int
_nc_parse_entry(struct entry *entryp, int literal, bool silent)
{
    int token_type;
    struct name_table_entry const *entry_ptr;
    char *ptr, *base;

    token_type = _nc_get_token();

    if (token_type == EOF)
	return (EOF);
    if (token_type != NAMES)
	_nc_err_abort("Entry does not start with terminal names in column one");

    _nc_init_entry(&entryp->tterm);

    entryp->cstart = _nc_comment_start;
    entryp->cend = _nc_comment_end;
    entryp->startline = _nc_start_line;
    DEBUG(2, ("Comment range is %ld to %ld", entryp->cstart, entryp->cend));

    /* junk the 2-character termcap name, if present */
    ptr = _nc_curr_token.tk_name;
    if (ptr[2] == '|') {
	ptr = _nc_curr_token.tk_name + 3;
	_nc_curr_token.tk_name[2] = '\0';
    }

    entryp->tterm.str_table = entryp->tterm.term_names = _nc_save_str(ptr);

    DEBUG(1, ("Starting '%s'", ptr));

    /*
     * We do this because the one-token lookahead in the parse loop
     * results in the terminal type getting prematurely set to correspond
     * to that of the next entry.
     */
    _nc_set_type(_nc_first_name(entryp->tterm.term_names));

    /* check for overly-long names and aliases */
    for (base = entryp->tterm.term_names; (ptr = strchr(base, '|')) != 0;
	 base = ptr + 1) {
	if (ptr - base > MAX_ALIAS) {
	    _nc_warning("%s `%.*s' may be too long",
			(base == entryp->tterm.term_names)
			? "primary name"
			: "alias",
			ptr - base, base);
	}
    }

    entryp->nuses = 0;

    for (token_type = _nc_get_token();
	 token_type != EOF && token_type != NAMES;
	 token_type = _nc_get_token()) {
	if (strcmp(_nc_curr_token.tk_name, "use") == 0
	    || strcmp(_nc_curr_token.tk_name, "tc") == 0) {
	    entryp->uses[entryp->nuses].name = _nc_save_str(_nc_curr_token.tk_valstring);
	    entryp->uses[entryp->nuses].line = _nc_curr_line;
	    entryp->nuses++;
	} else {
	    /* normal token lookup */
	    entry_ptr = _nc_find_entry(_nc_curr_token.tk_name,
				       _nc_syntax ? _nc_cap_hash_table : _nc_info_hash_table);

	    /*
	     * Our kluge to handle aliasing.  The reason it's done
	     * this ugly way, with a linear search, is so the hashing
	     * machinery doesn't have to be made really complicated
	     * (also we get better warnings this way).  No point in
	     * making this case fast, aliased caps aren't common now
	     * and will get rarer.
	     */
	    if (entry_ptr == NOTFOUND) {
		const struct alias *ap;

		if (_nc_syntax == SYN_TERMCAP) {
		    for (ap = _nc_capalias_table; ap->from; ap++)
			if (strcmp(ap->from, _nc_curr_token.tk_name) == 0) {
			    if (ap->to == (char *) 0) {
				_nc_warning("%s (%s termcap extension) ignored",
					    ap->from, ap->source);
				goto nexttok;
			    }

			    entry_ptr = _nc_find_entry(ap->to, _nc_cap_hash_table);
			    if (entry_ptr && !silent)
				_nc_warning("%s (%s termcap extension) aliased to %s",
					    ap->from, ap->source, ap->to);
			    break;
			}
		} else {	/* if (_nc_syntax == SYN_TERMINFO) */
		    for (ap = _nc_infoalias_table; ap->from; ap++)
			if (strcmp(ap->from, _nc_curr_token.tk_name) == 0) {
			    if (ap->to == (char *) 0) {
				_nc_warning("%s (%s terminfo extension) ignored",
					    ap->from, ap->source);
				goto nexttok;
			    }

			    entry_ptr = _nc_find_entry(ap->to, _nc_info_hash_table);
			    if (entry_ptr && !silent)
				_nc_warning("%s (%s terminfo extension) aliased to %s",
					    ap->from, ap->source, ap->to);
			    break;
			}

		    if (entry_ptr == NOTFOUND) {
			entry_ptr = lookup_fullname(_nc_curr_token.tk_name);
		    }
		}
	    }
#if NCURSES_XNAMES
	    /*
	     * If we have extended-names active, we will automatically
	     * define a name based on its context.
	     */
	    if (entry_ptr == NOTFOUND
		&& _nc_user_definable
		&& (entry_ptr = _nc_extend_names(entryp,
						 _nc_curr_token.tk_name,
						 token_type)) != 0) {
		if (_nc_tracing >= DEBUG_LEVEL(1))
		    _nc_warning("extended capability '%s'", _nc_curr_token.tk_name);
	    }
#endif /* NCURSES_XNAMES */

	    /* can't find this cap name, not even as an alias */
	    if (entry_ptr == NOTFOUND) {
		if (!silent)
		    _nc_warning("unknown capability '%s'",
				_nc_curr_token.tk_name);
		continue;
	    }

	    /* deal with bad type/value combinations. */
	    if (token_type != CANCEL && entry_ptr->nte_type != token_type) {
		/*
		 * Nasty special cases here handle situations in which type
		 * information can resolve name clashes.  Normal lookup
		 * finds the last instance in the capability table of a
		 * given name, regardless of type.  find_type_entry looks
		 * for a first matching instance with given type.  So as
		 * long as all ambiguous names occur in pairs of distinct
		 * type, this will do the job.
		 */

		/* tell max_attributes from arrow_key_map */
		if (token_type == NUMBER && !strcmp("ma", _nc_curr_token.tk_name))
		    entry_ptr = _nc_find_type_entry("ma", NUMBER,
						    _nc_get_table(_nc_syntax
								  != 0));

		/* map terminfo's string MT to MT */
		else if (token_type == STRING && !strcmp("MT", _nc_curr_token.tk_name))
		    entry_ptr = _nc_find_type_entry("MT", STRING,
						    _nc_get_table(_nc_syntax
								  != 0));

		/* treat strings without following "=" as empty strings */
		else if (token_type == BOOLEAN && entry_ptr->nte_type == STRING)
		    token_type = STRING;
		/* we couldn't recover; skip this token */
		else {
		    if (!silent) {
			const char *type_name;
			switch (entry_ptr->nte_type) {
			case BOOLEAN:
			    type_name = "boolean";
			    break;
			case STRING:
			    type_name = "string";
			    break;
			case NUMBER:
			    type_name = "numeric";
			    break;
			default:
			    type_name = "unknown";
			    break;
			}
			_nc_warning("wrong type used for %s capability '%s'",
				    type_name, _nc_curr_token.tk_name);
		    }
		    continue;
		}
	    }

	    /* now we know that the type/value combination is OK */
	    switch (token_type) {
	    case CANCEL:
		switch (entry_ptr->nte_type) {
		case BOOLEAN:
		    entryp->tterm.Booleans[entry_ptr->nte_index] = CANCELLED_BOOLEAN;
		    break;

		case NUMBER:
		    entryp->tterm.Numbers[entry_ptr->nte_index] = CANCELLED_NUMERIC;
		    break;

		case STRING:
		    entryp->tterm.Strings[entry_ptr->nte_index] = CANCELLED_STRING;
		    break;
		}
		break;

	    case BOOLEAN:
		entryp->tterm.Booleans[entry_ptr->nte_index] = TRUE;
		break;

	    case NUMBER:
		entryp->tterm.Numbers[entry_ptr->nte_index] =
		    _nc_curr_token.tk_valnumber;
		break;

	    case STRING:
		ptr = _nc_curr_token.tk_valstring;
		if (_nc_syntax == SYN_TERMCAP)
		    ptr = _nc_captoinfo(_nc_curr_token.tk_name,
					ptr,
					parametrized[entry_ptr->nte_index]);
		entryp->tterm.Strings[entry_ptr->nte_index] = _nc_save_str(ptr);
		break;

	    default:
		if (!silent)
		    _nc_warning("unknown token type");
		_nc_panic_mode((_nc_syntax == SYN_TERMCAP) ? ':' : ',');
		continue;
	    }
	}			/* end else cur_token.name != "use" */
      nexttok:
	continue;		/* cannot have a label w/o statement */
    }				/* endwhile (not EOF and not NAMES) */

    _nc_push_token(token_type);
    _nc_set_type(_nc_first_name(entryp->tterm.term_names));

    /*
     * Try to deduce as much as possible from extension capabilities
     * (this includes obsolete BSD capabilities).  Sigh...it would be more
     * space-efficient to call this after use resolution, but it has
     * to be done before entry allocation is wrapped up.
     */
    if (!literal) {
	if (_nc_syntax == SYN_TERMCAP) {
	    bool has_base_entry = FALSE;
	    int i;

	    /*
	     * Don't insert defaults if this is a `+' entry meant only
	     * for inclusion in other entries (not sure termcap ever
	     * had these, actually).
	     */
	    if (strchr(entryp->tterm.term_names, '+'))
		has_base_entry = TRUE;
	    else
		/*
		 * Otherwise, look for a base entry that will already
		 * have picked up defaults via translation.
		 */
		for (i = 0; i < entryp->nuses; i++)
		    if (!strchr((char *) entryp->uses[i].name, '+'))
			has_base_entry = TRUE;

	    postprocess_termcap(&entryp->tterm, has_base_entry);
	} else
	    postprocess_terminfo(&entryp->tterm);
    }
    _nc_wrap_entry(entryp);

    return (OK);
}
Example #2
0
_nc_parse_entry(struct entry *entryp, int literal, bool silent)
{
    int token_type;
    struct name_table_entry const *entry_ptr;
    char *ptr, *base;
    bool bad_tc_usage = FALSE;

    token_type = _nc_get_token(silent);

    if (token_type == EOF)
	return (EOF);
    if (token_type != NAMES)
	_nc_err_abort("Entry does not start with terminal names in column one");

    _nc_init_entry(&entryp->tterm);

    entryp->cstart = _nc_comment_start;
    entryp->cend = _nc_comment_end;
    entryp->startline = _nc_start_line;
    DEBUG(2, ("Comment range is %ld to %ld", entryp->cstart, entryp->cend));

    /*
     * Strip off the 2-character termcap name, if present.  Originally termcap
     * used that as an indexing aid.  We can retain 2-character terminfo names,
     * but note that they would be lost if we translate to/from termcap.  This
     * feature is supposedly obsolete since "newer" BSD implementations do not
     * use it; however our reference for this feature is SunOS 4.x, which
     * implemented it.  Note that the resulting terminal type was never the
     * 2-character name, but was instead the first alias after that.
     */
    ptr = _nc_curr_token.tk_name;
    if (_nc_syntax == SYN_TERMCAP
#if NCURSES_XNAMES
	&& !_nc_user_definable
#endif
	) {
	if (ptr[2] == '|') {
	    ptr += 3;
	    _nc_curr_token.tk_name[2] = '\0';
	}
    }

    entryp->tterm.str_table = entryp->tterm.term_names = _nc_save_str(ptr);

    DEBUG(1, ("Starting '%s'", ptr));

    /*
     * We do this because the one-token lookahead in the parse loop
     * results in the terminal type getting prematurely set to correspond
     * to that of the next entry.
     */
    _nc_set_type(_nc_first_name(entryp->tterm.term_names));

    /* check for overly-long names and aliases */
    for (base = entryp->tterm.term_names; (ptr = strchr(base, '|')) != 0;
	 base = ptr + 1) {
	if (ptr - base > MAX_ALIAS) {
	    _nc_warning("%s `%.*s' may be too long",
			(base == entryp->tterm.term_names)
			? "primary name"
			: "alias",
			(int) (ptr - base), base);
	}
    }

    entryp->nuses = 0;

    for (token_type = _nc_get_token(silent);
	 token_type != EOF && token_type != NAMES;
	 token_type = _nc_get_token(silent)) {
	bool is_use = (strcmp(_nc_curr_token.tk_name, "use") == 0);
	bool is_tc = !is_use && (strcmp(_nc_curr_token.tk_name, "tc") == 0);
	if (is_use || is_tc) {
	    entryp->uses[entryp->nuses].name = _nc_save_str(_nc_curr_token.tk_valstring);
	    entryp->uses[entryp->nuses].line = _nc_curr_line;
	    entryp->nuses++;
	    if (entryp->nuses > 1 && is_tc) {
		BAD_TC_USAGE
	    }
	} else {
Example #3
0
_nc_get_token(bool silent)
{
    static const char terminfo_punct[] = "@%&*!#";
    long number;
    int type;
    int ch;
    char *numchk;
    char numbuf[80];
    unsigned found;
    static char buffer[MAX_ENTRY_SIZE];
    char *ptr;
    int dot_flag = FALSE;
    long token_start;

    if (pushtype != NO_PUSHBACK) {
	int retval = pushtype;

	_nc_set_type(pushname);
	DEBUG(3, ("pushed-back token: `%s', class %d",
		  _nc_curr_token.tk_name, pushtype));

	pushtype = NO_PUSHBACK;
	pushname[0] = '\0';

	/* currtok wasn't altered by _nc_push_token() */
	return (retval);
    }

    if (end_of_stream())
	return (EOF);

  start_token:
    token_start = stream_pos();
    while ((ch = next_char()) == '\n' || iswhite(ch))
	continue;

    ch = eat_escaped_newline(ch);

    if (ch == EOF)
	type = EOF;
    else {
	/* if this is a termcap entry, skip a leading separator */
	if (separator == ':' && ch == ':')
	    ch = next_char();

	if (ch == '.'
#if NCURSES_EXT_FUNCS
	    && !_nc_disable_period
#endif
	    ) {
	    dot_flag = TRUE;
	    DEBUG(8, ("dot-flag set"));

	    while ((ch = next_char()) == '.' || iswhite(ch))
		continue;
	}

	if (ch == EOF) {
	    type = EOF;
	    goto end_of_token;
	}

	/* have to make some punctuation chars legal for terminfo */
	if (!isalnum(ch)
#if NCURSES_EXT_FUNCS
	    && !(ch == '.' && _nc_disable_period)
#endif
	    && !strchr(terminfo_punct, (char) ch)) {
	    if (!silent)
		_nc_warning("Illegal character (expected alphanumeric or %s) - %s",
			    terminfo_punct, unctrl((chtype) ch));
	    _nc_panic_mode(separator);
	    goto start_token;
	}

	ptr = buffer;
	*(ptr++) = ch;

	if (first_column) {
	    char *desc;

	    _nc_comment_start = token_start;
	    _nc_comment_end = _nc_curr_file_pos;
	    _nc_start_line = _nc_curr_line;

	    _nc_syntax = ERR;
	    while ((ch = next_char()) != '\n') {
		if (ch == EOF)
		    _nc_err_abort("premature EOF");
		else if (ch == ':' && last_char() != ',') {
		    _nc_syntax = SYN_TERMCAP;
		    separator = ':';
		    break;
		} else if (ch == ',') {
		    _nc_syntax = SYN_TERMINFO;
		    separator = ',';
		    /*
		     * Fall-through here is not an accident.  The idea is that
		     * if we see a comma, we figure this is terminfo unless we
		     * subsequently run into a colon -- but we don't stop
		     * looking for that colon until hitting a newline.  This
		     * allows commas to be embedded in description fields of
		     * either syntax.
		     */
		    /* FALLTHRU */
		} else
		    ch = eat_escaped_newline(ch);

		*ptr++ = ch;
	    }
	    ptr[0] = '\0';
	    if (_nc_syntax == ERR) {
		/*
		 * Grrr...what we ought to do here is barf, complaining that
		 * the entry is malformed.  But because a couple of name fields
		 * in the 8.2 termcap file end with |\, we just have to assume
		 * it's termcap syntax.
		 */
		_nc_syntax = SYN_TERMCAP;
		separator = ':';
	    } else if (_nc_syntax == SYN_TERMINFO) {
		/* throw away trailing /, *$/ */
		for (--ptr; iswhite(*ptr) || *ptr == ','; ptr--)
		    continue;
		ptr[1] = '\0';
	    }

	    /*
	     * This is the soonest we have the terminal name fetched.  Set up
	     * for following warning messages.
	     */
	    ptr = strchr(buffer, '|');
	    if (ptr == (char *) NULL)
		ptr = buffer + strlen(buffer);
	    ch = *ptr;
	    *ptr = '\0';
	    _nc_set_type(buffer);
	    *ptr = ch;

	    /*
	     * Compute the boundary between the aliases and the description
	     * field for syntax-checking purposes.
	     */
	    desc = strrchr(buffer, '|');
	    if (!silent && desc) {
		if (*desc == '\0')
		    _nc_warning("empty longname field");
		else if (strchr(desc, ' ') == (char *) NULL)
		    _nc_warning("older tic versions may treat the description field as an alias");
	    }
	    if (!desc)
		desc = buffer + strlen(buffer);

	    /*
	     * Whitespace in a name field other than the long name can confuse
	     * rdist and some termcap tools.  Slashes are a no-no.  Other
	     * special characters can be dangerous due to shell expansion.
	     */
	    for (ptr = buffer; ptr < desc; ptr++) {
		if (isspace(CharOf(*ptr))) {
		    if (!silent)
			_nc_warning("whitespace in name or alias field");
		    break;
		} else if (*ptr == '/') {
		    if (!silent)
			_nc_warning("slashes aren't allowed in names or aliases");
		    break;
		} else if (strchr("$[]!*?", *ptr)) {
		    if (!silent)
			_nc_warning("dubious character `%c' in name or alias field", *ptr);
		    break;
		}
	    }

	    ptr = buffer;

	    _nc_curr_token.tk_name = buffer;
	    type = NAMES;
	} else {
	    while ((ch = next_char()) != EOF) {
		if (!isalnum(ch)) {
		    if (_nc_syntax == SYN_TERMINFO) {
			if (ch != '_')
			    break;
		    } else {	/* allow ';' for "k;" */
			if (ch != ';')
			    break;
		    }
		}
		*(ptr++) = ch;
	    }

	    *ptr++ = '\0';
	    switch (ch) {
	    case ',':
	    case ':':
		if (ch != separator)
		    _nc_err_abort("Separator inconsistent with syntax");
		_nc_curr_token.tk_name = buffer;
		type = BOOLEAN;
		break;
	    case '@':
		if ((ch = next_char()) != separator && !silent)
		    _nc_warning("Missing separator after `%s', have %s",
				buffer, unctrl((chtype) ch));
		_nc_curr_token.tk_name = buffer;
		type = CANCEL;
		break;

	    case '#':
		found = 0;
		while (isalnum(ch = next_char())) {
		    numbuf[found++] = ch;
		    if (found >= sizeof(numbuf) - 1)
			break;
		}
		numbuf[found] = '\0';
		number = strtol(numbuf, &numchk, 0);
		if (!silent) {
		    if (numchk == numbuf)
			_nc_warning("no value given for `%s'", buffer);
		    if ((*numchk != '\0') || (ch != separator))
			_nc_warning("Missing separator");
		}
		_nc_curr_token.tk_name = buffer;
		_nc_curr_token.tk_valnumber = number;
		type = NUMBER;
		break;

	    case '=':
		ch = _nc_trans_string(ptr, buffer + sizeof(buffer));
		if (!silent && ch != separator)
		    _nc_warning("Missing separator");
		_nc_curr_token.tk_name = buffer;
		_nc_curr_token.tk_valstring = ptr;
		type = STRING;
		break;

	    case EOF:
		type = EOF;
		break;
	    default:
		/* just to get rid of the compiler warning */
		type = UNDEF;
		if (!silent)
		    _nc_warning("Illegal character - %s", unctrl((chtype) ch));
	    }
	}			/* end else (first_column == FALSE) */
    }				/* end else (ch != EOF) */

  end_of_token:

#ifdef TRACE
    if (dot_flag == TRUE)
	DEBUG(8, ("Commented out "));

    if (_nc_tracing >= DEBUG_LEVEL(7)) {
	switch (type) {
	case BOOLEAN:
	    _tracef("Token: Boolean; name='%s'",
		    _nc_curr_token.tk_name);
	    break;

	case NUMBER:
	    _tracef("Token: Number;  name='%s', value=%d",
		    _nc_curr_token.tk_name,
		    _nc_curr_token.tk_valnumber);
	    break;

	case STRING:
	    _tracef("Token: String;  name='%s', value=%s",
		    _nc_curr_token.tk_name,
		    _nc_visbuf(_nc_curr_token.tk_valstring));
	    break;

	case CANCEL:
	    _tracef("Token: Cancel; name='%s'",
		    _nc_curr_token.tk_name);
	    break;

	case NAMES:

	    _tracef("Token: Names; value='%s'",
		    _nc_curr_token.tk_name);
	    break;

	case EOF:
	    _tracef("Token: End of file");
	    break;

	default:
	    _nc_warning("Bad token type");
	}
    }
#endif

    if (dot_flag == TRUE)	/* if commented out, use the next one */
	type = _nc_get_token(silent);

    DEBUG(3, ("token: `%s', class %d",
	      _nc_curr_token.tk_name != 0 ? _nc_curr_token.tk_name :
	      "<null>",
	      type));

    return (type);
}