2018-04-29  Bert Bos  <bert@w3.org>

	* dtd.hash: Don't include the arguments in the forward declaration
	of lookup_element(), because those arguments differ slightly
	depending on which version of gperf is used to generate dtd.c:
	"unsigned int" in gperf 3.0 vs "size_t" in gperf 3.1.

	* conficgure.ac: Check for libiconv with the AM_ICONV macro (from
	gettext) instead of AC_SEARCH_LIBS. (Advice from Christian
	Weisgerber, received via Anthony Bentley). That in turn requires
	adding config.rpath to the EXTRA_DIST in Makefile.am.

2018-02-23  Bert Bos  <bert@w3.org>

	* tests/index10.sh: Added a test for a term database.

	* hxindex.c: Replaced fgets() by getline(), to allow lines of
	arbitrary length in the term database.

2018-02-15  Bert Bos  <bert@w3.org>

	* tests/index9.sh: Added a test for hxindex's option -N.

	* tests/index5.sh: Updated for hxindex's option -N.

	* tests/cite5.sh: Added the example from hxmkbib.1 as a test.

	* tests/normalize12.sh: Added a test for the NOSCRIPT element.

	* hxindex.1: Added option -N.

	* hxmkbib.1: The example was missing some "%" signs.

	* hxmkdir.c: Read the template before reading the auxfile, so that
	a pipe like "hxcite -a auxfile bib template | hxmkbib -a auxfile
	bib" becomes possible. hxcite will have the time to finishe
	writing the auxfile before hxmkbib reads it.

2018-02-12  Bert Bos  <bert@w3.org>

	* hxindex.c: Added option -N.

	* dtd.hash: Since HTML5, the NOSCRIPT element is phrase-level
	instead of block-level.

2018-02-09  Bert Bos  <bert@w3.org>

	* Version 7.6 released.

	* selmatch.c: matches_sel() could read uninitialized memory
	(h->parent where h is the root of a tree) potentially leading to a
	SIGBUS. Thanks to Anthony J. Bentley for finding this bug.

2018-01-20  Bert Bos  <bert@w3.org>

	* Version 7.5 released.

	* hxcite.1: Spelling error.

	* hxincl.1: Spelling error.

	* url.c: Fixed unsafe way to call warnx(). Thanks to Robin
	Naundorf for these three corrections.

2017-12-08  Bert Bos  <bert@w3.org>

	* tests/normalize11.sh: Added to test for "main".

	* dtd.hash: Added the HTML5 element "main".

2017-12-07  Bert Bos  <bert@w3.org>

	* hxwls.c: Added support for HTML5 elements (source, audio, video)
	and for the classid and codebase attribute. Fixed an error: input
	has a src attribute, not href.

2017-11-24  Bert Bos  <bert@w3.org>

	* Version 7.4 released.

	* hxremove.1, hxselect.1: Updated.

	* tests/remove4.sh, tests/select18.sh, tests/select19.sh,
	tests/select20.sh, tests/select21.sh, tests/select22.sh,
	tests/remove5.sh tests/remove6.sh tests/remove7.sh,
	tests/select23.sh tests/select24.sh: Added.

	* selmatch.c: Now handles the ':not()' pseudo-class and
	comma-separated selectors. Fixed a bug in matching the ':root'
	pseudo-class.

	* hxremove.c, hxselect.c: parse_selector() has different
	arguments.

	* selector.c: Now parses the ':not()' pseudo-class and
	comma-separated selectors. Errors are now printed with a newline
	at the end. Added new functions to dump (serialize) a selector to
	a file, useful for debugging.

	* openurl.c: Functions now either return CURLcode or CURLMcode,
	but not some mixture.

2017-11-11  Bert Bos  <bert@w3.org>

	* dtd.hash: Added some HTML5 elements: source, track, wbr (because
	they are empty) and audio and video (as parents for source and
	track).

	* toc4.sh, normalize10.sh: Added tests for the <source> element.

	* Released version 7.3.

2017-10-16  Bert Bos  <bert@w3.org>

	* normalize9.sh: Added test for A inside FIGCAPTION.

	* dtd.hash: An A can now be inside a FIGCAPTION. Made source
	shorter and more readable with some macros.

2017-10-06  Bert Bos  <bert@w3.org>

	* Released version 7.2.

	* tests/extract1.sh: Added option -N to nc, to close the
	connection after EOF on stdin, because hxextract (with libcurl)
	seems to wait for the closing of the connection.

	* tests/unent1.ssh, tests/unent2.ssh, tests/unent3.ssh,
	tests/unent4.ssh, tests/unent5.ssh: Added tests for hxunent.

	* hxunentmain.c: Fixed bug reported by Laurent Martelli: When a
	line was longer than the 4096-byte buffer, an entity could be on
	the edge between two buffers and not be recognized.

2017-08-13  Bert Bos  <bert@w3.org>

	* extract1.sh: return code 77 (skip) rather than 0 when netcat is
	unavailable.

	* tests/pipe4.sh, tests/select01.sh, ..., tests/select17.sh,
	tests/unpipe6.sh: New tests.

	* scan.l: Accept non-ASCII characters in names (element names,
	attribute names, etc.)

	* hxselect.c, hxremove.c: Put their common code in
	selmatch.c. hxselect now uses the data structures from tree.c,
	just like hxremove.c. The non-element selector '::attr(...)' is
	now also supported.

	* selector.c, selmatch.c: Added support for more structural
	pseudo-classes (last-child, nth-last-of-type, etc.). Generalized
	support for backslash escapes. Support hexadecimal backslash
	escapes also outside ASCII range (such characters are converted to
	UTF-8). Bug fixes.

2016-10-17  Bert Bos  <bert@w3.org>

	* Released version 7.1.

	* textwrap.c, hxnormalize.c, normalize7.sh, normalize8.sh: Allow a
	line break before the final ">" of a tag. Line length (option -l)
	is now counted in characters, not bytes, at least if it is UTF-8.
	Added two tests for these new features.

2016-05-10  Bert Bos  <bert@w3.org>

	* hxunpipe.c: Added option -b to escape SGML delimiters.

2016-04-14  Bert Bos  <bert@w3.org>

	* Released version 7.0.

	* hxwls.c: Added option -a to output URLs in ASCII (using punycode
	and %-escaping if needed). Converting Internationalized Domain
	Names to ASCII (with punycode) requires libidn or libidn2.

2016-04-08  Bert Bos  <bert@w3.org>

	* wls3.sh, wls5.sh: Tests for hxwls (entity expansion, ASCII
	conversion)

	* hxwls.c: Expand entities in URLs, into ASCII characters if
	possible, or into %-escaped UTF-8 octets otherwise.

2016-03-10  Bert Bos  <bert@w3.org>

	* tree.c (main): dumptree() now takes a FILE argument.

2016-03-07  Bert Bos  <bert@w3.org>

	* hxnormalize.c: Attempt at inserting <![CDATA[...]]> into <style>
	and <script> when going from HTML to XML and removing it again
	when going from XML to HTML. (Problem: We know whether the output
	is XML or not, but not whether the input is supposed to be XML...)

2016-03-04  Bert Bos  <bert@w3.org>

	* hxindex.c: Added options -O and -X to restrict the elements to
	index.

2015-07-21  Bert Bos  <bert@w3.org>

	* tree.c: When a CDATA element (i.e., <script> or <style>) is
	opened, instruct the lexical scanner to only recognize the
	element's end tag and treat everything else as text. (Note that
	this matches HTML5, but not SGML: in SGML, any "</foo" or even
	"</>" would end the CDATA element.)

	* normalize3.sh: new test to check parsing of CDATA elements.

	* normalize4.sh: new test to check parsing of CDATA elements.

	* scan.l: Added an initial condition CDATA that can be started
	from the tree builder (tree.c) when a CDATA element is opened.

	* dtd.hash: Added a field to indicate that an element is defined
	as CDATA

2015-03-12  Bert Bos  <bert@w3.org>

	* Makefile.am: Added another test: normalize2.sh. Added header2.c
	to the sources for hxprintlinks. Use groff instead of man2html to
	make HTML versions of the man pages.

	* dtd.hash: RDFA, in contrast to HTML, says that <meta> may be
	used as phrase-level element as well as in the <head>. The HTML
	tree now allows meta as RDFA says, at the cost of less error
	correction.

2015-01-23  Bert Bos  <bert@w3.org>

	* hxprintlinks: Replaced the Perl version by a (better) C
	version. Added corresponding tests.

2015-01-22  Bert Bos  <bert@w3.org>

	* hxtoc.c: Added option -f (flatten) to remove element structure
	from headers when they are copied to the ToC.

2014-10-21  Bert Bos  <bert@w3.org>

	* Released version 6.9.

	* scan.l: Don't declare strndup() as a function if it is in fact a
	macro. (If _GNU_SOURCE is defined and the compiler is called with
	-O, strndup() may be a macro instead of a function.)

2014-10-06  Bert Bos  <bert@w3.org>

	* hxtoc.c: HTML5 now has a "header" element instead of an
	"hgroup."

	* dtd.hash: Added HTML5 elements header, aside, section, article,
	nav and figure. They currently behave very much like a div. Added
	figcaption, which is allowed inside figure.

2014-10-02  Bert Bos  <bert@w3.org>

	* extract1.sh: Explicitly use 127.0.0.1 (an IPv4 address) instead
	of localhost, because the latter may resolve to ::1 (an IPv6
	address), which recent libcurl will use, but nc by default only
	listens to IPV4 traffic.

2014-09-02  Bert Bos  <bert@w3.org>

	* hxnormalize.c: A system literal is only required in the DOCTYPE
	of an XML document if there also is a public literal.

2014-08-18  Bert Bos  <bert@w3.org>

	* types.c: removed modifier "inline" from min() and max()
	functions so that hxindex compiles also with clang. clang, unlike
	gcc, but in conformance with the C99 standard, considers inline
	functions as static function (i.e., not exported).

	* openurl.c: fixed bug, where wait_for_data() was meant to return
	a code of type CURLcode, but actually returned CURLMcode.

	* Added option -s (rewrite links to self) to hxcopy.

2014-07-21  Bert Bos  <bert@w3.org>

	* Makefile.am: "make install" now also installs the COPYING file
	in docdir (by default /usr/local/share/doc/html-xml-utils/)

2014-04-02  Bert Bos  <bert@w3.org>

	* hxnormalize.c: Option -x now also creates an (empty) system
	literal in the doctype if the input didn't have one.

	* scan.l: Also escape '<' and '>' in attributes. Allowed but not
	needed for SGML (and hence HTML4), but required by XML.

2014-03-14  Bert Bos  <bert@w3.org>

	* tests/extract1.sh: Now trying with and without a -p option to
	nc, as that option is required on some versions of nc and
	disallowed on others. Its absence caused the test to hang on some
	platforms. Also removed bash-specific syntax, for more
	portability.
	(Thanks to Anthony J. Bentley <anthony [at] cathet.us>)

2014-03-12  Bert Bos  <bert@w3.org>

	* Released 6.7

	* openurl.c: Include fopencookie.h instead of fopencookie.e

	* fopencookie.c: Don't use cexport to generate a header file
	fopencookie.e but use a static fopencookie.h. (We need to
	distribute this header file, but the .e file may be empty.)

	* bootstrap.sh: No need for "-I m4" anymore, because the option is
	now in the Makefile.am.

	* Makefile.am: Need to make cexport before .e files can be
	generated. (The fix for distcheck wasn't complete yet.) Don't
	generate fopencookie.e, but provide a static fopencookie.h and
	make most programs depend on that.

2014-03-11  Bert Bos  <bert@w3.org>

	* Released 6.6

	* Makefile.am: (Finally :-) ) fixed an error reported by "make
	distcheck" (files left after distclean), by depending .e files on
	cexport.c, rather than on the binary cexport.

	* Makefile.am: Added "ACLOCAL_AMFLAGS = -I m4" to make aclocal
	search configuration macros in the directory m4.

	* hxcite.c, hxindex.c, hxmkbib.c, genid.c: Removed duplicate
	inclusion of <errno.h>

	* openurl.c, hxcopy.c: Include <errno.h> instead of <sys/errno.h>
	so that they compile on OpenBSD. (Thanks to Anthony J. Bentley
	<anthony [at] cathet.us>)

2014-01-15  Bert Bos  <bert@w3.org>

	* Released 6.5

2013-10-08  Bert Bos  <bert@w3.org>

	* openurl.c: Removed a forgotten occurrence of "Boolean" in the
	code path for compilations without libcurl.

	* hxcite.1: Documented the use of %K lines.

	* hxcite.c: Now looks for labels also on %K lines.

2013-07-27  Bert Bos  <bert@w3.org>

	* Makefile.am: No longer uses $(wildcard) to find all tests, which
	doesn't work anymore in automake since version 1.14.

	* hkmkbib.c: Now uses getopt() so that invalid options no longer
	cause a crash.

	* hkwls.c: Now uses getopt() so that invalid options no longer
	cause a crash.

	* All programs now include "config.h" instead of <config.h>.
	Somebody found it easier to compile them that way.

	* configure.ac: Renamed configure.in to configure.ac. Removed
	acinclude.m4 and put all macro definitions in separate files in
	directory m4.

	* hxnum.c: now using getopt() to parse and verify the command
	line.

2013-07-25  Bert Bos  <bert@w3.org>

	* hxindex.c: fixed crash when argument to trim() is NULL.

	* acinclude.m4: removed definition of LIBCURL_CHECK_CONFIG, there
	is one already in /usr/share/aclocal/libcurl.m4


2013-04-19  Bert Bos  <bert@w3.org>

	* Released 6.4

2013-03-23  Bert Bos  <bert@w3.org>

	* toc1.sh: Added a first test for hxtoc.

	* Removed type Boolean in favor of bool from stdbool.h.

	* hxtoc.c: Implemented the HTML5 algorithm which says that a Hn
	that is the first child of a sectioning element (or of a HGROUP
	that is itself the first child of a sectioning element) doesn't
	have level n, but a level equal to the number of ancestors that
	are sectioning elements. Where a sectioning element is a DIV,
	SECTION, ARTICLE, ASIDE or NAV whose first child is either an
	HGROUP or an Hn.

2013-03-18  Bert Bos  <bert@w3.org>

	* hxtoc.c: option -d not only looks for DIV elements, but now also
	for ARTICLE, ASIDE, NAV and SECTION.

2013-01-09  Bert Bos  <bert@w3.org>

	* Released 6.3

	* scan.l: A lonely "<" is allowed in SGML (although not in
	XML). Accept it and turn it into a "&lt;".

2012-11-16  Bert Bos  <bert@w3.org>

	* tests: Added incl10.sh, incl11.sh and index5.sh.

	* hxindex.c: Fixed bug in command line parsing that caused hxindex
	not to accept the -b option. Added options -s and -u to control
	the title attribute that is added on links when -n is used. The
	title attribute now contains something like "section 3.2" or, if
	-b is also used, something like "Working with cork, section
	3.2". The format of the index database for option -i changed and
	now includes the document title. The program now uses errx()
	instead of errexit(), so that error messages include the program
	name.

2012-10-23  Bert Bos  <bert@w3.org>

	* openurl.c: Added some assert() statements.

	* hxincl.c: Added options -v and -M.

	* hxaddid.c hxcite.c hxcopy.c hxextract.c hxpipe.c hxref.c
	hxremove.c hxselect.c: Added option -v to print the version
	number.

	* hxtabletrans.c: Added.

2012-10-21  Bert Bos  <bert@w3.org>

	* selector.c: Renamed Root to RootSel to avoid clash with Root in tree.c

	* hxremove.c: Added

	* tree.c: Added some routines to add nodes to a tree without
	checking against the HTML DTD (useful for generic XML documents
	and used by hxremove)

2012-10-19  Bert Bos  <bert@w3.org>

	* scan.l: added set_yyin() and get_yyin_name() to store/retrieve
	the name of the file that is going to parsed. Also added the file
	name as a parameter to include_file(). This is used by hxincl.c

	* hxincl.c: Now remembers the name of the included file so that
	the path of a recursively included file is relative to the file
	that includes it, rather than to the top level file.

2012-10-18  Bert Bos  <bert@w3.org>

	* hxincl.c: Now allows %-delimited variables in the file name that
	is to be included.

2012-10-09  Bert Bos  <bert@w3.org>

	* extract1.sh: Added a "sleep 1" after starting the servers to
	allow them more time to start and open some sockets.

	* hxref.c: "@", "(", ")" and "_" are now treated as significant
	letters when comparing terms.

	* genid.c: "@" is now used in the generated ID replaced by the
	letters "at" and "_" is kept as is.

2012-10-02  Bert Bos  <bert@w3.org>

	* Released version 6.2

	* html.y: fixed type error in call() macro: void != NULL. (Bug
	found by Michael Tautschnig of Debian)

2012-08-31  Bert Bos  <bert@w3.org>

	* hxnormalize.c: When checking whether an optional end tag can be
	omitted, skip any text nodes with only spaces. In HTML mode, don't
	try to guess the attribute name for a given attribute value (i.e.,
	dont replace <table border> by <table border=border>), but leave
	the attribute value alone.

	* textwrap.c: fixed bug that created lines indented with spaces
	and no further content.

	* dtd.hash: insert line break before <meta>, <link> and <style>
	instead of after. Better if such an element immediately follows
	the <head>.

2012-05-17  Bert Bos  <bert@w3.org>

	* Released version 6.1

2012-02-14  Bert Bos  <bert@w3.org>

	* hxref.c: Added option -l to apply language-specific
	normalizations when searching for a string. Currently only for
	English and only to search for terms without a plural "s".

2011-12-27  Bert Bos  <bert@w3.org>

	* hxunpipe.c: Now also handles \# (which becomes &#)

2011-08-24  Bert Bos  <bert@w3.org>

	* openurl.c: Store HTTP status line in the Dictionary of response
	headers (under key ":status"). Added fopenurl3() to open a URL
	with another method than GET. Use libcurl's built-in support for
	Content-Encoding. fopenurl() now takes an int *status parameter
	for better error reporting. Added http_strerror() function. More
	error checking and error reporting.

	* hxaddid.c hxcopy.c hxcount.c hxextract.c hxincl.c hxindex.c
	hxmultitoc.c hxname2id.c hxnormalize.c hxnsxml.c hxnum.c hxpipe.c
	hxprune.c hxref.c hxtoc.c hxunpipe.c hxunxmlns.c hxwls.c
	hxxmlns.c: () now takes an int *status parameter.

2011-08-23  Bert Bos  <bert@w3.org>

	* fopencookie.c: Avoid compiler warnings by wrapping functions
	calls in other functions calls that do typecasts.

	* openurl.c: Added a callback function to be called by exit(3) and
	which cleans libcurl connections. Also check if a new connection
	is closed immediately at the first action inside
	fopenurl2(). Return NULL in that case. (Probably the target
	doesn't exist.)

2011-08-23  Bert Bos  <bert@w3.org>

	* acinclude.m4: Added LIBCURL_CHECK_CONFIG

	* configure.in: Added LIBCURL_CHECK_CONFIG and check for fopencookie()

	* fopencookie.c: New file: implemented Linux-style fopencookie()
	with BSD-style funopen()

	* Makefile.am: Added fopencookie.c to most of the programs. Added
	libcurl flags.

	* hxunpipe.c: Added some {} to avoid a compiler warning.

	* openurl.c: Added an alternative implemention of this module
	using libcurl.

2011-08-17  Bert Bos  <bert@w3.org>

	* dtd.hash: No newline should be created before or after the
	BUTTON element, it is an inline element.

2011-07-10  Bert Bos  <bert@w3.org>

	* hxunpipe.c: Handle \ddd (where d is octal) as output by onsgmls.
	Handle input lines of arbitrary length. Remove unused option -x.

	* hxnormalize.c: Don't add extra spaces in attribute values.

2011-07-01  Bert Bos  <bert@w3.org>

	* hxnormalize.c: Added option -L to remove redundant lang and
	xml:lang attributes.

2011-04-07  Bert Bos  <bert@w3.org>

	* hxnormalize.c: Magic comments now recognized also inside
	elements with mixed content (except preformatted)

	* Now also handles a DOCTYPE with neither an FPI nor a system ID.

2011-03-17  Bert Bos  <bert@w3.org>

	* hxindex.c: Added option -r to suppress the removal of trailing
	whitespace and punctuation. Added logic to keep a final '.' from
	being removed if it is part of an abbreviaton or ellipsis.

2011-01-18  Bert Bos  <bert@w3.org>

	* connectsock.c: now uses getaddrinfo(), which handles IPv6 addresses.

	* url.c: regexp pattern now also deals with IPv6 addresses
	("http://[1234::1]/etc.")

2010-11-18  Bert Bos  <bert@w3.org>

	* hxcite.1: Added sections for example, bugs and history; improved
	wording.

	* hxcite.c: Updated some comments. Corrected usage message.

2010-11-17  Bert Bos  <bert@w3.org>

	* Released 5.8.

	* hxcite.c: Added option -c to no longer recognize [[...]] inside
	comments.

2010-07-12  Bert Bos  <bert@w3.org>

	* hxnsxml.c: Added hxnsxml as the complement to hxxmlns.

2010-06-24  Bert Bos  <bert@w3.org>

	* unent.hash: Entity "apos" was missing.

2010-05-04  Bert Bos  <bert@w3.org>

	* Added option -c to hxnormalize to make some comments start on a
	new line.

2010-04-27  Bert Bos  <bert@w3.org>

	* hxindex.c: Fixed bug: char replaced by int, because char is
	unsigned on PowerPC.

	* Released 5.7.

	* openurl.c: Fixed bug: only call dict_delete() if the dictionary
	actually exists.

	* xhref.c: Fixed bug: after wrapping an element in an A, we still
	need to look at its sister elements for more cross-references.

2010-04-26  Bert Bos  <bert@w3.org>

	* Released 5.6.

	* tree.c: Added wrap_elt() function. wrap_content() no longer
	converts element name to lowercase. (But wrap_contents() isn't
	used anywhere anymore.)

	* hxincl.c: Fixed bug: only look in substitutions dictionary if
	the dictionary has been initialized.

	* dict.c: Added assert() to ensure that Dictionary arguments are
	not NULL.

	* hxref.c: The <a> element is now put around the element to
	cross-link, instead of inside it.

2010-01-15  Bert Bos  <bert@w3.org>

	* hxindex.c: Now uses iconv to convert from UTF-8 to wchar_t
	strings and then uses setlocale and wcscoll to sort index terms
	according to user's language.

2009-12-07  Bert Bos  <bert@w3.org>

	* xml2asc.c: Added full error checking for invalid UTF-8.

2009-11-09  Bert Bos  <bert@w3.org>

	* scan.l: unquoted attribute values may now even contain '='. 

2009-10-05  Bert Bos  <bert@w3.org>

	* xml2asc.c: now returns a non-zero exit code if the input was not
	correct UTF-8.

2009-09-14  Bert Bos  <bert@w3.org>

	* hxnormalize.c: Added option -s.

2009-08-14  Bert Bos  <bert@w3.org>

	* hxcopy.c: URL references to the document itself, i.e., those
	that are of the form "", "#foo", and "?foo", are no longer
	rewritten.

2009-07-15  Bert Bos  <bert@w3.org>

	* Released 5.4

	* url.c: Bug in removal of /./ fixed. Now leaves one / instead of
	none.

2009-03-25  Bert Bos  <bert@w3.org>

	* hxindex.c: Bug if the first index term has no alphanumeric
	characters, so that it sorts to be equal to globalprevious. Fixed
	by making globalprevious a non-empty, but unlikely string.

2009-03-08  Bert Bos  <bert@w3.org>

	* hxnormalize.c: The characters '_' and ':' in attributes *do*
	require quote marks.

2009-03-02  Bert Bos  <bert@w3.org>

	* hxunentmain.c Error message now refers to the right command line
	argument.

	* hxincl.c: Added option -s to override which files are included.

2009-02-12  Bert Bos  <bert@w3.org>

	* textwrap.c: Buffer is now allocated on heap, so line length
	limit is now unlimited (limited by memory), instead of 32K.

2009-02-09  Bert Bos  <bert@w3.org>

	* openurl.c: Different way to check max # of redirects. Added a
	few assignments to errno to get slightly better than random error
	messages.

2009-01-28  Bert Bos  <bert@w3.org>

	* Released 5.3

	* hxcopy.c: Fixed bug in calculating path_from_url_to_url: Added a
	"/" at the end of the name of the current directory, because
	URL_s_absolutize needs to know the segment is indeed a directory.

2009-01-19  Bert Bos  <bert@w3.org>

	* hxtoc.c: Removed some unnecessary "EXPORT". Removed some
	duplicate code and used routines from types.c instead.

	* types.c: Added only_space(), pairlist_unset(), pairlist_set()
	and pairlist_get().

	* tree.c: get_attrib() now returns the attribute value, rather
	than a Boolean. Removed some duplicate code by using routines in
	types.c instead.

2009-01-15  Bert Bos  <bert@w3.org>

	* tree.c: Added function delete_attrib()

	* hxindex.c: Added option -f to remove used title attributes and
	magic comments.

	* genid.c: Now takes whole contents (up to MAXIDLEN) to construct
	an ID from, instead of only the first word.

2009-01-12  Bert Bos  <bert@w3.org>

	* url.c: replaced a comparison (<) that always fails when char is
	unsigned (as is the case, e.g., in gcc on PowerPC-based systems)
	by a bitwise and (&).

2009-01-09  Bert Bos  <bert@w3.org>

	* Released version 5.2.

	* hxnum.c: Removed some routines to test class attributes and
	re-used those from class.c instead.

	* hxindex.c: Added option -n to use section numbers as anchor text
	in the index instead of "#". Also improved sorting: Entities (such
	as "&lt;" or "&amp;") are now skipped when comparing terms. (As
	the manual explains, for best sorting, other entities should be
	expanded first with hxunent(1).)

2008-12-16  Bert Bos  <bert@w3.org>

	* tree.c: Fixed bug in lookup(), added missing '\0' at end of
	string for long tag names.

2008-12-15  Bert Bos  <bert@w3.org>

	* scan.l: Fixed bug: "<_foo" was tokenized as TEXT instead of
	START, because {data} didn't exclude a "<" followed by a "_".

2008-12-09  Bert Bos  <bert@w3.org>

	* url.c: Rewrote remove_dot_segments() to no longer remove ../
	from the start of a path.

	* openurl.c: Local files can be opened with mode "w" (remote ones
	still not)

	* Makefile.am: Added hxcopy utility.

2008-11-29  Bert Bos  <bert@w3.org>

	* scan.l: Fixed regexp for cdata.

2008-11-28  Bert Bos  <bert@w3.org>

	* hxincl.c: Added option -f.

2008-11-20  Bert Bos  <bert@w3.org>

	* Released version 5.1.

	* Updates to the manual pages: the "hx" prefix was missing from
	many occurrences of commnd names in the manual pages. (Thanks
	again to Daniel Leidert for pointing that out.)

2008-11-05  Bert Bos  <bert@w3.org>

	* Most programs renamed with a "hx" prefix.

	* hxtoc.c: Added option -d to use the DIV that encloses a header
	as the target of the links in the ToC, instead of the header
	itself.

2008-08-17  Bert Bos  <bert@w3.org>

	* Released version 4.7.

	* unentmain.c: Added information for gcc that usage() never
	returns.

	* url.c: Inverted condition (caused by incorrect replacement of
	strncmp() by hasprefix())

	* unentmain.c: Missing #include <ctype.h>

	* pipe.c: Now using lineno instead of yylineno.

	* Makefile.am: Replaced %-rules by old-style rules, added missing
	built sources and reordered them, so they are built in the right
	sequence. (Thanks to Daniel Leidert for these changes)

2008-08-10  Bert Bos  <bert@w3.org>

	* openurl.c: Now handles redirects and can store HTTP headers.

	* dict.c: New module to store HTTP headers (or any other key/value
	pairs, where both key and value are strings).

	* headers.c: New module to parse HTTP headers.

2008-07-30  Bert Bos  <bert@w3.org>

	* genid.c: made the search for existing IDs case-insentive. That's
	necessary for HTML. It wastes a few IDs for XML, but not too much.

2008-07-07  Bert Bos  <bert@w3.org>

	* toc.c: Added option -c to allow class=toc to be replaced by
	another class name.

2008-02-12  Bert Bos  <bert@w3.org>

	* Released version 4.5

	* Fixes to man pages, fix to uncdata (suppressed an incorrect
	newline in the output, new AC_INIT macro in configure.in, typo in
	README. (Thanks to Daniel Leidert)

2008-01-24  Bert Bos  <bert@w3.org>

	* index.c: The links now have a TITLE attribute for better
	accessibility.

2007-12-11  Bert Bos  <bert@w3.org>

	* scan.l: Now accepts "<![CDATA[...]>".

2007-10-23  Bert Bos  <bert@w3.org>

	* url.c: Query component is now separate from path component. 
	Making relative URLs absolute is now conformant to RFC 3986. This
	fixes in particular the incorrect output of wls.

2007-04-25  Bert Bos  <bert@w3.org>

	* genid.c: fixed bug with string not ending with \0 if len is 0.

2007-04-18  Bert Bos  <bert@w3.org>

	* unent.hash: Added option -f to "fix" unrecognized entities by
	replacing their & by &amp;

2007-03-22  Bert Bos  <bert@w3.org>

	* toc.c: DFN elements are no longer copied to the ToC, only their
	content is.

2007-03-15  Bert Bos  <bert@w3.org>

	* Released version 3.9

	* tree.c: fixed bug in create(), which left t->sister uninitialized

2006-10-04  Bert Bos  <bert@w3.org>

	* Released version 3.8

	* toc.c: Now allows spaces around "toc" in "<!--toc-->"

	* index.c: No longer destroys the title attribute.

2006-08-10  Bert Bos  <bert@w3.org>

	* xmlns.c: A local attribute will not get "{}" prefixed.

	* types.c: If strapp() is called with a NULL as first arg, a new
	string will be allocated on the heap.

	* index.c: Better sorting of terms with sub-terms.

	* addid.c: Made element matching case-sensitive when -x is used.

	* addadd.1: Removed reference to non-existent -a option.

2005-12-16  Bert Bos  <bert@Candy.local>

	* xref.c: If the element to link is a <span>, the span is replaced
	by an <a>, instead of the <a> being added, as for other kinds of
	elements.

	* The string type is now an array of chars instead of unsigned chars.

2005-11-09  Bert Bos  <bert@w3.org>

	* xmlns.c: An element without a Namespace prefix and without a
	default Namespace is no longer reported as an error.

	* xref.c: <cite> is no longer treated as en element whose content
	can be cross-referenced.

	* dtd.hash: No empty lines between <link> elements when
	normalizing. <Textarea> is treated like <pre>.

	* Makefile.am: Added unxmlns.

	* unxmlns.c: new program, undoes the effect of xmlns, i.e.,
	converts names like <{namespace-url}foo> to <foo
	xmlns="namespace-url">

	* pipe.c: No longer treats attribute "id" specially, but
	recognizes "xml:id" and gives it a type of TOKEN instead of CDATA.

2005-05-05  Bert Bos  <bert@w3.org>

	* Published version 3.7.

	* index.c: Now accepts a list of class names on the command line
	for elements that also will be added to the index, in addition the
	standard ones. Thanks to Francesco Cosoleto <cosoleto.francesco
	[at] sitoverde.com>

	* textwrap.c: The marker for which spaces to keep is now a
	character 0x1 instead of 0x80, because 0x80 can occur in UTF-8
	encoded XML data. 0x1 can occur as well, but is less likely. Need
	to fix the algorithm better than this...

2005-04-27  Bert Bos  <bert@w3.org>

	* normalize.c: Attribute values are printed without quotes when
	possible.

2005-03-11  Bert Bos  <bert@w3.org>

	* extract.1, printlinks.1, uncdata.1, xselect.1: new manpages,
	thanks to Beverly Davis <ponyfeathers [at] mailhaven.com>

2005-01-18  Bert Bos  <bert@w3.org>

	* multitoc.c: class attributes from header tags are now copied to
	generated LI tags (thanks to howcome)

2005-01-09  Bert Bos  <bert@w3.org>

	* scan.l: the scanner didn't accept the curly braces in the output
	of xmlns. Not sure what was wrong with the old regexp for "name",
	but after simplifying it, it worked...

2004-12-20  Bert Bos  <bert@w3.org>

	* name2id.c: new program: moves ID= and NAME= from A elements to
	their parents.

2004-11-08  Bert Bos  <bert@w3.org>

	* addid.1: added (thanks to James Rowe <Jay [at]
	jnrowe.ukfsn.org>)

	* incl.1: description corrected (thanks to James Rowe <Jay [at]
	jnrowe.ukfsn.org>)

2004-09-07  Bert Bos  <bert@w3.org>

	* toc.c: now copies any class attribute found the heading elements
	to the generated LI element in the ToC.

2004-09-01  Bert Bos  <bert@w3.org>

	* cite.c: Added %m as a possible part of the template (to stand
	for the value of the -m option), made cite recognize {{...}} as
	well as [[...]] (but without expanding it), and changed the
	default template to generate <!--{{%m%L}}-->, so that cite can be
	run again on it's own output.

	* Released version 3.4.

	* normalize.c: the check if an endtag was needed omitted to test
	for other things than text or an element. But a comment or procins
	can also follow an element (and requires that the end tag is
	inserted).

2004-08-30  Bert Bos  <bert@w3.org>

	* normalize.c, tree.c: an end tag cannot be omitted if the next
	element is a valid child of the current one. E.g., the </p> cannot
	be omitted in <p>...</p><ins>...</ins>. To test for this in
	normalize.c, the existing static function has_parent() in tree.c
	had to become exported.

2004-08-29  Bert Bos  <bert@w3.org>

	* pipe.c: bug: when & is *not* followed by #, it should still be
	printed.

2004-08-27  Bert Bos  <bert@w3.org>

	* Published version 3.2

	* types.s: fixed bug in eq() macro: missing parentheses caused
	some arguments to be misinterpreted.

	* pipe.c and unpipe.c: pipe outputs "&#" as "\#" (just like
	nsgmls) and unpipe translates it back. unpipe now also accepts a
	"C" line.

2004-08-10  Bert Bos  <bert@w3.org>

	* xref.c: TITLE attribute of DFN element was normalized too soon,
	which removed any "|" that was needed later.

2004-04-29  Bert Bos  <bert@w3.org>

	* Released version 3.0

	* cite.c: Fixed bug: after a [[...]] without a marker, the program
	would not look for more [[...]] on the same line.

2004-04-28  Bert Bos  <bert@w3.org>

	* unent.hash now considers a "&" that is not followed by "#" or an
	alphanumeric character to mean the same as "&amp;" (as required by
	SGML, although invalid in XML)

2004-04-26  Bert Bos  <bert@w3.org>

	* textwrap.c: increased MAXLINE, since I have an application that
	uses 20000-character lines.

	* scan.l: Scanner now returns the omitted ">" in "<foo<bar>",
	since this is valid in SGML (although not in XML, but we're not
	validating anyway).

2004-04-22  Bert Bos  <bert@w3.org>

	* Released version 2.9.

	* mkbib.c: If a field ends with "." and the template would put a
	"." after it, only one of the two periods is printed. Added option
	-n and -r, to specify how many authors are printed at most and
	what string is printed instead if there are more. Defaults are 3
	and "et al."

2004-04-21  Bert Bos  <bert@w3.org>

	* cite.c, mkbib.c: increased size of hash table from 200 to 4096.

2003-12-02  Bert Bos  <bert@w3.org>

	* normalize.c: added logic to avoid that an end tag is omitted if
	there is text after it.

2003-09-15  Bert Bos  <bert@w3.org>

	* tree.c: if the currently open element doesn't accept text,
	append_text no longer starts by closing all open elements that
	have omittable end tags and don't accept text, but only closes as
	many as needed to arrive at one that accepts the text's preferred
	parent (which is P in HTML). This fixes a bug that text inside
	BODY closes the BODY and HTML and then reopens them again to
	insert a P. It now stops before closing BODY, because BODY will
	accept the P.

2003-09-04  Bert Bos  <bert@w3.org>

	* incl.c: printed one "?" too many at the end of a PI.

2003-08-11  Bert Bos  <bert@w3.org>

	* xref.c: now splits the key into |-separated parts, just like
	index.c does.

2003-08-07  Bert Bos  <bert@w3.org>

	* addid.c: removed reference to search.h, since it is not used.

	* mkbib.c, xref.c, cite.c: now use our own implementation of hash
	tables if search.h does not exist.

2003-07-15  Bert Bos  <bert@w3.org>

	* mkbib.c: fixed bug that caused last entry in the database to not
	be properly stored in the hash table.

2003-04-09  Bert Bos  <bert@w3.org>

	* Renamed search.h to search-freebsd.h and arranged for it to be
	used only if the system doesn't have a search.h of its own.

	* Added sources for twalk, tfind and tsearch.

	* Added search.h and adapted configure.in.

	* Changed export.h to quote the argument of EXPORTDEF, to deal
	with the C preprocessor in gcc 3.2, which otherwise inserts extra
	spaces. Adapted cexport.c to unquote the argument of EXPORTDEF
	again before writing it to a .e file.

2003-03-12  Bert Bos  <bert@w3.org>

	* configure.in: removed some unneeded rules (LN_S, MAN2HTML,
	ERROR_AT_LINE, REALLOC) and dropped requirement from autocon 2.57
	to autocon 2.52

2003-01-21  Bert Bos  <bert@bert.inria.fr>

	* heap.c: changed the __FILE__ magic so that it compiles with
	gcc3. (It gives a harmless warning still :-( )

2002-11-06  Bert Bos  <bert@w3.org>

	* scan.l: added recognition for UTF-8 Byte Order Mark at start.
	(The BOM is subsequently ignored, we don't currently do anything
	with the knowledge that the file is UTF-8.)

2002-10-30  Bert Bos  <bert@w3.org>

	* asc2xml, xml2asc, xmlrecode: removed xmlrecode and made asc2xml
	and xml2asc independent programs. They were so small that
	symlinking made no sense.

	* tree.c: added routine need_stag() that check if the start tag is
	required and added build_path() that tries to build a path of
	omitted start tags from the current tree to the element that is to
	be added.

	* dtd.hash: added a column for omittable start tags, to help the
	heuristics in html_push (tree.c) to infer the intended structure
	better.

2002-10-29  Bert Bos  <bert@w3.org>

	* xselect.c: now supports case-insensitive matching

	* xref.c: when finding a match, now ignores all characters except
	letters, digits and dashes.

	* uncdata.c: Fixed bug: forgot to print "<!" when starting a
	comment.

	* toc.1: Added option -t

	* Released version 2.4

	* dtd.hash: added table as possible parent for tr. It's not
	correct, but it avoids an incorrect correction with tr inside a
	nested table: it closed the nested table and considered the tr to
	be part of the outer table. A better solution is needed...

2002-02-20  Bert Bos  <bert@w3.org>

	* uncdata.c: new program: removes CDATA sections from XML files.

2002-02-18  Bert Bos  <bert@w3.org>

	* xselect.c: Added options -s (separator) and -i
	(case-insensitive). Added some documentation.

2002-02-14  Bert Bos  <bert@w3.org>

	* selector.c: fixed bug in recognition of attribute selectors:
	pointer was incremented before the character was stored, instead
	of the reverse.

2002-02-05  Bert Bos  <bert@w3.org>

	* Released version 2.2.

	* cite.c: Now accepts a -m option to flag references. This allows
	to find differently flagged references in different passes.

	* multitoc.c: Fixed two bugs: a missing ! before strncmp and
	missing printf(">"), which caused headings with child elements to
	be copied incorrectly.

2001-09-25  Bert Bos  <bert@w3.org>

	* index.c: Fixed bug: no longer tries to index empty elements,
	such as "<dfn></dfn>"

2001-08-01  Bert Bos  <bert@w3.org>

	* selector.c: Bug fixed: some occurrences of START_EQ were renamed
	to EQ, but not all.

	* select.c: renamed to xselect.c, because select is a built-in
	command of certain shells.

2001-07-17  Bert Bos  <bert@w3.org>

	* unent.hash: fixed bug that copied a whole line instead of a
	single unexpanded entity. Added option -b ("built-in") to leave
	standard entities (amp, gt, lt, quot) untocuhed.

	* Released version 2.0

	* normalize.c: Added option -d to omit the doctype from the
	output.

2001-07-16  Bert Bos  <bert@w3.org>

	* selector.c: Fixed bug in parsing of "[name]"; fixed missing
	case in attribs_to_string()

	* select.c: Now implements :lang and :nth...(an+b)

2001-07-10  Bert Bos  <bert@w3.org>

	* Released version 1.8

	* mkbib.c: fixed bug in read_entries(). On Linux, calling
	hsearch() with an action of ENTER replaces any existing data
	with new data, if it is different, On Solaris the existing record
	is kept. The routine now puts the data in the existing record.

2001-07-09  Bert Bos  <bert@w3.org>

	* Released version 1.7

	* selector.c: Added (types and a parser for CSS selector)

	* select.c: Added (implementation of CSS selectors)

2001-05-23  Bert Bos  <bert@w3.org>

	* Released version 1.6

2001-05-21  Bert Bos  <bert@w3.org>

	* toc.c: In expand(), if keep-anchors is false, IDs are now not
	copied either.

2001-04-19  Bert Bos  <bert@w3.org>

	* textwrap.c: Fixed bug when space fell exactly at end of line.
	But logic still is not perfect, since it creates spaces (indents)
	on empty lines.

2001-03-22  Bert Bos  <bert@w3.org>

	* toc.c: </li> was not printed, even if xml was True

2001-02-04  Bert Bos  <bert@w3.org>

	* printlinks: added. 

2000-09-14  Bert Bos  <bert@w3.org>

	* normalize.c: Fixed bug that caused text not to be preformatted
 	inside an element nested inside a PRE.

	* textwrap.c: Fixed bug that caused lines not to be broken at all
 	if there was no break point before the target length.

2000-09-04  Bert Bos  <bert@w3.org>

	* xref.c: now skips quotes when comparing terms

2000-08-29  Bert Bos  <bert@w3.org>

	* Released version 1.3

2000-08-26  Bert Bos  <bert@w3.org>

	* index.c: termcmp() now correctly compares keys that end with
 	punctuation.

	* normalize.1: now describes the -e (add endtags) option.

	* xref.c: now uses title attribute in instances as well

2000-08-21  Bert Bos  <bbos@lanalana.inria.fr>

	* Released version 1.2

	* All occurrences of "DOCTYPE" are now in uppercase.

2000-08-20  Bert Bos  <bert@w3.org>

	* tree.c: Now checks if there is a parent open before popping
	elements from the stack.

	* textwrap.c: Now flushes at \n or \r in preformatted, because
	long preformatted elements could too easily overflow the buffer.

	* html.y: Added some const's to formal parameters

	* genid.c: minlen now works even if there are child elements; no
 	longer skips '-' and '.' when it generates an ID; no longer
	generates zero-length IDs.

	* cexport.c: Now recognizes quoted file names after "#line"

	* class.c: Put contains() and has_class() in new file

	* Makefile.am: Added addid and extract

Version 1.1 (2 August 2000)

    Better (?) heuristic for closing and generating elements in
    html_push() in tree.c. It no longer closes and re-opens the <html> 
    element in case both </head> and <body> are missing.

    Different path for bash (/usr/local/bin/bash instead of
    /usr/local/gnu/bin/bash) in cite-mkbib.

    cite-mkbib was missing from the tar file.

    Program incl changed. It now looks for <!--include "file"-->.
    Because of that, a callback endincl() was added in html.y

    Some programs now interpret argument "-" as meaning "read from
    standard input."

    Fixed bug in toc: it didn't correctly guard against re-use of IDs, 
    since it stored a random string rather than an ID in the binary
    tree.

    Added some handy functions to types.c: stapp(), chomp(), min(),
    max().