 	state[level].syntax = initialsyntax;
1155 	state[level].parenlevel = 0;
1156 	state[level].category = TSTATE_TOP;
1157 
1158 	STARTSTACKSTR(out);
1159 	loop: {	/* for each line, until end of word */
1160 		CHECKEND();	/* set c to PEOF if at end of here document */
1161 		for (;;) {	/* until end of line or end of word */
1162 			CHECKSTRSPACE(3, out);	/* permit 3 calls to USTPUTC */
1163 
1164 			synentry = state[level].syntax[c];
1165 
1166 			switch(synentry) {
1167 			case CNL:	/* '\n' */
1168 				if (state[level].syntax == BASESYNTAX)
1169 					goto endword;	/* exit outer loop */
1170 				USTPUTC(c, out);
1171 				plinno++;
1172 				if (doprompt)
1173 					setprompt(2);
1174 				else
1175 					setprompt(0);
1176 				c = pgetc();
1177 				goto loop;		/* continue outer loop */
1178 			case CWORD:
1179 				USTPUTC(c, out);
1180 				break;
1181 			case CCTL:
1182 				if (eofmark == NULL || initialsyntax != SQSYNTAX)
1183 					USTPUTC(CTLESC, out);
1184 				USTPUTC(c, out);
1185 				break;
1186 			case CBACK:	/* backslash */
1187 				c = pgetc();
1188 				if (c == PEOF) {
1189 					USTPUTC('\\', out);
1190 					pungetc();
1191 				} else if (c == '\n') {
1192 					plinno++;
1193 					if (doprompt)
1194 						setprompt(2);
1195 					else
1196 						setprompt(0);
1197 				} else {
1198 					if (state[level].syntax == DQSYNTAX &&
1199 					    c != '\\' && c != '`' && c != '$' &&
1200 					    (c != '"' || (eofmark != NULL &&
1201 						newvarnest == 0)) &&
1202 					    (c != '}' || state[level].category != TSTATE_VAR_OLD))
1203 						USTPUTC('\\', out);
1204 					if (SQSYNTAX[c] == CCTL)
1205 						USTPUTC(CTLESC, out);
1206 					else if (eofmark == NULL ||
1207 					    newvarnest > 0)
1208 						USTPUTC(CTLQUOTEMARK, out);
1209 					USTPUTC(c, out);
1210 					quotef++;
1211 				}
1212 				break;
1213 			case CSQUOTE:
1214 				USTPUTC(CTLQUOTEMARK, out);
1215 				state[level].syntax = SQSYNTAX;
1216 				break;
1217 			case CDQUOTE:
1218 				USTPUTC(CTLQUOTEMARK, out);
1219 				state[level].syntax = DQSYNTAX;
1220 				break;
1221 			case CENDQUOTE:
1222 				if (eofmark != NULL && newvarnest == 0)
1223 					USTPUTC(c, out);
1224 				else {
1225 					if (state[level].category == TSTATE_ARITH)
1226 						state[level].syntax = ARISYNTAX;
1227 					else
1228 						state[level].syntax = BASESYNTAX;
1229 					quotef++;
1230 				}
1231 				break;
1232 			case CVAR:	/* '$' */
1233 				PARSESUB();		/* parse substitution */
1234 				break;
1235 			case CENDVAR:	/* '}' */
1236 				if (level > 0 &&
1237 				    (state[level].category == TSTATE_VAR_OLD ||
1238 				    state[level].category == TSTATE_VAR_NEW)) {
1239 					if (state[level].category == TSTATE_VAR_OLD)
1240 						state[level - 1].syntax = state[level].syntax;
1241 					else
1242 						newvarnest--;
1243 					level--;
1244 					USTPUTC(CTLENDVAR, out);
1245 				} else {
1246 					USTPUTC(c, out);
1247 				}
1248 				break;
1249 			case CLP:	/* '(' in arithmetic */
1250 				state[level].parenlevel++;
1251 				USTPUTC(c, out);
1252 				break;
1253 			case CRP:	/* ')' in arithmetic */
1254 				if (state[level].parenlevel > 0) {
1255 					USTPUTC(c, out);
1256 					--state[level].parenlevel;
1257 				} else {
1258 					if (pgetc() == ')') {
1259 						if (level > 0 &&
1260 						    state[level].category == TSTATE_ARITH) {
1261 							level--;
1262 							USTPUTC(CTLENDARI, out);
1263 						} else
1264 							USTPUTC(')', out);
1265 					} else {
1266 						/*
1267 						 * unbalanced parens
1268 						 *  (don't 2nd guess - no error)
1269 						 */
1270 						pungetc();
1271 						USTPUTC(')', out);
1272 					}
1273 				}
1274 				break;
1275 			case CBQUOTE:	/* '`' */
1276 				out = parsebackq(out, &bqlist, 1,
1277 				    state[level].syntax == DQSYNTAX &&
1278 				    (eofmark == NULL || newvarnest > 0),
1279 				    state[level].syntax == DQSYNTAX || state[level].syntax == ARISYNTAX);
1280 				break;
1281 			case CEOF:
1282 				goto endword;		/* exit outer loop */
1283 			default:
1284 				if (level == 0)
1285 					goto endword;	/* exit outer loop */
1286 				USTPUTC(c, out);
1287 			}
1288 			c = pgetc_macro();
1289 		}
1290 	}
1291 endword:
1292 	if (state[level].syntax == ARISYNTAX)
1293 		synerror("Missing '))'");
1294 	if (state[level].syntax != BASESYNTAX && eofmark == NULL)
1295 		synerror("Unterminated quoted string");
1296 	if (state[level].category == TSTATE_VAR_OLD ||
1297 	    state[level].category == TSTATE_VAR_NEW) {
1298 		startlinno = plinno;
1299 		synerror("Missing '}'");
1300 	}
1301 	if (state != state_static)
1302 		parser_temp_free_upto(state);
1303 	USTPUTC('\0', out);
1304 	len = out - stackblock();
1305 	out = stackblock();
1306 	if (eofmark == NULL) {
1307 		if ((c == '>' || c == '<')
1308 		 && quotef == 0
1309 		 && len <= 2
1310 		 && (*out == '\0' || is_digit(*out))) {
1311 			PARSEREDIR();
1312 			return lasttoken = TREDIR;
1313 		} else {
1314 			pungetc();
1315 		}
1316 	}
1317 	quoteflag = quotef;
1318 	backquotelist = bqlist;
1319 	grabstackblock(len);
1320 	wordtext = out;
1321 	return lasttoken = TWORD;
1322 /* end of readtoken routine */
1323 
1324 
1325 /*
1326  * Check to see whether we are at the end of the here document.  When this
1327  * is called, c is set to the first character of the next input line.  If
1328  * we are at the end of the here document, this routine sets the c to PEOF.
1329  */
1330 
1331 checkend: {
1332 	if (eofmark) {
1333 		if (striptabs) {
1334 			while (c == '\t')
1335 				c = pgetc();
1336 		}
1337 		if (c == *eofmark) {
1338 			if (pfgets(line, sizeof line) != NULL) {
1339 				char *p, *q;
1340 
1341 				p = line;
1342 				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
1343 				if (*p == '\n' && *q == '\0') {
1344 					c = PEOF;
1345 					plinno++;
1346 					needprompt = doprompt;
1347 				} else {
1348 					pushstring(line, strlen(line), NULL);
1349 				}
1350 			}
1351 		}
1352 	}
1353 	goto checkend_return;
1354 }
1355 
1356 
1357 /*
1358  * Parse a redirection operator.  The variable "out" points to a string
1359  * specifying the fd to be redirected.  The variable "c" contains the
1360  * first character of the redirection operator.
1361  */
1362 
1363 parseredir: {
1364 	char fd = *out;
1365 	union node *np;
1366 
1367 	np = (union node *)stalloc(sizeof (struct nfile));
1368 	if (c == '>') {
1369 		np->nfile.fd = 1;
1370 		c = pgetc();
1371 		if (c == '>')
1372 			np->type = NAPPEND;
1373 		else if (c == '&')
1374 			np->type = NTOFD;
1375 		else if (c == '|')
1376 			np->type = NCLOBBER;
1377 		else {
1378 			np->type = NTO;
1379 			pungetc();
1380 		}
1381 	} else {	/* c == '<' */
1382 		np->nfile.fd = 0;
1383 		c = pgetc();
1384 		if (c == '<') {
1385 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
1386 				np = (union node *)stalloc(sizeof (struct nhere));
1387 				np->nfile.fd = 0;
1388 			}
1389 			np->type = NHERE;
1390 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
1391 			heredoc->here = np;
1392 			if ((c = pgetc()) == '-') {
1393 				heredoc->striptabs = 1;
1394 			} else {
1395 				heredoc->striptabs = 0;
1396 				pungetc();
1397 			}
1398 		} else if (c == '&')
1399 			np->type = NFROMFD;
1400 		else if (c == '>')
1401 			np->type = NFROMTO;
1402 		else {
1403 			np->type = NFROM;
1404 			pungetc();
1405 		}
1406 	}
1407 	if (fd != '\0')
1408 		np->nfile.fd = digit_val(fd);
1409 	redirnode = np;
1410 	goto parseredir_return;
1411 }
1412 
1413 
1414 /*
1415  * Parse a substitution.  At this point, we have read the dollar sign
1416  * and nothing else.
1417  */
1418 
1419 parsesub: {
1420 	char buf[10];
1421 	int subtype;
1422 	int typeloc;
1423 	int flags;
1424 	char *p;
1425 	static const char types[] = "}-+?=";
1426 	int bracketed_name = 0; /* used to handle ${[0-9]*} variables */
1427 	int i;
1428 	int linno;
1429 	int length;
1430 
1431 	c = pgetc();
1432 	if (c != '(' && c != '{' && (is_eof(c) || !is_name(c)) &&
1433 	    !is_special(c)) {
1434 		USTPUTC('$', out);
1435 		pungetc();
1436 	} else if (c == '(') {	/* $(command) or $((arith)) */
1437 		if (pgetc() == '(') {
1438 			PARSEARITH();
1439 		} else {
1440 			pungetc();
1441 			out = parsebackq(out, &bqlist, 0,
1442 			    state[level].syntax == DQSYNTAX &&
1443 			    (eofmark == NULL || newvarnest > 0),
1444 			    state[level].syntax == DQSYNTAX ||
1445 			    state[level].syntax == ARISYNTAX);
1446 		}
1447 	} else {
1448 		USTPUTC(CTLVAR, out);
1449 		typeloc = out - stackblock();
1450 		USTPUTC(VSNORMAL, out);
1451 		subtype = VSNORMAL;
1452 		flags = 0;
1453 		if (c == '{') {
1454 			bracketed_name = 1;
1455 			c = pgetc();
1456 			if (c == '#') {
1457 				if ((c = pgetc()) == '}')
1458 					c = '#';
1459 				else
1460 					subtype = VSLENGTH;
1461 			}
1462 			else
1463 				subtype = 0;
1464 		}
1465 		if (!is_eof(c) && is_name(c)) {
1466 			length = 0;
1467 			do {
1468 				STPUTC(c, out);
1469 				c = pgetc();
1470 				length++;
1471 			} while (!is_eof(c) && is_in_name(c));
1472 			if (length == 6 &&
1473 			    strncmp(out - length, "LINENO", length) == 0) {
1474 				/* Replace the variable name with the
1475 				 * current line number. */
1476 				linno = plinno;
1477 				if (funclinno != 0)
1478 					linno -= funclinno - 1;
1479 				snprintf(buf, sizeof(buf), "%d", linno);
1480 				STADJUST(-6, out);
1481 				for (i = 0; buf[i] != '\0'; i++)
1482 					STPUTC(buf[i], out);
1483 				flags |= VSLINENO;
1484 			}
1485 		} else if (is_digit(c)) {
1486 			if (bracketed_name) {
1487 				do {
1488 					STPUTC(c, out);
1489 					c = pgetc();
1490 				} while (is_digit(c));
1491 			} else {
1492 				STPUTC(c, out);
1493 				c = pgetc();
1494 			}
1495 		} else {
1496 			if (! is_special(c)) {
1497 				subtype = VSERROR;
1498 				if (c == '}')
1499 					pungetc();
1500 				else if (c == '\n' || c == PEOF)
1501 					synerror("Unexpected end of line in substitution");
1502 				else
1503 					USTPUTC(c, out);
1504 			} else {
1505 				USTPUTC(c, out);
1506 				c = pgetc();
1507 			}
1508 		}
1509 		if (subtype == 0) {
1510 			switch (c) {
1511 			case ':':
1512 				flags |= VSNUL;
1513 				c = pgetc();
1514 				/*FALLTHROUGH*/
1515 			default:
1516 				p = strchr(types, c);
1517 				if (p == NULL) {
1518 					if (c == '\n' || c == PEOF)
1519 						synerror("Unexpected end of line in substitution");
1520 					if (flags == VSNUL)
1521 						STPUTC(':', out);
1522 					STPUTC(c, out);
1523 					subtype = VSERROR;
1524 				} else
1525 					subtype = p - types + VSNORMAL;
1526 				break;
1527 			case '%':
1528 			case '#':
1529 				{
1530 					int cc = c;
1531 					subtype = c == '#' ? VSTRIMLEFT :
1532 							     VSTRIMRIGHT;
1533 					c = pgetc();
1534 					if (c == cc)
1535 						subtype++;
1536 					else
1537 						pungetc();
1538 					break;
1539 				}
1540 			}
1541 		} else if (subtype != VSERROR) {
1542 			pungetc();
1543 		}
1544 		STPUTC('=', out);
1545 		if (subtype != VSLENGTH && (state[level].syntax == DQSYNTAX ||
1546 		    state[level].syntax == ARISYNTAX))
1547 			flags |= VSQUOTE;
1548 		*(stackblock() + typeloc) = subtype | flags;
1549 		if (subtype != VSNORMAL) {
1550 			if (level + 1 >= maxnest) {
1551 				maxnest *= 2;
1552 				if (state == state_static) {
1553 					state = parser_temp_alloc(
1554 					    maxnest * sizeof(*state));
1555 					memcpy(state, state_static,
1556 					    MAXNEST_STATIC * sizeof(*state));
1557 				} else
1558 					state = parser_temp_realloc(state,
1559 					    maxnest * sizeof(*state));
1560 			}
1561 			level++;
1562 			state[level].parenlevel = 0;
1563 			if (subtype == VSMINUS || subtype == VSPLUS ||
1564 			    subtype == VSQUESTION || subtype == VSASSIGN) {
1565 				/*
1566 				 * For operators that were in the Bourne shell,
1567 				 * inherit the double-quote state.
1568 				 */
1569 				state[level].syntax = state[level - 1].syntax;
1570 				state[level].category = TSTATE_VAR_OLD;
1571 			} else {
1572 				/*
1573 				 * The other operators take a pattern,
1574 				 * so go to BASESYNTAX.
1575 				 * Also, ' and " are now special, even
1576 				 * in here documents.
1577 				 */
1578 				state[level].syntax = BASESYNTAX;
1579 				state[level].category = TSTATE_VAR_NEW;
1580 				newvarnest++;
1581 			}
1582 		}
1583 	}
1584 	goto parsesub_return;
1585 }
1586 
1587 
1588 /*
1589  * Parse an arithmetic expansion (indicate start of one and set state)
1590  */
1591 parsearith: {
1592 
1593 	if (level + 1 >= maxnest) {
1594 		maxnest *= 2;
1595 		if (state == state_static) {
1596 			state = parser_temp_alloc(
1597 			    maxnest * sizeof(*state));
1598 			memcpy(state, state_static,
1599 			    MAXNEST_STATIC * sizeof(*state));
1600 		} else
1601 			state = parser_temp_realloc(state,
1602 			    maxnest * sizeof(*state));
1603 	}
1604 	level++;
1605 	state[level].syntax = ARISYNTAX;
1606 	state[level].parenlevel = 0;
1607 	state[level].category = TSTATE_ARITH;
1608 	USTPUTC(CTLARI, out);
1609 	if (state[level - 1].syntax == DQSYNTAX)
1610 		USTPUTC('"',out);
1611 	else
1612 		USTPUTC(' ',out);
1613 	goto parsearith_return;
1614 }
1615 
1616 } /* end of readtoken */
1617 
1618 
1619 
1620 #ifdef mkinit
1621 RESET {
1622 	tokpushback = 0;
1623 	checkkwd = 0;
1624 }
1625 #endif
1626 
1627 /*
1628  * Returns true if the text contains nothing to expand (no dollar signs
1629  * or backquotes).
1630  */
1631 
1632 STATIC int
1633 noexpand(char *text)
1634 {
1635 	char *p;
1636 	char c;
1637 
1638 	p = text;
1639 	while ((c = *p++) != '\0') {
1640 		if ( c == CTLQUOTEMARK)
1641 			continue;
1642 		if (c == CTLESC)
1643 			p++;
1644 		else if (BASESYNTAX[(int)c] == CCTL)
1645 			return 0;
1646 	}
1647 	return 1;
1648 }
1649 
1650 
1651 /*
1652  * Return true if the argument is a legal variable name (a letter or
1653  * underscore followed by zero or more letters, underscores, and digits).
1654  */
1655 
1656 int
1657 goodname(const char *name)
1658 {
1659 	const char *p;
1660 
1661 	p = name;
1662 	if (! is_name(*p))
1663 		return 0;
1664 	while (*++p) {
1665 		if (! is_in_name(*p))
1666 			return 0;
1667 	}
1668 	return 1;
1669 }
1670 
1671 
1672 /*
1673  * Called when an unexpected token is read during the parse.  The argument
1674  * is the token that is expected, or -1 if more than one type of token can
1675  * occur at this point.
1676  */
1677 
1678 STATIC void
1679 synexpect(int token)
1680 {
1681 	char msg[64];
1682 
1683 	if (token >= 0) {
1684 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1685 			tokname[lasttoken], tokname[token]);
1686 	} else {
1687 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1688 	}
1689 	synerror(msg);
1690 }
1691 
1692 
1693 STATIC void
1694 synerror(const char *msg)
1695 {
1696 	if (commandname)
1697 		outfmt(out2, "%s: %d: ", commandname, startlinno);
1698 	outfmt(out2, "Syntax error: %s\n", msg);
1699 	error((char *)NULL);
1700 }
1701 
1702 STATIC void
1703 setprompt(int which)
1704 {
1705 	whichprompt = which;
1706 
1707 #ifndef NO_HISTORY
1708 	if (!el)
1709 #endif
1710 	{
1711 		out2str(getprompt(NULL));
1712 		flushout(out2);
1713 	}
1714 }
1715 
1716 /*
1717  * called by editline -- any expansions to the prompt
1718  *    should be added here.
1719  */
1720 char *
1721 getprompt(void *unused __unused)
1722 {
1723 	static char ps[PROMPTLEN];
1724 	char *fmt;
1725 	const char *pwd;
1726 	int i, trim;
1727 	static char internal_error[] = "<internal prompt error>";
1728 
1729 	/*
1730 	 * Select prompt format.
1731 	 */
1732 	switch (whichprompt) {
1733 	case 0:
1734 		fmt = nullstr;
1735 		break;
1736 	case 1:
1737 		fmt = ps1val();
1738 		break;
1739 	case 2:
1740 		fmt = ps2val();
1741 		break;
1742 	default:
1743 		return internal_error;
1744 	}
1745 
1746 	/*
1747 	 * Format prompt string.
1748 	 */
1749 	for (i = 0; (i < 127) && (*fmt != '\0'); i++, fmt++)
1750 		if (*fmt == '\\')
1751 			switch (*++fmt) {
1752 
1753 				/*
1754 				 * Hostname.
1755 				 *
1756 				 * \h specifies just the local hostname,
1757 				 * \H specifies fully-qualified hostname.
1758 				 */
1759 			case 'h':
1760 			case 'H':
1761 				ps[i] = '\0';
1762 				gethostname(&ps[i], PROMPTLEN - i);
1763 				/* Skip to end of hostname. */
1764 				trim = (*fmt == 'h') ? '.' : '\0';
1765 				while ((ps[i+1] != '\0') && (ps[i+1] != trim))
1766 					i++;
1767 				break;
1768 
1769 				/*
1770 				 * Working directory.
1771 				 *
1772 				 * \W specifies just the final component,
1773 				 * \w specifies the entire path.
1774 				 */
1775 			case 'W':
1776 			case 'w':
1777 				pwd = lookupvar("PWD");
1778 				if (pwd == NULL)
1779 					pwd = "?";
1780 				if (*fmt == 'W' &&
1781 				    *pwd == '/' && pwd[1] != '\0')
1782 					strlcpy(&ps[i], strrchr(pwd, '/') + 1,
1783 					    PROMPTLEN - i);
1784 				else
1785 					strlcpy(&ps[i], pwd, PROMPTLEN - i);
1786 				/* Skip to end of path. */
1787 				while (ps[i + 1] != '\0')
1788 					i++;
1789 				break;
1790 
1791 				/*
1792 				 * Superuser status.
1793 				 *
1794 				 * '$' for normal users, '#' for root.
1795 				 */
1796 			case '$':
1797 				ps[i] = (geteuid() != 0) ? '$' : '#';
1798 				break;
1799 
1800 				/*
1801 				 * A literal \.
1802 				 */
1803 			case '\\':
1804 				ps[i] = '\\';
1805 				break;
1806 
1807 				/*
1808 				 * Emit unrecognized formats verbatim.
1809 				 */
1810 			default:
1811 				ps[i++] = '\\';
1812 				ps[i] = *fmt;
1813 				break;
1814 			}
1815 		else
1816 			ps[i] = *fmt;
1817 	ps[i] = '\0';
1818 	return (ps);
1819 }
1820