xref: /freebsd/contrib/ncurses/ncurses/tinfo/parse_entry.c (revision 68ad2b0d7af2a3571c4abac9afa712f9b09b721c)
1 /****************************************************************************
2  * Copyright 2018-2024,2025 Thomas E. Dickey                                *
3  * Copyright 1998-2016,2017 Free Software Foundation, Inc.                  *
4  *                                                                          *
5  * Permission is hereby granted, free of charge, to any person obtaining a  *
6  * copy of this software and associated documentation files (the            *
7  * "Software"), to deal in the Software without restriction, including      *
8  * without limitation the rights to use, copy, modify, merge, publish,      *
9  * distribute, distribute with modifications, sublicense, and/or sell       *
10  * copies of the Software, and to permit persons to whom the Software is    *
11  * furnished to do so, subject to the following conditions:                 *
12  *                                                                          *
13  * The above copyright notice and this permission notice shall be included  *
14  * in all copies or substantial portions of the Software.                   *
15  *                                                                          *
16  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS  *
17  * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF               *
18  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.   *
19  * IN NO EVENT SHALL THE ABOVE COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM,   *
20  * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR    *
21  * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR    *
22  * THE USE OR OTHER DEALINGS IN THE SOFTWARE.                               *
23  *                                                                          *
24  * Except as contained in this notice, the name(s) of the above copyright   *
25  * holders shall not be used in advertising or otherwise to promote the     *
26  * sale, use or other dealings in this Software without prior written       *
27  * authorization.                                                           *
28  ****************************************************************************/
29 
30 /****************************************************************************
31  *  Author: Zeyd M. Ben-Halim <zmbenhal@netcom.com> 1992,1995               *
32  *     and: Eric S. Raymond <esr@snark.thyrsus.com>                         *
33  *     and: Thomas E. Dickey                        1996-on                 *
34  ****************************************************************************/
35 
36 /*
37  *	parse_entry.c -- compile one terminfo or termcap entry
38  *
39  *	Get an exact in-core representation of an entry.  Don't
40  *	try to resolve use or tc capabilities, that is someone
41  *	else's job.  Depends on the lexical analyzer to get tokens
42  *	from the input stream.
43  */
44 
45 #define __INTERNAL_CAPS_VISIBLE
46 #include <curses.priv.h>
47 
48 #include <ctype.h>
49 #include <tic.h>
50 
51 MODULE_ID("$Id: parse_entry.c,v 1.117 2025/11/23 20:25:15 tom Exp $")
52 
53 #ifdef LINT
54 static short const parametrized[] =
55 {0};
56 #else
57 #include <parametrized.h>
58 #endif
59 
60 static void postprocess_termcap(TERMTYPE2 *, bool);
61 static void postprocess_terminfo(TERMTYPE2 *);
62 static struct name_table_entry const *lookup_fullname(const char *name);
63 
64 #if NCURSES_XNAMES
65 
66 static struct name_table_entry const *
_nc_extend_names(ENTRY * entryp,const char * name,int token_type)67 _nc_extend_names(ENTRY * entryp, const char *name, int token_type)
68 {
69     static struct name_table_entry temp;
70     TERMTYPE2 *tp = &(entryp->tterm);
71     unsigned offset = 0;
72     unsigned actual;
73     unsigned tindex;
74     unsigned first, last, n;
75     bool found;
76 
77     switch (token_type) {
78     case BOOLEAN:
79 	first = 0;
80 	last = tp->ext_Booleans;
81 	offset = tp->ext_Booleans;
82 	tindex = tp->num_Booleans;
83 	break;
84     case NUMBER:
85 	first = tp->ext_Booleans;
86 	last = tp->ext_Numbers + first;
87 	offset = (unsigned) (tp->ext_Booleans + tp->ext_Numbers);
88 	tindex = tp->num_Numbers;
89 	break;
90     case STRING:
91 	first = (unsigned) (tp->ext_Booleans + tp->ext_Numbers);
92 	last = tp->ext_Strings + first;
93 	offset = (unsigned) (tp->ext_Booleans + tp->ext_Numbers + tp->ext_Strings);
94 	tindex = tp->num_Strings;
95 	break;
96     case CANCEL:
97 	actual = NUM_EXT_NAMES(tp);
98 	for (n = 0; n < actual; n++) {
99 	    if (!strcmp(name, tp->ext_Names[n])) {
100 		if (n > (unsigned) (tp->ext_Booleans + tp->ext_Numbers)) {
101 		    token_type = STRING;
102 		} else if (n > tp->ext_Booleans) {
103 		    token_type = NUMBER;
104 		} else {
105 		    token_type = BOOLEAN;
106 		}
107 		return _nc_extend_names(entryp, name, token_type);
108 	    }
109 	}
110 	/* Well, we are given a cancel for a name that we don't recognize */
111 	return _nc_extend_names(entryp, name, STRING);
112     default:
113 	return NULL;
114     }
115 
116     /* Adjust the 'offset' (insertion-point) to keep the lists of extended
117      * names sorted.
118      */
119     for (n = first, found = FALSE; n < last; n++) {
120 	int cmp = strcmp(tp->ext_Names[n], name);
121 	if (cmp == 0)
122 	    found = TRUE;
123 	if (cmp >= 0) {
124 	    offset = n;
125 	    tindex = n - first;
126 	    switch (token_type) {
127 	    case BOOLEAN:
128 		tindex += BOOLCOUNT;
129 		break;
130 	    case NUMBER:
131 		tindex += NUMCOUNT;
132 		break;
133 	    case STRING:
134 		tindex += STRCOUNT;
135 		break;
136 	    }
137 	    break;
138 	}
139     }
140 
141 #define for_each_value(max) \
142 	for (last = (unsigned) (max - 1); last > tindex; last--)
143 
144     if (!found) {
145 	char *saved;
146 
147 	if ((saved = _nc_save_str(name)) == NULL)
148 	    return NULL;
149 
150 	switch (token_type) {
151 	case BOOLEAN:
152 	    tp->ext_Booleans++;
153 	    tp->num_Booleans++;
154 	    TYPE_REALLOC(NCURSES_SBOOL, tp->num_Booleans, tp->Booleans);
155 	    for_each_value(tp->num_Booleans)
156 		tp->Booleans[last] = tp->Booleans[last - 1];
157 	    break;
158 	case NUMBER:
159 	    tp->ext_Numbers++;
160 	    tp->num_Numbers++;
161 	    TYPE_REALLOC(NCURSES_INT2, tp->num_Numbers, tp->Numbers);
162 	    for_each_value(tp->num_Numbers)
163 		tp->Numbers[last] = tp->Numbers[last - 1];
164 	    break;
165 	case STRING:
166 	    tp->ext_Strings++;
167 	    tp->num_Strings++;
168 	    TYPE_REALLOC(char *, tp->num_Strings, tp->Strings);
169 	    for_each_value(tp->num_Strings)
170 		tp->Strings[last] = tp->Strings[last - 1];
171 	    break;
172 	}
173 	actual = NUM_EXT_NAMES(tp);
174 	TYPE_REALLOC(char *, actual, tp->ext_Names);
175 	while (--actual > offset)
176 	    tp->ext_Names[actual] = tp->ext_Names[actual - 1];
177 	tp->ext_Names[offset] = saved;
178     }
179 
180     temp.nte_name = tp->ext_Names[offset];
181     temp.nte_type = token_type;
182     temp.nte_index = (short) tindex;
183     temp.nte_link = -1;
184 
185     return &temp;
186 }
187 
188 static const char *
usertype2s(int mask)189 usertype2s(int mask)
190 {
191     const char *result = "unknown";
192     if (mask & (1 << BOOLEAN)) {
193 	result = "boolean";
194     } else if (mask & (1 << NUMBER)) {
195 	result = "number";
196     } else if (mask & (1 << STRING)) {
197 	result = "string";
198     }
199     return result;
200 }
201 
202 static bool
expected_type(const char * name,int token_type,bool silent)203 expected_type(const char *name, int token_type, bool silent)
204 {
205     struct user_table_entry const *entry = _nc_find_user_entry(name);
206     bool result = TRUE;
207     if ((entry != NULL) && (token_type != CANCEL)) {
208 	int have_type = (1 << token_type);
209 	if (!(entry->ute_type & have_type)) {
210 	    if (!silent)
211 		_nc_warning("expected %s-type for %s, have %s",
212 			    usertype2s(entry->ute_type),
213 			    name,
214 			    usertype2s(have_type));
215 	    result = FALSE;
216 	}
217     }
218     return result;
219 }
220 #endif /* NCURSES_XNAMES */
221 
222 /*
223  * A valid entry name uses characters from the "portable character set"
224  * (more commonly referred to as US-ASCII), and disallows some of the
225  * punctuation characters:
226  *
227  * '/' is a pathname separator
228  * '\' may be a pathname separator, but more important, is an escape
229  * '|' delimits names and description
230  * '#' denotes a numeric value
231  * '=' denotes a string value
232  * '@' denotes a cancelled symbol
233  * ',' separates terminfo capabilities
234  * ':' separates termcap capabilities
235  *
236  * Termcap capability names may begin with a '#' or '@' (since they have
237  * exactly two characters).
238  */
239 static bool
valid_entryname(const char * name)240 valid_entryname(const char *name)
241 {
242     bool result = TRUE;
243     bool first = TRUE;
244     int ch;
245     while ((ch = UChar(*name++)) != '\0') {
246 	if (ch <= ' ' || ch > '~' || strchr("/\\|=,:", ch) != NULL) {
247 	    result = FALSE;
248 	    break;
249 	}
250 	if (!first && strchr("#@", ch) != NULL) {
251 	    result = FALSE;
252 	    break;
253 	}
254 	first = FALSE;
255     }
256     return result;
257 }
258 
259 /*
260  *	int
261  *	_nc_parse_entry(entry, literal, silent)
262  *
263  *	Compile one entry.  Doesn't try to resolve use or tc capabilities.
264  *
265  *	found-forward-use = FALSE
266  *	re-initialise internal arrays
267  *	get_token();
268  *	if the token was not a name in column 1, complain and die
269  *	save names in entry's string table
270  *	while (get_token() is not EOF and not NAMES)
271  *	        check for existence and type-correctness
272  *	        enter cap into structure
273  *	        if STRING
274  *	            save string in entry's string table
275  *	push back token
276  */
277 
278 #define BAD_TC_USAGE if (!bad_tc_usage) \
279  	{ bad_tc_usage = TRUE; \
280 	 _nc_warning("Legacy termcap allows only a trailing tc= clause"); }
281 
282 #define MAX_NUMBER MAX_OF_TYPE(NCURSES_INT2)
283 
284 NCURSES_EXPORT(int)
_nc_parse_entry(ENTRY * entryp,int literal,bool silent)285 _nc_parse_entry(ENTRY * entryp, int literal, bool silent)
286 {
287     int token_type;
288     struct name_table_entry const *entry_ptr;
289     char *ptr, *base;
290     const char *name;
291     bool bad_tc_usage = FALSE;
292 
293     TR(TRACE_DATABASE,
294        (T_CALLED("_nc_parse_entry(entry=%p, literal=%d, silent=%d)"),
295 	(void *) entryp, literal, silent));
296 
297     token_type = _nc_get_token(silent);
298 
299     if (token_type == EOF)
300 	returnDB(EOF);
301     if (token_type != NAMES)
302 	_nc_err_abort("Entry does not start with terminal names in column one");
303 
304     _nc_init_entry(entryp);
305 
306     entryp->cstart = _nc_comment_start;
307     entryp->cend = _nc_comment_end;
308     entryp->startline = _nc_start_line;
309     DEBUG(2, ("Comment range is %ld to %ld", entryp->cstart, entryp->cend));
310 
311     /*
312      * Strip off the 2-character termcap name, if present.  Originally termcap
313      * used that as an indexing aid.  We can retain 2-character terminfo names,
314      * but note that they would be lost if we translate to/from termcap.  This
315      * feature is supposedly obsolete since "newer" BSD implementations do not
316      * use it; however our reference for this feature is SunOS 4.x, which
317      * implemented it.  Note that the resulting terminal type was never the
318      * 2-character name, but was instead the first alias after that.
319      */
320 #define ok_TC2(s) (isgraph(UChar(s)) && (s) != '|')
321     ptr = _nc_curr_token.tk_name;
322     if (_nc_syntax == SYN_TERMCAP
323 #if NCURSES_XNAMES
324 	&& !_nc_user_definable
325 #endif
326 	) {
327 	if (ok_TC2(ptr[0]) && ok_TC2(ptr[1]) && (ptr[2] == '|')) {
328 	    ptr += 3;
329 	    _nc_curr_token.tk_name[2] = '\0';
330 	}
331     }
332 
333     entryp->tterm.str_table = entryp->tterm.term_names = _nc_save_str(ptr);
334 
335     if (entryp->tterm.str_table == NULL)
336 	returnDB(ERR);
337 
338     DEBUG(2, ("Starting '%s'", ptr));
339 
340     /*
341      * We do this because the one-token lookahead in the parse loop
342      * results in the terminal type getting prematurely set to correspond
343      * to that of the next entry.
344      */
345     name = _nc_first_name(entryp->tterm.term_names);
346     if (!valid_entryname(name)) {
347 	_nc_warning("invalid entry name \"%s\"", name);
348 	name = "invalid";
349     }
350     _nc_set_type(name);
351 
352     /* check for overly-long names and aliases */
353     for (base = entryp->tterm.term_names; (ptr = strchr(base, '|')) != NULL;
354 	 base = ptr + 1) {
355 	if (ptr - base > MAX_ALIAS) {
356 	    _nc_warning("%s `%.*s' may be too long",
357 			(base == entryp->tterm.term_names)
358 			? "primary name"
359 			: "alias",
360 			(int) (ptr - base), base);
361 	}
362     }
363 
364     entryp->nuses = 0;
365 
366     for (token_type = _nc_get_token(silent);
367 	 token_type != EOF && token_type != NAMES;
368 	 token_type = _nc_get_token(silent)) {
369 	bool is_use = (strcmp(_nc_curr_token.tk_name, "use") == 0);
370 	bool is_tc = !is_use && (strcmp(_nc_curr_token.tk_name, "tc") == 0);
371 	if (is_use || is_tc) {
372 	    char *saved;
373 
374 	    if (!VALID_STRING(_nc_curr_token.tk_valstring)
375 		|| _nc_curr_token.tk_valstring[0] == '\0') {
376 		_nc_warning("missing name for use-clause");
377 		continue;
378 	    } else if (!valid_entryname(_nc_curr_token.tk_valstring)) {
379 		_nc_warning("invalid name for use-clause \"%s\"",
380 			    _nc_curr_token.tk_valstring);
381 		continue;
382 	    } else if (entryp->nuses >= HARD_MAX_USES) {
383 		_nc_warning("too many use-clauses, ignored \"%s\"",
384 			    _nc_curr_token.tk_valstring);
385 		continue;
386 	    } else if (entryp->nuses >= WARN_MAX_USES) {
387 		_nc_warning("possibly too many use-clauses (%d vs %d), \"%s\"",
388 			    entryp->nuses,
389 			    WARN_MAX_USES,
390 			    _nc_curr_token.tk_valstring);
391 	    }
392 	    if ((saved = _nc_save_str(_nc_curr_token.tk_valstring)) != NULL) {
393 		entryp->uses[entryp->nuses].name = saved;
394 		entryp->uses[entryp->nuses].line = _nc_curr_line;
395 		entryp->nuses++;
396 		if (entryp->nuses > 1 && is_tc) {
397 		    BAD_TC_USAGE
398 		}
399 	    }
400 	} else {
401 	    /* normal token lookup */
402 	    entry_ptr = _nc_find_entry(_nc_curr_token.tk_name,
403 				       _nc_get_hash_table(_nc_syntax == SYN_TERMCAP));
404 
405 	    /*
406 	     * Our kluge to handle aliasing.  The reason it is done
407 	     * this ugly way, with a linear search, is so the hashing
408 	     * machinery doesn't have to be made really complicated
409 	     * (also we get better warnings this way).  No point in
410 	     * making this case fast, aliased caps aren't common now
411 	     * and will get rarer.
412 	     */
413 	    if (entry_ptr == NOTFOUND) {
414 		const struct alias *ap;
415 
416 		if (_nc_syntax == SYN_TERMCAP) {
417 		    if (entryp->nuses != 0) {
418 			BAD_TC_USAGE
419 		    }
420 		    for (ap = _nc_get_alias_table(TRUE); ap->from; ap++)
421 			if (strcmp(ap->from, _nc_curr_token.tk_name) == 0) {
422 			    if (ap->to == (char *) 0) {
423 				_nc_warning("%s (%s termcap extension) ignored",
424 					    ap->from, ap->source);
425 				goto nexttok;
426 			    }
427 
428 			    entry_ptr = _nc_find_entry(ap->to,
429 						       _nc_get_hash_table(TRUE));
430 			    if (entry_ptr && !silent)
431 				_nc_warning("%s (%s termcap extension) aliased to %s",
432 					    ap->from, ap->source, ap->to);
433 			    break;
434 			}
435 		} else {	/* if (_nc_syntax == SYN_TERMINFO) */
436 		    for (ap = _nc_get_alias_table(FALSE); ap->from; ap++)
437 			if (strcmp(ap->from, _nc_curr_token.tk_name) == 0) {
438 			    if (ap->to == (char *) 0) {
439 				_nc_warning("%s (%s terminfo extension) ignored",
440 					    ap->from, ap->source);
441 				goto nexttok;
442 			    }
443 
444 			    entry_ptr = _nc_find_entry(ap->to,
445 						       _nc_get_hash_table(FALSE));
446 			    if (entry_ptr && !silent)
447 				_nc_warning("%s (%s terminfo extension) aliased to %s",
448 					    ap->from, ap->source, ap->to);
449 			    break;
450 			}
451 
452 		    if (entry_ptr == NOTFOUND) {
453 			entry_ptr = lookup_fullname(_nc_curr_token.tk_name);
454 		    }
455 		}
456 	    }
457 #if NCURSES_XNAMES
458 	    /*
459 	     * If we have extended-names active, we will automatically
460 	     * define a name based on its context.
461 	     */
462 	    if (entry_ptr == NOTFOUND
463 		&& _nc_user_definable) {
464 		if (expected_type(_nc_curr_token.tk_name, token_type, silent)) {
465 		    if ((entry_ptr = _nc_extend_names(entryp,
466 						      _nc_curr_token.tk_name,
467 						      token_type)) != NULL) {
468 			if (_nc_tracing >= DEBUG_LEVEL(1)) {
469 			    _nc_warning("extended capability '%s'",
470 					_nc_curr_token.tk_name);
471 			}
472 		    }
473 		} else {
474 		    /* ignore it: we have already printed error message */
475 		    continue;
476 		}
477 	    }
478 #endif /* NCURSES_XNAMES */
479 
480 	    /* can't find this cap name, not even as an alias */
481 	    if (entry_ptr == NOTFOUND) {
482 		if (!silent)
483 		    _nc_warning("unknown capability '%s'",
484 				_nc_curr_token.tk_name);
485 		continue;
486 	    }
487 
488 	    /* deal with bad type/value combinations. */
489 	    if (token_type == CANCEL) {
490 		/*
491 		 * Prefer terminfo in this (long-obsolete) ambiguity:
492 		 */
493 		if (!strcmp("ma", _nc_curr_token.tk_name)) {
494 		    entry_ptr = _nc_find_type_entry("ma", NUMBER,
495 						    _nc_syntax != SYN_TERMINFO);
496 		    assert(entry_ptr != NULL);
497 		}
498 	    } else if (entry_ptr->nte_type != token_type) {
499 		/*
500 		 * Nasty special cases here handle situations in which type
501 		 * information can resolve name clashes.  Normal lookup
502 		 * finds the last instance in the capability table of a
503 		 * given name, regardless of type.  find_type_entry looks
504 		 * for a first matching instance with given type.  So as
505 		 * long as all ambiguous names occur in pairs of distinct
506 		 * type, this will do the job.
507 		 */
508 
509 		if (token_type == NUMBER
510 		    && !strcmp("ma", _nc_curr_token.tk_name)) {
511 		    /* tell max_attributes from arrow_key_map */
512 		    entry_ptr = _nc_find_type_entry("ma", NUMBER,
513 						    _nc_syntax != SYN_TERMINFO);
514 		    assert(entry_ptr != NULL);
515 
516 		} else if (token_type == STRING
517 			   && !strcmp("MT", _nc_curr_token.tk_name)) {
518 		    /* map terminfo's string MT to MT */
519 		    entry_ptr = _nc_find_type_entry("MT", STRING,
520 						    _nc_syntax != SYN_TERMINFO);
521 		    assert(entry_ptr != NULL);
522 
523 		} else if (token_type == BOOLEAN
524 			   && entry_ptr->nte_type == STRING) {
525 		    /* treat strings without following "=" as empty strings */
526 		    token_type = STRING;
527 		} else {
528 		    /* we couldn't recover; skip this token */
529 		    if (!silent) {
530 			const char *type_name;
531 			switch (entry_ptr->nte_type) {
532 			case BOOLEAN:
533 			    type_name = "boolean";
534 			    break;
535 			case STRING:
536 			    type_name = "string";
537 			    break;
538 			case NUMBER:
539 			    type_name = "numeric";
540 			    break;
541 			default:
542 			    type_name = "unknown";
543 			    break;
544 			}
545 			_nc_warning("wrong type used for %s capability '%s'",
546 				    type_name, _nc_curr_token.tk_name);
547 		    }
548 		    continue;
549 		}
550 	    }
551 
552 	    /* now we know that the type/value combination is OK */
553 	    switch (token_type) {
554 	    case CANCEL:
555 		switch (entry_ptr->nte_type) {
556 		case BOOLEAN:
557 		    entryp->tterm.Booleans[entry_ptr->nte_index] =
558 			(NCURSES_SBOOL) CANCELLED_BOOLEAN;
559 		    break;
560 
561 		case NUMBER:
562 		    entryp->tterm.Numbers[entry_ptr->nte_index] = CANCELLED_NUMERIC;
563 		    break;
564 
565 		case STRING:
566 		    entryp->tterm.Strings[entry_ptr->nte_index] = CANCELLED_STRING;
567 		    break;
568 		}
569 		break;
570 
571 	    case BOOLEAN:
572 		entryp->tterm.Booleans[entry_ptr->nte_index] = TRUE;
573 		break;
574 
575 	    case NUMBER:
576 #if !NCURSES_EXT_NUMBERS
577 		if (_nc_curr_token.tk_valnumber > MAX_NUMBER) {
578 		    entryp->tterm.Numbers[entry_ptr->nte_index] = MAX_NUMBER;
579 		} else
580 #endif
581 		{
582 		    entryp->tterm.Numbers[entry_ptr->nte_index] =
583 			(NCURSES_INT2) _nc_curr_token.tk_valnumber;
584 		}
585 		break;
586 
587 	    case STRING:
588 		ptr = _nc_curr_token.tk_valstring;
589 		if (_nc_syntax == SYN_TERMCAP) {
590 		    int n = entry_ptr->nte_index;
591 		    ptr = _nc_captoinfo(_nc_curr_token.tk_name,
592 					ptr,
593 					(n < (int) SIZEOF(parametrized))
594 					? parametrized[n]
595 					: 0);
596 		}
597 		entryp->tterm.Strings[entry_ptr->nte_index] = _nc_save_str(ptr);
598 		break;
599 
600 	    default:
601 		if (!silent)
602 		    _nc_warning("unknown token type");
603 		_nc_panic_mode((char) ((_nc_syntax == SYN_TERMCAP) ? ':' : ','));
604 		continue;
605 	    }
606 	}			/* end else cur_token.name != "use" */
607       nexttok:
608 	continue;		/* cannot have a label w/o statement */
609     }				/* endwhile (not EOF and not NAMES) */
610 
611     _nc_push_token(token_type);
612     _nc_set_type(_nc_first_name(entryp->tterm.term_names));
613 
614     /*
615      * Try to deduce as much as possible from extension capabilities
616      * (this includes obsolete BSD capabilities).  Sigh...it would be more
617      * space-efficient to call this after use resolution, but it has
618      * to be done before entry allocation is wrapped up.
619      */
620     if (!literal) {
621 	if (_nc_syntax == SYN_TERMCAP) {
622 	    bool has_base_entry = FALSE;
623 
624 	    /*
625 	     * Don't insert defaults if this is a `+' entry meant only
626 	     * for inclusion in other entries (not sure termcap ever
627 	     * had these, actually).
628 	     */
629 	    if (strchr(entryp->tterm.term_names, '+')) {
630 		has_base_entry = TRUE;
631 	    } else {
632 		unsigned i;
633 		/*
634 		 * Otherwise, look for a base entry that will already
635 		 * have picked up defaults via translation.
636 		 */
637 		for (i = 0; i < entryp->nuses; i++) {
638 		    if (entryp->uses[i].name != NULL
639 			&& !strchr(entryp->uses[i].name, '+'))
640 			has_base_entry = TRUE;
641 		}
642 	    }
643 
644 	    postprocess_termcap(&entryp->tterm, has_base_entry);
645 	} else
646 	    postprocess_terminfo(&entryp->tterm);
647     }
648     _nc_wrap_entry(entryp, FALSE);
649 
650     returnDB(OK);
651 }
652 
653 NCURSES_EXPORT(int)
_nc_capcmp(const char * s,const char * t)654 _nc_capcmp(const char *s, const char *t)
655 /* compare two string capabilities, stripping out padding */
656 {
657     bool ok_s = VALID_STRING(s);
658     bool ok_t = VALID_STRING(t);
659 
660     if (ok_s && ok_t) {
661 	for (;;) {
662 	    if (s[0] == '$' && s[1] == '<') {
663 		for (s += 2;; s++) {
664 		    if (!(isdigit(UChar(*s))
665 			  || *s == '.'
666 			  || *s == '*'
667 			  || *s == '/'
668 			  || *s == '>')) {
669 			break;
670 		    }
671 		}
672 	    }
673 
674 	    if (t[0] == '$' && t[1] == '<') {
675 		for (t += 2;; t++) {
676 		    if (!(isdigit(UChar(*t))
677 			  || *t == '.'
678 			  || *t == '*'
679 			  || *t == '/'
680 			  || *t == '>')) {
681 			break;
682 		    }
683 		}
684 	    }
685 
686 	    /* we've now pushed s and t past any padding they pointed at */
687 
688 	    if (*s == '\0' && *t == '\0')
689 		return (0);
690 
691 	    if (*s != *t)
692 		return (*t - *s);
693 
694 	    /* else *s == *t but one is not NUL, so continue */
695 	    s++, t++;
696 	}
697     } else if (ok_s || ok_t) {
698 	return 1;
699     }
700     return 0;
701 }
702 
703 static void
append_acs0(string_desc * dst,int code,const char * src,size_t off)704 append_acs0(string_desc * dst, int code, const char *src, size_t off)
705 {
706     if (src != NULL && off < strlen(src)) {
707 	char temp[3];
708 	temp[0] = (char) code;
709 	temp[1] = src[off];
710 	temp[2] = 0;
711 	_nc_safe_strcat(dst, temp);
712     }
713 }
714 
715 static void
append_acs(string_desc * dst,int code,const char * src)716 append_acs(string_desc * dst, int code, const char *src)
717 {
718     if (VALID_STRING(src) && strlen(src) == 1) {
719 	append_acs0(dst, code, src, 0);
720     }
721 }
722 
723 /*
724  * The ko capability, if present, consists of a comma-separated capability
725  * list.  For each capability, we may assume there is a keycap that sends the
726  * string which is the value of that capability.
727  */
728 #define DATA(from, to) { { from }, { to } }
729 typedef struct {
730     const char from[3];
731     const char to[6];
732 } assoc;
733 static assoc const ko_xlate[] =
734 {
735     DATA("al", "kil1"),		/* insert line key  -> KEY_IL    */
736     DATA("bt", "kcbt"),		/* back tab         -> KEY_BTAB  */
737     DATA("cd", "ked"),		/* clear-to-eos key -> KEY_EOL   */
738     DATA("ce", "kel"),		/* clear-to-eol key -> KEY_EOS   */
739     DATA("cl", "kclr"),		/* clear key        -> KEY_CLEAR */
740     DATA("ct", "tbc"),		/* clear all tabs   -> KEY_CATAB */
741     DATA("dc", "kdch1"),	/* delete char      -> KEY_DC    */
742     DATA("dl", "kdl1"),		/* delete line      -> KEY_DL    */
743     DATA("do", "kcud1"),	/* down key         -> KEY_DOWN  */
744     DATA("ei", "krmir"),	/* exit insert key  -> KEY_EIC   */
745     DATA("ho", "khome"),	/* home key         -> KEY_HOME  */
746     DATA("ic", "kich1"),	/* insert char key  -> KEY_IC    */
747     DATA("im", "kIC"),		/* insert-mode key  -> KEY_SIC   */
748     DATA("le", "kcub1"),	/* le key           -> KEY_LEFT  */
749     DATA("nd", "kcuf1"),	/* nd key           -> KEY_RIGHT */
750     DATA("nl", "kent"),		/* new line key     -> KEY_ENTER */
751     DATA("st", "khts"),		/* set-tab key      -> KEY_STAB  */
752     DATA("ta", ""),
753     DATA("up", "kcuu1"),	/* up-arrow key     -> KEY_UP    */
754 };
755 
756 /*
757  * This routine fills in string caps that either had defaults under
758  * termcap or can be manufactured from obsolete termcap capabilities.
759  * It was lifted from Ross Ridge's mytinfo package.
760  */
761 
762 static const char C_CR[] = "\r";
763 static const char C_LF[] = "\n";
764 static const char C_BS[] = "\b";
765 static const char C_HT[] = "\t";
766 
767 /*
768  * This bit of legerdemain turns all the terminfo variable names into
769  * references to locations in the arrays Booleans, Numbers, and Strings ---
770  * precisely what's needed.
771  */
772 
773 #undef CUR
774 #define CUR tp->
775 
776 static void
postprocess_termcap(TERMTYPE2 * tp,bool has_base)777 postprocess_termcap(TERMTYPE2 *tp, bool has_base)
778 {
779     char buf[MAX_LINE * 2 + 2];
780     string_desc result;
781 
782     TR(TRACE_DATABASE,
783        (T_CALLED("postprocess_termcap(tp=%p, has_base=%d)"),
784 	(void *) tp, has_base));
785 
786     /*
787      * TERMCAP DEFAULTS AND OBSOLETE-CAPABILITY TRANSLATIONS
788      *
789      * This first part of the code is the functional inverse of the
790      * fragment in capdefaults.c.
791      * ----------------------------------------------------------------------
792      */
793 
794     /* if there was a tc entry, assume we picked up defaults via that */
795     if (!has_base) {
796 	if (WANTED(init_3string) && PRESENT(termcap_init2))
797 	    init_3string = _nc_save_str(termcap_init2);
798 
799 	if (WANTED(reset_2string) && PRESENT(termcap_reset))
800 	    reset_2string = _nc_save_str(termcap_reset);
801 
802 	if (WANTED(carriage_return)) {
803 	    if (carriage_return_delay > 0) {
804 		_nc_SPRINTF(buf, _nc_SLIMIT(sizeof(buf))
805 			    "%s$<%d>", C_CR, carriage_return_delay);
806 		carriage_return = _nc_save_str(buf);
807 	    } else
808 		carriage_return = _nc_save_str(C_CR);
809 	}
810 	if (WANTED(cursor_left)) {
811 	    if (backspace_delay > 0) {
812 		_nc_SPRINTF(buf, _nc_SLIMIT(sizeof(buf))
813 			    "%s$<%d>", C_BS, backspace_delay);
814 		cursor_left = _nc_save_str(buf);
815 	    } else if (backspaces_with_bs == 1)
816 		cursor_left = _nc_save_str(C_BS);
817 	    else if (PRESENT(backspace_if_not_bs))
818 		cursor_left = backspace_if_not_bs;
819 	}
820 	/* vi doesn't use "do", but it does seem to use nl (or '\n') instead */
821 	if (WANTED(cursor_down)) {
822 	    if (PRESENT(linefeed_if_not_lf))
823 		cursor_down = linefeed_if_not_lf;
824 	    else if (linefeed_is_newline != 1) {
825 		if (new_line_delay > 0) {
826 		    _nc_SPRINTF(buf, _nc_SLIMIT(sizeof(buf))
827 				"%s$<%d>", C_LF, new_line_delay);
828 		    cursor_down = _nc_save_str(buf);
829 		} else
830 		    cursor_down = _nc_save_str(C_LF);
831 	    }
832 	}
833 	if (WANTED(scroll_forward) && crt_no_scrolling != 1) {
834 	    if (PRESENT(linefeed_if_not_lf))
835 		cursor_down = linefeed_if_not_lf;
836 	    else if (linefeed_is_newline != 1) {
837 		if (new_line_delay > 0) {
838 		    _nc_SPRINTF(buf, _nc_SLIMIT(sizeof(buf))
839 				"%s$<%d>", C_LF, new_line_delay);
840 		    scroll_forward = _nc_save_str(buf);
841 		} else
842 		    scroll_forward = _nc_save_str(C_LF);
843 	    }
844 	}
845 	if (WANTED(newline)) {
846 	    if (linefeed_is_newline == 1) {
847 		if (new_line_delay > 0) {
848 		    _nc_SPRINTF(buf, _nc_SLIMIT(sizeof(buf))
849 				"%s$<%d>", C_LF, new_line_delay);
850 		    newline = _nc_save_str(buf);
851 		} else
852 		    newline = _nc_save_str(C_LF);
853 	    } else if (PRESENT(carriage_return) && PRESENT(scroll_forward)) {
854 		_nc_str_init(&result, buf, sizeof(buf));
855 		if (_nc_safe_strcat(&result, carriage_return)
856 		    && _nc_safe_strcat(&result, scroll_forward))
857 		    newline = _nc_save_str(buf);
858 	    } else if (PRESENT(carriage_return) && PRESENT(cursor_down)) {
859 		_nc_str_init(&result, buf, sizeof(buf));
860 		if (_nc_safe_strcat(&result, carriage_return)
861 		    && _nc_safe_strcat(&result, cursor_down))
862 		    newline = _nc_save_str(buf);
863 	    }
864 	}
865     }
866 
867     /*
868      * Inverse of capdefaults.c code ends here.
869      * ----------------------------------------------------------------------
870      *
871      * TERMCAP-TO TERMINFO MAPPINGS FOR SOURCE TRANSLATION
872      *
873      * These translations will *not* be inverted by tgetent().
874      */
875 
876     if (!has_base) {
877 	/*
878 	 * We wait until now to decide if we've got a working cr because even
879 	 * one that doesn't work can be used for newline. Unfortunately the
880 	 * space allocated for it is wasted.
881 	 */
882 	if (return_does_clr_eol == 1 || no_correctly_working_cr == 1)
883 	    carriage_return = ABSENT_STRING;
884 
885 	/*
886 	 * Supposedly most termcap entries have ta now and '\t' is no longer a
887 	 * default, but it doesn't seem to be true...
888 	 */
889 	if (WANTED(tab)) {
890 	    if (horizontal_tab_delay > 0) {
891 		_nc_SPRINTF(buf, _nc_SLIMIT(sizeof(buf))
892 			    "%s$<%d>", C_HT, horizontal_tab_delay);
893 		tab = _nc_save_str(buf);
894 	    } else
895 		tab = _nc_save_str(C_HT);
896 	}
897 	if (init_tabs == ABSENT_NUMERIC && has_hardware_tabs == TRUE)
898 	    init_tabs = 8;
899 
900 	/*
901 	 * Assume we can beep with ^G unless we're given bl@.
902 	 */
903 	if (WANTED(bell))
904 	    bell = _nc_save_str("\007");
905     }
906 
907     /*
908      * Translate the old termcap :pt: capability to it#8 + ht=\t
909      */
910     if (has_hardware_tabs == TRUE) {
911 	if (init_tabs != 8 && init_tabs != ABSENT_NUMERIC)
912 	    _nc_warning("hardware tabs with a width other than 8: %d", init_tabs);
913 	else {
914 	    if (PRESENT(tab) && _nc_capcmp(tab, C_HT))
915 		_nc_warning("hardware tabs with a non-^I tab string %s",
916 			    _nc_visbuf(tab));
917 	    else {
918 		if (WANTED(tab))
919 		    tab = _nc_save_str(C_HT);
920 		init_tabs = 8;
921 	    }
922 	}
923     }
924     /*
925      * Now translate the ko capability, if there is one.  This
926      * isn't from mytinfo...
927      */
928     if (PRESENT(other_non_function_keys)) {
929 	char *base;
930 	char *bp, *cp, *dp;
931 	struct name_table_entry const *from_ptr;
932 	struct name_table_entry const *to_ptr;
933 	char buf2[MAX_TERMINFO_LENGTH];
934 	bool foundim;
935 
936 	/* we're going to use this for a special case later */
937 	dp = strchr(other_non_function_keys, 'i');
938 	foundim = (dp != NULL) && (dp[1] == 'm');
939 
940 	/* look at each comma-separated capability in the ko string... */
941 	for (base = other_non_function_keys;
942 	     (cp = strchr(base, ',')) != NULL;
943 	     base = cp + 1) {
944 	    size_t len = (unsigned) (cp - base);
945 	    size_t n;
946 	    assoc const *ap = NULL;
947 
948 	    for (n = 0; n < SIZEOF(ko_xlate); ++n) {
949 		if (len == strlen(ko_xlate[n].from)
950 		    && strncmp(ko_xlate[n].from, base, len) == 0) {
951 		    ap = ko_xlate + n;
952 		    break;
953 		}
954 	    }
955 	    if (ap == NULL) {
956 		_nc_warning("unknown capability `%.*s' in ko string",
957 			    (int) len, base);
958 		continue;
959 	    } else if (ap->to[0] == '\0')	/* ignore it */
960 		continue;
961 
962 	    /* now we know we found a match in ko_table, so... */
963 
964 	    from_ptr = _nc_find_entry(ap->from, _nc_get_hash_table(TRUE));
965 	    to_ptr = _nc_find_entry(ap->to, _nc_get_hash_table(FALSE));
966 
967 	    if (!from_ptr || !to_ptr)	/* should never happen! */
968 		_nc_err_abort("ko translation table is invalid, I give up");
969 
970 	    if (WANTED(tp->Strings[from_ptr->nte_index])) {
971 		_nc_warning("no value for ko capability %s", ap->from);
972 		continue;
973 	    }
974 
975 	    if (tp->Strings[to_ptr->nte_index]) {
976 		const char *s = tp->Strings[from_ptr->nte_index];
977 		const char *t = tp->Strings[to_ptr->nte_index];
978 		/* There's no point in warning about it if it is the same
979 		 * string; that's just an inefficiency.
980 		 */
981 		if (VALID_STRING(s) && VALID_STRING(t) && strcmp(s, t) != 0)
982 		    _nc_warning("%s (%s) already has an explicit value %s, ignoring ko",
983 				ap->to, ap->from, t);
984 		continue;
985 	    }
986 
987 	    /*
988 	     * The magic moment -- copy the mapped key string over,
989 	     * stripping out padding.
990 	     */
991 	    bp = tp->Strings[from_ptr->nte_index];
992 	    if (VALID_STRING(bp)) {
993 		for (dp = buf2; *bp; bp++) {
994 		    if ((size_t) (dp - buf2) >= (sizeof(buf2) - sizeof(TERMTYPE2)))
995 			  break;
996 		    if (bp[0] == '$' && bp[1] == '<') {
997 			while (*bp && *bp != '>') {
998 			    ++bp;
999 			}
1000 		    } else
1001 			*dp++ = *bp;
1002 		}
1003 		*dp = '\0';
1004 
1005 		tp->Strings[to_ptr->nte_index] = _nc_save_str(buf2);
1006 	    } else {
1007 		tp->Strings[to_ptr->nte_index] = bp;
1008 	    }
1009 	}
1010 
1011 	/*
1012 	 * Note: ko=im and ko=ic both want to grab the `Insert'
1013 	 * keycap.  There's a kich1 but no ksmir, so the ic capability
1014 	 * got mapped to kich1 and im to kIC to avoid a collision.
1015 	 * If the description has im but not ic, hack kIC back to kich1.
1016 	 */
1017 	if (foundim && WANTED(key_ic) && PRESENT(key_sic)) {
1018 	    key_ic = key_sic;
1019 	    key_sic = ABSENT_STRING;
1020 	}
1021     }
1022 
1023     if (!has_base) {
1024 	if (!hard_copy) {
1025 	    if (WANTED(key_backspace))
1026 		key_backspace = _nc_save_str(C_BS);
1027 	    if (WANTED(key_left))
1028 		key_left = _nc_save_str(C_BS);
1029 	    if (WANTED(key_down))
1030 		key_down = _nc_save_str(C_LF);
1031 	}
1032     }
1033 
1034     /*
1035      * Translate XENIX forms characters.
1036      */
1037     if (PRESENT(acs_ulcorner) ||
1038 	PRESENT(acs_llcorner) ||
1039 	PRESENT(acs_urcorner) ||
1040 	PRESENT(acs_lrcorner) ||
1041 	PRESENT(acs_ltee) ||
1042 	PRESENT(acs_rtee) ||
1043 	PRESENT(acs_btee) ||
1044 	PRESENT(acs_ttee) ||
1045 	PRESENT(acs_hline) ||
1046 	PRESENT(acs_vline) ||
1047 	PRESENT(acs_plus)) {
1048 	char buf2[MAX_TERMCAP_LENGTH];
1049 
1050 	_nc_str_init(&result, buf2, sizeof(buf2));
1051 	_nc_safe_strcat(&result, acs_chars);
1052 
1053 	append_acs(&result, 'j', acs_lrcorner);
1054 	append_acs(&result, 'k', acs_urcorner);
1055 	append_acs(&result, 'l', acs_ulcorner);
1056 	append_acs(&result, 'm', acs_llcorner);
1057 	append_acs(&result, 'n', acs_plus);
1058 	append_acs(&result, 'q', acs_hline);
1059 	append_acs(&result, 't', acs_ltee);
1060 	append_acs(&result, 'u', acs_rtee);
1061 	append_acs(&result, 'v', acs_btee);
1062 	append_acs(&result, 'w', acs_ttee);
1063 	append_acs(&result, 'x', acs_vline);
1064 
1065 	if (buf2[0]) {
1066 	    acs_chars = _nc_save_str(buf2);
1067 	    _nc_warning("acsc string synthesized from XENIX capabilities");
1068 	}
1069     } else if (acs_chars == ABSENT_STRING
1070 	       && PRESENT(enter_alt_charset_mode)
1071 	       && PRESENT(exit_alt_charset_mode)) {
1072 	acs_chars = _nc_save_str(VT_ACSC);
1073     }
1074     returnVoidDB;
1075 }
1076 
1077 static void
postprocess_terminfo(TERMTYPE2 * tp)1078 postprocess_terminfo(TERMTYPE2 *tp)
1079 {
1080     TR(TRACE_DATABASE,
1081        (T_CALLED("postprocess_terminfo(tp=%p)"),
1082 	(void *) tp));
1083 
1084     /*
1085      * TERMINFO-TO-TERMINFO MAPPINGS FOR SOURCE TRANSLATION
1086      * ----------------------------------------------------------------------
1087      */
1088 
1089     /*
1090      * Translate AIX forms characters.
1091      */
1092     if (PRESENT(box_chars_1)) {
1093 	char buf2[MAX_TERMCAP_LENGTH];
1094 	string_desc result;
1095 
1096 	_nc_str_init(&result, buf2, sizeof(buf2));
1097 	_nc_safe_strcat(&result, acs_chars);
1098 
1099 	append_acs0(&result, 'l', box_chars_1, 0);	/* ACS_ULCORNER */
1100 	append_acs0(&result, 'q', box_chars_1, 1);	/* ACS_HLINE */
1101 	append_acs0(&result, 'k', box_chars_1, 2);	/* ACS_URCORNER */
1102 	append_acs0(&result, 'x', box_chars_1, 3);	/* ACS_VLINE */
1103 	append_acs0(&result, 'j', box_chars_1, 4);	/* ACS_LRCORNER */
1104 	append_acs0(&result, 'm', box_chars_1, 5);	/* ACS_LLCORNER */
1105 	append_acs0(&result, 'w', box_chars_1, 6);	/* ACS_TTEE */
1106 	append_acs0(&result, 'u', box_chars_1, 7);	/* ACS_RTEE */
1107 	append_acs0(&result, 'v', box_chars_1, 8);	/* ACS_BTEE */
1108 	append_acs0(&result, 't', box_chars_1, 9);	/* ACS_LTEE */
1109 	append_acs0(&result, 'n', box_chars_1, 10);	/* ACS_PLUS */
1110 
1111 	if (buf2[0]) {
1112 	    acs_chars = _nc_save_str(buf2);
1113 	    _nc_warning("acsc string synthesized from AIX capabilities");
1114 	    box_chars_1 = ABSENT_STRING;
1115 	}
1116     }
1117     /*
1118      * ----------------------------------------------------------------------
1119      */
1120     returnVoidDB;
1121 }
1122 
1123 /*
1124  * Do a linear search through the terminfo tables to find a given full-name.
1125  * We don't expect to do this often, so there's no hashing function.
1126  *
1127  * In effect, this scans through the 3 lists of full-names, and looks them
1128  * up in _nc_info_table, which is organized so that the nte_index fields are
1129  * sorted, but the nte_type fields are not necessarily grouped together.
1130  */
1131 static struct name_table_entry const *
lookup_fullname(const char * find)1132 lookup_fullname(const char *find)
1133 {
1134     int state = -1;
1135 
1136     for (;;) {
1137 	int count = 0;
1138 	NCURSES_CONST char *const *names;
1139 
1140 	switch (++state) {
1141 	case BOOLEAN:
1142 	    names = boolfnames;
1143 	    break;
1144 	case STRING:
1145 	    names = strfnames;
1146 	    break;
1147 	case NUMBER:
1148 	    names = numfnames;
1149 	    break;
1150 	default:
1151 	    return NOTFOUND;
1152 	}
1153 
1154 	for (count = 0; names[count] != NULL; count++) {
1155 	    if (!strcmp(names[count], find)) {
1156 		struct name_table_entry const *entry_ptr = _nc_get_table(FALSE);
1157 		while (entry_ptr->nte_type != state
1158 		       || entry_ptr->nte_index != count)
1159 		    entry_ptr++;
1160 		return entry_ptr;
1161 	    }
1162 	}
1163     }
1164 }
1165 
1166 /* parse_entry.c ends here */
1167