xref: /original-bsd/bin/sh/parser.c (revision 6a39c8ab)
1 /*-
2  * Copyright (c) 1991 The Regents of the University of California.
3  * All rights reserved.
4  *
5  * This code is derived from software contributed to Berkeley by
6  * Kenneth Almquist.
7  *
8  * %sccs.include.redist.c%
9  */
10 
11 #ifndef lint
12 static char sccsid[] = "@(#)parser.c	5.7 (Berkeley) 06/23/92";
13 #endif /* not lint */
14 
15 #include "shell.h"
16 #include "parser.h"
17 #include "nodes.h"
18 #include "expand.h"	/* defines rmescapes() */
19 #include "redir.h"	/* defines copyfd() */
20 #include "syntax.h"
21 #include "options.h"
22 #include "input.h"
23 #include "output.h"
24 #include "var.h"
25 #include "error.h"
26 #include "memalloc.h"
27 #include "mystring.h"
28 #include "alias.h"
29 #include "myhistedit.h"
30 
31 
32 /*
33  * Shell command parser.
34  */
35 
36 #define EOFMARKLEN 79
37 
38 /* values returned by readtoken */
39 #include "token.def"
40 
41 
42 
43 struct heredoc {
44 	struct heredoc *next;	/* next here document in list */
45 	union node *here;		/* redirection node */
46 	char *eofmark;		/* string indicating end of input */
47 	int striptabs;		/* if set, strip leading tabs */
48 };
49 
50 
51 
52 struct heredoc *heredoclist;	/* list of here documents to read */
53 int parsebackquote;		/* nonzero if we are inside backquotes */
54 int doprompt;			/* if set, prompt the user */
55 int needprompt;			/* true if interactive and at start of line */
56 int lasttoken;			/* last token read */
57 MKINIT int tokpushback;		/* last token pushed back */
58 char *wordtext;			/* text of last word returned by readtoken */
59 MKINIT int checkkwd;            /* 1 == check for kwds, 2 == also eat newlines */
60 struct nodelist *backquotelist;
61 union node *redirnode;
62 struct heredoc *heredoc;
63 int quoteflag;			/* set if (part of) last token was quoted */
64 int startlinno;			/* line # where last token started */
65 
66 
67 #define GDB_HACK 1 /* avoid local declarations which gdb can't handle */
68 #ifdef GDB_HACK
69 static const char argvars[5] = {CTLVAR, VSNORMAL|VSQUOTE, '@', '=', '\0'};
70 static const char types[] = "}-+?=";
71 #endif
72 
73 
74 STATIC union node *list __P((int));
75 STATIC union node *andor __P((void));
76 STATIC union node *pipeline __P((void));
77 STATIC union node *command __P((void));
78 STATIC union node *simplecmd __P((void));
79 STATIC void parsefname __P((void));
80 STATIC void parseheredoc __P((void));
81 STATIC int readtoken __P((void));
82 STATIC int readtoken1 __P((int, char const *, char *, int));
83 STATIC void attyline __P((void));
84 STATIC int noexpand __P((char *));
85 STATIC void synexpect __P((int));
86 STATIC void synerror __P((char *));
87 STATIC void setprompt __P((int));
88 
89 /*
90  * Read and parse a command.  Returns NEOF on end of file.  (NULL is a
91  * valid parse tree indicating a blank line.)
92  */
93 
94 union node *
95 parsecmd(interact) {
96 	int t;
97 
98 	doprompt = interact;
99 	if (doprompt)
100 		setprompt(1);
101 	else
102 		setprompt(0);
103 	needprompt = 0;
104 	t = readtoken();
105 	if (t == TEOF)
106 		return NEOF;
107 	if (t == TNL)
108 		return NULL;
109 	tokpushback++;
110 	return list(1);
111 }
112 
113 
114 STATIC union node *
115 list(nlflag) {
116 	union node *n1, *n2, *n3;
117 
118 	checkkwd = 2;
119 	if (nlflag == 0 && tokendlist[peektoken()])
120 		return NULL;
121 	n1 = andor();
122 	for (;;) {
123 		switch (readtoken()) {
124 		case TBACKGND:
125 			if (n1->type == NCMD || n1->type == NPIPE) {
126 				n1->ncmd.backgnd = 1;
127 			} else if (n1->type == NREDIR) {
128 				n1->type = NBACKGND;
129 			} else {
130 				n3 = (union node *)stalloc(sizeof (struct nredir));
131 				n3->type = NBACKGND;
132 				n3->nredir.n = n1;
133 				n3->nredir.redirect = NULL;
134 				n1 = n3;
135 			}
136 			goto tsemi;
137 		case TNL:
138 			tokpushback++;
139 			/* fall through */
140 tsemi:	    case TSEMI:
141 			if (readtoken() == TNL) {
142 				parseheredoc();
143 				if (nlflag)
144 					return n1;
145 			} else {
146 				tokpushback++;
147 			}
148 			checkkwd = 2;
149 			if (tokendlist[peektoken()])
150 				return n1;
151 			n2 = andor();
152 			n3 = (union node *)stalloc(sizeof (struct nbinary));
153 			n3->type = NSEMI;
154 			n3->nbinary.ch1 = n1;
155 			n3->nbinary.ch2 = n2;
156 			n1 = n3;
157 			break;
158 		case TEOF:
159 			if (heredoclist)
160 				parseheredoc();
161 			else
162 				pungetc();		/* push back EOF on input */
163 			return n1;
164 		default:
165 			if (nlflag)
166 				synexpect(-1);
167 			tokpushback++;
168 			return n1;
169 		}
170 	}
171 }
172 
173 
174 
175 STATIC union node *
176 andor() {
177 	union node *n1, *n2, *n3;
178 	int t;
179 
180 	n1 = pipeline();
181 	for (;;) {
182 		if ((t = readtoken()) == TAND) {
183 			t = NAND;
184 		} else if (t == TOR) {
185 			t = NOR;
186 		} else {
187 			tokpushback++;
188 			return n1;
189 		}
190 		n2 = pipeline();
191 		n3 = (union node *)stalloc(sizeof (struct nbinary));
192 		n3->type = t;
193 		n3->nbinary.ch1 = n1;
194 		n3->nbinary.ch2 = n2;
195 		n1 = n3;
196 	}
197 }
198 
199 
200 
201 STATIC union node *
202 pipeline() {
203 	union node *n1, *pipenode, *notnode;
204 	struct nodelist *lp, *prev;
205 	int negate = 0;
206 
207 	TRACE(("pipeline: entered\n"));
208 	while (readtoken() == TNOT) {
209 		TRACE(("pipeline: TNOT recognized\n"));
210 		negate = !negate;
211 	}
212 	tokpushback++;
213 	n1 = command();
214 	if (readtoken() == TPIPE) {
215 		pipenode = (union node *)stalloc(sizeof (struct npipe));
216 		pipenode->type = NPIPE;
217 		pipenode->npipe.backgnd = 0;
218 		lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
219 		pipenode->npipe.cmdlist = lp;
220 		lp->n = n1;
221 		do {
222 			prev = lp;
223 			lp = (struct nodelist *)stalloc(sizeof (struct nodelist));
224 			lp->n = command();
225 			prev->next = lp;
226 		} while (readtoken() == TPIPE);
227 		lp->next = NULL;
228 		n1 = pipenode;
229 	}
230 	tokpushback++;
231 	if (negate) {
232 		notnode = (union node *)stalloc(sizeof (struct nnot));
233 		notnode->type = NNOT;
234 		notnode->nnot.com = n1;
235 		n1 = notnode;
236 	}
237 	return n1;
238 }
239 
240 
241 
242 STATIC union node *
243 command() {
244 	union node *n1, *n2;
245 	union node *ap, **app;
246 	union node *cp, **cpp;
247 	union node *redir, **rpp;
248 	int t;
249 
250 	checkkwd = 2;
251 	switch (readtoken()) {
252 	case TIF:
253 		n1 = (union node *)stalloc(sizeof (struct nif));
254 		n1->type = NIF;
255 		n1->nif.test = list(0);
256 		if (readtoken() != TTHEN)
257 			synexpect(TTHEN);
258 		n1->nif.ifpart = list(0);
259 		n2 = n1;
260 		while (readtoken() == TELIF) {
261 			n2->nif.elsepart = (union node *)stalloc(sizeof (struct nif));
262 			n2 = n2->nif.elsepart;
263 			n2->type = NIF;
264 			n2->nif.test = list(0);
265 			if (readtoken() != TTHEN)
266 				synexpect(TTHEN);
267 			n2->nif.ifpart = list(0);
268 		}
269 		if (lasttoken == TELSE)
270 			n2->nif.elsepart = list(0);
271 		else {
272 			n2->nif.elsepart = NULL;
273 			tokpushback++;
274 		}
275 		if (readtoken() != TFI)
276 			synexpect(TFI);
277 		checkkwd = 1;
278 		break;
279 	case TWHILE:
280 	case TUNTIL: {
281 		int got;
282 		n1 = (union node *)stalloc(sizeof (struct nbinary));
283 		n1->type = (lasttoken == TWHILE)? NWHILE : NUNTIL;
284 		n1->nbinary.ch1 = list(0);
285 		if ((got=readtoken()) != TDO) {
286 TRACE(("expecting DO got %s %s\n", tokname[got], got == TWORD ? wordtext : ""));
287 			synexpect(TDO);
288 		}
289 		n1->nbinary.ch2 = list(0);
290 		if (readtoken() != TDONE)
291 			synexpect(TDONE);
292 		checkkwd = 1;
293 		break;
294 	}
295 	case TFOR:
296 		if (readtoken() != TWORD || quoteflag || ! goodname(wordtext))
297 			synerror("Bad for loop variable");
298 		n1 = (union node *)stalloc(sizeof (struct nfor));
299 		n1->type = NFOR;
300 		n1->nfor.var = wordtext;
301 		if (readtoken() == TWORD && ! quoteflag && equal(wordtext, "in")) {
302 			app = ≈
303 			while (readtoken() == TWORD) {
304 				n2 = (union node *)stalloc(sizeof (struct narg));
305 				n2->type = NARG;
306 				n2->narg.text = wordtext;
307 				n2->narg.backquote = backquotelist;
308 				*app = n2;
309 				app = &n2->narg.next;
310 			}
311 			*app = NULL;
312 			n1->nfor.args = ap;
313 		} else {
314 #ifndef GDB_HACK
315 			static const char argvars[5] = {CTLVAR, VSNORMAL|VSQUOTE,
316 								   '@', '=', '\0'};
317 #endif
318 			n2 = (union node *)stalloc(sizeof (struct narg));
319 			n2->type = NARG;
320 			n2->narg.text = (char *)argvars;
321 			n2->narg.backquote = NULL;
322 			n2->narg.next = NULL;
323 			n1->nfor.args = n2;
324 		}
325 		if (lasttoken != TNL && lasttoken != TSEMI)
326 			synexpect(-1);
327 		checkkwd = 2;
328 		if ((t = readtoken()) == TDO)
329 			t = TDONE;
330 		else if (t == TBEGIN)
331 			t = TEND;
332 		else
333 			synexpect(-1);
334 		n1->nfor.body = list(0);
335 		if (readtoken() != t)
336 			synexpect(t);
337 		checkkwd = 1;
338 		break;
339 	case TCASE:
340 		n1 = (union node *)stalloc(sizeof (struct ncase));
341 		n1->type = NCASE;
342 		if (readtoken() != TWORD)
343 			synexpect(TWORD);
344 		n1->ncase.expr = n2 = (union node *)stalloc(sizeof (struct narg));
345 		n2->type = NARG;
346 		n2->narg.text = wordtext;
347 		n2->narg.backquote = backquotelist;
348 		n2->narg.next = NULL;
349 		while (readtoken() == TNL);
350 		if (lasttoken != TWORD || ! equal(wordtext, "in"))
351 			synerror("expecting \"in\"");
352 		cpp = &n1->ncase.cases;
353 		while (checkkwd = 2, readtoken() == TWORD) {
354 			*cpp = cp = (union node *)stalloc(sizeof (struct nclist));
355 			cp->type = NCLIST;
356 			app = &cp->nclist.pattern;
357 			for (;;) {
358 				*app = ap = (union node *)stalloc(sizeof (struct narg));
359 				ap->type = NARG;
360 				ap->narg.text = wordtext;
361 				ap->narg.backquote = backquotelist;
362 				if (readtoken() != TPIPE)
363 					break;
364 				app = &ap->narg.next;
365 				if (readtoken() != TWORD)
366 					synexpect(TWORD);
367 			}
368 			ap->narg.next = NULL;
369 			if (lasttoken != TRP)
370 				synexpect(TRP);
371 			cp->nclist.body = list(0);
372 			if ((t = readtoken()) == TESAC)
373 				tokpushback++;
374 			else if (t != TENDCASE)
375 				synexpect(TENDCASE);
376 			cpp = &cp->nclist.next;
377 		}
378 		*cpp = NULL;
379 		if (lasttoken != TESAC)
380 			synexpect(TESAC);
381 		checkkwd = 1;
382 		break;
383 	case TLP:
384 		n1 = (union node *)stalloc(sizeof (struct nredir));
385 		n1->type = NSUBSHELL;
386 		n1->nredir.n = list(0);
387 		n1->nredir.redirect = NULL;
388 		if (readtoken() != TRP)
389 			synexpect(TRP);
390 		checkkwd = 1;
391 		break;
392 	case TBEGIN:
393 		n1 = list(0);
394 		if (readtoken() != TEND)
395 			synexpect(TEND);
396 		checkkwd = 1;
397 		break;
398 	case TWORD:
399 	case TREDIR:
400 		tokpushback++;
401 		return simplecmd();
402 	default:
403 		synexpect(-1);
404 	}
405 
406 	/* Now check for redirection which may follow command */
407 	rpp = &redir;
408 	while (readtoken() == TREDIR) {
409 		*rpp = n2 = redirnode;
410 		rpp = &n2->nfile.next;
411 		parsefname();
412 	}
413 	tokpushback++;
414 	*rpp = NULL;
415 	if (redir) {
416 		if (n1->type != NSUBSHELL) {
417 			n2 = (union node *)stalloc(sizeof (struct nredir));
418 			n2->type = NREDIR;
419 			n2->nredir.n = n1;
420 			n1 = n2;
421 		}
422 		n1->nredir.redirect = redir;
423 	}
424 	return n1;
425 }
426 
427 
428 STATIC union node *
429 simplecmd() {
430 	union node *args, **app;
431 	union node *redir, **rpp;
432 	union node *n;
433 
434 	args = NULL;
435 	app = &args;
436 	rpp = &redir;
437 	for (;;) {
438 		if (readtoken() == TWORD) {
439 			n = (union node *)stalloc(sizeof (struct narg));
440 			n->type = NARG;
441 			n->narg.text = wordtext;
442 			n->narg.backquote = backquotelist;
443 			*app = n;
444 			app = &n->narg.next;
445 		} else if (lasttoken == TREDIR) {
446 			*rpp = n = redirnode;
447 			rpp = &n->nfile.next;
448 			parsefname();	/* read name of redirection file */
449 		} else if (lasttoken == TLP && app == &args->narg.next
450 					    && rpp == &redir) {
451 			/* We have a function */
452 			if (readtoken() != TRP)
453 				synexpect(TRP);
454 #ifdef notdef
455 			if (! goodname(n->narg.text))
456 				synerror("Bad function name");
457 #endif
458 			n->type = NDEFUN;
459 			n->narg.next = command();
460 			return n;
461 		} else {
462 			tokpushback++;
463 			break;
464 		}
465 	}
466 	*app = NULL;
467 	*rpp = NULL;
468 	n = (union node *)stalloc(sizeof (struct ncmd));
469 	n->type = NCMD;
470 	n->ncmd.backgnd = 0;
471 	n->ncmd.args = args;
472 	n->ncmd.redirect = redir;
473 	return n;
474 }
475 
476 
477 STATIC void
478 parsefname() {
479 	union node *n = redirnode;
480 
481 	if (readtoken() != TWORD)
482 		synexpect(-1);
483 	if (n->type == NHERE) {
484 		struct heredoc *here = heredoc;
485 		struct heredoc *p;
486 		int i;
487 
488 		if (quoteflag == 0)
489 			n->type = NXHERE;
490 		TRACE(("Here document %d\n", n->type));
491 		if (here->striptabs) {
492 			while (*wordtext == '\t')
493 				wordtext++;
494 		}
495 		if (! noexpand(wordtext) || (i = strlen(wordtext)) == 0 || i > EOFMARKLEN)
496 			synerror("Illegal eof marker for << redirection");
497 		rmescapes(wordtext);
498 		here->eofmark = wordtext;
499 		here->next = NULL;
500 		if (heredoclist == NULL)
501 			heredoclist = here;
502 		else {
503 			for (p = heredoclist ; p->next ; p = p->next);
504 			p->next = here;
505 		}
506 	} else if (n->type == NTOFD || n->type == NFROMFD) {
507 		if (is_digit(wordtext[0]))
508 			n->ndup.dupfd = digit_val(wordtext[0]);
509 		else if (wordtext[0] == '-')
510 			n->ndup.dupfd = -1;
511 		else
512 			goto bad;
513 		if (wordtext[1] != '\0') {
514 bad:
515 			synerror("Bad fd number");
516 		}
517 	} else {
518 		n->nfile.fname = (union node *)stalloc(sizeof (struct narg));
519 		n = n->nfile.fname;
520 		n->type = NARG;
521 		n->narg.next = NULL;
522 		n->narg.text = wordtext;
523 		n->narg.backquote = backquotelist;
524 	}
525 }
526 
527 
528 /*
529  * Input any here documents.
530  */
531 
532 STATIC void
533 parseheredoc() {
534 	struct heredoc *here;
535 	union node *n;
536 
537 	while (heredoclist) {
538 		here = heredoclist;
539 		heredoclist = here->next;
540 		if (needprompt) {
541 			setprompt(2);
542 			needprompt = 0;
543 		}
544 		readtoken1(pgetc(), here->here->type == NHERE? SQSYNTAX : DQSYNTAX,
545 				here->eofmark, here->striptabs);
546 		n = (union node *)stalloc(sizeof (struct narg));
547 		n->narg.type = NARG;
548 		n->narg.next = NULL;
549 		n->narg.text = wordtext;
550 		n->narg.backquote = backquotelist;
551 		here->here->nhere.doc = n;
552 	}
553 }
554 
555 STATIC int
556 peektoken() {
557 	int t;
558 
559 	t = readtoken();
560 	tokpushback++;
561 	return (t);
562 }
563 
564 STATIC int xxreadtoken();
565 
566 STATIC int
567 readtoken() {
568 	int t;
569 	int savecheckkwd = checkkwd;
570 	struct alias *ap;
571 #ifdef DEBUG
572 	int alreadyseen = tokpushback;
573 #endif
574 
575 	top:
576 	t = xxreadtoken();
577 
578 	if (checkkwd) {
579 		/*
580 		 * eat newlines
581 		 */
582 		if (checkkwd == 2) {
583 			checkkwd = 0;
584 			while (t == TNL) {
585 				parseheredoc();
586 				t = xxreadtoken();
587 			}
588 		} else
589 			checkkwd = 0;
590 		/*
591 		 * check for keywords and aliases
592 		 */
593 		if (t == TWORD && !quoteflag) {
594 			register char * const *pp, *s;
595 
596 			for (pp = parsekwd; *pp; pp++) {
597 				if (**pp == *wordtext && equal(*pp, wordtext)) {
598 					lasttoken = t = pp - parsekwd + KWDOFFSET;
599 					TRACE(("keyword %s recognized\n", tokname[t]));
600 					goto out;
601 				}
602 			}
603 			if (ap = lookupalias(wordtext, 1)) {
604 				pushstring(ap->val, strlen(ap->val), ap);
605 				checkkwd = savecheckkwd;
606 				goto top;
607 			}
608 		}
609 out:
610 		checkkwd = 0;
611 	}
612 #ifdef DEBUG
613 	if (!alreadyseen)
614 	    TRACE(("token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
615 	else
616 	    TRACE(("reread token %s %s\n", tokname[t], t == TWORD ? wordtext : ""));
617 #endif
618 	return (t);
619 }
620 
621 
622 /*
623  * Read the next input token.
624  * If the token is a word, we set backquotelist to the list of cmds in
625  *	backquotes.  We set quoteflag to true if any part of the word was
626  *	quoted.
627  * If the token is TREDIR, then we set redirnode to a structure containing
628  *	the redirection.
629  * In all cases, the variable startlinno is set to the number of the line
630  *	on which the token starts.
631  *
632  * [Change comment:  here documents and internal procedures]
633  * [Readtoken shouldn't have any arguments.  Perhaps we should make the
634  *  word parsing code into a separate routine.  In this case, readtoken
635  *  doesn't need to have any internal procedures, but parseword does.
636  *  We could also make parseoperator in essence the main routine, and
637  *  have parseword (readtoken1?) handle both words and redirection.]
638  */
639 
640 #define RETURN(token)	return lasttoken = token
641 
642 STATIC int
643 xxreadtoken() {
644 	register c;
645 
646 	if (tokpushback) {
647 		tokpushback = 0;
648 		return lasttoken;
649 	}
650 	if (needprompt) {
651 		setprompt(2);
652 		needprompt = 0;
653 	}
654 	startlinno = plinno;
655 	for (;;) {	/* until token or start of word found */
656 		c = pgetc_macro();
657 		if (c == ' ' || c == '\t')
658 			continue;		/* quick check for white space first */
659 		switch (c) {
660 		case ' ': case '\t':
661 			continue;
662 		case '#':
663 			while ((c = pgetc()) != '\n' && c != PEOF);
664 			pungetc();
665 			continue;
666 		case '\\':
667 			if (pgetc() == '\n') {
668 				startlinno = ++plinno;
669 				if (doprompt)
670 					setprompt(2);
671 				else
672 					setprompt(0);
673 				continue;
674 			}
675 			pungetc();
676 			goto breakloop;
677 		case '\n':
678 			plinno++;
679 			needprompt = doprompt;
680 			RETURN(TNL);
681 		case PEOF:
682 			RETURN(TEOF);
683 		case '&':
684 			if (pgetc() == '&')
685 				RETURN(TAND);
686 			pungetc();
687 			RETURN(TBACKGND);
688 		case '|':
689 			if (pgetc() == '|')
690 				RETURN(TOR);
691 			pungetc();
692 			RETURN(TPIPE);
693 		case ';':
694 			if (pgetc() == ';')
695 				RETURN(TENDCASE);
696 			pungetc();
697 			RETURN(TSEMI);
698 		case '(':
699 			RETURN(TLP);
700 		case ')':
701 			RETURN(TRP);
702 		default:
703 			goto breakloop;
704 		}
705 	}
706 breakloop:
707 	return readtoken1(c, BASESYNTAX, (char *)NULL, 0);
708 #undef RETURN
709 }
710 
711 
712 
713 /*
714  * If eofmark is NULL, read a word or a redirection symbol.  If eofmark
715  * is not NULL, read a here document.  In the latter case, eofmark is the
716  * word which marks the end of the document and striptabs is true if
717  * leading tabs should be stripped from the document.  The argument firstc
718  * is the first character of the input token or document.
719  *
720  * Because C does not have internal subroutines, I have simulated them
721  * using goto's to implement the subroutine linkage.  The following macros
722  * will run code that appears at the end of readtoken1.
723  */
724 
725 #define CHECKEND()	{goto checkend; checkend_return:;}
726 #define PARSEREDIR()	{goto parseredir; parseredir_return:;}
727 #define PARSESUB()	{goto parsesub; parsesub_return:;}
728 #define PARSEBACKQOLD()	{oldstyle = 1; goto parsebackq; parsebackq_oldreturn:;}
729 #define PARSEBACKQNEW()	{oldstyle = 0; goto parsebackq; parsebackq_newreturn:;}
730 #define	PARSEARITH()	{goto parsearith; parsearith_return:;}
731 
732 STATIC int
733 readtoken1(firstc, syntax, eofmark, striptabs)
734 	int firstc;
735 	char const *syntax;
736 	char *eofmark;
737 	int striptabs;
738 	{
739 	register c = firstc;
740 	register char *out;
741 	int len;
742 	char line[EOFMARKLEN + 1];
743 	struct nodelist *bqlist;
744 	int quotef;
745 	int dblquote;
746 	int varnest;	/* levels of variables expansion */
747 	int arinest;	/* levels of arithmetic expansion */
748 	int parenlevel;	/* levels of parens in arithmetic */
749 	int oldstyle;
750 	char const *prevsyntax;	/* syntax before arithmetic */
751 
752 	startlinno = plinno;
753 	dblquote = 0;
754 	if (syntax == DQSYNTAX)
755 		dblquote = 1;
756 	quotef = 0;
757 	bqlist = NULL;
758 	varnest = 0;
759 	arinest = 0;
760 	parenlevel = 0;
761 
762 	STARTSTACKSTR(out);
763 	loop: {	/* for each line, until end of word */
764 #if ATTY
765 		if (c == '\034' && doprompt
766 		 && attyset() && ! equal(termval(), "emacs")) {
767 			attyline();
768 			if (syntax == BASESYNTAX)
769 				return readtoken();
770 			c = pgetc();
771 			goto loop;
772 		}
773 #endif
774 		CHECKEND();	/* set c to PEOF if at end of here document */
775 		for (;;) {	/* until end of line or end of word */
776 			CHECKSTRSPACE(3, out);	/* permit 3 calls to USTPUTC */
777 			switch(syntax[c]) {
778 			case CNL:	/* '\n' */
779 				if (syntax == BASESYNTAX)
780 					goto endword;	/* exit outer loop */
781 				USTPUTC(c, out);
782 				plinno++;
783 				if (doprompt)
784 					setprompt(2);
785 				else
786 					setprompt(0);
787 				c = pgetc();
788 				goto loop;		/* continue outer loop */
789 			case CWORD:
790 				USTPUTC(c, out);
791 				break;
792 			case CCTL:
793 				if (eofmark == NULL || dblquote)
794 					USTPUTC(CTLESC, out);
795 				USTPUTC(c, out);
796 				break;
797 			case CBACK:	/* backslash */
798 				c = pgetc();
799 				if (c == PEOF) {
800 					USTPUTC('\\', out);
801 					pungetc();
802 				} else if (c == '\n') {
803 					if (doprompt)
804 						setprompt(2);
805 					else
806 						setprompt(0);
807 				} else {
808 					if (dblquote && c != '\\' && c != '`' && c != '$'
809 							 && (c != '"' || eofmark != NULL))
810 						USTPUTC('\\', out);
811 					if (SQSYNTAX[c] == CCTL)
812 						USTPUTC(CTLESC, out);
813 					USTPUTC(c, out);
814 					quotef++;
815 				}
816 				break;
817 			case CSQUOTE:
818 				syntax = SQSYNTAX;
819 				break;
820 			case CDQUOTE:
821 				syntax = DQSYNTAX;
822 				dblquote = 1;
823 				break;
824 			case CENDQUOTE:
825 				if (eofmark) {
826 					USTPUTC(c, out);
827 				} else {
828 					if (arinest)
829 						syntax = ARISYNTAX;
830 					else
831 						syntax = BASESYNTAX;
832 					quotef++;
833 					dblquote = 0;
834 				}
835 				break;
836 			case CVAR:	/* '$' */
837 				PARSESUB();		/* parse substitution */
838 				break;
839 			case CENDVAR:	/* '}' */
840 				if (varnest > 0) {
841 					varnest--;
842 					USTPUTC(CTLENDVAR, out);
843 				} else {
844 					USTPUTC(c, out);
845 				}
846 				break;
847 			case CLP:	/* '(' in arithmetic */
848 				parenlevel++;
849 				USTPUTC(c, out);
850 				break;
851 			case CRP:	/* ')' in arithmetic */
852 				if (parenlevel > 0) {
853 					USTPUTC(c, out);
854 					--parenlevel;
855 				} else {
856 					if (pgetc() == ')') {
857 						if (--arinest == 0) {
858 							USTPUTC(CTLENDARI, out);
859 							syntax = prevsyntax;
860 						} else
861 							USTPUTC(')', out);
862 					} else {
863 						/*
864 						 * unbalanced parens
865 						 *  (don't 2nd guess - no error)
866 						 */
867 						pungetc();
868 						USTPUTC(')', out);
869 					}
870 				}
871 				break;
872 			case CBQUOTE:	/* '`' */
873 				if (parsebackquote && syntax == BASESYNTAX) {
874 					if (out == stackblock())
875 						return lasttoken = TENDBQUOTE;
876 					else
877 						goto endword;	/* exit outer loop */
878 				}
879 				PARSEBACKQOLD();
880 				break;
881 			case CEOF:
882 				goto endword;		/* exit outer loop */
883 			default:
884 				if (varnest == 0)
885 					goto endword;	/* exit outer loop */
886 				USTPUTC(c, out);
887 			}
888 			c = pgetc_macro();
889 		}
890 	}
891 endword:
892 	if (syntax == ARISYNTAX)
893 		synerror("Missing '))'");
894 	if (syntax != BASESYNTAX && eofmark == NULL)
895 		synerror("Unterminated quoted string");
896 	if (varnest != 0) {
897 		startlinno = plinno;
898 		synerror("Missing '}'");
899 	}
900 	USTPUTC('\0', out);
901 	len = out - stackblock();
902 	out = stackblock();
903 	if (eofmark == NULL) {
904 		if ((c == '>' || c == '<')
905 		 && quotef == 0
906 		 && len <= 2
907 		 && (*out == '\0' || is_digit(*out))) {
908 			PARSEREDIR();
909 			return lasttoken = TREDIR;
910 		} else {
911 			pungetc();
912 		}
913 	}
914 	quoteflag = quotef;
915 	backquotelist = bqlist;
916 	grabstackblock(len);
917 	wordtext = out;
918 	return lasttoken = TWORD;
919 /* end of readtoken routine */
920 
921 
922 
923 /*
924  * Check to see whether we are at the end of the here document.  When this
925  * is called, c is set to the first character of the next input line.  If
926  * we are at the end of the here document, this routine sets the c to PEOF.
927  */
928 
929 checkend: {
930 	if (eofmark) {
931 		if (striptabs) {
932 			while (c == '\t')
933 				c = pgetc();
934 		}
935 		if (c == *eofmark) {
936 			if (pfgets(line, sizeof line) != NULL) {
937 				register char *p, *q;
938 
939 				p = line;
940 				for (q = eofmark + 1 ; *q && *p == *q ; p++, q++);
941 				if (*p == '\n' && *q == '\0') {
942 					c = PEOF;
943 					plinno++;
944 					needprompt = doprompt;
945 				} else {
946 					pushstring(line, strlen(line), NULL);
947 				}
948 			}
949 		}
950 	}
951 	goto checkend_return;
952 }
953 
954 
955 /*
956  * Parse a redirection operator.  The variable "out" points to a string
957  * specifying the fd to be redirected.  The variable "c" contains the
958  * first character of the redirection operator.
959  */
960 
961 parseredir: {
962 	char fd = *out;
963 	union node *np;
964 
965 	np = (union node *)stalloc(sizeof (struct nfile));
966 	if (c == '>') {
967 		np->nfile.fd = 1;
968 		c = pgetc();
969 		if (c == '>')
970 			np->type = NAPPEND;
971 		else if (c == '&')
972 			np->type = NTOFD;
973 		else {
974 			np->type = NTO;
975 			pungetc();
976 		}
977 	} else {	/* c == '<' */
978 		np->nfile.fd = 0;
979 		c = pgetc();
980 		if (c == '<') {
981 			if (sizeof (struct nfile) != sizeof (struct nhere)) {
982 				np = (union node *)stalloc(sizeof (struct nhere));
983 				np->nfile.fd = 0;
984 			}
985 			np->type = NHERE;
986 			heredoc = (struct heredoc *)stalloc(sizeof (struct heredoc));
987 			heredoc->here = np;
988 			if ((c = pgetc()) == '-') {
989 				heredoc->striptabs = 1;
990 			} else {
991 				heredoc->striptabs = 0;
992 				pungetc();
993 			}
994 		} else if (c == '&')
995 			np->type = NFROMFD;
996 		else {
997 			np->type = NFROM;
998 			pungetc();
999 		}
1000 	}
1001 	if (fd != '\0')
1002 		np->nfile.fd = digit_val(fd);
1003 	redirnode = np;
1004 	goto parseredir_return;
1005 }
1006 
1007 
1008 /*
1009  * Parse a substitution.  At this point, we have read the dollar sign
1010  * and nothing else.
1011  */
1012 
1013 parsesub: {
1014 	int subtype;
1015 	int typeloc;
1016 	int flags;
1017 	char *p;
1018 #ifndef GDB_HACK
1019 	static const char types[] = "}-+?=";
1020 #endif
1021 
1022 	c = pgetc();
1023 	if (c != '(' && c != '{' && !is_name(c) && !is_special(c)) {
1024 		USTPUTC('$', out);
1025 		pungetc();
1026 	} else if (c == '(') {	/* $(command) or $((arith)) */
1027 		if (pgetc() == '(') {
1028 			PARSEARITH();
1029 		} else {
1030 			pungetc();
1031 			PARSEBACKQNEW();
1032 		}
1033 	} else {
1034 		USTPUTC(CTLVAR, out);
1035 		typeloc = out - stackblock();
1036 		USTPUTC(VSNORMAL, out);
1037 		subtype = VSNORMAL;
1038 		if (c == '{') {
1039 			c = pgetc();
1040 			subtype = 0;
1041 		}
1042 		if (is_name(c)) {
1043 			do {
1044 				STPUTC(c, out);
1045 				c = pgetc();
1046 			} while (is_in_name(c));
1047 		} else {
1048 			if (! is_special(c))
1049 badsub:				synerror("Bad substitution");
1050 			USTPUTC(c, out);
1051 			c = pgetc();
1052 		}
1053 		STPUTC('=', out);
1054 		flags = 0;
1055 		if (subtype == 0) {
1056 			if (c == ':') {
1057 				flags = VSNUL;
1058 				c = pgetc();
1059 			}
1060 			p = strchr(types, c);
1061 			if (p == NULL)
1062 				goto badsub;
1063 			subtype = p - types + VSNORMAL;
1064 		} else {
1065 			pungetc();
1066 		}
1067 		if (dblquote || arinest)
1068 			flags |= VSQUOTE;
1069 		*(stackblock() + typeloc) = subtype | flags;
1070 		if (subtype != VSNORMAL)
1071 			varnest++;
1072 	}
1073 	goto parsesub_return;
1074 }
1075 
1076 
1077 /*
1078  * Called to parse command substitutions.  Newstyle is set if the command
1079  * is enclosed inside $(...); nlpp is a pointer to the head of the linked
1080  * list of commands (passed by reference), and savelen is the number of
1081  * characters on the top of the stack which must be preserved.
1082  */
1083 
1084 parsebackq: {
1085 	struct nodelist **nlpp;
1086 	int savepbq;
1087 	union node *n;
1088 	char *volatile str;
1089 	struct jmploc jmploc;
1090 	struct jmploc *volatile savehandler;
1091 	int savelen;
1092 	int t;
1093 
1094 	savepbq = parsebackquote;
1095 	if (setjmp(jmploc.loc)) {
1096 		if (str)
1097 			ckfree(str);
1098 		parsebackquote = 0;
1099 		handler = savehandler;
1100 		longjmp(handler->loc, 1);
1101 	}
1102 	INTOFF;
1103 	str = NULL;
1104 	savelen = out - stackblock();
1105 	if (savelen > 0) {
1106 		str = ckmalloc(savelen);
1107 		bcopy(stackblock(), str, savelen);
1108 	}
1109 	savehandler = handler;
1110 	handler = &jmploc;
1111 	INTON;
1112 	nlpp = &bqlist;
1113 	while (*nlpp)
1114 		nlpp = &(*nlpp)->next;
1115 	*nlpp = (struct nodelist *)stalloc(sizeof (struct nodelist));
1116 	(*nlpp)->next = NULL;
1117 	parsebackquote = oldstyle;
1118 	n = list(0);
1119 	t = oldstyle? TENDBQUOTE : TRP;
1120 	if (readtoken() != t)
1121 		synexpect(t);
1122 	(*nlpp)->n = n;
1123 	while (stackblocksize() <= savelen)
1124 		growstackblock();
1125 	STARTSTACKSTR(out);
1126 	if (str) {
1127 		bcopy(str, out, savelen);
1128 		STADJUST(savelen, out);
1129 		INTOFF;
1130 		ckfree(str);
1131 		str = NULL;
1132 		INTON;
1133 	}
1134 	parsebackquote = savepbq;
1135 	handler = savehandler;
1136 	if (arinest || dblquote)
1137 		USTPUTC(CTLBACKQ | CTLQUOTE, out);
1138 	else
1139 		USTPUTC(CTLBACKQ, out);
1140 	if (oldstyle)
1141 		goto parsebackq_oldreturn;
1142 	else
1143 		goto parsebackq_newreturn;
1144 }
1145 
1146 /*
1147  * Parse an arithmetic expansion (indicate start of one and set state)
1148  */
1149 parsearith: {
1150 
1151 	if (++arinest == 1) {
1152 		prevsyntax = syntax;
1153 		syntax = ARISYNTAX;
1154 		USTPUTC(CTLARI, out);
1155 	} else {
1156 		/*
1157 		 * we collapse embedded arithmetic expansion to
1158 		 * parenthesis, which should be equivalent
1159 		 */
1160 		USTPUTC('(', out);
1161 	}
1162 	goto parsearith_return;
1163 }
1164 
1165 } /* end of readtoken */
1166 
1167 
1168 
1169 #ifdef mkinit
1170 RESET {
1171 	tokpushback = 0;
1172 	checkkwd = 0;
1173 }
1174 #endif
1175 
1176 /*
1177  * Returns true if the text contains nothing to expand (no dollar signs
1178  * or backquotes).
1179  */
1180 
1181 STATIC int
1182 noexpand(text)
1183 	char *text;
1184 	{
1185 	register char *p;
1186 	register char c;
1187 
1188 	p = text;
1189 	while ((c = *p++) != '\0') {
1190 		if (c == CTLESC)
1191 			p++;
1192 		else if (BASESYNTAX[c] == CCTL)
1193 			return 0;
1194 	}
1195 	return 1;
1196 }
1197 
1198 
1199 /*
1200  * Return true if the argument is a legal variable name (a letter or
1201  * underscore followed by zero or more letters, underscores, and digits).
1202  */
1203 
1204 int
1205 goodname(name)
1206 	char *name;
1207 	{
1208 	register char *p;
1209 
1210 	p = name;
1211 	if (! is_name(*p))
1212 		return 0;
1213 	while (*++p) {
1214 		if (! is_in_name(*p))
1215 			return 0;
1216 	}
1217 	return 1;
1218 }
1219 
1220 
1221 /*
1222  * Called when an unexpected token is read during the parse.  The argument
1223  * is the token that is expected, or -1 if more than one type of token can
1224  * occur at this point.
1225  */
1226 
1227 STATIC void
1228 synexpect(token) {
1229 	char msg[64];
1230 
1231 	if (token >= 0) {
1232 		fmtstr(msg, 64, "%s unexpected (expecting %s)",
1233 			tokname[lasttoken], tokname[token]);
1234 	} else {
1235 		fmtstr(msg, 64, "%s unexpected", tokname[lasttoken]);
1236 	}
1237 	synerror(msg);
1238 }
1239 
1240 
1241 STATIC void
1242 synerror(msg)
1243 	char *msg;
1244 	{
1245 	if (commandname)
1246 		outfmt(&errout, "%s: %d: ", commandname, startlinno);
1247 	outfmt(&errout, "Syntax error: %s\n", msg);
1248 	error((char *)NULL);
1249 }
1250 
1251 STATIC void
1252 setprompt(which)
1253 	int which;
1254 	{
1255 	whichprompt = which;
1256 
1257 	if (!el)
1258 		out2str(getprompt(NULL));
1259 }
1260 
1261 /*
1262  * called by editline -- any expansions to the prompt
1263  *    should be added here.
1264  */
1265 char *
1266 getprompt(unused)
1267 	void *unused;
1268 	{
1269 	switch (whichprompt) {
1270 	case 0:
1271 		return "";
1272 	case 1:
1273 		return ps1val();
1274 	case 2:
1275 		return ps2val();
1276 	default:
1277 		return "<internal prompt error>";
1278 	}
1279 }
1280