Sun May 10 06:28:55 1998  Nelson H. F. Beebe  <beebe@math.utah.edu>

	* Version 2.11.4 completed.

	* Files changed with this release: ChangeLog Makefile.in README
	bibclean.c bibclean.h bibclean.html bibclean.man custom.h fix.c 
	isbn.c option.c.

	* New files with this release: testbib6.bok testbib6.eok
	testbib6.org.

	* Makefile.in: Add testbib6.org to the BIBTEX-TESTS list to
	exercise the bug fix in fix_month() described below.

	Change modes of installed files to permit group write access.

	Change the mode of the installed .bibcleanrc file to ensure
	readability.

	Save any old installed version of bibclean as bibclean.old during
	"make install".

	Create a hard link from the installed bibclean to bibclean-2.11.4,
	and similarly, for the installed manual page.  That way, if a
	later version comes out and is installed, then bibclean-2.11.4
	will still remain accessible.  [I'm doing this for all of my
	software distributions; extensive experience with GNUware, and
	even commercial packages like Matlab and Maple, on ten
	architectures, has shown the desirability of having multiple
	versions available as a check when a suspected bug turns up.]

	The "make uninstall" command will remove the
	version-number-specific installed files, if you wish to back out
	of an install.

	Change symbolic links to hard links; they are all used in a
	context where this has no side effects.

	Change CP to mean "rcp -p", so as to preserve file time stamps on
	installation.

	* README: Update my mailing addresses.

	* bibclean.c: Update my mailing addresses in the file header.

	* bibclean.h: Automatically generated from the formatted output of
	bibclean.man to include that file's changes.

	* bibclean.hhtml: Automatically generated from bibclean.man to
	include that file's changes.

	* bibclean.man: Update my mailing addresses in the file header and
	the AUTHOR section.  Document the new exit behavior for -author
	and -version.  Documen the new GNU/POSIX --option support, and add
	a paragraph on the handling of filenames that would be confused
	with option names.  Add cross references to bibcheck(1),
	bibdup(1), bibjoin(1), biblabel(), and biborder(1).

	* custom.h: Change type of sleep() on NeXT from int to unsigned
	int.
	
	* fix.c: Fix a rarely-encountered, but long-standing, bug in
	fix_month(). Prior to this version, a value

	    mar # "\slash" # apr

	would be incorrectly transformed

	    month = mar # "\slash" # " # apr # "

	because in_quoted_string was incorrect for the remaineder of
	the value.

	If the input value was changed to

	    mar # "\slash " # apr

	then that space inside the quoted string preserved the correctness
	of in_quoted_string, and the output was correct.

	* isbn.c: Add several countries to the list of recognized ISBN
	prefixes.  Similar updates have been added to the Emacs Lisp file,
	isbn.el, available in a separate software distribution.  Update my
	mailing addresses in the file header.

	* option.c: : Update my mailing addresses on opt_author().  Add
	support for GNU/POSIX-style --options.  Update bibclean version
	number.  Make -author and -version handling exit with a success
	code after their output on stderr.

Sat May  4 07:52:26 1996  Nelson H. F. Beebe  <beebe@math.utah.edu>

	* Version 2.11.3 (final edits before public release)

	* Build and test bibclean under 8 IBM PC DOS C and C++
 	compilers, and under VAX VMS 6.1.  bibclean already builds and
 	tests successfully under about 40 C and C++ compilers on 10
 	different UNIX architectures.

	* Update README, ibmpc/dos/README and vms/vax/README files.

	* In bibclean.c, move sanity checks on HAVE_xxx pattern matching
 	flags to custom.h.

	* In bibclean.man, update version release date to match that set
 	in option.c.

	* In chek.c, update computation of stdlog_on_stdout to be
 	stricter.  Fix off-by-one error in loop termination condition in
 	u72copy_element().  Add code in copy_element() to skip trailing
 	space and hyphens, to avoid generating bogus warnings about
 	checksum mismatches.

	* In configure.in, remove -g from CFLAGS if we are compiling
 	with lcc because it produces bad debug symbol tables on Sun
 	Solaris 2.x.

	* In do.c, remove unused LAST_SCREEN_LINE macro.  Change open of
	bibliography files to use binary mode in OS_PCDOS.

	* In fix.c, change types of n and nupper in fix_title() from int
	to size_t to avoid type conflict warnings from some compilers.

	* In fndfil.c, remove unused NEWLINE() macro.  Update remainder
 	of code to match fndfil.c in DVI 3.0 development tree.

	* In isbn.c, add workaround for bug in Watcom 10.0 C++ compilers
	in ISBN_hyphenate().  Fix serious bug in squeeze_ISBN() that had
	gone undetected for a very long time: the loop exit condition
	referred to in_ISBN instead of *in_ISBN.

	* In keybrd.c, add #include <starlet.h> to get some needed
 	library function headers.  Remove some old unused macros.
  	Update get_screen_lines() for VAX VMS 6.x.

	* In Makefile.in, remove some obsolete files and targets, update
 	the DIST-FILES list, and add missing testbib5.org file to
 	BIBTEX-TESTS list.  Remove DIST-FILES-BIN list, since the share
 	bundle option is no longer provided, and DIST-FILES now has all
 	files listed.  Add match, romtol, and subdist targets.  Add
 	dependency on subdist of distribution file targets.  Add
 	test-match and test-romtol targets.

	* In match.c, add month_patterns[], pages_patterns[], and
 	volume_patterns[].  Update messages to include input line
 	number.  Modify process() to accept ``key = "value",'' lines
 	that can be easily extracted from bibliography file collections.
  	A test suite has been prepared in match.dat, and the test-match
 	and match targets added to Makefile.in to support this testing.

	* In option.c, update release date in BIBCLEAN_VERSION.

	* In vaxvms.c, include <lib$routines.h> and <starlet.h> for more
	function prototypes.  Update for VAX VMS 6.x.  Add conditionals
	around memxxx() functions, so as to avoid linker warnings about
	duplicate symbols, and add missing return values in the memxxx()
	functions.  Change LIB$SPAWN to lib$spawn.

	* In vmswild.c, remove unused LINSIZ macro.

Sun Apr 28 08:54:38 1996  Nelson H. F. Beebe  <beebe@math.utah.edu>

	* Version 2.11.3 (continued):

	* Makefile.in (test-bibtex): Add test-bibtex-7 with test files
	testcodn.{org,bok,eok}, to test CODEN handling.

Thu Apr 25 11:24:18 1996  Nelson H. F. Beebe  <beebe@math.utah.edu>

	* Version 2.11.3:

	* In bibclean.c, remove declarations of unused functions that
 	were moved to new file chek.c in 2.11.0.  Change one test (n <=
 	0) to (n == 0) to avoid compiler warnings about test of unsigned
 	values for being negative.

	* Add CODEN validation support to chek.c, and completely rewrite
 	the ISBN and ISSN handling.  The reason for the latter is that
 	by defining a simple list grammar for CODEN, ISBN, and ISSN
 	string values, the code becomes cleaner and more rigorous, and
 	parts of it can be shared for the checking of all three types.
  	Future versions of bibclean may add support for additional
 	key/value pairs for which the string values can also be
 	validated, and such support will then be very simple to supply.
  	New functions added: check_CODEN(), bad_CODEN(), copy_element(),
 	incomplete_CODEN(), is_CODEN_char(), is_ISBN_char(),
 	is_ISSN_char(), parse_list(), parse_element(),
 	parse_separator(), validate_CODEN(), validate_ISBN(),
 	validate_ISSN().  New typedef added: parse_data.

	These improvements necessitated a change in testisxn.eok, since
 	the recognition of ISBN values (and also CODEN and ISSN values)
 	is now quite strict, so some of the erroneous ISBNs in
 	testisxn.org are now handled slightly differently.  The same
 	number of error messages are issued in both 2.11.2 and 2.11.3
 	however.

	* Add mention of the CODEN validations to bibclean.man.

	* Add CODEN entry in out_value() checks[] array in do.c.

Mon Mar  4 16:23:20 1996  Nelson H. F. Beebe  <beebe@math.utah.edu>

	* Version 2.11.2: This version adds six new command-line
	options: -[no]-align-equals, -[no]-keep-preamble-spaces,
	and -[no]-keep-string-spaces.

 	The first pair of these was suggested by Matthew Morley
 	<Matthew.Morley@gmd.de> who contributed an implementation for a
 	much older version of bibclean (2.05); the idea has been used, but
 	not the contributed code.

	The second and third pairs supply a need that has developed as
	the TUG and BibNet bibliography collections have grown: namely,
	sometimes bibliographies have rather long, and carefully spaced,
	@Preamble{} and @String{} entries, and these new options allow
	the internal spacing in those entries to be preserved.

	* bibclean.c: Add the six new options to the file header comments.
  	Add function out_verbatim() to handle output without space
 	fiddling.  Add new macros KEEP_PREAMBLE_SPACES() and
 	KEEP_STRING_SPACES(), and use them in get_next_non_blank() and
 	out_string().

	* bibclean.man: Document the new -[no]-align-equals,
	-[no]-keep-preamble-spaces, and -[no]-keep-string-spaces options.

	* do.c: Add new macros KEEP_PREAMBLE_SPACES() and
 	KEEP_STRING_SPACES(), and use them in several places.  Add new
 	global variables keep_preamble_spaces and keep_string_spaces.
  	Make in_preamble instead of static, since it is now needed in
 	bibclean.c too.  Add new functions do_preamble_2() and
 	do_string_2() to hold old bodies of do_preamble() and do_string(),
 	to simplify setting of in_preamble and in_string flags.  Add
 	support in out_equals() for -align-equals option.  The code for
 	-keep-preamble-spaces and -keep-string-spaces turned out to be
 	quite hard to get right, largely because of the design flaw (for
 	which I alone hold responsibility) that input space is discarded
 	at a rather low level, rather than being collected as a legitimate
 	token that is carried up to the highest levels, and then
 	discarded, or otherwise modified.

	* fix.c: Add additional checks in fix_pages() to avoid inserting
	en-dashes where they don't belong.

	* keybrd.c: Add MIN() macro.  Add new functions beep() and
 	erase_characters() to shorten coding.  Move search code out of
 	do_more() into separate function do_search().  Add support for
 	word-erase, line-erase, and line-reprint in do_search().  When a
 	search string is reused, print it again.  Beep if too many
 	characters are entered in a search string.  Add support for
 	forward-paragraph and backward-paragraph commands.  Correct
 	binding of M-> to be KEYBOARD_END instead of KEYBOARD_HOME.  Add
 	support for X Window System arrow keys, and PgUP and PgDn keys.

	* option.c: Add new functions opt_align_equals(),
	opt_keep_preamble_spaces(), and opt_keep_string_spaces(), plus
	corresponding entries in the options[] array, and document them
	in the opt_help() help_lines[] and usage() usage_lines[] arrays.

Sat Oct  7 08:59:48 1995  Nelson H. F. Beebe  <beebe@math.utah.edu>

	* Version 2.11: The major change for this new version is the
	conversion of the source code to use the GNU autoconf system so
	that on UNIX systems, the configure script can automatically
	prepare a header file, config.h, that deals with the variations
	in UNIX and compiler implementations.  This entailed:

		(a) the creation of 3 new files: Makefile.in,
 		config.hin, and configure.in;

		(b) elimination of 12 header files: machdefs.h os.h
 		osatari.h ospcdos.h osprimos.h osrmx.h ostops20.h
 		osunix.h osvaxvms.h osvmcms.h unixlib.h xstddef.h;

		(c) adjustments of #include statements, protecting most
 		with #if HAVE_xxx ... #endif;

		(d) change of the NEW_STYLE and STDC macros to
 		HAVE_STDC;

		(e) replacement of free() calls by macro FREE(), so as
 		to be able to hide some typecasting to deal with
 		defective C/C++ implementations.

	* bibclean.c: The code in this file became too large for IBM PC
 	DOS compilers, so it has been split into 10 new header files:
		ch.h config.h custom.h delete.h keybrd.h pattern.h
 		token.h toklst.h xunistd.h yesorno.h
	and 5 additional source code files:
		chek.c do.c fix.c keybrd.c option.c.
	[The name is chek.c, rather than check.c, to avoid a conflict
 	with the GNU standard Makefile target, check.]  The 683-line
 	revision history has been moved from bibclean.c into this new
 	ChangeLog file (see below), following GNU Project conventions;
 	I've not bothered to revise the recording style to match that
 	supported by the GNU Emacs add-change-log-entry-other-window
 	command.

	This file splitting reduced bibclean.c from 8K lines to 2K
 	lines, and the file do.c, at 2.9K lines, is now the largest
 	source code file.  Here is a current source code line count:

		% wc -l *.c | sort +0nr -1
		  11626 total
		   2885 do.c
		   2094 bibclean.c
		   1051 vmswild.c
		    880 vaxvms.c
		    790 isbn.c
		    774 chek.c
		    606 keybrd.c
		    572 fix.c
		    569 fndfil.c
		    551 option.c
		    395 match.c
		    240 strtol.c
		    174 romtol.c
		     45 strist.c

		% wc -l *.h | sort +0nr -1
		    835 total
		    260 bibclean.h
		    178 config.h
		     81 custom.h
		     51 toklst.h
		     47 keybrd.h
		     37 ch.h
		     26 xctype.h
		     18 xstring.h
		     17 xstdlib.h
		     16 xstat.h
		     15 token.h
		     12 xtypes.h
		     12 xunistd.h
		     11 pattern.h
		     10 delete.h
		     10 match.h
		      9 xerrno.h
		      9 xlimits.h
		      9 xpwd.h
		      7 yesorno.h

	Despite the increased number of source files between versions
 	2.10 and 2.11, the total source code has been reduced from 16.6K
 	lines to 12.5K lines, largely due to the use of GNU configure.
  	Of course, one must factor in about 1K new lines in Makefile.in,
 	config.hin, and configure.in, and 1.6K lines in the
 	automatically-generated configure script.

	In summary, although this was my first use of autoconf, and the
 	development effort stretched over 3 otherwise very busy weeks, I
 	believe the payoff for other programs will be significant,
 	because their config.in and configure.in files will be very
 	similar to those for bibclean, so the conversion effort will be
 	much smaller.

	Once config.h files are manually generated generated for other
 	operating systems, they too will be largely reusable for most of
 	my other programs in C/C++.

	* do.c: In get_line(), change
		if (*(p-1) == '\\')
	to
		if ((p > &line[0]) && (*(p-1) == '\\'))
	in order to avoid possible out-of-bounds pointer reference.

	* fix.c: In fix_pages(), add code
	* match.c: In next_s(), change
		for (++s; ; )
	to
		for (++s; *s; )
	to eliminate compiler warnings about the final return() being
 	unreachable.  The additional test of *s against '\0' is of
	no significance for execution time.

	* Makefile.in: remove vaxvms.c and vmswild.c, since they will be
 	handled now by a separate Makefile for VAX VMS and OpenVMS.

	* README: Completely rewrite, with new installation instructions
 	for UNIX using the configure script, and new subsections
 	containing status reports of installation attempts on all major
 	UNIX variants.

Revision history (reverse time order):

[08-Mar-1995]	2.10.1
		In get_braced_string(), handle overlooked case of \"{x}.


[04-Jun-1994]--[18-Oct-1994]	2.10
		Add support for simple forward and backward searches
		in help display in do_more().

		Add World-Wide Web URI, URL, and URN names to the list
		of field names that are forced to uppercase.  Disable
		code in out_s() that breaks lines at punctuation
		characters, because this can introduce unwanted line
		breaks in file names and WWW names.

		Add code in out_c() to preserve last line in output
		buffer, so that a subsequent DELETE_LINE operation
		will have a complete line to delete.  Previously, with
		-delete-empty, if the output buffer filled up in the
		middle of a line that was going to be deleted, the
		initial part of the line would be (incorrectly)
		output.  This was a hard bug to track down, because it
		happened very rarely.  Extend out_flush() to force out
		all buffered output.

		Change bad_ISBN() to not put incorrect hyphens back
		into ISBNs.  ISBN hyphen positions vary, because large
		publishers get short prefixes, and small publishers
		get long prefixes.  The ISBN is a 10-character value
		of the form

		countrygroupnumber-publishernumber-booknumber-checkdigit

		where the first three fields are of variable width and
		contain only the digits [0-9], and the last
		single-character checkdigit field contains [0-9X].
		Larger country groups or publishers have smaller
		numbers, and correspondingly larger booknumbers.
		Thus, the TeXbook ISBN, 0-201-13447-0, consists of 0
		for English-language countries, 201 for
		Addison-Wesley, 13447 for the book number, and 0 for
		the check digit.  Although it is permissible to use
		spaces instead of hyphends, we prefer the latter in
		order to have unbreakable text for improved
		readability and editing convenience.

		Add code in fix_title() to supply braces around
		brace-level-zero TeX control words containing
		upper-case letters, to obtain conversions like these:
		    "\TeX{} for the Impatient" -> "{\TeX} for the Impatient"
		    "\TeX\ for the Impatient"  -> "{\TeX} for the Impatient"
		in order to protect the control words from downcasing
		in some BibTeX styles.

		Change DELETE_CHAR and DELETE_LINE values to be
		outside the range of possible signed or unsigned
		character values, so as not to interfere with handling
		of 8-bit character data.

		Add new functions bcolumn(), bdelc(), bdelline(),
		bflush(), blastc(), bpeekc(), and bputc() to localize
		access to the output buffer, buf[], and the variables
		that record line number, column number, and output
		byte position.  Completely rewrite out_c() to use
		these new functions, substantially simplifying the
		code.

		Add original_file initialized to the_file in error()
		and warning(), and used in out_status(), to more
		accurately reflect output file positions.  Previously,
		the error messages themselves resulted in modification
		of these values.

		In new_entry(), initialize non_white_chars to zero.
		Previously, if multiple input files were specified,
		the first entry in files after the first was not
		correctly recognized if it started on the first line.
		This bug went undetected for so long because bibclean
		is rarely used with multiple input files.

		At most calls of out_with_parbreak_error(), put the
		current character back into the input stream so that
		it is output after the error message instead of before.
		This is more likely to avoid splitting an input token.

		The remaining fixes are for bugs and problems caught
		by Alf-Christian Achilles <bibservadmin@ira.uka.de>
		who maintains a very large bibliography collection on
		ftp.ira.uka.de in /pub/bibliography consisting of
		about 300 bibliographies, and 800 bibcleaned files,
		occupying about 100MB.

		In trim_value(), check for final "\<space>" control
		sequence, and discard it; previously, only the space
		was discarded, leaving an erroneous bare backslash.

		In get_braced_string(), add missing additional
		boundary check in loop to convert braced string to
		quoted string.

		Add DELETE_WHITESPACE support.  In out_c(), this
		requests discarding of all trailing whitespace in the
		output buffer.  In do_BibTeX_entry(), do_group(),
		do_preamble(), and do_Scribe_entry(), if
		prettyprinting, call out_c(DELETE_WHITESPACE) after
		parsing optional space following the @ and the name.
		Previously, a newline following the @ or the name
		would be preserved, which is legal, but not desirable
		in prettyprinted output.  At end of do_one_file(),
		call out_c(DELETE_WHITESPACE) to discard trailing
		space at end of file.

		Add -[no-]German-style option to provide for special
		treatment of quote characters inside braced strings,
		following the conventions of german.sty, which
		overloads quote to simplify input and representation
		of German umlaut accents, sharp-s (es-zet), ligature
		separators, invisible hyphens, raised/lowered quotes,
		French guillemets, and discretionary hyphens.

		Add external file isbn.c, and call to ISBN_hyphenate()
		in bad_ISBN() and check_ISBN(), so that correct
		hyphenation of ISBNs is automatically supplied.


[01-Mar-1994]	2.09
		Add -[no-]keep-linebreaks, -[no-]keep-parbreaks, and
		-[no-]keep-spaces.  This required usurping two control
		characters, Ctl-N and Ctl-P, for LINEBREAK and
		PARBREAK respectively; should such characters occur in
		an input string value, they will be misinterpreted,
		and will not be preserved in the output.

		This restriction could be eliminated if value strings
		had such characters converted to escape sequences
		which were converted back again on output; however,
		this would pose a substantial complication for column
		position tracking, and carry a run-time penalty from
		the extra conversions.  Since control characters are
		not expected to be present in value strings, loss of
		two of them is not expected to be a problem.

[24-Sep-1993]	2.08
		Update bibclean.c to handle characters in range
		128..255 when char is a signed data type, and extend
		testbib1.bib with more tests of the complete character
		set.  Augment Makefile with convenience targets for
		each host environment.  Add private input character
		pushback support, because ANSI/ISO Standard C only
		guarantees one character of pushback, and bibclean
		needs at least 3.

[04-Jun-1993]	2.07
		Update *.h files from DVI development to include support
		for DEC Alpha 3000/500x OSF 1.2 (cc, c89, cxx), and HP
		9000/735 HP-UX A.09.01 (cc, c89, CC, gcc, g++).  Private
		memset() below is not compiled on DEC Alpha to avoid
		conflicts with vendor-supplied version.

[30-Nov-1992 -- 29-Jan-1993]	2.06
		[1] Extend fix_author() to handle conversion of
		"Smith, Jr., A. B." to "A. B. {Smith, Jr.}", and
		"Smith Jr., A. B." to "A. B. {Smith Jr.}".  Introduce
		new auxiliary function check_junior() called by
		fix_author().

		[2] Extend month_pair[] table to include abbreviated month
		names.  Add new function month_token() and rewrite
		fix_month() to use it to do a token-based parse of the
		current_value[] string in order to be able to handle
		conversions like these:

			"January"		--> jan
			"Jan."			--> jan
			"January 24"		--> jan # " 24"
			"24 Jan."		--> "24 " # jan
			"May/June"		--> may # "/" # jun
			"February and May"	--> feb # " and " # may

		New test input files have been added, with new
		expected output and error files.

		[3] Add -[no-]fix-font-changes switch and new function
		opt_fix_font_changes() to supply additional braces
		around font changes in titles to prevent letter case
		conversion by some BibTeX stylos.  The real work is
		done in the new function brace_font_changes() called
		at the end of fix_title().

		[4] Fix bug in out_value() with -delete-empty-fields
		requested.  The code incorrectly assumed that a string
		value of 2 or fewer characters was an empty string.
		That is correct if the string value is a quoted
		string, but wrong if the string value is a 1- or
		2-character macro.  The code now correctly checks
		explicitly for empty string, rather than using its
		length to make the deletion decision. Thanks to
		Manfred Aben <manfred@swi.psy.uva.nl> for reporting
		this bug!

		[5] Fix bug in get_token(); the code must test for a
		non-NULL pointer before calling SKIP_SPACE().  Thanks
		to Gil Webster <gil@dap.csiro.au> for reporting this
		bug!  [And thanks to the Internet funding agencies,
		who make worldwide collaboration like this possible.]

		[6] Add support for Roman numeral matching in match.c.
		New file romtol.c contains romtol(), isroman(), and a
		test program.

		[7] Add calls to perror() at file open failures.

		[8] Fix stupid error in brace_font_changes(); s[] was not
		NUL-terminated before the final call to strcpy().

		[9] In format(), make newmsg[] an internal static array
		instead of reusing shared_string; otherwise, a warning
		message from check_length() wipes out the current
		string value.

		[10] In main(), initialize the_file.input.filename to an
		empty string, to avoid dereferencing a NULL pointer in
		warning() during command-line option parsing.

		[11] In do_args(), add code to display erroneous option
		switches.

		[12] In apply_function(), make string comparisons use
		longer of the minimum match length and the option
		switch length, so that all characters of the option
		switch are tested.  Previously, option misspellings
		after the minimum match length went undetected.

		[13] Add -[no-]prettyprint support, with new functions
		do_string(), opt_prettyprint(), out_at(),
		out_close_brace(), out_comma(), out_newline(),
		out_open_brace(), out_other(), and out_token(), and
		new data type token_t.  Move do_comma() processing
		out of do_field_value_pair() so that it can be used
		by do_string().

		[14] Add BYTE_VAL() macro for printing of characters with
		octal formats.

		[15] Change tag/key to key/field to agree with Appendix B
		of LaTeX User's Guide and Reference Manual.  Change
		format items %t/%k to %k/%f to match key/field
		terminology.  The user impact of this rather large
		source code and documentation change should be
		minimal, and the removal of the disagreement with the
		LaTeX book needs to be done now, rather than later.
		bibclean.ini does not use either of the changed format
		items.

		[16] Change keyboard uses of key to keyboard, so key now
		refers exclusively to bibliography citation keys.

		[17] Change MAX_COLUMN to MAX_WIDTH and add -max-width
		support, with new function opt_max_width() and new
		variable max_width.  Include "xlimits.h".

		[18] Add BIBCLEAN_EXT and BIBCLEAN_INI to define
		environment variables that can supply alternate
		initialization file extension and name.  Add
		GETDEFAULT() macro to simplify coding.

		[19] Rename put_char() to out_c(), and remove macro out_c().

		[20] Add line wrapping support for lexical analysis output
		in out_c().

		[21] Add ``# line nnn "filename"'' output from out_token()
		for lexical analysis output.

		[22] Add strdup macro to redefine that name as
		strdup_private to avoid problems with incorrect
		<string.h> declarations of that function (for DEC
		Alpha OSF/1).

		[23] Add out_input_position() and token_start so that
		out_token() can record both starting and ending line
		numbers of a multi-line value token in lexical
		analyzer output.

		[24] Add check for non-zero brace level at end-of-file in
		do_one_file().

		[25] Add support for @Include{...}, a proposed extension
		of BibTeX.

		[26] Add append_value(), do_newline(),
		do_optional_inline_comment(), do_optional_space(),
		do_space(), get_inline_comment(),
		get_optional_space(), and out_complex_value() so that
		intervening space can be output when lexing, and so
		that we can support in-line comments as well as
		horizontal and vertical space between lexical items,
		according to the proposed grammar for BibTeX.  Add
		do_preamble() to do rigorous parse of @Preamble{...}.
		Revise do_BibTeX_value() to support recognition of
		optional space between tokens of a string expression,
		splitting it into two separate functions
		do_BibTeX_value_1() and do_BibTeX_value_2() to handle
		the two cases of prettyprinting and lexical analysis.
		Add out_string() to localize test for output style.

		[27] Add checks for end-of-file in quoted and braced
		strings so these errors get reported.  Suppress
		pattern value checking for empty values; some of the
		check_xxx() functions did this already, but some did
		not.  Now the test is localized in one place, in
		out_value(), for all of them.


[16-Nov-1992 -- 24-Nov-1992]	2.05
		Add Makefile steps to automatically extract help() text
		from output of manual pages into new file bibclean.h, so
		the built-in documentation stays up-to-date.  The usage
		messages still need manual adjustment if switches are
		added or changed.

		Add missing test of check_values in check_patterns().

		Add support for optional warning messages with
		patterns from initialization files.  New function:
		get_token().  New parsing code in do_new_pattern() to
		handle optional warning message strings.  Add message
		argument to add_pattern().

		Remove strip_comments() since comment processing is
		now handled by get_token() and do_new_pattern().  This
		permits unescaped comment characters inside quoted
		strings.

		Write bibclean.reg, an initialization file similar to
		bibclean.ini, but with regular expressions.

		Replace cascaded if statements for regular expression
		testing with loop over patterns in check_patterns().

		Move inclusion of match.h to after definition of
		typedef YESorNO, and change type of match_pattern()
		from int to YESorNO.

		Add do_fileinit() and code in main() to call
		do_fileinit() for each named input file with an
		extension, replacing that extension with INITFILE_EXT
		(default .ini).  This adds a bibliography-specific
		initialization capability to the system-wide,
		user-wide, and job-wide files already supported.

		Change -keep-initials and -keep-names to -fix-initials
		and -fix-names, making them positive, rather than
		negative, options.  Also, make them independent by
		moving invocations of fix_periods() outside of
		fix_author(), and by checking fix_names in
		fix_author() instead of at start of fix_namelist().

		Add -[no-]read-init-files option to allow control over
		which initialization files are read.

		Add -[no-]trace-file-opening option to allow easy
		tracing of file opening attempts by the program.  A
		similar feature in my DVI drivers has proved enormously
		valuable in tracking down problems of missing files.

		Rename entry_name[] to current_entry_name[], key[] to
		current_key[], tag[] to current_tag[], and value[] to
		current_value[] to get more distinctive names for
		those global variables.

		Include the value string matching code selection in the
		version() message; this is needed so that users can
		prepare initialization files with the correct pattern
		syntax.

		Make several MAX_xxx symbolic constants definable
		at compile time.

		Add MAX_PATTERN_NAMES constant, and increase
		pattern_names[] table to that size, leaving empty slots
		for expansion.  Extend add_pattern() so that
		unrecognized key names result in creation of new entries
		in pattern_names[], making the set of key/value pairs
		extensible without modification of the bibclean source
		code.  Add check_other() to handle checking of other
		keywords.

		Add unexpected() to localize issuing of unexpected value
		warnings.

		Repair next_s() in match.c to skip past <backslash><non-letter>
		TeX control sequence; it was stopping one character
		early.

		Revise upper-case letter bracing code in fix_title()
		to handle more cases.

		Rewrite space collapsing code in fix_pages() to only
		collapse space around en-dashes.  The previous code
		was too aggressive, so that "319 with 30 illustrations"
		became "319 with30illustrations".

		Add check_tag() called from do_tag_name(), and add
		second argument, value, to check_patterns().

		Add format() called from error() and warning() to
		expand %e (@entry name), %k (key), %t (tag), %v
		(value), and %% (percent) format items in messages.
		This feature is needed so user-defined messages in
		initialization files can get key, tag, and value into
		messages.  It also simplifies, and improves, calls to
		warning() and error().

		Add some missing (void) typecasts before str***() calls.

		Change word_length() to return one more than true
		length at end of string.  Change tests in out_s() to
		> MAX_COLUMN instead of >= MAX_COLUMN.  Previously, if
		a line ended exactly at column MAX_COLUMN, bibclean
		could produce a spurious blank line, and would
		sometimes wrap a line earlier than necessary.  Add
		additional punctuation wrap points in out_s(), and
		remove tests for non-blank whitespace in switch()
		statement.

		Change type of all string index variables from int to
		size_t.

		In get_simple_string(), use enum type for type codes if
		NEW_STYLE.

		In check_year(), validate all sequences of 1 or more
		digits.

		Use the C preprocessor to define memmove() to be
		Memmove(), so we always use our own version.  Too many
		C and C++ implementations were found to be lacking it,
		sigh...  Similarly, we provide our own version of
		strtol() (in a separate file) from the DVI 3.0
		development, because it too is missing from older UNIX
		systems.

		Complete port to IBM PC DOS with Turbo C 2.0, and
		Turbo C and C++ 3.0.  This required economization of
		storage for arrays of size [MAX_TOKEN_SIZE] to get
		global data below 64KB without having to reduce
		MAX_TOKEN.

		Added code in do_more() and preprocessor conditionals
		in out_lines() to handle character-at-a-time input for
		help paging on IBM PC DOS. Keyboard function keys
		PgUp, PgDn, End, Home, Up arrow and Down arrow are
		also recognized.  This was easy to do because most PC
		DOS C compilers provide getch() to get a keyboard
		character without echo.  No fiddling of terminal modes
		is needed like it is on other systems.

		The IBM PC DOS port exposed a problem in findfile(),
		where it was assumed that an environment variable
		would not be longer than the longest filename.  Turbo
		C sets the latter to 80 characters, but environment
		variables can be set that are almost 128 characters
		long.  Microsoft C 5.0 also sets it to 80, but C 5.1
		sets it to 144, and C 6.0 and C and C++ 7.0 set it to
		260.  This has been handled by defining MAXPATHLEN at
		compile time, overriding the built-in defaults.

		Add support for character-at-a-time input for help
		paging on VAX VMS, and for getting the screen size in
		get_screen_lines().

		Rename do_more_init() to kbopen(), do_more_term() to
		kbclose(), and use kbget() in do_more() to conceal the
		heavily-O/S dependent details of the kbxxx()
		functions.

		Introduce STREQUAL() macro to simplify coding.

		Introduce KEY_FUNCTION_ENTRY type and apply_function()
		to simplify coding, and use it in do_args(),
		do_preargs(), and out_value().  Argument actions are
		moved into separate functions, opt_xxx().  Rename
		show_author() to opt_author(), and help() to
		opt_help().  Rename do_file() to do_one_file(), and
		move file loop code from main() into new do_files().

		Split large body of get_simple_string() into four new
		functions, get_braced_string(), get_digit_string(),
		get_quoted_string(), and get_identifier_string().

		Add check_inodes() to determine whether stdlog and
		stdout are the same file.  If so, we need to ensure
		that each warning message begins a new line, without
		double spacing unnecessarily when they are different
		files.

		Add memset() implementation for SunOS 4.1.1 CC (C++)
		and BSD 4.3 UNIX because it is missing from their
		run-time libraries.

		Replace fopen() by macro FOPEN() to work around
		erroneous fopen() prototype for SunOS 4.1.1 CC (C++).

		Complete port to IBM PC DOS with Microsoft C 5.1 and
		6.0 compilers.  Minor source changes (the CONST macro
		below) needed to work around compiler errors.


[15-Nov-1992]	2.04
		Minor changes to complete successful VAX VMS
		installation and test.


[15-Nov-1992]	2.03
		Add match_pattern() support for consistent pattern
		matching in the check_xxx() functions, using new code
		defined separately in match.c.

		Add support for run-time redefinition of patterns via
		one or more initialization file(s) found in the PATH
		(system-defined) and BIBINPUTS (user-defined) search
		paths.  New functions: add_pattern(),
		check_patterns(), do_initfile(), do_new_pattern(),
		do_single_arg(), enlarge_table(), get_line(),
		strdup(), strip_comments(), and trim_value().  New C
		preprocessor symbols: HAVE_OLDCODE, HAVE_PATTERNS,
		HAVE_RECOMP, and HAVE_REGEXP.  One of these should be
		defined at compile time; if none are, then
		HAVE_PATTERNS is the default.

		Since options can now be specified in initialization
		files, they each need negations so the command line
		can override values from an initialization file.

		Change all YES/NO flags to new type, YESorNO, for
		better type checking.

		Add do_more(), do_more_init(), and do_more_term(), for
		pausing during help output; a private version of
		screen paging is used instead of a pager invoked by
		system() for better portability across systems.  Set
		SCREEN_LINES to 0 at compile time to suppress this
		feature.

		In fix_title(), add code to brace upper-case letters
		for cases like:
			"X11" -> "{X11}"
			"Standard C Library" -> "Standard {C} Library"
			"C++ Book" -> "{C}++ Book"
		leaving
			"A xxx"
		unchanged.


[11-Nov-1992]	2.02
		Add bad_ISBN(), bad_ISSN(), check_ISBN(), and
		check_ISSN() for validation of ISBN and ISSN values.
		ISBN == "International Standard Book Number", and ISSN
		= "International Standard Serial Number".

		Add testisxn.bib and testisxn.bok to the test
		collection, with steps in the Makefile to run the
		test.

		Add support for embedded \" in Scribe value strings
		(forgotten in 2.01 revision); they are converted from
		\"x to {\"x}.


[10-Nov-1992]	2.01
		Add support for conversion of level-0 \"x to {\"x} and
		x"y to x{"}y in value strings.  Such input is illegal
		for BibTeX, and causes hard-to-find errors, since
		BibTeX raises an error at the line where it runs out
		of string collection space, rather than at the
		beginning of the collection point.


[06-Nov-1992]	2.00
		Add full Scribe .bib file input compatibility with
		-scribe command-line option.

		Add support for multiple .bib file arguments on
		command line, with new do_file() function to process
		them.

		Allow slash as well as hyphen for introducing
		command-line options on VAX VMS and IBM PC DOS.

		Add argument summary to help() (text extracted
		verbatim from the manual pages).

		Add new -delete-empty-fields, -keep-names,
		-no-parbreaks, -remove-OPT-prefixes, and -no-warnings
		command-line options and support code.

		Add new out_with_error() and out_with_parbreak_error()
		functions, and APPEND_CHAR() and EMPTY_STRING() macros
		to shorten and clarify coding.

		Add flush_inter_entry_space() function to standardize
		line spacing.

		Increase array sizes to MAX_TOKEN_SIZE (= MAX_TOKEN +
		3) to reduce array bounds checking in inner loops.

		Add additional file position tracking to enhance error
		localization (structures IO_PAIR and POSITION, and
		functions new_io_pair(), new_position(),
		out_position(), and out_status()).  Error messages are
		parsable by GNU Emacs M-x next-error (C-x `) when
		bibclean is run from Emacs by the command
		M-x compile<RET>bibclean foo.bib >foo.new

		Use arrays of constant strings for multiple string
		output via new function out_lines(), instead of multiple
		calls to fprintf().

		Add additional checking via check_chapter(),
		check_month(), check_number(), check_pages(),
		check_volume(), check_year(), and match_regexp().

		Supply implementation of memmove() library function
		missing from g++ 2.2.2 library.


[03-Oct-1992]	1.06
		Correct logic error in do_comma() that prevented correct
		recognition of @name(key = "value") where the last
		key/value pair did not have a trailing comma.

		Add C++ support.

		Add key_pair[] and entry_pair[] tables for
		standardization of letter case usage, and use the new
		NAME_PAIR type in fix_month().

		Update author address.

		Rename author() to show_author() to avoid shadowing
		global names.

		Fix two assignments of constant strings to char*
		pointers.

		Remove variable at_line_number which was defined, but
		never used.

[01-Aug-1992]	1.05
		Add -keep-initials switch support (thanks to Karl Berry
		<karl@cs.umb.edu>).  Internationalize telephone and FAX
		numbers.

[02-Jan-1992]	1.04
		Modify fix_title() to ignore macros.  Modify
		fix_author()) to ignore author lists with parentheses
		(e.g.  author = "P. D. Q. Bach (113 MozartStrasse,
		Vienna, Austria)").

[31-Dec-1991]	1.03
		Add fix_title() to supply braces around unbraced
		upper-case acronyms in titles, and add private
		definition of MAX().

[15-Nov-1991]	1.02
		Handle @String(...) and @Preamble(...), converting
		outer parentheses to braces.  Insert spaces after
		author and editor initials, and normalize names to
		form "P. D. Q. Bach" instead of "Bach, P. D. Q.".

[10-Oct-1991]	1.01
		Increase MAX_TOKEN to match enlarged BibTeX, and add
		check against STD_MAX_TOKEN.
		Output ISBN and ISSN in upper case.
		Always surround = by blanks in key = "value".

[19-Dec-1990]	1.00 (version number unchanged)
		Install Sun386i bug fix.

[08-Oct-1990]	1.00
		Original version.