1 /****************************************************************************
2  * Copyright (c) 1998,1999,2000,2001 Free Software Foundation, Inc.         *
3  *                                                                          *
4  * Permission is hereby granted, free of charge, to any person obtaining a  *
5  * copy of this software and associated documentation files (the            *
6  * "Software"), to deal in the Software without restriction, including      *
7  * without limitation the rights to use, copy, modify, merge, publish,      *
8  * distribute, distribute with modifications, sublicense, and/or sell       *
9  * copies of the Software, and to permit persons to whom the Software is    *
10  * furnished to do so, subject to the following conditions:                 *
11  *                                                                          *
12  * The above copyright notice and this permission notice shall be included  *
13  * in all copies or substantial portions of the Software.                   *
14  *                                                                          *
15  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS  *
16  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF               *
17  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.   *
18  * IN NO EVENT SHALL THE ABOVE COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,   *
19  * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR    *
20  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR    *
21  * THE USE OR OTHER DEALINGS IN THE SOFTWARE.                               *
22  *                                                                          *
23  * Except as contained in this notice, the name(s) of the above copyright   *
24  * holders shall not be used in advertising or otherwise to promote the     *
25  * sale, use or other dealings in this Software without prior written       *
26  * authorization.                                                           *
27  ****************************************************************************/
28 
29 /****************************************************************************
30  *  Author: Zeyd M. Ben-Halim <zmbenhal@netcom.com> 1992,1995               *
31  *     and: Eric S. Raymond <esr@snark.thyrsus.com>                         *
32  ****************************************************************************/
33 
34 /*
35  *	comp_scan.c --- Lexical scanner for terminfo compiler.
36  *
37  *	_nc_reset_input()
38  *	_nc_get_token()
39  *	_nc_panic_mode()
40  *	int _nc_syntax;
41  *	int _nc_curr_line;
42  *	long _nc_curr_file_pos;
43  *	long _nc_comment_start;
44  *	long _nc_comment_end;
45  */
46 
47 #include <curses.priv.h>
48 
49 #include <ctype.h>
50 #include <term_entry.h>
51 #include <tic.h>
52 
53 MODULE_ID("$Id: comp_scan.c,v 1.59 2001/09/23 00:56:29 tom Exp $")
54 
55 /*
56  * Maximum length of string capability we'll accept before raising an error.
57  * Yes, there is a real capability in /etc/termcap this long, an "is".
58  */
59 #define MAXCAPLEN	600
60 
61 #define iswhite(ch)	(ch == ' '  ||  ch == '\t')
62 
63 NCURSES_EXPORT_VAR(int)
64 _nc_syntax = 0;			/* termcap or terminfo? */
65 NCURSES_EXPORT_VAR(long)
66 _nc_curr_file_pos = 0;		/* file offset of current line */
67 NCURSES_EXPORT_VAR(long)
68 _nc_comment_start = 0;		/* start of comment range before name */
69 NCURSES_EXPORT_VAR(long)
70 _nc_comment_end = 0;		/* end of comment range before name */
71 NCURSES_EXPORT_VAR(long)
72 _nc_start_line = 0;		/* start line of current entry */
73 
74 NCURSES_EXPORT_VAR(struct token)
75 _nc_curr_token =
76 {
77     0, 0, 0
78 };
79 
80 /*****************************************************************************
81  *
82  * Token-grabbing machinery
83  *
84  *****************************************************************************/
85 
86 static bool first_column;	/* See 'next_char()' below */
87 static char separator;		/* capability separator */
88 static int pushtype;		/* type of pushback token */
89 static char *pushname;
90 
91 #if NCURSES_EXT_FUNCS
92 NCURSES_EXPORT_VAR(bool)
93 _nc_disable_period = FALSE;	/* used by tic -a option */
94 #endif
95 
96 static int last_char(void);
97 static int next_char(void);
98 static long stream_pos(void);
99 static bool end_of_stream(void);
100 static void push_back(char c);
101 
102 /* Assume we may be looking at a termcap-style continuation */
103 static inline int
104 eat_escaped_newline(int ch)
105 {
106     if (ch == '\\')
107 	while ((ch = next_char()) == '\n' || iswhite(ch))
108 	    continue;
109     return ch;
110 }
111 
112 /*
113  *	int
114  *	get_token()
115  *
116  *	Scans the input for the next token, storing the specifics in the
117  *	global structure 'curr_token' and returning one of the following:
118  *
119  *		NAMES		A line beginning in column 1.  'name'
120  *				will be set to point to everything up to but
121  *				not including the first separator on the line.
122  *		BOOLEAN		An entry consisting of a name followed by
123  *				a separator.  'name' will be set to point to
124  *				the name of the capability.
125  *		NUMBER		An entry of the form
126  *					name#digits,
127  *				'name' will be set to point to the capability
128  *				name and 'valnumber' to the number given.
129  *		STRING		An entry of the form
130  *					name=characters,
131  *				'name' is set to the capability name and
132  *				'valstring' to the string of characters, with
133  *				input translations done.
134  *		CANCEL		An entry of the form
135  *					name@,
136  *				'name' is set to the capability name and
137  *				'valnumber' to -1.
138  *		EOF		The end of the file has been reached.
139  *
140  *	A `separator' is either a comma or a semicolon, depending on whether
141  *	we are in termcap or terminfo mode.
142  *
143  */
144 
145 NCURSES_EXPORT(int)
146 _nc_get_token(bool silent)
147 {
148     static const char terminfo_punct[] = "@%&*!#";
149     static char *buffer;
150 
151     char *numchk;
152     char *ptr;
153     char numbuf[80];
154     int ch;
155     int dot_flag = FALSE;
156     int type;
157     long number;
158     long token_start;
159     unsigned found;
160 
161     if (pushtype != NO_PUSHBACK) {
162 	int retval = pushtype;
163 
164 	_nc_set_type(pushname != 0 ? pushname : "");
165 	DEBUG(3, ("pushed-back token: `%s', class %d",
166 		  _nc_curr_token.tk_name, pushtype));
167 
168 	pushtype = NO_PUSHBACK;
169 	if (pushname != 0)
170 	    pushname[0] = '\0';
171 
172 	/* currtok wasn't altered by _nc_push_token() */
173 	return (retval);
174     }
175 
176     if (end_of_stream())
177 	return (EOF);
178 
179   start_token:
180     token_start = stream_pos();
181     while ((ch = next_char()) == '\n' || iswhite(ch))
182 	continue;
183 
184     ch = eat_escaped_newline(ch);
185 
186     if (ch == EOF)
187 	type = EOF;
188     else {
189 	/* if this is a termcap entry, skip a leading separator */
190 	if (separator == ':' && ch == ':')
191 	    ch = next_char();
192 
193 	if (ch == '.'
194 #if NCURSES_EXT_FUNCS
195 	    && !_nc_disable_period
196 #endif
197 	    ) {
198 	    dot_flag = TRUE;
199 	    DEBUG(8, ("dot-flag set"));
200 
201 	    while ((ch = next_char()) == '.' || iswhite(ch))
202 		continue;
203 	}
204 
205 	if (ch == EOF) {
206 	    type = EOF;
207 	    goto end_of_token;
208 	}
209 
210 	/* have to make some punctuation chars legal for terminfo */
211 	if (!isalnum(ch)
212 #if NCURSES_EXT_FUNCS
213 	    && !(ch == '.' && _nc_disable_period)
214 #endif
215 	    && !strchr(terminfo_punct, (char) ch)) {
216 	    if (!silent)
217 		_nc_warning("Illegal character (expected alphanumeric or %s) - %s",
218 			    terminfo_punct, unctrl((chtype) ch));
219 	    _nc_panic_mode(separator);
220 	    goto start_token;
221 	}
222 
223 	if (buffer == 0)
224 	    buffer = _nc_doalloc(buffer, MAX_ENTRY_SIZE);
225 
226 	ptr = buffer;
227 	*(ptr++) = ch;
228 
229 	if (first_column) {
230 	    char *desc;
231 
232 	    _nc_comment_start = token_start;
233 	    _nc_comment_end = _nc_curr_file_pos;
234 	    _nc_start_line = _nc_curr_line;
235 
236 	    _nc_syntax = ERR;
237 	    while ((ch = next_char()) != '\n') {
238 		if (ch == EOF)
239 		    _nc_err_abort("premature EOF");
240 		else if (ch == ':' && last_char() != ',') {
241 		    _nc_syntax = SYN_TERMCAP;
242 		    separator = ':';
243 		    break;
244 		} else if (ch == ',') {
245 		    _nc_syntax = SYN_TERMINFO;
246 		    separator = ',';
247 		    /*
248 		     * Fall-through here is not an accident.  The idea is that
249 		     * if we see a comma, we figure this is terminfo unless we
250 		     * subsequently run into a colon -- but we don't stop
251 		     * looking for that colon until hitting a newline.  This
252 		     * allows commas to be embedded in description fields of
253 		     * either syntax.
254 		     */
255 		    /* FALLTHRU */
256 		} else
257 		    ch = eat_escaped_newline(ch);
258 
259 		*ptr++ = ch;
260 	    }
261 	    ptr[0] = '\0';
262 	    if (_nc_syntax == ERR) {
263 		/*
264 		 * Grrr...what we ought to do here is barf, complaining that
265 		 * the entry is malformed.  But because a couple of name fields
266 		 * in the 8.2 termcap file end with |\, we just have to assume
267 		 * it's termcap syntax.
268 		 */
269 		_nc_syntax = SYN_TERMCAP;
270 		separator = ':';
271 	    } else if (_nc_syntax == SYN_TERMINFO) {
272 		/* throw away trailing /, *$/ */
273 		for (--ptr; iswhite(*ptr) || *ptr == ','; ptr--)
274 		    continue;
275 		ptr[1] = '\0';
276 	    }
277 
278 	    /*
279 	     * This is the soonest we have the terminal name fetched.  Set up
280 	     * for following warning messages.
281 	     */
282 	    ptr = strchr(buffer, '|');
283 	    if (ptr == (char *) NULL)
284 		ptr = buffer + strlen(buffer);
285 	    ch = *ptr;
286 	    *ptr = '\0';
287 	    _nc_set_type(buffer);
288 	    *ptr = ch;
289 
290 	    /*
291 	     * Compute the boundary between the aliases and the description
292 	     * field for syntax-checking purposes.
293 	     */
294 	    desc = strrchr(buffer, '|');
295 	    if (!silent && desc) {
296 		if (*desc == '\0')
297 		    _nc_warning("empty longname field");
298 		else if (strchr(desc, ' ') == (char *) NULL)
299 		    _nc_warning("older tic versions may treat the description field as an alias");
300 	    }
301 	    if (!desc)
302 		desc = buffer + strlen(buffer);
303 
304 	    /*
305 	     * Whitespace in a name field other than the long name can confuse
306 	     * rdist and some termcap tools.  Slashes are a no-no.  Other
307 	     * special characters can be dangerous due to shell expansion.
308 	     */
309 	    for (ptr = buffer; ptr < desc; ptr++) {
310 		if (isspace(UChar(*ptr))) {
311 		    if (!silent)
312 			_nc_warning("whitespace in name or alias field");
313 		    break;
314 		} else if (*ptr == '/') {
315 		    if (!silent)
316 			_nc_warning("slashes aren't allowed in names or aliases");
317 		    break;
318 		} else if (strchr("$[]!*?", *ptr)) {
319 		    if (!silent)
320 			_nc_warning("dubious character `%c' in name or alias field", *ptr);
321 		    break;
322 		}
323 	    }
324 
325 	    ptr = buffer;
326 
327 	    _nc_curr_token.tk_name = buffer;
328 	    type = NAMES;
329 	} else {
330 	    while ((ch = next_char()) != EOF) {
331 		if (!isalnum(ch)) {
332 		    if (_nc_syntax == SYN_TERMINFO) {
333 			if (ch != '_')
334 			    break;
335 		    } else {	/* allow ';' for "k;" */
336 			if (ch != ';')
337 			    break;
338 		    }
339 		}
340 		*(ptr++) = ch;
341 	    }
342 
343 	    *ptr++ = '\0';
344 	    switch (ch) {
345 	    case ',':
346 	    case ':':
347 		if (ch != separator)
348 		    _nc_err_abort("Separator inconsistent with syntax");
349 		_nc_curr_token.tk_name = buffer;
350 		type = BOOLEAN;
351 		break;
352 	    case '@':
353 		if ((ch = next_char()) != separator && !silent)
354 		    _nc_warning("Missing separator after `%s', have %s",
355 				buffer, unctrl((chtype) ch));
356 		_nc_curr_token.tk_name = buffer;
357 		type = CANCEL;
358 		break;
359 
360 	    case '#':
361 		found = 0;
362 		while (isalnum(ch = next_char())) {
363 		    numbuf[found++] = ch;
364 		    if (found >= sizeof(numbuf) - 1)
365 			break;
366 		}
367 		numbuf[found] = '\0';
368 		number = strtol(numbuf, &numchk, 0);
369 		if (!silent) {
370 		    if (numchk == numbuf)
371 			_nc_warning("no value given for `%s'", buffer);
372 		    if ((*numchk != '\0') || (ch != separator))
373 			_nc_warning("Missing separator");
374 		}
375 		_nc_curr_token.tk_name = buffer;
376 		_nc_curr_token.tk_valnumber = number;
377 		type = NUMBER;
378 		break;
379 
380 	    case '=':
381 		ch = _nc_trans_string(ptr, buffer + MAX_ENTRY_SIZE);
382 		if (!silent && ch != separator)
383 		    _nc_warning("Missing separator");
384 		_nc_curr_token.tk_name = buffer;
385 		_nc_curr_token.tk_valstring = ptr;
386 		type = STRING;
387 		break;
388 
389 	    case EOF:
390 		type = EOF;
391 		break;
392 	    default:
393 		/* just to get rid of the compiler warning */
394 		type = UNDEF;
395 		if (!silent)
396 		    _nc_warning("Illegal character - %s", unctrl((chtype) ch));
397 	    }
398 	}			/* end else (first_column == FALSE) */
399     }				/* end else (ch != EOF) */
400 
401   end_of_token:
402 
403 #ifdef TRACE
404     if (dot_flag == TRUE)
405 	DEBUG(8, ("Commented out "));
406 
407     if (_nc_tracing >= DEBUG_LEVEL(7)) {
408 	switch (type) {
409 	case BOOLEAN:
410 	    _tracef("Token: Boolean; name='%s'",
411 		    _nc_curr_token.tk_name);
412 	    break;
413 
414 	case NUMBER:
415 	    _tracef("Token: Number;  name='%s', value=%d",
416 		    _nc_curr_token.tk_name,
417 		    _nc_curr_token.tk_valnumber);
418 	    break;
419 
420 	case STRING:
421 	    _tracef("Token: String;  name='%s', value=%s",
422 		    _nc_curr_token.tk_name,
423 		    _nc_visbuf(_nc_curr_token.tk_valstring));
424 	    break;
425 
426 	case CANCEL:
427 	    _tracef("Token: Cancel; name='%s'",
428 		    _nc_curr_token.tk_name);
429 	    break;
430 
431 	case NAMES:
432 
433 	    _tracef("Token: Names; value='%s'",
434 		    _nc_curr_token.tk_name);
435 	    break;
436 
437 	case EOF:
438 	    _tracef("Token: End of file");
439 	    break;
440 
441 	default:
442 	    _nc_warning("Bad token type");
443 	}
444     }
445 #endif
446 
447     if (dot_flag == TRUE)	/* if commented out, use the next one */
448 	type = _nc_get_token(silent);
449 
450     DEBUG(3, ("token: `%s', class %d",
451 	      _nc_curr_token.tk_name != 0 ? _nc_curr_token.tk_name :
452 	      "<null>",
453 	      type));
454 
455     return (type);
456 }
457 
458 /*
459  *	char
460  *	trans_string(ptr)
461  *
462  *	Reads characters using next_char() until encountering a separator, nl,
463  *	or end-of-file.  The returned value is the character which caused
464  *	reading to stop.  The following translations are done on the input:
465  *
466  *		^X  goes to  ctrl-X (i.e. X & 037)
467  *		{\E,\n,\r,\b,\t,\f}  go to
468  *			{ESCAPE,newline,carriage-return,backspace,tab,formfeed}
469  *		{\^,\\}  go to  {carat,backslash}
470  *		\ddd (for ddd = up to three octal digits)  goes to the character ddd
471  *
472  *		\e == \E
473  *		\0 == \200
474  *
475  */
476 
477 NCURSES_EXPORT(char)
478 _nc_trans_string(char *ptr, char *last)
479 {
480     int count = 0;
481     int number = 0;
482     int i, c;
483     chtype ch, last_ch = '\0';
484     bool ignored = FALSE;
485     bool long_warning = FALSE;
486 
487     while ((ch = c = next_char()) != (chtype) separator && c != EOF) {
488 	if (ptr == (last - 1))
489 	    break;
490 	if ((_nc_syntax == SYN_TERMCAP) && c == '\n')
491 	    break;
492 	if (ch == '^' && last_ch != '%') {
493 	    ch = c = next_char();
494 	    if (c == EOF)
495 		_nc_err_abort("Premature EOF");
496 
497 	    if (!(is7bits(ch) && isprint(ch))) {
498 		_nc_warning("Illegal ^ character - %s", unctrl(ch));
499 	    }
500 	    if (ch == '?') {
501 		*(ptr++) = '\177';
502 		if (_nc_tracing)
503 		    _nc_warning("Allow ^? as synonym for \\177");
504 	    } else {
505 		if ((ch &= 037) == 0)
506 		    ch = 128;
507 		*(ptr++) = (char) (ch);
508 	    }
509 	} else if (ch == '\\') {
510 	    ch = c = next_char();
511 	    if (c == EOF)
512 		_nc_err_abort("Premature EOF");
513 
514 	    if (ch >= '0' && ch <= '7') {
515 		number = ch - '0';
516 		for (i = 0; i < 2; i++) {
517 		    ch = c = next_char();
518 		    if (c == EOF)
519 			_nc_err_abort("Premature EOF");
520 
521 		    if (c < '0' || c > '7') {
522 			if (isdigit(c)) {
523 			    _nc_warning("Non-octal digit `%c' in \\ sequence", c);
524 			    /* allow the digit; it'll do less harm */
525 			} else {
526 			    push_back((char) c);
527 			    break;
528 			}
529 		    }
530 
531 		    number = number * 8 + c - '0';
532 		}
533 
534 		if (number == 0)
535 		    number = 0200;
536 		*(ptr++) = (char) number;
537 	    } else {
538 		switch (c) {
539 		case 'E':
540 		case 'e':
541 		    *(ptr++) = '\033';
542 		    break;
543 
544 		case 'a':
545 		    *(ptr++) = '\007';
546 		    break;
547 
548 		case 'l':
549 		case 'n':
550 		    *(ptr++) = '\n';
551 		    break;
552 
553 		case 'r':
554 		    *(ptr++) = '\r';
555 		    break;
556 
557 		case 'b':
558 		    *(ptr++) = '\010';
559 		    break;
560 
561 		case 's':
562 		    *(ptr++) = ' ';
563 		    break;
564 
565 		case 'f':
566 		    *(ptr++) = '\014';
567 		    break;
568 
569 		case 't':
570 		    *(ptr++) = '\t';
571 		    break;
572 
573 		case '\\':
574 		    *(ptr++) = '\\';
575 		    break;
576 
577 		case '^':
578 		    *(ptr++) = '^';
579 		    break;
580 
581 		case ',':
582 		    *(ptr++) = ',';
583 		    break;
584 
585 		case ':':
586 		    *(ptr++) = ':';
587 		    break;
588 
589 		case '\n':
590 		    continue;
591 
592 		default:
593 		    _nc_warning("Illegal character %s in \\ sequence",
594 				unctrl(ch));
595 		    *(ptr++) = (char) ch;
596 		}		/* endswitch (ch) */
597 	    }			/* endelse (ch < '0' ||  ch > '7') */
598 	}
599 	/* end else if (ch == '\\') */
600 	else if (ch == '\n' && (_nc_syntax == SYN_TERMINFO)) {
601 	    /* newlines embedded in a terminfo string are ignored */
602 	    ignored = TRUE;
603 	} else {
604 	    *(ptr++) = (char) ch;
605 	}
606 
607 	if (!ignored) {
608 	    last_ch = ch;
609 	    count++;
610 	}
611 	ignored = FALSE;
612 
613 	if (count > MAXCAPLEN && !long_warning) {
614 	    _nc_warning("Very long string found.  Missing separator?");
615 	    long_warning = TRUE;
616 	}
617     }				/* end while */
618 
619     *ptr = '\0';
620 
621     return (ch);
622 }
623 
624 /*
625  *	_nc_push_token()
626  *
627  *	Push a token of given type so that it will be reread by the next
628  *	get_token() call.
629  */
630 
631 NCURSES_EXPORT(void)
632 _nc_push_token(int tokclass)
633 {
634     /*
635      * This implementation is kind of bogus, it will fail if we ever do more
636      * than one pushback at a time between get_token() calls.  It relies on the
637      * fact that _nc_curr_token is static storage that nothing but
638      * _nc_get_token() touches.
639      */
640     pushtype = tokclass;
641     if (pushname == 0)
642 	pushname = _nc_doalloc(pushname, MAX_NAME_SIZE + 1);
643     _nc_get_type(pushname);
644 
645     DEBUG(3, ("pushing token: `%s', class %d",
646 	      _nc_curr_token.tk_name, pushtype));
647 }
648 
649 /*
650  * Panic mode error recovery - skip everything until a "ch" is found.
651  */
652 NCURSES_EXPORT(void)
653 _nc_panic_mode(char ch)
654 {
655     int c;
656 
657     for (;;) {
658 	c = next_char();
659 	if (c == ch)
660 	    return;
661 	if (c == EOF)
662 	    return;
663     }
664 }
665 
666 /*****************************************************************************
667  *
668  * Character-stream handling
669  *
670  *****************************************************************************/
671 
672 #define LEXBUFSIZ	1024
673 
674 static char *bufptr;		/* otherwise, the input buffer pointer */
675 static char *bufstart;		/* start of buffer so we can compute offsets */
676 static FILE *yyin;		/* scanner's input file descriptor */
677 
678 /*
679  *	_nc_reset_input()
680  *
681  *	Resets the input-reading routines.  Used on initialization,
682  *	or after a seek has been done.  Exactly one argument must be
683  *	non-null.
684  */
685 
686 NCURSES_EXPORT(void)
687 _nc_reset_input(FILE * fp, char *buf)
688 {
689     pushtype = NO_PUSHBACK;
690     if (pushname != 0)
691 	pushname[0] = '\0';
692     yyin = fp;
693     bufstart = bufptr = buf;
694     _nc_curr_file_pos = 0L;
695     if (fp != 0)
696 	_nc_curr_line = 0;
697     _nc_curr_col = 0;
698 }
699 
700 /*
701  *	int last_char()
702  *
703  *	Returns the final nonblank character on the current input buffer
704  */
705 static int
706 last_char(void)
707 {
708     size_t len = strlen(bufptr);
709     while (len--) {
710 	if (!isspace(UChar(bufptr[len])))
711 	    return bufptr[len];
712     }
713     return 0;
714 }
715 
716 /*
717  *	int next_char()
718  *
719  *	Returns the next character in the input stream.  Comments and leading
720  *	white space are stripped.
721  *
722  *	The global state variable 'firstcolumn' is set TRUE if the character
723  *	returned is from the first column of the input line.
724  *
725  *	The global variable _nc_curr_line is incremented for each new line.
726  *	The global variable _nc_curr_file_pos is set to the file offset of the
727  *	beginning of each line.
728  */
729 
730 static int
731 next_char(void)
732 {
733     if (!yyin) {
734 	/*
735 	 * An string with an embedded null will truncate the input.  This is
736 	 * intentional (we don't read binary files here).
737 	 */
738 	if (*bufptr == '\0')
739 	    return (EOF);
740 	if (*bufptr == '\n') {
741 	    _nc_curr_line++;
742 	    _nc_curr_col = 0;
743 	}
744     } else if (!bufptr || !*bufptr) {
745 	/*
746 	 * In theory this could be recoded to do its I/O one character at a
747 	 * time, saving the buffer space.  In practice, this turns out to be
748 	 * quite hard to get completely right.  Try it and see.  If you
749 	 * succeed, don't forget to hack push_back() correspondingly.
750 	 */
751 	static char *result;
752 	static size_t allocated;
753 	size_t used;
754 	size_t len;
755 
756 	do {
757 	    bufstart = 0;
758 	    used = 0;
759 	    do {
760 		if (used + (LEXBUFSIZ / 4) >= allocated) {
761 		    allocated += (allocated + LEXBUFSIZ);
762 		    result = _nc_doalloc(result, allocated);
763 		    if (result == 0)
764 			return (EOF);
765 		}
766 		if (used == 0)
767 		    _nc_curr_file_pos = ftell(yyin);
768 
769 		if (fgets(result + used, allocated - used, yyin) != NULL) {
770 		    bufstart = result;
771 		    if (used == 0) {
772 			_nc_curr_line++;
773 			_nc_curr_col = 0;
774 		    }
775 		} else {
776 		    if (used != 0)
777 			strcat(result, "\n");
778 		}
779 		if ((bufptr = bufstart) != 0) {
780 		    used = strlen(bufptr);
781 		    while (iswhite(*bufptr))
782 			bufptr++;
783 
784 		    /*
785 		     * Treat a trailing <cr><lf> the same as a <newline> so we
786 		     * can read files on OS/2, etc.
787 		     */
788 		    if ((len = strlen(bufptr)) > 1) {
789 			if (bufptr[len - 1] == '\n'
790 			    && bufptr[len - 2] == '\r') {
791 			    len--;
792 			    bufptr[len - 1] = '\n';
793 			    bufptr[len] = '\0';
794 			}
795 		    }
796 		} else {
797 		    return (EOF);
798 		}
799 	    } while (bufptr[len - 1] != '\n');	/* complete a line */
800 	} while (result[0] == '#');	/* ignore comments */
801     }
802 
803     first_column = (bufptr == bufstart);
804 
805     _nc_curr_col++;
806     return (*bufptr++);
807 }
808 
809 static void
810 push_back(char c)
811 /* push a character back onto the input stream */
812 {
813     if (bufptr == bufstart)
814 	_nc_syserr_abort("Can't backspace off beginning of line");
815     *--bufptr = c;
816 }
817 
818 static long
819 stream_pos(void)
820 /* return our current character position in the input stream */
821 {
822     return (yyin ? ftell(yyin) : (bufptr ? bufptr - bufstart : 0));
823 }
824 
825 static bool
826 end_of_stream(void)
827 /* are we at end of input? */
828 {
829     return ((yyin ? feof(yyin) : (bufptr && *bufptr == '\0'))
830 	    ? TRUE : FALSE);
831 }
832