xref: /titanic_41/usr/src/cmd/fm/eversholt/common/esclex.c (revision 70025d765b044c6d8594bb965a2247a61e991a99)
1 /*
2  * CDDL HEADER START
3  *
4  * The contents of this file are subject to the terms of the
5  * Common Development and Distribution License, Version 1.0 only
6  * (the "License").  You may not use this file except in compliance
7  * with the License.
8  *
9  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10  * or http://www.opensolaris.org/os/licensing.
11  * See the License for the specific language governing permissions
12  * and limitations under the License.
13  *
14  * When distributing Covered Code, include this CDDL HEADER in each
15  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16  * If applicable, add the following below this CDDL HEADER, with the
17  * fields enclosed by brackets "[]" replaced with your own identifying
18  * information: Portions Copyright [yyyy] [name of copyright owner]
19  *
20  * CDDL HEADER END
21  */
22 /*
23  * Copyright 2005 Sun Microsystems, Inc.  All rights reserved.
24  * Use is subject to license terms.
25  *
26  * esclex.c -- lexer for esc
27  *
28  * this module provides lexical analysis and error handling routine
29  * expected by the yacc-generated parser (i.e. yylex() and yyerror()).
30  * it also does lots of tracking of things like filenames, line numbers,
31  * and what tokens are seen on a line up to the point where a syntax error
32  * was found.  this module also arranges for the input source files to
33  * be run through cpp.
34  */
35 
36 #pragma ident	"%Z%%M%	%I%	%E% SMI"
37 
38 #include <stdio.h>
39 #include <ctype.h>
40 #include <string.h>
41 #include <stdlib.h>
42 #include <unistd.h>
43 #include <time.h>
44 #include <errno.h>
45 #include "out.h"
46 #include "alloc.h"
47 #include "stats.h"
48 #include "stable.h"
49 #include "lut.h"
50 #include "literals.h"
51 #include "tree.h"
52 #include "esclex.h"
53 #include "eftread.h"
54 #include "check.h"
55 #include "y.tab.h"
56 
57 /* ridiculously long token buffer -- disallow any token longer than this */
58 #define	MAXTOK	8192
59 static char Tok[MAXTOK];
60 
61 /* some misc stats we keep on the lexer & parser */
62 static struct stats *Tokcount;
63 static struct stats *Lexelapse;
64 struct stats *Filecount;
65 struct filestats {
66 	struct filestats *next;
67 	struct stats *stats;
68 } *Fstats;
69 
70 static int Errcount;
71 
72 /* input file state */
73 static char **Files;
74 static const char *Fileopened;
75 static FILE *Fp;
76 static int Line;
77 static const char *File;
78 static const char *Cpp = "/usr/bin/cpp";
79 #ifdef	ESC
80 static const char *Cppargs;
81 static const char *Cppstdargs = "-undef -Y.";
82 #endif	/* ESC */
83 
84 /* for debugging */
85 static int Lexecho;	/* echo tokens as we read them */
86 
87 /* forward declarations of our internal routines */
88 static int record(int tok, const char *s);
89 static void dumpline(int flags);
90 static void doident();
91 static void dopragma(const char *tok);
92 
93 /*
94  * table of reserved words.  this table is only used by lex_init()
95  * to intialize the Rwords lookup table.
96  */
97 static const struct {
98 	const char *word;
99 	const int val;
100 } Rwords[] = {
101 	{ "asru", ASRU },
102 	{ "div", DIV },
103 	{ "engine", ENGINE },
104 	{ "event", EVENT },
105 	{ "fru", FRU },
106 	{ "if", IF },
107 	{ "mask", MASK },
108 	{ "prop", PROP },
109 	{ "config", CONFIG },
110 	/*
111 	 * PATHFUNC indicates functions that operate only on paths
112 	 * and quotes
113 	 */
114 	{ "is_connected", PATHFUNC },
115 	{ "is_under", PATHFUNC },
116 };
117 
118 /*
119  * Rwordslut is a lookup table of reserved words.  lhs is the word
120  * (in the string table) and the rhs is the token value returned
121  * by the yylex() for that word.
122  */
123 static struct lut *Rwordslut;
124 
125 static const struct {
126 	const char *suffix;
127 	const unsigned long long nsec;
128 } Timesuffix[] = {
129 	{ "nanosecond",		1ULL },
130 	{ "nanoseconds",	1ULL },
131 	{ "nsec",		1ULL },
132 	{ "nsecs",		1ULL },
133 	{ "ns",			1ULL },
134 	{ "microsecond",	1000ULL },
135 	{ "microseconds",	1000ULL },
136 	{ "usec",		1000ULL },
137 	{ "usecs",		1000ULL },
138 	{ "us",			1000ULL },
139 	{ "millisecond",	1000000ULL },
140 	{ "milliseconds",	1000000ULL },
141 	{ "msec",		1000000ULL },
142 	{ "msecs",		1000000ULL },
143 	{ "ms",			1000000ULL },
144 	{ "second",		1000000000ULL },
145 	{ "seconds",		1000000000ULL },
146 	{ "s",			1000000000ULL },
147 	{ "minute",		1000000000ULL * 60 },
148 	{ "minutes",		1000000000ULL * 60 },
149 	{ "min",		1000000000ULL * 60 },
150 	{ "mins",		1000000000ULL * 60 },
151 	{ "m",			1000000000ULL * 60 },
152 	{ "hour",		1000000000ULL * 60 * 60 },
153 	{ "hours",		1000000000ULL * 60 * 60 },
154 	{ "hr",			1000000000ULL * 60 * 60 },
155 	{ "hrs",		1000000000ULL * 60 * 60 },
156 	{ "h",			1000000000ULL * 60 * 60 },
157 	{ "day",		1000000000ULL * 60 * 60 * 24 },
158 	{ "days",		1000000000ULL * 60 * 60 * 24 },
159 	{ "d",			1000000000ULL * 60 * 60 * 24 },
160 	{ "week",		1000000000ULL * 60 * 60 * 24 * 7 },
161 	{ "weeks",		1000000000ULL * 60 * 60 * 24 * 7 },
162 	{ "wk",			1000000000ULL * 60 * 60 * 24 * 7 },
163 	{ "wks",		1000000000ULL * 60 * 60 * 24 * 7 },
164 	{ "month",		1000000000ULL * 60 * 60 * 24 * 30 },
165 	{ "months",		1000000000ULL * 60 * 60 * 24 * 30 },
166 	{ "year",		1000000000ULL * 60 * 60 * 24 * 365 },
167 	{ "years",		1000000000ULL * 60 * 60 * 24 * 365 },
168 	{ "yr",			1000000000ULL * 60 * 60 * 24 * 365 },
169 	{ "yrs",		1000000000ULL * 60 * 60 * 24 * 365 },
170 };
171 
172 /*
173  * some wrappers around the general lut functions to provide type checking...
174  */
175 
176 static struct lut *
177 lex_s2i_lut_add(struct lut *root, const char *s, int i)
178 {
179 	return (lut_add(root, (void *)s, (void *)i, NULL));
180 }
181 
182 static int
183 lex_s2i_lut_lookup(struct lut *root, const char *s)
184 {
185 	return ((int)lut_lookup(root, (void *)s, NULL));
186 }
187 
188 static struct lut *
189 lex_s2ullp_lut_add(struct lut *root, const char *s,
190     const unsigned long long *ullp)
191 {
192 	return (lut_add(root, (void *)s, (void *)ullp, NULL));
193 }
194 
195 const unsigned long long *
196 lex_s2ullp_lut_lookup(struct lut *root, const char *s)
197 {
198 	return ((unsigned long long *)lut_lookup(root, (void *)s, NULL));
199 }
200 
201 /*
202  * lex_init -- initialize the lexer with appropriate filenames & debug flags
203  */
204 
205 /*ARGSUSED*/
206 void
207 lex_init(char **av, const char *cppargs, int lexecho)
208 {
209 	int i;
210 #ifdef	ESC
211 	const char *ptr;
212 #endif	/* ESC */
213 
214 	Lexecho = lexecho;
215 	Tokcount = stats_new_counter("lex.tokens", "total tokens in", 1);
216 	Filecount = stats_new_counter("lex.files", "total files read", 0);
217 	Lexelapse = stats_new_elapse("lex.time", "elapsed lex/parse time", 1);
218 
219 #ifdef	ESC
220 	Cppargs = cppargs;
221 
222 	/* allow user to tell us where cpp is if it is some weird place */
223 	if (ptr = getenv("_ESC_CPP"))
224 		Cpp = ptr;
225 
226 	/* and in case it takes some special stdargs */
227 	if (ptr = getenv("_ESC_CPP_STDARGS"))
228 		Cppstdargs = ptr;
229 
230 	/* verify we can find cpp */
231 	if (access(Cpp, X_OK) < 0) {
232 		Cpp = "/usr/lib/cpp";
233 		if (access(Cpp, X_OK) < 0)
234 			out(O_DIE, "can't locate cpp");
235 	}
236 #endif	/* ESC */
237 
238 	Files = av;
239 
240 	/* verify we can find all the input files */
241 	while (*av) {
242 		if (strlen(*av) >= MAXTOK - strlen(Cpp) - 3)
243 			out(O_DIE, "filename too long: %.100s...", *av);
244 		if (access(*av, R_OK) < 0)
245 			out(O_DIE|O_SYS, "%s", *av);
246 		av++;
247 		stats_counter_bump(Filecount);
248 	}
249 
250 	/* put reserved words into the string table & a lookup table */
251 	for (i = 0; i < sizeof (Rwords) / sizeof (*Rwords); i++)
252 		Rwordslut = lex_s2i_lut_add(Rwordslut,
253 		    stable(Rwords[i].word), Rwords[i].val);
254 
255 	/* initialize table of timeval suffixes */
256 	for (i = 0; i < sizeof (Timesuffix) / sizeof (*Timesuffix); i++) {
257 		Timesuffixlut = lex_s2ullp_lut_add(Timesuffixlut,
258 		    stable(Timesuffix[i].suffix), &Timesuffix[i].nsec);
259 	}
260 
261 	/* record start time */
262 	stats_elapse_start(Lexelapse);
263 }
264 
265 void
266 closefile(void)
267 {
268 	if (Fp != NULL) {
269 #ifdef	ESC
270 		if (pclose(Fp) > 0)
271 			out(O_DIE, "cpp errors while reading \"%s\", "
272 			    "bailing out.", Fileopened);
273 #else
274 		(void) fclose(Fp);
275 #endif	/* ESC */
276 	}
277 	Fp = NULL;
278 }
279 
280 /*
281  * yylex -- the lexer, called yylex() because that's what yacc wants
282  */
283 
284 int
285 yylex()
286 {
287 	int c;
288 	int nextc;
289 	char *ptr = Tok;
290 	char *eptr = &Tok[MAXTOK];
291 	const char *cptr;
292 	int startline;
293 	int val;
294 	static int bol = 1;	/* true if we're at beginning of line */
295 
296 	for (;;) {
297 		while (Fp == NULL) {
298 			if (*Files == NULL)
299 				return (record(EOF, NULL));
300 			Fileopened = stable(*Files++);
301 #ifdef	ESC
302 			sprintf(Tok, "%s %s %s %s",
303 			    Cpp, Cppstdargs, Cppargs, Fileopened);
304 			if ((Fp = popen(Tok, "r")) == NULL)
305 				out(O_DIE|O_SYS, "%s", Tok);
306 #else
307 			Fp = eftread_fopen(Fileopened);
308 #endif	/* ESC */
309 			Line = 1;
310 			bol = 1;
311 
312 			/* add name to stats for visibility */
313 			if (Fp != NULL) {
314 				static int fnum;
315 				char nbuf[100];
316 				struct filestats *nfs = MALLOC(sizeof (*nfs));
317 				(void) sprintf(nbuf, "lex.file%d", fnum++);
318 				nfs->stats = stats_new_string(nbuf, "", 0);
319 				stats_string_set(nfs->stats, Fileopened);
320 				nfs->next = Fstats;
321 				Fstats = nfs;
322 			}
323 		}
324 
325 		switch (c = getc(Fp)) {
326 		case '#':
327 			/* enforce that we're at beginning of line */
328 			if (!bol)
329 				return (record(c, NULL));
330 
331 			while ((c = getc(Fp)) != EOF &&
332 			    (c == ' ' || c == '\t'))
333 				;
334 			if (!isdigit(c)) {
335 				/*
336 				 * three cases here:
337 				 *	#pragma
338 				 *	#ident
339 				 *	#something-we-don't-understand
340 				 * anything we don't expect we just ignore.
341 				 */
342 				*ptr++ = c;
343 				while ((c = getc(Fp)) != EOF && isalnum(c))
344 					if (ptr < eptr - 1)
345 						*ptr++ = c;
346 				*ptr++ = '\0';
347 				if (strcmp(Tok, "pragma") == 0) {
348 					/* skip white space */
349 					while ((c = getc(Fp)) != EOF &&
350 					    (c == ' ' || c == '\t'))
351 						;
352 
353 					if (c == EOF || c == '\n')
354 						outfl(O_DIE, File, Line,
355 						    "bad #pragma");
356 
357 					/* pull in next token */
358 					ptr = Tok;
359 					*ptr++ = c;
360 					while ((c = getc(Fp)) != EOF &&
361 					    !isspace(c))
362 						if (ptr < eptr - 1)
363 							*ptr++ = c;
364 					*ptr++ = '\0';
365 					(void) ungetc(c, Fp);
366 
367 					dopragma(Tok);
368 				} else if (strcmp(Tok, "ident") == 0)
369 					doident();
370 			} else {
371 				/* handle file & line info from cpp */
372 				Line = 0;
373 				do {
374 					if (!isdigit(c))
375 						break;
376 					Line = Line * 10 + c - '0';
377 				} while ((c = getc(Fp)) != EOF);
378 				Line--;		/* newline will increment it */
379 				while (c != EOF && isspace(c))
380 					c = getc(Fp);
381 				if (c != '"')
382 					outfl(O_DIE, File, Line,
383 					    "bad # statement (file name)");
384 				while ((c = getc(Fp)) != EOF && c != '"')
385 					if (ptr < eptr - 1)
386 						*ptr++ = c;
387 				*ptr++ = '\0';
388 				if (c != '"')
389 					outfl(O_DIE, File, Line,
390 					    "bad # statement (quotes)");
391 				File = stable(Tok);
392 			}
393 			/* skip the rest of the cpp line */
394 			while ((c = getc(Fp)) != EOF && c != '\n' && c != '\r')
395 				;
396 			if (c == EOF)
397 				return (record(c, NULL));
398 			else
399 				(void) ungetc(c, Fp);
400 			ptr = Tok;
401 			break;
402 
403 		case EOF:
404 			closefile();
405 			continue;
406 
407 		case '\n':
408 			Line++;
409 			bol = 1;
410 			break;
411 
412 		case '\r':
413 		case ' ':
414 		case '\t':
415 			bol = 0;
416 			break;
417 
418 		case '/':
419 			bol = 0;
420 			/* comment handling */
421 			if ((nextc = getc(Fp)) == EOF)
422 				outfl(O_DIE, File, Line, "unexpected EOF");
423 			else if (nextc == '*') {
424 				startline = Line;
425 				while ((c = getc(Fp)) != EOF) {
426 					if (c == '\n')
427 						Line++;
428 					else if (c == '*' &&
429 						(((c = getc(Fp)) == EOF) ||
430 						(c == '/')))
431 						break;
432 				}
433 				if (c == EOF) {
434 					outfl(O_DIE, File, Line,
435 					    "end of comment not seen "
436 					    "(started on line %d)",
437 					    startline);
438 				}
439 			} else {
440 				/* wasn't a comment, return the '/' token */
441 				(void) ungetc(nextc, Fp);
442 				return (record(c, NULL));
443 			}
444 			break;
445 
446 		case '"': {
447 			int prevc;
448 
449 			bol = 0;
450 			prevc = '\0';
451 			/* quoted string handling */
452 			startline = Line;
453 			for (;;) {
454 				c = getc(Fp);
455 				if (c == EOF)
456 					outfl(O_DIE, File, Line,
457 					    "end of string not seen "
458 					    "(started on line %d)",
459 					    startline);
460 				else if (c == '\n')
461 					Line++;
462 				else if (c == '"' && prevc != '\\')
463 					break;
464 				else if (ptr < eptr)
465 					*ptr++ = c;
466 				prevc = c;
467 			}
468 			if (ptr >= eptr)
469 				out(O_DIE, File, Line, "string too long");
470 			*ptr++ = '\0';
471 			return (record(QUOTE, stable(Tok)));
472 		}
473 		case '&':
474 			bol = 0;
475 			/* && */
476 			if ((nextc = getc(Fp)) == '&')
477 				return (record(AND, NULL));
478 			else {
479 				(void) ungetc(nextc, Fp);
480 				return (record(c, NULL));
481 			}
482 			/*NOTREACHED*/
483 			break;
484 
485 		case '|':
486 			bol = 0;
487 			/* || */
488 			if ((nextc = getc(Fp)) == '|')
489 				return (record(OR, NULL));
490 			else {
491 				(void) ungetc(nextc, Fp);
492 				return (record(c, NULL));
493 			}
494 			/*NOTREACHED*/
495 			break;
496 
497 		case '!':
498 			bol = 0;
499 			/* ! or != */
500 			if ((nextc = getc(Fp)) == '=')
501 				return (record(NE, NULL));
502 			else {
503 				(void) ungetc(nextc, Fp);
504 				return (record(c, NULL));
505 			}
506 			/*NOTREACHED*/
507 			break;
508 
509 		case '=':
510 			bol = 0;
511 			/* == */
512 			if ((nextc = getc(Fp)) == '=')
513 				return (record(EQ, NULL));
514 			else {
515 				(void) ungetc(nextc, Fp);
516 				return (record(c, NULL));
517 			}
518 			/*NOTREACHED*/
519 			break;
520 
521 		case '-':
522 			bol = 0;
523 			/* -> */
524 			if ((nextc = getc(Fp)) == '>')
525 				return (record(ARROW, stable(Tok)));
526 			else {
527 				(void) ungetc(nextc, Fp);
528 				return (record(c, NULL));
529 			}
530 			/*NOTREACHED*/
531 			break;
532 
533 		case '<':
534 			bol = 0;
535 			if ((nextc = getc(Fp)) == '=')
536 				/* <= */
537 				return (record(LE, NULL));
538 			else if (nextc == '<')
539 				/* << */
540 				return (record(LSHIFT, NULL));
541 			else {
542 				(void) ungetc(nextc, Fp);
543 				return (record(c, NULL));
544 			}
545 			/*NOTREACHED*/
546 			break;
547 
548 		case '>':
549 			bol = 0;
550 			if ((nextc = getc(Fp)) == '=')
551 				/* >= */
552 				return (record(GE, NULL));
553 			else if (nextc == '>')
554 				/* >> */
555 				return (record(RSHIFT, NULL));
556 			else {
557 				(void) ungetc(nextc, Fp);
558 				return (record(c, NULL));
559 			}
560 			/*NOTREACHED*/
561 			break;
562 
563 		default:
564 			bol = 0;
565 			if (isdigit(c)) {
566 				int base;
567 
568 				/* collect rest of number */
569 				if (c == '0') {
570 					*ptr++ = c;
571 					if ((c = getc(Fp)) == EOF) {
572 						*ptr++ = '\0';
573 						return (record(NUMBER,
574 						    stable(Tok)));
575 					} else if (c == 'x' || c == 'X') {
576 						*ptr++ = c;
577 						base = 16;
578 					} else {
579 						(void) ungetc(c, Fp);
580 						base = 8;
581 					}
582 				} else {
583 					*ptr++ = c;
584 					base = 10;
585 				}
586 				while ((c = getc(Fp)) != EOF) {
587 					if (ptr >= eptr)
588 						out(O_DIE, File, Line,
589 						    "number too long");
590 
591 					switch (base) {
592 					case 16:
593 						if (c >= 'a' && c <= 'f' ||
594 						    c >= 'A' && c <= 'F') {
595 							*ptr++ = c;
596 							continue;
597 						}
598 						/*FALLTHRU*/
599 					case 10:
600 						if (c >= '8' && c <= '9') {
601 							*ptr++ = c;
602 							continue;
603 						}
604 						/*FALLTHRU*/
605 					case 8:
606 						if (c >= '0' && c <= '7') {
607 							*ptr++ = c;
608 							continue;
609 						}
610 						/* not valid for this base */
611 						*ptr++ = '\0';
612 						(void) ungetc(c, Fp);
613 						return (record(NUMBER,
614 						    stable(Tok)));
615 					}
616 				}
617 				*ptr++ = '\0';
618 				return (record(NUMBER, stable(Tok)));
619 			} else if (isalpha(c)) {
620 				/* collect identifier */
621 				*ptr++ = c;
622 				for (;;) {
623 					c = getc(Fp);
624 					if ((isalnum(c) || c == '_') &&
625 						ptr < eptr)
626 						*ptr++ = c;
627 					else {
628 						(void) ungetc(c, Fp);
629 						break;
630 					}
631 				}
632 				if (ptr >= eptr)
633 					out(O_DIE, File, Line,
634 					    "identifier too long");
635 				*ptr++ = '\0';
636 				cptr = stable(Tok);
637 				if (val = lex_s2i_lut_lookup(Rwordslut, cptr)) {
638 					return (record(val, cptr));
639 				}
640 				return (record(ID, cptr));
641 			} else
642 				return (record(c, NULL));
643 		}
644 		/*NOTREACHED*/
645 	}
646 }
647 
648 /*
649  * the record()/dumpline() routines are used to track & report
650  * the list of tokens seen on a given line.  this is used in two ways.
651  * first, syntax errors found by the parser are reported by us (via
652  * yyerror()) and we tack on the tokens processed so far on the current
653  * line to help indicate exactly where the error is.  second, if "lexecho"
654  * debugging is turned on, these routines provide it.
655  */
656 #define	MAXRECORD 1000
657 static int Recordedline;
658 static struct {
659 	int tok;
660 	const char *s;
661 } Recorded[MAXRECORD];
662 static int Recordnext;
663 
664 static int
665 record(int tok, const char *s)
666 {
667 	stats_counter_bump(Tokcount);
668 	if (Line != Recordedline) {
669 		/* starting new line, dump out the previous line */
670 		if (Lexecho && Recordedline) {
671 			outfl(O_NONL, File, Recordedline, "lex: ");
672 			dumpline(O_OK);
673 		}
674 		Recordedline = Line;
675 		Recordnext = 0;
676 	}
677 	if (Recordnext >= MAXRECORD)
678 		outfl(O_DIE, File, Line, "line too long, bailing out");
679 	Recorded[Recordnext].tok = tok;
680 	Recorded[Recordnext++].s = s;
681 
682 	yylval.tok.s = s;
683 	yylval.tok.file = File;
684 	yylval.tok.line = Line;
685 	return (tok);
686 }
687 
688 /*ARGSUSED*/
689 static void
690 dumpline(int flags)
691 {
692 	int i;
693 
694 	for (i = 0; i < Recordnext; i++)
695 		if (Recorded[i].s && Recorded[i].tok != ARROW)
696 			switch (Recorded[i].tok) {
697 			case T_QUOTE:
698 				out(flags|O_NONL, " \"%s\"",
699 					Recorded[i].s);
700 				break;
701 
702 			default:
703 				out(flags|O_NONL, " %s",
704 					Recorded[i].s);
705 				break;
706 			}
707 		else
708 			switch (Recorded[i].tok) {
709 			case EOF:
710 				out(flags|O_NONL, " EOF");
711 				break;
712 			case ARROW:
713 				out(flags|O_NONL, " ->%s",
714 					Recorded[i].s);
715 				break;
716 			case EQ:
717 				out(flags|O_NONL, " ==");
718 				break;
719 			case NE:
720 				out(flags|O_NONL, " !=");
721 				break;
722 			case OR:
723 				out(flags|O_NONL, " ||");
724 				break;
725 			case AND:
726 				out(flags|O_NONL, " &&");
727 				break;
728 			case LE:
729 				out(flags|O_NONL, " <=");
730 				break;
731 			case GE:
732 				out(flags|O_NONL, " >=");
733 				break;
734 			case LSHIFT:
735 				out(flags|O_NONL, " <<");
736 				break;
737 			case RSHIFT:
738 				out(flags|O_NONL, " >>");
739 				break;
740 			default:
741 				if (isprint(Recorded[i].tok))
742 					out(flags|O_NONL, " %c",
743 						Recorded[i].tok);
744 				else
745 					out(flags|O_NONL, " '\\%03o'",
746 						Recorded[i].tok);
747 				break;
748 			}
749 	out(flags, NULL);
750 }
751 
752 /*
753  * yyerror -- report a pareser error, called yyerror because yacc wants it
754  */
755 
756 void
757 yyerror(const char *s)
758 {
759 	Errcount++;
760 	outfl(O_ERR|O_NONL, File, Line, "%s, tokens: ", s);
761 	dumpline(O_ERR);
762 }
763 
764 /*
765  * doident -- handle "#pragma ident" directives
766  */
767 static void
768 doident()
769 {
770 	int c;
771 	char *ptr = Tok;
772 	char *eptr = &Tok[MAXTOK];
773 
774 	/* skip white space and quotes */
775 	while ((c = getc(Fp)) != EOF &&
776 	    (c == ' ' || c == '\t' || c == '"'))
777 		;
778 
779 	if (c == EOF || c == '\n')
780 		outfl(O_DIE, File, Line, "bad ident");
781 
782 	/* pull in next token */
783 	ptr = Tok;
784 	*ptr++ = c;
785 	while ((c = getc(Fp)) != EOF && c != '"' && c != '\n')
786 		if (ptr < eptr - 1)
787 			*ptr++ = c;
788 	*ptr++ = '\0';
789 	if (c != '\n') {
790 		/* skip to end of line (including close quote, if any) */
791 		while ((c = getc(Fp)) != EOF && c != '\n')
792 			;
793 	}
794 	(void) ungetc(c, Fp);
795 	Ident = lut_add(Ident, (void *)stable(Tok), (void *)0, NULL);
796 
797 	outfl(O_VERB, File, Line, "pragma set: ident \"%s\"", Tok);
798 }
799 
800 /*
801  * dodictionary -- handle "#pragma dictionary" directives
802  */
803 static void
804 dodictionary()
805 {
806 	int c;
807 	char *ptr = Tok;
808 	char *eptr = &Tok[MAXTOK];
809 
810 	/* skip white space and quotes */
811 	while ((c = getc(Fp)) != EOF &&
812 	    (c == ' ' || c == '\t' || c == '"'))
813 		;
814 
815 	if (c == EOF || c == '\n')
816 		outfl(O_DIE, File, Line, "bad dictionary");
817 
818 	/* pull in next token */
819 	ptr = Tok;
820 	*ptr++ = c;
821 	while ((c = getc(Fp)) != EOF && c != '"' && c != '\n')
822 		if (ptr < eptr - 1)
823 			*ptr++ = c;
824 	*ptr++ = '\0';
825 	if (c != '\n') {
826 		/* skip to end of line (including close quote, if any) */
827 		while ((c = getc(Fp)) != EOF && c != '\n')
828 			;
829 	}
830 	(void) ungetc(c, Fp);
831 	Dicts = lut_add(Dicts, (void *)stable(Tok), (void *)0, NULL);
832 
833 	outfl(O_VERB, File, Line, "pragma set: dictionary \"%s\"", Tok);
834 }
835 
836 /*
837  * doallow_cycles -- handle "#pragma allow_cycles" directives
838  */
839 static void
840 doallow_cycles()
841 {
842 	int c;
843 	char *ptr = Tok;
844 	char *eptr = &Tok[MAXTOK];
845 	unsigned long long newlevel;
846 
847 	/*
848 	 * by default the compiler does not allow cycles or loops
849 	 * in propagations.  when cycles are encountered, the
850 	 * compiler prints out an error message.
851 	 *
852 	 *   "#pragma allow_cycles" and
853 	 *   "#pragma allow_cycles 0"
854 	 * allow cycles, but any such cycle will produce a warning
855 	 * message.
856 	 *
857 	 *   "#pragma allow_cycles N"
858 	 * with N > 0 will allow cycles and not produce any
859 	 * warning messages.
860 	 */
861 
862 	/* skip white space and quotes */
863 	while ((c = getc(Fp)) != EOF &&
864 	    (c == ' ' || c == '\t' || c == '"'))
865 		;
866 
867 	if (c == EOF || c == '\n')
868 		newlevel = 0ULL;
869 	else {
870 
871 		/* pull in next token */
872 		ptr = Tok;
873 		*ptr++ = c;
874 		while ((c = getc(Fp)) != EOF && c != '"' && c != '\n')
875 			if (ptr < eptr - 1)
876 				*ptr++ = c;
877 		*ptr++ = '\0';
878 		if (c != '\n') {
879 			/* skip to end of line */
880 			while ((c = getc(Fp)) != EOF && c != '\n')
881 				;
882 		}
883 		newlevel = strtoll(Tok, NULL, 0);
884 	}
885 	(void) ungetc(c, Fp);
886 
887 	(void) check_cycle_level(newlevel);
888 	outfl(O_VERB, File, Line,
889 	    "pragma set: allow_cycles (%s)",
890 		    newlevel ? "no warnings" : "with warnings");
891 }
892 
893 /*
894  * dopragma -- handle #pragma directives
895  */
896 static void
897 dopragma(const char *tok)
898 {
899 	if (strcmp(tok, "ident") == 0)
900 		doident();
901 	else if (strcmp(tok, "dictionary") == 0)
902 		dodictionary();
903 	else if (strcmp(tok, "new_errors_only") == 0) {
904 		if (Pragma_new_errors_only++ == 0)
905 			outfl(O_VERB, File, Line,
906 			    "pragma set: new_errors_only");
907 	} else if (strcmp(tok, "trust_ereports") == 0) {
908 		if (Pragma_trust_ereports++ == 0)
909 			outfl(O_VERB, File, Line,
910 			    "pragma set: trust_ereports");
911 	} else if (strcmp(tok, "allow_cycles") == 0)
912 		doallow_cycles();
913 	else
914 		outfl(O_VERB, File, Line,
915 		    "unknown pragma ignored: \"%s\"", tok);
916 }
917 
918 /*
919  * lex_fini -- finalize the lexer
920  */
921 
922 int
923 lex_fini(void)
924 {
925 	stats_elapse_stop(Lexelapse);
926 	closefile();
927 	if (Lexecho) {
928 		outfl(O_OK, File, Line, "lex: ");
929 		dumpline(O_OK);
930 	}
931 	return (Errcount);
932 }
933 
934 void
935 lex_free(void)
936 {
937 	struct filestats *nfstats = Fstats;
938 
939 	/*
940 	 * Free up memory consumed by the lexer
941 	 */
942 	stats_delete(Tokcount);
943 	stats_delete(Filecount);
944 	stats_delete(Lexelapse);
945 	while (nfstats != NULL) {
946 		Fstats = nfstats->next;
947 		stats_delete(nfstats->stats);
948 		FREE(nfstats);
949 		nfstats = Fstats;
950 	}
951 	lut_free(Timesuffixlut, NULL, NULL);
952 	lut_free(Rwordslut, NULL, NULL);
953 	lut_free(Ident, NULL, NULL);
954 	lut_free(Dicts, NULL, NULL);
955 }
956