build: Remove expat dependency

* DEPENDENCIES: Suggest libxml2 instead of expat.
* gnulib-local/lib/markup.c: New file.
* gnulib-local/lib/markup.h: New file.
* gnulib-local/modules/markup: New file.
* autogen.sh (GNULIB_MODULES_LIBGETTEXTPO): Add markup module.
* gettext-tools/configure.ac: Remove checks for expat.
* gettext-tools/gnulib-lib/.gitignore: Ignore modules pulled by
gnulib-tool due to the markup module usage.
* gettext-tools/gnulib-tests/.gitignore: Likewise.
* gettext-tools/libgettextpo/.gitignore: Likewise.
* gettext-tools/libgettextpo/Makefile.am (libgettextpo_la_AUXSOURCES):
Remove ../src/libexpat-compat.c.
(libgettextpo_la_LDFLAGS): Remove @LTLIBEXPAT@.
* gettext-tools/src/Makefile.am (noinst_HEADERS): Remove
libexpat-compat.h.
(libgettextsrc_la_SOURCES): Remove libexpat-compat.c.
(libgettextsrc_la_LDFLAGS): Remove @LTLIBEXPAT@.
* gettext-tools/src/format-kde-kuit.c: Use functions from markup.h, when
the file is being compiled as part of libgettextpo.  Otherwise use
libxml2.
* gettext-tools/src/libexpat-compat.c: Remove.
* gettext-tools/src/libexpat-compat.h: Remove.

15 files changed, 1899 insertions, 586 deletions

diff --git a/.gitignore b/.gitignore
index 11be874..2fdf4f6 100644
--- a/.gitignore
+++ b/.gitignore
@@ -43,6 +43,7 @@ Makefile
 /gettext-tools/examples/ChangeLog
 /gettext-tools/po/ChangeLog
 
+/build-aux/ar-lib
 /build-aux/git-version-gen
 /build-aux/gitlog-to-changelog
 /build-aux/snippet
diff --git a/DEPENDENCIES b/DEPENDENCIES
index e11eb94..4a26690 100644
--- a/DEPENDENCIES
+++ b/DEPENDENCIES
@@ -28,15 +28,15 @@ The following packages should be installed before GNU gettext is installed
   + If it is installed in a nonstandard directory, pass the option
     --with-ncurses-prefix=DIR or --with-libtermcap-prefix to 'configure'.
 
-* expat 1.95 or newer
+* libxml
   + Recommended.
-    Needed for 'xgettext', so that it can parse Glade XML files.
+    Needed for 'xgettext' and 'msgfmt', so that it can parse XML files.
   + Homepage:
-    http://expat.sourceforge.net/
+    http://xmlsoft.org/
   + Download:
-    http://sourceforge.net/project/showfiles.php?group_id=10127
+    ftp://xmlsoft.org/
   + If it is installed in a nonstandard directory, pass the option
-    --with-libexpat-prefix to 'configure'.
+    --with-libxml2-prefix to 'configure'.
 
 * A Java runtime and compiler (e.g. GNU gcj or kaffe).
   + Recommended.
diff --git a/autogen.sh b/autogen.sh
index d498be8..604a059 100755
--- a/autogen.sh
+++ b/autogen.sh
@@ -342,6 +342,7 @@ if ! $skip_gnulib; then
       hash
       iconv
       libunistring-optional
+      markup
       minmax
       open
       ostream
diff --git a/gettext-tools/configure.ac b/gettext-tools/configure.ac
index fdf156e..dba0775 100644
--- a/gettext-tools/configure.ac
+++ b/gettext-tools/configure.ac
@@ -291,25 +291,6 @@ AH_BOTTOM([
 #endif
 ])
 
-dnl Check for the expat XML parser.
-dnl On operating systems where binary distribution vendors are likely to
-dnl ship both gettext and expat, we use dynamic loading to avoid a hard
-dnl dependency from gettext to expat.
-case "$host_os" in
-  linux*)
-    AC_DEFINE([DYNLOAD_LIBEXPAT], [1],
-      [Define to 1 if libexpat shall be dynamically loaded via dlopen().])
-    LIBEXPAT="-ldl"
-    LTLIBEXPAT="-ldl"
-    AC_SUBST([LIBEXPAT])
-    AC_SUBST([LTLIBEXPAT])
-    ;;
-  *)
-    AC_LIB_HAVE_LINKFLAGS([expat], [],
-      [#include <expat.h>], [XML_ExpatVersion();])
-    ;;
-esac
-
 dnl Check for nm output filter that yields the exported symbols.
 gt_GLOBAL_SYMBOL_PIPE
 
diff --git a/gettext-tools/gnulib-lib/.gitignore b/gettext-tools/gnulib-lib/.gitignore
index 70ff6a8..5ff51b2 100644
--- a/gettext-tools/gnulib-lib/.gitignore
+++ b/gettext-tools/gnulib-lib/.gitignore
@@ -10,7 +10,14 @@
 /uniwidth
 
 # Files brought in by gnulib-tool:
+
+
+
+# Files generated by gperf:
+# Files generated by moopp:
+# Files generated by the autotools:
 /Makefile.gnulib
+/Makefile.in
 /acl-errno-valid.c
 /acl-internal.c
 /acl-internal.h
@@ -90,13 +97,21 @@
 /fcntl.in.h
 /fd-hook.c
 /fd-hook.h
+/fd-ostream.c
+/fd-ostream.h
 /fd-ostream.oo.c
 /fd-ostream.oo.h
 /fd-safer-flag.c
 /fd-safer.c
+/fd_ostream.priv.h
+/fd_ostream.vt.h
 /file-has-acl.c
+/file-ostream.c
+/file-ostream.h
 /file-ostream.oo.c
 /file-ostream.oo.h
+/file_ostream.priv.h
+/file_ostream.vt.h
 /filename.h
 /findprog.c
 /findprog.h
@@ -132,6 +147,8 @@
 /gl_anylinked_list2.h
 /gl_array_list.c
 /gl_array_list.h
+/gl_linked_list.c
+/gl_linked_list.h
 /gl_linkedhash_list.c
 /gl_linkedhash_list.h
 /gl_list.c
@@ -142,18 +159,31 @@
 /glibconfig.in.h
 /hash.c
 /hash.h
+/html-ostream.c
+/html-ostream.h
 /html-ostream.oo.c
 /html-ostream.oo.h
+/html-styled-ostream.c
+/html-styled-ostream.h
 /html-styled-ostream.oo.c
 /html-styled-ostream.oo.h
+/html_ostream.priv.h
+/html_ostream.vt.h
+/html_styled_ostream.priv.h
+/html_styled_ostream.vt.h
 /iconv.c
 /iconv.in.h
 /iconv_close.c
 /iconv_open-aix.gperf
+/iconv_open-aix.h
 /iconv_open-hpux.gperf
+/iconv_open-hpux.h
 /iconv_open-irix.gperf
+/iconv_open-irix.h
 /iconv_open-osf.gperf
+/iconv_open-osf.h
 /iconv_open-solaris.gperf
+/iconv_open-solaris.h
 /iconv_open.c
 /iconveh.h
 /ignore-value.h
@@ -180,6 +210,8 @@
 /malloca.c
 /malloca.h
 /malloca.valgrind
+/markup.c
+/markup.h
 /mbchar.c
 /mbchar.h
 /mbiter.c
@@ -210,8 +242,12 @@
 /obstack.h
 /open.c
 /opendir.c
+/ostream.c
+/ostream.h
 /ostream.oo.c
 /ostream.oo.h
+/ostream.priv.h
+/ostream.vt.h
 /pathmax.h
 /pipe-filter-aux.c
 /pipe-filter-aux.h
@@ -319,8 +355,12 @@
 /strstr.c
 /strtol.c
 /strtoul.c
+/styled-ostream.c
+/styled-ostream.h
 /styled-ostream.oo.c
 /styled-ostream.oo.h
+/styled_ostream.priv.h
+/styled_ostream.vt.h
 /sys_select.in.h
 /sys_socket.in.h
 /sys_stat.in.h
@@ -329,10 +369,18 @@
 /sys_wait.in.h
 /tempname.c
 /tempname.h
+/term-ostream.c
+/term-ostream.h
 /term-ostream.oo.c
 /term-ostream.oo.h
+/term-styled-ostream.c
+/term-styled-ostream.h
 /term-styled-ostream.oo.c
 /term-styled-ostream.oo.h
+/term_ostream.priv.h
+/term_ostream.vt.h
+/term_styled_ostream.priv.h
+/term_styled_ostream.vt.h
 /terminfo.h
 /time.in.h
 /tmpdir.c
@@ -389,47 +437,3 @@
 /xstriconveh.h
 /xvasprintf.c
 /xvasprintf.h
-
-# Files generated by the autotools:
-/Makefile.in
-
-# Files generated by gperf:
-/iconv_open-aix.h
-/iconv_open-hpux.h
-/iconv_open-irix.h
-/iconv_open-osf.h
-/iconv_open-solaris.h
-
-# Files generated by moopp:
-/fd-ostream.c
-/fd-ostream.h
-/fd_ostream.priv.h
-/fd_ostream.vt.h
-/file-ostream.c
-/file-ostream.h
-/file_ostream.priv.h
-/file_ostream.vt.h
-/html-ostream.c
-/html-ostream.h
-/html-styled-ostream.c
-/html-styled-ostream.h
-/html_ostream.priv.h
-/html_ostream.vt.h
-/html_styled_ostream.priv.h
-/html_styled_ostream.vt.h
-/ostream.c
-/ostream.h
-/ostream.priv.h
-/ostream.vt.h
-/styled-ostream.c
-/styled-ostream.h
-/styled_ostream.priv.h
-/styled_ostream.vt.h
-/term-ostream.c
-/term-ostream.h
-/term-styled-ostream.c
-/term-styled-ostream.h
-/term_ostream.priv.h
-/term_ostream.vt.h
-/term_styled_ostream.priv.h
-/term_styled_ostream.vt.h
diff --git a/gettext-tools/gnulib-tests/.gitignore b/gettext-tools/gnulib-tests/.gitignore
index 5f93dbc..d66e6d0 100644
--- a/gettext-tools/gnulib-tests/.gitignore
+++ b/gettext-tools/gnulib-tests/.gitignore
@@ -116,6 +116,7 @@
 /test-inttypes.c
 /test-iswblank.c
 /test-langinfo.c
+/test-linked_list.c
 /test-linkedhash_list.c
 /test-locale.c
 /test-localename.c
diff --git a/gettext-tools/libgettextpo/.gitignore b/gettext-tools/libgettextpo/.gitignore
index a7d1bb1..41424a8 100644
--- a/gettext-tools/libgettextpo/.gitignore
+++ b/gettext-tools/libgettextpo/.gitignore
@@ -11,15 +11,13 @@
 # Files brought in by gnulib-tool:
 /Makefile.gnulib
 /alignof.h
+/alloca.h
 /alloca.in.h
 /arg-nonnull.h
 /asnprintf.c
 /asprintf.c
 /basename.c
 /basename.h
-/charset.alias
-/close.c
-/configmake.h
 /c++defs.h
 /c-ctype.c
 /c-ctype.h
@@ -29,9 +27,12 @@
 /c-strncasecmp.c
 /c-strstr.c
 /c-strstr.h
+/charset.alias
+/close.c
 /concat-filename.c
 /concat-filename.h
 /config.charset
+/configmake.h
 /diffseq.h
 /dosname.h
 /errno.in.h
@@ -42,6 +43,7 @@
 /exitfail.c
 /exitfail.h
 /exported.sh
+/fcntl.h
 /fcntl.in.h
 /fd-hook.c
 /fd-hook.h
@@ -61,12 +63,21 @@
 /gcd.h
 /getdelim.c
 /getline.c
-/gettext.h
 /gettext-po.h
+/gettext.h
 /gettimeofday.c
+/gl_anylinked_list1.h
+/gl_anylinked_list2.h
+/gl_linked_list.c
+/gl_linked_list.h
+/gl_list.c
+/gl_list.h
+/gl_xlist.c
+/gl_xlist.h
 /hash.c
 /hash.h
 /iconv.c
+/iconv.h
 /iconv.in.h
 /iconv_close.c
 /iconv_open-aix.gperf
@@ -85,6 +96,8 @@
 /malloca.c
 /malloca.h
 /malloca.valgrind
+/markup.c
+/markup.h
 /mbrtowc.c
 /mbsinit.c
 /mbswidth.c
@@ -113,10 +126,13 @@
 /rawmemchr.c
 /rawmemchr.valgrind
 /realloc.c
+/ref-add.sed
 /ref-add.sin
+/ref-del.sed
 /ref-del.sin
 /relocatable.c
 /relocatable.h
+/signal.h
 /signal.in.h
 /sigprocmask.c
 /size_max.h
@@ -126,7 +142,10 @@
 /stddef.in.h
 /stdint.in.h
 /stdio-write.c
+/stdio.c
+/stdio.h
 /stdio.in.h
+/stdlib.h
 /stdlib.in.h
 /stpcpy.c
 /stpncpy.c
@@ -143,15 +162,19 @@
 /striconveh.h
 /striconveha.c
 /striconveha.h
+/string.h
 /string.in.h
 /strstr.c
 /sys_stat.in.h
 /sys_time.in.h
 /sys_types.in.h
+/time.h
 /time.in.h
 /uniconv.in.h
 /unictype.in.h
 /unilbrk.in.h
+/unistd.c
+/unistd.h
 /unistd.in.h
 /unistr.in.h
 /unitypes.in.h
@@ -163,9 +186,11 @@
 /vasprintf.c
 /verify.h
 /warn-on-use.h
+/wchar.h
 /wchar.in.h
-/wctype.in.h
 /wctype-h.c
+/wctype.h
+/wctype.in.h
 /wcwidth.c
 /xalloc.h
 /xasprintf.c
@@ -185,21 +210,12 @@
 
 # Files generated by the autotools:
 /Makefile.in
-/alloca.h
-/fcntl.h
-/iconv.h
-/ref-add.sed
-/ref-del.sed
-/signal.h
-/stdio.c
-/stdio.h
-/stdlib.h
-/string.h
-/time.h
-/unistd.c
-/unistd.h
-/wchar.h
-/wctype.h
+/uniconv.h
+/unictype.h
+/unilbrk.h
+/unistr.h
+/unitypes.h
+/uniwidth.h
 
 # Files generated by gperf:
 /iconv_open-aix.h
diff --git a/gettext-tools/libgettextpo/Makefile.am b/gettext-tools/libgettextpo/Makefile.am
index 5a4404d..dfbf40c 100644
--- a/gettext-tools/libgettextpo/Makefile.am
+++ b/gettext-tools/libgettextpo/Makefile.am
@@ -41,6 +41,8 @@ AM_CPPFLAGS = \
   -I../src -I$(top_srcdir)/src \
   -I../intl -I$(top_srcdir)/../gettext-runtime/intl
 
+DEFS = -DIN_LIBGETTEXTPO=1 @DEFS@
+
 # libgettextpo contains the public API for PO files.
 libgettextpo_la_SOURCES = \
   gettext-po.c \
@@ -93,8 +95,7 @@ libgettextpo_la_AUXSOURCES = \
   ../src/plural-exp.c \
   ../src/plural-eval.c \
   ../src/msgl-check.c \
-  ../src/sentence.c \
-  ../src/libexpat-compat.c
+  ../src/sentence.c
 
 # Libtool's library version information for libgettextpo.
 # See the libtool documentation, section "Library interface versions".
@@ -109,7 +110,7 @@ libgettextpo_la_LIBADD = libgnu.la $(WOE32_LIBADD) $(LTLIBUNISTRING)
 libgettextpo_la_LDFLAGS = \
   -version-info $(LTV_CURRENT):$(LTV_REVISION):$(LTV_AGE) \
   -rpath $(libdir) \
-  @LTLIBINTL@ @LTLIBICONV@ @LTLIBEXPAT@ -lc -no-undefined
+  @LTLIBINTL@ @LTLIBICONV@ -lc -no-undefined
 
 # Tell the mingw or Cygwin linker which symbols to export.
 if WOE32DLL
@@ -155,7 +156,7 @@ config.h: $(BUILT_SOURCES)
 	          sf=`echo "$$f" | sed -e 's,\\.[^.]*$$,,'`.c; \
 	          test -f $$sf || sf=$(srcdir)/$$sf; \
 	          of=`echo "$$f" | sed -e 's,^.*/,,' -e 's,\\.[^.]*$$,,'`.$(OBJEXT); \
-	          $(COMPILE) -c $$sf || { rm -f config.h; exit 1; }; \
+	          $(COMPILE) $(DEFS) -c $$sf || { rm -f config.h; exit 1; }; \
 	          sh ./exported.sh $$of 1>&5; \
 	          rm -f $$of `echo "$$of" | sed -e 's,\\.$(OBJEXT)$$,.lo,'`; \
 	          ;; \
diff --git a/gettext-tools/src/Makefile.am b/gettext-tools/src/Makefile.am
index f3dab4c..7109072 100644
--- a/gettext-tools/src/Makefile.am
+++ b/gettext-tools/src/Makefile.am
@@ -56,7 +56,7 @@ po-time.h plural-table.h lang-table.h format.h filters.h \
 xgettext.h x-c.h x-po.h x-sh.h x-python.h x-lisp.h x-elisp.h x-librep.h \
 x-scheme.h x-smalltalk.h x-java.h x-properties.h x-csharp.h x-awk.h x-ycp.h \
 x-tcl.h x-perl.h x-php.h x-stringtable.h x-rst.h x-glade.h x-lua.h \
-x-javascript.h x-vala.h x-gsettings.h x-desktop.h x-appdata.h libexpat-compat.h
+x-javascript.h x-vala.h x-gsettings.h x-desktop.h x-appdata.h
 
 EXTRA_DIST += FILES project-id
 
@@ -152,7 +152,7 @@ $(COMMON_SOURCE) read-catalog.c \
 color.c write-catalog.c write-properties.c write-stringtable.c write-po.c \
 msgl-ascii.c msgl-iconv.c msgl-equal.c msgl-cat.c msgl-header.c msgl-english.c \
 msgl-check.c file-list.c msgl-charset.c po-time.c plural-exp.c plural-eval.c \
-plural-table.c quote.h sentence.h sentence.c libexpat-compat.c \
+plural-table.c quote.h sentence.h sentence.c \
 $(FORMAT_SOURCE) \
 read-desktop.c locating-rule.c its.c
 
@@ -254,7 +254,7 @@ cldr_plurals_LDADD = libgettextsrc.la $(LDADD)
 # use iconv().
 libgettextsrc_la_LDFLAGS = \
   -release @VERSION@ \
-  ../gnulib-lib/libgettextlib.la $(LTLIBUNISTRING) @LTLIBINTL@ @LTLIBICONV@ @LTLIBEXPAT@ -lc -no-undefined
+  ../gnulib-lib/libgettextlib.la $(LTLIBUNISTRING) @LTLIBINTL@ @LTLIBICONV@ -lc -no-undefined
 
 libgettextsrc_la_CPPFLAGS = $(AM_CPPFLAGS) $(INCXML)
 
diff --git a/gettext-tools/src/format-kde-kuit.c b/gettext-tools/src/format-kde-kuit.c
index 3e00697..afd6b15 100644
--- a/gettext-tools/src/format-kde-kuit.c
+++ b/gettext-tools/src/format-kde-kuit.c
@@ -24,12 +24,28 @@
 #include <stdlib.h>
 
 #include "format.h"
-#include "libexpat-compat.h"
 #include "unistr.h"
 #include "xalloc.h"
 #include "xvasprintf.h"
 #include "gettext.h"
 
+#if IN_LIBGETTEXTPO
+/* Use included markup parser to avoid extra dependency from
+   libgettextpo to libxml2.  */
+# ifndef FORMAT_KDE_KUIT_FALLBACK_MARKUP
+#  define FORMAT_KDE_KUIT_USE_FALLBACK_MARKUP 1
+# endif
+#else
+#  define FORMAT_KDE_KUIT_USE_LIBXML2 1
+#endif
+
+#if FORMAT_KDE_KUIT_USE_LIBXML2
+# include <libxml/parser.h>
+#elif FORMAT_KDE_KUIT_USE_FALLBACK_MARKUP
+# include "markup.h"
+#endif
+
+
 #define _(str) gettext (str)
 
 #define SIZEOF(a) (sizeof(a) / sizeof(a[0]))
@@ -48,26 +64,8 @@ struct spec
   void *base;
 };
 
-#if DYNLOAD_LIBEXPAT || HAVE_LIBEXPAT
-
 #define XML_NS "https://www.gnu.org/s/gettext/kde"
 
-/* Callback called when <element> is seen.  */
-static void
-start_element_handler (void *data, const char *name,
-                       const char **attributes)
-{
-  /* Nothing to do for now.  We could check text outside of a
-     structuring tag, etc.  */
-}
-
-/* Callback called when </element> is seen.  */
-static void
-end_element_handler (void *data, const char *name)
-{
-  /* Nothing to do.  */
-}
-
 struct char_range
 {
   ucs4_t start;
@@ -182,8 +180,6 @@ is_reference (const char *input)
   return false;
 }
 
-#endif
-
 
 static void *
 format_parse (const char *format, bool translated, char *fdi,
@@ -191,83 +187,99 @@ format_parse (const char *format, bool translated, char *fdi,
 {
   struct spec spec;
   struct spec *result;
+  const char *str;
+  const char *str_limit;
+  size_t amp_count;
+  char *buffer, *bp;
 
   spec.base = NULL;
 
-#if DYNLOAD_LIBEXPAT || HAVE_LIBEXPAT
-  if (LIBEXPAT_AVAILABLE ())
-    {
-      XML_Parser parser;
-      const char *str = format;
-      const char *str_limit = str + strlen (format);
-      size_t amp_count;
-      char *buffer, *bp;
+  /* Preprocess the input, putting the content in a <gt:kuit> element.  */
+  str = format;
+  str_limit = str + strlen (format);
 
-      for (amp_count = 0; str < str_limit; amp_count++)
-        {
-          const char *amp = strchrnul (str, '&');
-          if (*amp != '&')
-            break;
-          str = amp + 1;
-        }
+  for (amp_count = 0; str < str_limit; amp_count++)
+    {
+      const char *amp = strchrnul (str, '&');
+      if (*amp != '&')
+        break;
+      str = amp + 1;
+    }
 
-      buffer = xmalloc (amp_count * 4
-                        + strlen (format)
-                        + strlen ("<gt:kuit xmlns:gt=\"" XML_NS "\"></gt:kuit>")
-                        + 1);
-      *buffer = '\0';
+  buffer = xmalloc (amp_count * 4
+                    + strlen (format)
+                    + strlen ("<gt:kuit xmlns:gt=\"" XML_NS "\"></gt:kuit>")
+                    + 1);
+  *buffer = '\0';
 
-      bp = buffer;
-      bp = stpcpy (bp, "<gt:kuit xmlns:gt=\"" XML_NS "\">");
-      str = format;
-      while (str < str_limit)
-        {
-          const char *amp = strchrnul (str, '&');
+  bp = buffer;
+  bp = stpcpy (bp, "<gt:kuit xmlns:gt=\"" XML_NS "\">");
+  str = format;
+  while (str < str_limit)
+    {
+      const char *amp = strchrnul (str, '&');
 
-          bp = stpncpy (bp, str, amp - str);
-          if (*amp != '&')
-            break;
+      bp = stpncpy (bp, str, amp - str);
+      if (*amp != '&')
+        break;
 
-          bp = stpcpy (bp, is_reference (amp) ? "&" : "&amp;");
-          str = amp + 1;
-        }
-      stpcpy (bp, "</gt:kuit>");
+      bp = stpcpy (bp, is_reference (amp) ? "&" : "&amp;");
+      str = amp + 1;
+    }
+  stpcpy (bp, "</gt:kuit>");
 
-      parser = XML_ParserCreate (NULL);
-      if (parser == NULL)
+#if FORMAT_KDE_KUIT_USE_LIBXML2
+    {
+      xmlDocPtr doc;
+
+      doc = xmlReadMemory (buffer, strlen (buffer), "", NULL,
+                           XML_PARSE_NONET
+                           | XML_PARSE_NOWARNING
+                           | XML_PARSE_NOERROR
+                           | XML_PARSE_NOBLANKS);
+      if (doc == NULL)
         {
-          *invalid_reason = xasprintf (_("memory exhausted"));
-          free (buffer);
-          return NULL;
+          xmlError *err = xmlGetLastError ();
+          *invalid_reason =
+            xasprintf (_("error while parsing: %s"),
+                       err->message);
         }
 
-      XML_SetElementHandler (parser,
-                             start_element_handler,
-                             end_element_handler);
+      free (buffer);
+      xmlFreeDoc (doc);
+    }
+#elif FORMAT_KDE_KUIT_FALLBACK_MARKUP
+    {
+      markup_parser_ty parser;
+      markup_parse_context_ty *context;
 
-      if (XML_Parse (parser, buffer, strlen (buffer), 0) == 0)
+      memset (&parser, 0, sizeof (markup_parser_ty));
+      context = markup_parse_context_new (&parser, 0, NULL);
+      if (!markup_parse_context_parse (context, buffer, strlen (buffer)))
         {
           *invalid_reason =
             xasprintf (_("error while parsing: %s"),
-                       XML_ErrorString (XML_GetErrorCode (parser)));
+                       markup_parse_context_get_error (context));
           free (buffer);
-          XML_ParserFree (parser);
+          markup_parse_context_free (context);
           return NULL;
         }
 
-      if (XML_Parse (parser, NULL, 0, 1) == 0)
+      if (!markup_parse_context_end_parse (context))
         {
           *invalid_reason =
             xasprintf (_("error while parsing: %s"),
-                       XML_ErrorString (XML_GetErrorCode (parser)));
+                       markup_parse_context_get_error (context));
           free (buffer);
-          XML_ParserFree (parser);
+          markup_parse_context_free (context);
           return NULL;
         }
 
       free (buffer);
-      XML_ParserFree (parser);
+      markup_parse_context_free (context);
     }
+#else
+    /* No support for XML.  */
 #endif
 
   spec.base = formatstring_kde.parse (format, translated, fdi, invalid_reason);
diff --git a/gettext-tools/src/libexpat-compat.c b/gettext-tools/src/libexpat-compat.c
deleted file mode 100644
index 33d264a..0000000
--- a/gettext-tools/src/libexpat-compat.c
+++ /dev/null
@@ -1,327 +0,0 @@
-/* xgettext libexpat compatibility.
-   Copyright (C) 2002-2003, 2005-2009, 2013, 2015 Free Software
-   Foundation, Inc.
-
-   This file was written by Bruno Haible <haible@clisp.cons.org>, 2002.
-
-   This program is free software: you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published by
-   the Free Software Foundation; either version 3 of the License, or
-   (at your option) any later version.
-
-   This program is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-   GNU General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
-
-#ifdef HAVE_CONFIG_H
-# include "config.h"
-#endif
-
-#include <stdlib.h>
-#include <string.h>
-
-#if DYNLOAD_LIBEXPAT
-# include <dlfcn.h>
-#else
-# if HAVE_LIBEXPAT
-#  include <expat.h>
-# endif
-#endif
-
-/* Keep the references to XML_GetCurrent{Line,Column}Number symbols
-   before loading libexpat-compat.h, since they are redefined to
-   rpl_XML_GetCurrent{Line,Column}Number .  */
-#if !DYNLOAD_LIBEXPAT && XML_MAJOR_VERSION >= 2
-static void *p_XML_GetCurrentLineNumber = (void *) &XML_GetCurrentLineNumber;
-static void *p_XML_GetCurrentColumnNumber = (void *) &XML_GetCurrentColumnNumber;
-#endif
-
-#include "libexpat-compat.h"
-
-/* ======================= Different libexpat ABIs.  ======================= */
-
-/* There are three different ABIs of libexpat, regarding the functions
-   XML_GetCurrentLineNumber and XML_GetCurrentColumnNumber.
-   In expat < 2.0, they return an 'int'.
-   In expat >= 2.0, they return
-     - a 'long' if expat was compiled with the default flags, or
-     - a 'long long' if expat was compiled with -DXML_LARGE_SIZE.
-   But the <expat.h> include file does not contain the information whether
-   expat was compiled with -DXML_LARGE_SIZE; so the include file is lying!
-   For this information, we need to call XML_GetFeatureList(), for
-   expat >= 2.0.1; for expat = 2.0.0, we have to assume the default flags.  */
-
-#if !DYNLOAD_LIBEXPAT && XML_MAJOR_VERSION >= 2
-
-/* expat >= 2.0 -> Return type is 'int64_t' worst-case.  */
-
-/* Return true if libexpat was compiled with -DXML_LARGE_SIZE.  */
-static bool
-is_XML_LARGE_SIZE_ABI (void)
-{
-  static bool tested;
-  static bool is_large;
-
-  if (!tested)
-    {
-      const XML_Feature *features;
-
-      is_large = false;
-      for (features = XML_GetFeatureList (); features->name != NULL; features++)
-        if (strcmp (features->name, "XML_LARGE_SIZE") == 0)
-          {
-            is_large = true;
-            break;
-          }
-
-      tested = true;
-    }
-  return is_large;
-}
-
-int64_t
-rpl_XML_GetCurrentLineNumber (XML_Parser parser)
-{
-  if (is_XML_LARGE_SIZE_ABI ())
-    return ((int64_t (*) (XML_Parser)) p_XML_GetCurrentLineNumber) (parser);
-  else
-    return ((long (*) (XML_Parser)) p_XML_GetCurrentLineNumber) (parser);
-}
-
-int64_t
-rpl_XML_GetCurrentColumnNumber (XML_Parser parser)
-{
-  if (is_XML_LARGE_SIZE_ABI ())
-    return ((int64_t (*) (XML_Parser)) p_XML_GetCurrentColumnNumber) (parser);
-  else
-    return ((long (*) (XML_Parser)) p_XML_GetCurrentColumnNumber) (parser);
-}
-#endif
-
-
-/* ===================== Dynamic loading of libexpat.  ===================== */
-
-#if DYNLOAD_LIBEXPAT
-
-static XML_Expat_Version (*p_XML_ExpatVersionInfo) (void);
-
-XML_Expat_Version
-XML_ExpatVersionInfo (void)
-{
-  return (*p_XML_ExpatVersionInfo) ();
-}
-
-static const XML_Feature * (*p_XML_GetFeatureList) (void);
-
-const XML_Feature *
-XML_GetFeatureList (void)
-{
-  return (*p_XML_GetFeatureList) ();
-}
-
-enum XML_Size_ABI
-get_XML_Size_ABI (void)
-{
-  static bool tested;
-  static enum XML_Size_ABI abi;
-
-  if (!tested)
-    {
-      if (XML_ExpatVersionInfo () .major >= 2)
-        /* expat >= 2.0 -> XML_Size is 'int64_t' or 'long'.  */
-        {
-          const XML_Feature *features;
-
-          abi = is_long;
-          for (features = XML_GetFeatureList ();
-               features->name != NULL;
-               features++)
-            if (strcmp (features->name, "XML_LARGE_SIZE") == 0)
-              {
-                abi = is_int64_t;
-                break;
-              }
-        }
-      else
-        /* expat < 2.0 -> XML_Size is 'int'.  */
-        abi = is_int;
-      tested = true;
-    }
-  return abi;
-}
-
-static XML_Parser (*p_XML_ParserCreate) (const XML_Char *encoding);
-
-XML_Parser
-XML_ParserCreate (const XML_Char *encoding)
-{
-  return (*p_XML_ParserCreate) (encoding);
-}
-
-static void (*p_XML_SetElementHandler) (XML_Parser parser,
-                                        XML_StartElementHandler start,
-                                        XML_EndElementHandler end);
-
-void
-XML_SetElementHandler (XML_Parser parser,
-                       XML_StartElementHandler start,
-                       XML_EndElementHandler end)
-{
-  (*p_XML_SetElementHandler) (parser, start, end);
-}
-
-
-static void (*p_XML_SetCharacterDataHandler) (XML_Parser parser,
-                                              XML_CharacterDataHandler handler);
-
-void
-XML_SetCharacterDataHandler (XML_Parser parser,
-                             XML_CharacterDataHandler handler)
-{
-  (*p_XML_SetCharacterDataHandler) (parser, handler);
-}
-
-
-static void (*p_XML_SetCommentHandler) (XML_Parser parser,
-                                        XML_CommentHandler handler);
-
-void
-XML_SetCommentHandler (XML_Parser parser, XML_CommentHandler handler)
-{
-  (*p_XML_SetCommentHandler) (parser, handler);
-}
-
-
-static int (*p_XML_Parse) (XML_Parser parser, const char *s,
-                           int len, int isFinal);
-
-int
-XML_Parse (XML_Parser parser, const char *s, int len, int isFinal)
-{
-  return (*p_XML_Parse) (parser, s, len, isFinal);
-}
-
-
-static enum XML_Error (*p_XML_GetErrorCode) (XML_Parser parser);
-
-enum XML_Error
-XML_GetErrorCode (XML_Parser parser)
-{
-  return (*p_XML_GetErrorCode) (parser);
-}
-
-
-static void *p_XML_GetCurrentLineNumber;
-
-int64_t
-XML_GetCurrentLineNumber (XML_Parser parser)
-{
-  switch (get_XML_Size_ABI ())
-    {
-    case is_int:
-      return ((int (*) (XML_Parser)) p_XML_GetCurrentLineNumber) (parser);
-    case is_long:
-      return ((long (*) (XML_Parser)) p_XML_GetCurrentLineNumber) (parser);
-    case is_int64_t:
-      return ((int64_t (*) (XML_Parser)) p_XML_GetCurrentLineNumber) (parser);
-    default:
-      abort ();
-    }
-}
-
-static void *p_XML_GetCurrentColumnNumber;
-
-int64_t
-XML_GetCurrentColumnNumber (XML_Parser parser)
-{
-  switch (get_XML_Size_ABI ())
-    {
-    case is_int:
-      return ((int (*) (XML_Parser)) p_XML_GetCurrentColumnNumber) (parser);
-    case is_long:
-      return ((long (*) (XML_Parser)) p_XML_GetCurrentColumnNumber) (parser);
-    case is_int64_t:
-      return ((int64_t (*) (XML_Parser)) p_XML_GetCurrentColumnNumber) (parser);
-    default:
-      abort ();
-    }
-}
-
-
-static const XML_LChar * (*p_XML_ErrorString) (int code);
-
-const XML_LChar *
-XML_ErrorString (int code)
-{
-  return (*p_XML_ErrorString) (code);
-}
-
-static void (*p_XML_ParserFree) (XML_Parser parser);
-
-void
-XML_ParserFree (XML_Parser parser)
-{
-  return (*p_XML_ParserFree) (parser);
-}
-
-static int libexpat_loaded = 0;
-
-bool
-load_libexpat ()
-{
-  if (libexpat_loaded == 0)
-    {
-      void *handle;
-
-      /* Try to load libexpat-2.x.  */
-      handle = dlopen ("libexpat.so.1", RTLD_LAZY);
-      if (handle == NULL)
-        /* Try to load libexpat-1.x.  */
-        handle = dlopen ("libexpat.so.0", RTLD_LAZY);
-      if (handle != NULL
-          && (p_XML_ExpatVersionInfo =
-                (XML_Expat_Version (*) (void))
-                dlsym (handle, "XML_ExpatVersionInfo")) != NULL
-          && (p_XML_GetFeatureList =
-                (const XML_Feature * (*) (void))
-                dlsym (handle, "XML_GetFeatureList")) != NULL
-          && (p_XML_ParserCreate =
-                (XML_Parser (*) (const XML_Char *))
-                dlsym (handle, "XML_ParserCreate")) != NULL
-          && (p_XML_SetElementHandler =
-                (void (*) (XML_Parser, XML_StartElementHandler, XML_EndElementHandler))
-                dlsym (handle, "XML_SetElementHandler")) != NULL
-          && (p_XML_SetCharacterDataHandler =
-                (void (*) (XML_Parser, XML_CharacterDataHandler))
-                dlsym (handle, "XML_SetCharacterDataHandler")) != NULL
-          && (p_XML_SetCommentHandler =
-                (void (*) (XML_Parser, XML_CommentHandler))
-                dlsym (handle, "XML_SetCommentHandler")) != NULL
-          && (p_XML_Parse =
-                (int (*) (XML_Parser, const char *, int, int))
-                dlsym (handle, "XML_Parse")) != NULL
-          && (p_XML_GetErrorCode =
-                (enum XML_Error (*) (XML_Parser))
-                dlsym (handle, "XML_GetErrorCode")) != NULL
-          && (p_XML_GetCurrentLineNumber =
-                dlsym (handle, "XML_GetCurrentLineNumber")) != NULL
-          && (p_XML_GetCurrentColumnNumber =
-                dlsym (handle, "XML_GetCurrentColumnNumber")) != NULL
-          && (p_XML_ParserFree =
-                (void (*) (XML_Parser))
-                dlsym (handle, "XML_ParserFree")) != NULL
-          && (p_XML_ErrorString =
-                (const XML_LChar * (*) (int))
-                dlsym (handle, "XML_ErrorString")) != NULL)
-        libexpat_loaded = 1;
-      else
-        libexpat_loaded = -1;
-    }
-  return libexpat_loaded >= 0;
-}
-
-#endif
diff --git a/gettext-tools/src/libexpat-compat.h b/gettext-tools/src/libexpat-compat.h
deleted file mode 100644
index 3e41e82..0000000
--- a/gettext-tools/src/libexpat-compat.h
+++ /dev/null
@@ -1,95 +0,0 @@
-/* xgettext libexpat compatibility.
-   Copyright (C) 2002-2003, 2005-2009, 2013, 2015 Free Software
-   Foundation, Inc.
-
-   This file was written by Bruno Haible <haible@clisp.cons.org>, 2002.
-
-   This program is free software: you can redistribute it and/or modify
-   it under the terms of the GNU General Public License as published by
-   the Free Software Foundation; either version 3 of the License, or
-   (at your option) any later version.
-
-   This program is distributed in the hope that it will be useful,
-   but WITHOUT ANY WARRANTY; without even the implied warranty of
-   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-   GNU General Public License for more details.
-
-   You should have received a copy of the GNU General Public License
-   along with this program.  If not, see <http://www.gnu.org/licenses/>.  */
-
-#include <stdbool.h>
-#include <stdint.h>
-#if DYNLOAD_LIBEXPAT
-# include <dlfcn.h>
-#else
-# if HAVE_LIBEXPAT
-#  include <expat.h>
-# endif
-#endif
-
-#if !DYNLOAD_LIBEXPAT && XML_MAJOR_VERSION >= 2
-int64_t rpl_XML_GetCurrentLineNumber (XML_Parser parser);
-# undef XML_GetCurrentLineNumber
-# define XML_GetCurrentLineNumber rpl_XML_GetCurrentLineNumber
-
-int64_t rpl_XML_GetCurrentColumnNumber (XML_Parser parser);
-# undef XML_GetCurrentColumnNumber
-# define XML_GetCurrentColumnNumber rpl_XML_GetCurrentColumnNumber
-#endif
-
-/* ===================== Dynamic loading of libexpat.  ===================== */
-
-#if DYNLOAD_LIBEXPAT
-typedef struct
-  {
-    int major;
-    int minor;
-    int micro;
-  }
-  XML_Expat_Version;
-enum XML_FeatureEnum { XML_FEATURE_END = 0 };
-typedef struct
-  {
-    enum XML_FeatureEnum feature;
-    const char *name;
-    long int value;
-  }
-  XML_Feature;
-typedef void *XML_Parser;
-typedef char XML_Char;
-typedef char XML_LChar;
-enum XML_Error { XML_ERROR_NONE };
-typedef void (*XML_StartElementHandler) (void *userData, const XML_Char *name, const XML_Char **atts);
-typedef void (*XML_EndElementHandler) (void *userData, const XML_Char *name);
-typedef void (*XML_CharacterDataHandler) (void *userData, const XML_Char *s, int len);
-typedef void (*XML_CommentHandler) (void *userData, const XML_Char *data);
-
-XML_Expat_Version XML_ExpatVersionInfo (void);
-const XML_Feature * XML_GetFeatureList (void);
-
-enum XML_Size_ABI { is_int, is_long, is_int64_t };
-enum XML_Size_ABI get_XML_Size_ABI (void);
-
-XML_Parser XML_ParserCreate (const XML_Char *encoding);
-void XML_SetElementHandler (XML_Parser parser,
-                            XML_StartElementHandler start,
-                            XML_EndElementHandler end);
-void XML_SetCharacterDataHandler (XML_Parser parser,
-                                  XML_CharacterDataHandler handler);
-void XML_SetCommentHandler (XML_Parser parser, XML_CommentHandler handler);
-int XML_Parse (XML_Parser parser, const char *s, int len, int isFinal);
-enum XML_Error XML_GetErrorCode (XML_Parser parser);
-int64_t XML_GetCurrentLineNumber (XML_Parser parser);
-int64_t XML_GetCurrentColumnNumber (XML_Parser parser);
-const XML_LChar * XML_ErrorString (int code);
-void XML_ParserFree (XML_Parser parser);
-
-bool load_libexpat ();
-
-#define LIBEXPAT_AVAILABLE() (load_libexpat ())
-
-#elif HAVE_LIBEXPAT
-
-#define LIBEXPAT_AVAILABLE() true
-
-#endif
diff --git a/gnulib-local/lib/markup.c b/gnulib-local/lib/markup.c
new file mode 100644
index 0000000..a0f6856
--- /dev/null
+++ b/gnulib-local/lib/markup.c
@@ -0,0 +1,1523 @@
+/* markup.c -- simple XML-like parser
+   Copyright (C) 2015 Free Software Foundation, Inc.
+
+   This file is not part of the GNU gettext program, but is used with
+   GNU gettext.
+
+   This is a stripped down version of GLib's gmarkup.c.  The original
+   copyright notice is as follows:
+*/
+
+/* gmarkup.c - Simple XML-like parser
+ *
+ *  Copyright 2000, 2003 Red Hat, Inc.
+ *  Copyright 2007, 2008 Ryan Lortie <desrt@desrt.ca>
+ *
+ * GLib is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 3 of the
+ * License, or (at your option) any later version.
+ *
+ * GLib is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with GLib; see the file COPYING.LIB.  If not,
+ * see <http://www.gnu.org/licenses/>.
+ */
+
+#include "config.h"
+
+#include <assert.h>
+#include <stdarg.h>
+#include <string.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <errno.h>
+
+/* Specification */
+#include "markup.h"
+
+#include "c-ctype.h"
+#include "gettext.h"
+#include "gl_linked_list.h"
+#include "gl_xlist.h"
+#include "unictype.h"
+#include "unistr.h"
+#include "xalloc.h"
+#include "xvasprintf.h"
+
+#define _(s) gettext(s)
+
+/**
+ * The "markup" parser is intended to parse a simple markup format
+ * that's a subset of XML.  This is a small, efficient, easy-to-use
+ * parser.  It should not be used if you expect to interoperate with
+ * other applications generating full-scale XML.  However, it's very
+ * useful for application data files, config files, etc. where you
+ * know your application will be the only one writing the file.
+ * Full-scale XML parsers should be able to parse the subset used by
+ * markup, so you can easily migrate to full-scale XML at a later
+ * time if the need arises.
+ *
+ * The parser is not guaranteed to signal an error on all invalid XML;
+ * the parser may accept documents that an XML parser would not.
+ * However, XML documents which are not well-formed (which is a weaker
+ * condition than being valid.  See the XML specification
+ * <http://www.w3.org/TR/REC-xml/> for definitions of these terms.)
+ * are not considered valid GMarkup documents.
+ *
+ * Simplifications to XML include:
+ *
+ * - Only UTF-8 encoding is allowed
+ *
+ * - No user-defined entities
+ *
+ * - Processing instructions, comments and the doctype declaration
+ *   are "passed through" but are not interpreted in any way
+ *
+ * - No DTD or validation
+ *
+ * The markup format does support:
+ *
+ * - Elements
+ *
+ * - Attributes
+ *
+ * - 5 standard entities: &amp; &lt; &gt; &quot; &apos;
+ *
+ * - Character references
+ *
+ * - Sections marked as CDATA
+ */
+
+typedef enum
+{
+  STATE_START,
+  STATE_AFTER_OPEN_ANGLE,
+  STATE_AFTER_CLOSE_ANGLE,
+  STATE_AFTER_ELISION_SLASH, /* the slash that obviates need for end element */
+  STATE_INSIDE_OPEN_TAG_NAME,
+  STATE_INSIDE_ATTRIBUTE_NAME,
+  STATE_AFTER_ATTRIBUTE_NAME,
+  STATE_BETWEEN_ATTRIBUTES,
+  STATE_AFTER_ATTRIBUTE_EQUALS_SIGN,
+  STATE_INSIDE_ATTRIBUTE_VALUE_SQ,
+  STATE_INSIDE_ATTRIBUTE_VALUE_DQ,
+  STATE_INSIDE_TEXT,
+  STATE_AFTER_CLOSE_TAG_SLASH,
+  STATE_INSIDE_CLOSE_TAG_NAME,
+  STATE_AFTER_CLOSE_TAG_NAME,
+  STATE_INSIDE_PASSTHROUGH,
+  STATE_ERROR
+} markup_parse_state_ty;
+
+typedef struct
+{
+  const char *prev_element;
+  const markup_parser_ty *prev_parser;
+  void *prev_user_data;
+} markup_recursion_tracker_ty;
+
+typedef struct
+{
+  char *buffer;
+  size_t bufmax;
+  size_t buflen;
+} markup_string_ty;
+
+struct _markup_parse_context_ty
+{
+  const markup_parser_ty *parser;
+
+  markup_parse_flags_ty flags;
+
+  int line_number;
+  int char_number;
+
+  markup_parse_state_ty state;
+
+  void *user_data;
+
+  /* A piece of character data or an element that
+   * hasn't "ended" yet so we haven't yet called
+   * the callback for it.
+   */
+  markup_string_ty *partial_chunk;
+
+  gl_list_t tag_stack;          /* <markup_string_ty> */
+
+  char **attr_names;
+  char **attr_values;
+  int cur_attr;
+  int alloc_attrs;
+
+  const char *current_text;
+  ssize_t current_text_len;
+  const char *current_text_end;
+
+  /* used to save the start of the last interesting thingy */
+  const char *start;
+
+  const char *iter;
+
+  char *error_text;
+
+  unsigned int document_empty : 1;
+  unsigned int parsing : 1;
+  unsigned int awaiting_pop : 1;
+  int balance;
+
+  /* subparser support */
+  gl_list_t subparser_stack;    /* <markup_recursion_tracker_ty *> */
+  const char *subparser_element;
+};
+
+static markup_string_ty *
+markup_string_new (void)
+{
+  return XZALLOC (markup_string_ty);
+}
+
+static char *
+markup_string_free (markup_string_ty *string, bool free_segment)
+{
+  if (free_segment)
+    {
+      free (string->buffer);
+      free (string);
+      return NULL;
+    }
+  else
+    {
+      char *result = string->buffer;
+      free (string);
+      return result;
+    }
+}
+
+static void
+markup_string_free1 (markup_string_ty *string)
+{
+  markup_string_free (string, true);
+}
+
+static void
+markup_string_truncate (markup_string_ty *string, size_t length)
+{
+  assert (string && length < string->buflen - 1);
+  string->buffer[length] = '\0';
+  string->buflen = length;
+}
+
+static void
+markup_string_append (markup_string_ty *string, const char *to_append,
+                      size_t length)
+{
+  if (string->buflen + length + 1 > string->bufmax)
+    {
+      string->bufmax *= 2;
+      if (string->buflen + length + 1 > string->bufmax)
+        string->bufmax = string->buflen + length + 1;
+      string->buffer = xrealloc (string->buffer, string->bufmax);
+    }
+  memcpy (string->buffer + string->buflen, to_append, length);
+  string->buffer[length] = '\0';
+  string->buflen = length;
+}
+
+static inline void
+string_blank (markup_string_ty *string)
+{
+  if (string->bufmax > 0)
+    {
+      *string->buffer = '\0';
+      string->buflen = 0;
+    }
+}
+
+/* Creates a new parse context.  A parse context is used to parse
+   marked-up documents.  You can feed any number of documents into a
+   context, as long as no errors occur; once an error occurs, the
+   parse context can't continue to parse text (you have to free it and
+   create a new parse context).  */
+markup_parse_context_ty *
+markup_parse_context_new (const markup_parser_ty *parser,
+                          markup_parse_flags_ty flags,
+                          void *user_data)
+{
+  markup_parse_context_ty *context;
+
+  assert (parser != NULL);
+
+  context = XMALLOC (markup_parse_context_ty);
+
+  context->parser = parser;
+  context->flags = flags;
+  context->user_data = user_data;
+
+  context->line_number = 1;
+  context->char_number = 1;
+
+  context->partial_chunk = NULL;
+
+  context->state = STATE_START;
+  context->tag_stack =
+    gl_list_create_empty (GL_LINKED_LIST,
+                          NULL, NULL,
+                          (gl_listelement_dispose_fn) markup_string_free1,
+                          true);
+  context->attr_names = NULL;
+  context->attr_values = NULL;
+  context->cur_attr = -1;
+  context->alloc_attrs = 0;
+
+  context->current_text = NULL;
+  context->current_text_len = -1;
+  context->current_text_end = NULL;
+
+  context->start = NULL;
+  context->iter = NULL;
+
+  context->error_text = NULL;
+
+  context->document_empty = true;
+  context->parsing = false;
+
+  context->awaiting_pop = false;
+  context->subparser_stack =
+    gl_list_create_empty (GL_LINKED_LIST,
+                          NULL, NULL,
+                          (gl_listelement_dispose_fn) free,
+                          true);
+  context->subparser_element = NULL;
+
+  context->balance = 0;
+
+  return context;
+}
+
+static void clear_attributes (markup_parse_context_ty *context);
+
+/* Frees a parse context.  This function can't be called from inside
+   one of the markup_parser_ty functions or while a subparser is
+   pushed.  */
+void
+markup_parse_context_free (markup_parse_context_ty *context)
+{
+  assert (context != NULL);
+  assert (!context->parsing);
+  assert (gl_list_size (context->subparser_stack) == 0);
+  assert (!context->awaiting_pop);
+
+  clear_attributes (context);
+  free (context->attr_names);
+  free (context->attr_values);
+
+  gl_list_free (context->tag_stack);
+  gl_list_free (context->subparser_stack);
+
+  if (context->partial_chunk)
+    markup_string_free (context->partial_chunk, true);
+
+  free (context->error_text);
+
+  free (context);
+}
+
+static void pop_subparser_stack (markup_parse_context_ty *context);
+
+static void
+emit_error (markup_parse_context_ty *context, const char *error_text)
+{
+  context->state = STATE_ERROR;
+
+  if (context->parser->error)
+    (*context->parser->error) (context, error_text, context->user_data);
+
+  /* report the error all the way up to free all the user-data */
+  while (gl_list_size (context->subparser_stack) > 0)
+    {
+      pop_subparser_stack (context);
+      context->awaiting_pop = false; /* already been freed */
+
+      if (context->parser->error)
+        (*context->parser->error) (context, error_text, context->user_data);
+    }
+
+  if (context->error_text)
+    free (context->error_text);
+  context->error_text = xstrdup (error_text);
+}
+
+#define IS_COMMON_NAME_END_CHAR(c) \
+  ((c) == '=' || (c) == '/' || (c) == '>' || (c) == ' ')
+
+static bool
+slow_name_validate (markup_parse_context_ty *context, const char *name)
+{
+  const char *p = name;
+  ucs4_t uc;
+
+  if (u8_check ((uint8_t *) name, strlen (name)) != NULL)
+    {
+      emit_error (context, _("invalid UTF-8 sequence"));
+      return false;
+    }
+
+  if (!(c_isalpha (*p)
+        || (!IS_COMMON_NAME_END_CHAR (*p)
+            && (*p == '_'
+                || *p == ':'
+                || (u8_mbtouc (&uc, (uint8_t *) name, strlen (name)) > 0
+                    && uc_is_alpha (uc))))))
+    {
+      char *error_text = xasprintf (_("'%s' is not a valid name"), name);
+      emit_error (context, error_text);
+      free (error_text);
+      return false;
+    }
+
+  for (p = (char *) u8_next (&uc, (uint8_t *) name);
+       p != NULL;
+       p = (char *) u8_next (&uc, (uint8_t *) p))
+    {
+      /* is_name_char */
+      if (!(c_isalnum (*p) ||
+            (!IS_COMMON_NAME_END_CHAR (*p) &&
+             (*p == '.' ||
+              *p == '-' ||
+              *p == '_' ||
+              *p == ':' ||
+              uc_is_alpha (uc)))))
+        {
+          char *error_text = xasprintf (_("'%s' is not a valid name: '%c'"),
+                                        name, *p);
+          emit_error (context, error_text);
+          free (error_text);
+          return false;
+        }
+    }
+  return true;
+}
+
+/*
+ * Use me for elements, attributes etc.
+ */
+static bool
+name_validate (markup_parse_context_ty *context, const char *name)
+{
+  char mask;
+  const char *p;
+
+  /* name start char */
+  p = name;
+  if (IS_COMMON_NAME_END_CHAR (*p)
+      || !(c_isalpha (*p) || *p == '_' || *p == ':'))
+    goto slow_validate;
+
+  for (mask = *p++; *p != '\0'; p++)
+    {
+      mask |= *p;
+
+      /* is_name_char */
+      if (!(c_isalnum (*p)
+            || (!IS_COMMON_NAME_END_CHAR (*p)
+                && (*p == '.' || *p == '-' || *p == '_' || *p == ':'))))
+        goto slow_validate;
+    }
+
+  if (mask & 0x80) /* un-common / non-ascii */
+    goto slow_validate;
+
+  return true;
+
+ slow_validate:
+  return slow_name_validate (context, name);
+}
+
+static bool
+text_validate (markup_parse_context_ty *context,
+               const char *p,
+               int len)
+{
+  if (u8_check ((const uint8_t *) p, len) != NULL)
+    {
+      emit_error (context, _("invalid UTF-8 sequence"));
+      return false;
+    }
+  else
+    return true;
+}
+
+/*
+ * re-write the GString in-place, unescaping anything that escaped.
+ * most XML does not contain entities, or escaping.
+ */
+static bool
+unescape_string_inplace (markup_parse_context_ty *context,
+                         markup_string_ty *string,
+                         bool *is_ascii)
+{
+  char mask, *to;
+  const char *from;
+  bool normalize_attribute;
+
+  if (string->buflen == 0)
+    return true;
+
+  *is_ascii = false;
+
+  /* are we unescaping an attribute or not ? */
+  if (context->state == STATE_INSIDE_ATTRIBUTE_VALUE_SQ
+      || context->state == STATE_INSIDE_ATTRIBUTE_VALUE_DQ)
+    normalize_attribute = true;
+  else
+    normalize_attribute = false;
+
+  /*
+   * Meeks' theorem: unescaping can only shrink text.
+   * for &lt; etc. this is obvious, for &#xffff; more
+   * thought is required, but this is patently so.
+   */
+  mask = 0;
+  for (from = to = string->buffer; *from != '\0'; from++, to++)
+    {
+      *to = *from;
+
+      mask |= *to;
+      if (normalize_attribute && (*to == '\t' || *to == '\n'))
+        *to = ' ';
+      if (*to == '\r')
+        {
+          *to = normalize_attribute ? ' ' : '\n';
+          if (from[1] == '\n')
+            from++;
+        }
+      if (*from == '&')
+        {
+          from++;
+          if (*from == '#')
+            {
+              int base = 10;
+              unsigned long l;
+              char *end = NULL;
+
+              from++;
+
+              if (*from == 'x')
+                {
+                  base = 16;
+                  from++;
+                }
+
+              errno = 0;
+              l = strtoul (from, &end, base);
+
+              if (end == from || errno != 0)
+                {
+                  emit_error (context,
+                              _("out of range when resolving character ref"));
+                  return false;
+                }
+              else if (*end != ';')
+                {
+                  emit_error (context,
+                              _("character reference does not end with a ';'"));
+                  return false;
+                }
+              else
+                {
+                  /* characters XML 1.1 permits */
+                  if ((0 < l && l <= 0xD7FF) ||
+                      (0xE000 <= l && l <= 0xFFFD) ||
+                      (0x10000 <= l && l <= 0x10FFFF))
+                    {
+                      char buf[8];
+                      int length;
+                      length = u8_uctomb ((uint8_t *) buf, l, 8);
+                      memcpy (to, buf, length);
+                      to += length - 1;
+                      from = end;
+                      if (l >= 0x80) /* not ascii */
+                        mask |= 0x80;
+                    }
+                  else
+                    {
+                      emit_error (context, _("invalid character reference"));
+                      return false;
+                    }
+                }
+            }
+
+          else if (strncmp (from, "lt;", 3) == 0)
+            {
+              *to = '<';
+              from += 2;
+            }
+          else if (strncmp (from, "gt;", 3) == 0)
+            {
+              *to = '>';
+              from += 2;
+            }
+          else if (strncmp (from, "amp;", 4) == 0)
+            {
+              *to = '&';
+              from += 3;
+            }
+          else if (strncmp (from, "quot;", 5) == 0)
+            {
+              *to = '"';
+              from += 4;
+            }
+          else if (strncmp (from, "apos;", 5) == 0)
+            {
+              *to = '\'';
+              from += 4;
+            }
+          else
+            {
+              if (*from == ';')
+                emit_error (context, _("empty entity '&;'"));
+              else
+                {
+                  const char *end = strchr (from, ';');
+                  if (end)
+                    emit_error (context, _("unknown entity name"));
+                  else
+                    emit_error (context, _("entity does not end with a ';'"));
+                }
+              return false;
+            }
+        }
+    }
+
+  assert (to - string->buffer <= string->buflen);
+  if (to - string->buffer != string->buflen)
+    markup_string_truncate (string, to - string->buffer);
+
+  *is_ascii = !(mask & 0x80);
+
+  return true;
+}
+
+static inline bool
+advance_char (markup_parse_context_ty *context)
+{
+  context->iter++;
+  context->char_number++;
+
+  if (context->iter == context->current_text_end)
+      return false;
+
+  else if (*context->iter == '\n')
+    {
+      context->line_number++;
+      context->char_number = 1;
+    }
+
+  return true;
+}
+
+static inline bool
+xml_isspace (char c)
+{
+  return c == ' ' || c == '\t' || c == '\n' || c == '\r';
+}
+
+static void
+skip_spaces (markup_parse_context_ty *context)
+{
+  do
+    {
+      if (!xml_isspace (*context->iter))
+        return;
+    }
+  while (advance_char (context));
+}
+
+static void
+advance_to_name_end (markup_parse_context_ty *context)
+{
+  do
+    {
+      if (IS_COMMON_NAME_END_CHAR (*(context->iter)))
+        return;
+      if (xml_isspace (*(context->iter)))
+        return;
+    }
+  while (advance_char (context));
+}
+
+static void
+add_to_partial (markup_parse_context_ty *context,
+                const char         *text_start,
+                const char         *text_end)
+{
+  if (context->partial_chunk == NULL)
+    { /* allocate a new chunk to parse into */
+
+      context->partial_chunk = markup_string_new ();
+    }
+
+  if (text_start != text_end)
+    markup_string_append (context->partial_chunk,
+                          text_start, text_end - text_start);
+}
+
+static inline void
+truncate_partial (markup_parse_context_ty *context)
+{
+  if (context->partial_chunk != NULL)
+    string_blank (context->partial_chunk);
+}
+
+static inline const char*
+current_element (markup_parse_context_ty *context)
+{
+  const markup_string_ty *string = gl_list_get_at (context->tag_stack, 0);
+  return string->buffer;
+}
+
+static void
+pop_subparser_stack (markup_parse_context_ty *context)
+{
+  markup_recursion_tracker_ty *tracker;
+
+  assert (gl_list_size (context->subparser_stack) > 0);
+
+  tracker = (markup_recursion_tracker_ty *) gl_list_get_at (context->subparser_stack, 0);
+
+  context->awaiting_pop = true;
+
+  context->user_data = tracker->prev_user_data;
+  context->parser = tracker->prev_parser;
+  context->subparser_element = tracker->prev_element;
+  free (tracker);
+
+  gl_list_remove_at (context->subparser_stack, 0);
+}
+
+static void
+push_partial_as_tag (markup_parse_context_ty *context)
+{
+  gl_list_add_first (context->tag_stack, context->partial_chunk);
+  context->partial_chunk = NULL;
+}
+
+static void
+pop_tag (markup_parse_context_ty *context)
+{
+  gl_list_remove_at (context->tag_stack, 0);
+}
+
+static void
+possibly_finish_subparser (markup_parse_context_ty *context)
+{
+  if (current_element (context) == context->subparser_element)
+    pop_subparser_stack (context);
+}
+
+static void
+ensure_no_outstanding_subparser (markup_parse_context_ty *context)
+{
+  context->awaiting_pop = false;
+}
+
+static void
+add_attribute (markup_parse_context_ty *context, markup_string_ty *string)
+{
+  if (context->cur_attr + 2 >= context->alloc_attrs)
+    {
+      context->alloc_attrs += 5; /* silly magic number */
+      context->attr_names = xrealloc (context->attr_names, sizeof (char *) * context->alloc_attrs);
+      context->attr_values = xrealloc (context->attr_values, sizeof(char *) * context->alloc_attrs);
+    }
+  context->cur_attr++;
+  context->attr_names[context->cur_attr] = xstrdup (string->buffer);
+  context->attr_values[context->cur_attr] = NULL;
+  context->attr_names[context->cur_attr+1] = NULL;
+  context->attr_values[context->cur_attr+1] = NULL;
+}
+
+static void
+clear_attributes (markup_parse_context_ty *context)
+{
+  /* Go ahead and free the attributes. */
+  for (; context->cur_attr >= 0; context->cur_attr--)
+    {
+      int pos = context->cur_attr;
+      free (context->attr_names[pos]);
+      free (context->attr_values[pos]);
+      context->attr_names[pos] = context->attr_values[pos] = NULL;
+    }
+  assert (context->cur_attr == -1);
+  assert (context->attr_names == NULL ||
+          context->attr_names[0] == NULL);
+  assert (context->attr_values == NULL ||
+          context->attr_values[0] == NULL);
+}
+
+static void
+markup_parse_context_push (markup_parse_context_ty *context,
+                           const markup_parser_ty *parser,
+                           void *user_data)
+{
+  markup_recursion_tracker_ty *tracker;
+
+  tracker = XMALLOC (markup_recursion_tracker_ty);
+  tracker->prev_element = context->subparser_element;
+  tracker->prev_parser = context->parser;
+  tracker->prev_user_data = context->user_data;
+
+  context->subparser_element = current_element (context);
+  context->parser = parser;
+  context->user_data = user_data;
+
+  gl_list_add_first (context->subparser_stack, tracker);
+}
+
+static void
+markup_parse_context_pop (markup_parse_context_ty *context)
+{
+  if (!context->awaiting_pop)
+    possibly_finish_subparser (context);
+
+  assert (context->awaiting_pop);
+
+  context->awaiting_pop = false;
+}
+
+/* This has to be a separate function to ensure the alloca's
+ * are unwound on exit - otherwise we grow & blow the stack
+ * with large documents
+ */
+static inline void
+emit_start_element (markup_parse_context_ty *context)
+{
+  int i, j = 0;
+  const char *start_name;
+  const char **attr_names;
+  const char **attr_values;
+
+  /* In case we want to ignore qualified tags and we see that we have
+   * one here, we push a subparser.  This will ignore all tags inside of
+   * the qualified tag.
+   *
+   * We deal with the end of the subparser from emit_end_element.
+   */
+  if ((context->flags & MARKUP_IGNORE_QUALIFIED)
+      && strchr (current_element (context), ':'))
+    {
+      static const markup_parser_ty ignore_parser;
+      markup_parse_context_push (context, &ignore_parser, NULL);
+      clear_attributes (context);
+      return;
+    }
+
+  attr_names = XCALLOC (context->cur_attr + 2, const char *);
+  attr_values = XCALLOC (context->cur_attr + 2, const char *);
+  for (i = 0; i < context->cur_attr + 1; i++)
+    {
+      /* Possibly omit qualified attribute names from the list */
+      if ((context->flags & MARKUP_IGNORE_QUALIFIED)
+          && strchr (context->attr_names[i], ':'))
+        continue;
+
+      attr_names[j] = context->attr_names[i];
+      attr_values[j] = context->attr_values[i];
+      j++;
+    }
+  attr_names[j] = NULL;
+  attr_values[j] = NULL;
+
+  /* Call user callback for element start */
+  start_name = current_element (context);
+
+  if (context->parser->start_element && name_validate (context, start_name))
+    (* context->parser->start_element) (context,
+                                        start_name,
+                                        (const char **)attr_names,
+                                        (const char **)attr_values,
+                                        context->user_data);
+  free (attr_names);
+  free (attr_values);
+  clear_attributes (context);
+}
+
+static void
+emit_end_element (markup_parse_context_ty *context)
+{
+  assert (gl_list_size (context->tag_stack) != 0);
+
+  possibly_finish_subparser (context);
+
+  /* We might have just returned from our ignore subparser */
+  if ((context->flags & MARKUP_IGNORE_QUALIFIED)
+      && strchr (current_element (context), ':'))
+    {
+      markup_parse_context_pop (context);
+      pop_tag (context);
+      return;
+    }
+
+  if (context->parser->end_element)
+    (* context->parser->end_element) (context,
+                                      current_element (context),
+                                      context->user_data);
+
+  ensure_no_outstanding_subparser (context);
+
+  pop_tag (context);
+}
+
+/* Feed some data to the parse context.  The data need not be valid
+   UTF-8; an error will be signaled if it's invalid.  The data need
+   not be an entire document; you can feed a document into the parser
+   incrementally, via multiple calls to this function.  Typically, as
+   you receive data from a network connection or file, you feed each
+   received chunk of data into this function, aborting the process if
+   an error occurs. Once an error is reported, no further data may be
+   fed to the parse context; all errors are fatal.  */
+bool
+markup_parse_context_parse (markup_parse_context_ty *context,
+                            const char *text,
+                            ssize_t text_len)
+{
+  assert (context != NULL);
+  assert (text != NULL);
+  assert (context->state != STATE_ERROR);
+  assert (!context->parsing);
+
+  if (text_len < 0)
+    text_len = strlen (text);
+
+  if (text_len == 0)
+    return true;
+
+  context->parsing = true;
+
+
+  context->current_text = text;
+  context->current_text_len = text_len;
+  context->current_text_end = context->current_text + text_len;
+  context->iter = context->current_text;
+  context->start = context->iter;
+
+  while (context->iter != context->current_text_end)
+    {
+      switch (context->state)
+        {
+        case STATE_START:
+          /* Possible next state: AFTER_OPEN_ANGLE */
+
+          assert (gl_list_size (context->tag_stack) == 0);
+
+          /* whitespace is ignored outside of any elements */
+          skip_spaces (context);
+
+          if (context->iter != context->current_text_end)
+            {
+              if (*context->iter == '<')
+                {
+                  /* Move after the open angle */
+                  advance_char (context);
+
+                  context->state = STATE_AFTER_OPEN_ANGLE;
+
+                  /* this could start a passthrough */
+                  context->start = context->iter;
+
+                  /* document is now non-empty */
+                  context->document_empty = false;
+                }
+              else
+                {
+                  emit_error (context,
+                              _("document must begin with an element"));
+                }
+            }
+          break;
+
+        case STATE_AFTER_OPEN_ANGLE:
+          /* Possible next states: INSIDE_OPEN_TAG_NAME,
+           *  AFTER_CLOSE_TAG_SLASH, INSIDE_PASSTHROUGH
+           */
+          if (*context->iter == '?' ||
+              *context->iter == '!')
+            {
+              /* include < in the passthrough */
+              const char *openangle = "<";
+              add_to_partial (context, openangle, openangle + 1);
+              context->start = context->iter;
+              context->balance = 1;
+              context->state = STATE_INSIDE_PASSTHROUGH;
+            }
+          else if (*context->iter == '/')
+            {
+              /* move after it */
+              advance_char (context);
+
+              context->state = STATE_AFTER_CLOSE_TAG_SLASH;
+            }
+          else if (!IS_COMMON_NAME_END_CHAR (*(context->iter)))
+            {
+              context->state = STATE_INSIDE_OPEN_TAG_NAME;
+
+              /* start of tag name */
+              context->start = context->iter;
+            }
+          else
+            {
+              emit_error (context, _("invalid character after '<'"));
+            }
+          break;
+
+          /* The AFTER_CLOSE_ANGLE state is actually sort of
+           * broken, because it doesn't correspond to a range
+           * of characters in the input stream as the others do,
+           * and thus makes things harder to conceptualize
+           */
+        case STATE_AFTER_CLOSE_ANGLE:
+          /* Possible next states: INSIDE_TEXT, STATE_START */
+          if (gl_list_size (context->tag_stack) == 0)
+            {
+              context->start = NULL;
+              context->state = STATE_START;
+            }
+          else
+            {
+              context->start = context->iter;
+              context->state = STATE_INSIDE_TEXT;
+            }
+          break;
+
+        case STATE_AFTER_ELISION_SLASH:
+          /* Possible next state: AFTER_CLOSE_ANGLE */
+          if (*context->iter == '>')
+            {
+              /* move after the close angle */
+              advance_char (context);
+              context->state = STATE_AFTER_CLOSE_ANGLE;
+              emit_end_element (context);
+            }
+          else
+            {
+              emit_error (context, _("missing '>'"));
+            }
+          break;
+
+        case STATE_INSIDE_OPEN_TAG_NAME:
+          /* Possible next states: BETWEEN_ATTRIBUTES */
+
+          /* if there's a partial chunk then it's the first part of the
+           * tag name. If there's a context->start then it's the start
+           * of the tag name in current_text, the partial chunk goes
+           * before that start though.
+           */
+          advance_to_name_end (context);
+
+          if (context->iter == context->current_text_end)
+            {
+              /* The name hasn't necessarily ended. Merge with
+               * partial chunk, leave state unchanged.
+               */
+              add_to_partial (context, context->start, context->iter);
+            }
+          else
+            {
+              /* The name has ended. Combine it with the partial chunk
+               * if any; push it on the stack; enter next state.
+               */
+              add_to_partial (context, context->start, context->iter);
+              push_partial_as_tag (context);
+
+              context->state = STATE_BETWEEN_ATTRIBUTES;
+              context->start = NULL;
+            }
+          break;
+
+        case STATE_INSIDE_ATTRIBUTE_NAME:
+          /* Possible next states: AFTER_ATTRIBUTE_NAME */
+
+          advance_to_name_end (context);
+          add_to_partial (context, context->start, context->iter);
+
+          /* read the full name, if we enter the equals sign state
+           * then add the attribute to the list (without the value),
+           * otherwise store a partial chunk to be prepended later.
+           */
+          if (context->iter != context->current_text_end)
+            context->state = STATE_AFTER_ATTRIBUTE_NAME;
+          break;
+
+        case STATE_AFTER_ATTRIBUTE_NAME:
+          /* Possible next states: AFTER_ATTRIBUTE_EQUALS_SIGN */
+
+          skip_spaces (context);
+
+          if (context->iter != context->current_text_end)
+            {
+              /* The name has ended. Combine it with the partial chunk
+               * if any; push it on the stack; enter next state.
+               */
+              if (!name_validate (context, context->partial_chunk->buffer))
+                break;
+
+              add_attribute (context, context->partial_chunk);
+
+              markup_string_free (context->partial_chunk, true);
+              context->partial_chunk = NULL;
+              context->start = NULL;
+
+              if (*context->iter == '=')
+                {
+                  advance_char (context);
+                  context->state = STATE_AFTER_ATTRIBUTE_EQUALS_SIGN;
+                }
+              else
+                {
+                  emit_error (context, _("missing '='"));
+                }
+            }
+          break;
+
+        case STATE_BETWEEN_ATTRIBUTES:
+          /* Possible next states: AFTER_CLOSE_ANGLE,
+           * AFTER_ELISION_SLASH, INSIDE_ATTRIBUTE_NAME
+           */
+          skip_spaces (context);
+
+          if (context->iter != context->current_text_end)
+            {
+              if (*context->iter == '/')
+                {
+                  advance_char (context);
+                  context->state = STATE_AFTER_ELISION_SLASH;
+                }
+              else if (*context->iter == '>')
+                {
+                  advance_char (context);
+                  context->state = STATE_AFTER_CLOSE_ANGLE;
+                }
+              else if (!IS_COMMON_NAME_END_CHAR (*(context->iter)))
+                {
+                  context->state = STATE_INSIDE_ATTRIBUTE_NAME;
+                  /* start of attribute name */
+                  context->start = context->iter;
+                }
+              else
+                {
+                  emit_error (context, _("missing '>' or '/'"));
+                }
+
+              /* If we're done with attributes, invoke
+               * the start_element callback
+               */
+              if (context->state == STATE_AFTER_ELISION_SLASH ||
+                  context->state == STATE_AFTER_CLOSE_ANGLE)
+                emit_start_element (context);
+            }
+          break;
+
+        case STATE_AFTER_ATTRIBUTE_EQUALS_SIGN:
+          /* Possible next state: INSIDE_ATTRIBUTE_VALUE_[SQ/DQ] */
+
+          skip_spaces (context);
+
+          if (context->iter != context->current_text_end)
+            {
+              if (*context->iter == '"')
+                {
+                  advance_char (context);
+                  context->state = STATE_INSIDE_ATTRIBUTE_VALUE_DQ;
+                  context->start = context->iter;
+                }
+              else if (*context->iter == '\'')
+                {
+                  advance_char (context);
+                  context->state = STATE_INSIDE_ATTRIBUTE_VALUE_SQ;
+                  context->start = context->iter;
+                }
+              else
+                {
+                  emit_error (context, _("missing opening quote"));
+                }
+            }
+          break;
+
+        case STATE_INSIDE_ATTRIBUTE_VALUE_SQ:
+        case STATE_INSIDE_ATTRIBUTE_VALUE_DQ:
+          /* Possible next states: BETWEEN_ATTRIBUTES */
+          {
+            char delim;
+
+            if (context->state == STATE_INSIDE_ATTRIBUTE_VALUE_SQ)
+              {
+                delim = '\'';
+              }
+            else
+              {
+                delim = '"';
+              }
+
+            do
+              {
+                if (*context->iter == delim)
+                  break;
+              }
+            while (advance_char (context));
+          }
+          if (context->iter == context->current_text_end)
+            {
+              /* The value hasn't necessarily ended. Merge with
+               * partial chunk, leave state unchanged.
+               */
+              add_to_partial (context, context->start, context->iter);
+            }
+          else
+            {
+              bool is_ascii;
+              /* The value has ended at the quote mark. Combine it
+               * with the partial chunk if any; set it for the current
+               * attribute.
+               */
+              add_to_partial (context, context->start, context->iter);
+
+              assert (context->cur_attr >= 0);
+
+              if (unescape_string_inplace (context, context->partial_chunk,
+                                           &is_ascii)
+                  && (is_ascii
+                      || text_validate (context,
+                                        context->partial_chunk->buffer,
+                                        context->partial_chunk->buflen)))
+                {
+                  /* success, advance past quote and set state. */
+                  context->attr_values[context->cur_attr] =
+                    markup_string_free (context->partial_chunk, false);
+                  context->partial_chunk = NULL;
+                  advance_char (context);
+                  context->state = STATE_BETWEEN_ATTRIBUTES;
+                  context->start = NULL;
+                }
+
+              truncate_partial (context);
+            }
+          break;
+
+        case STATE_INSIDE_TEXT:
+          /* Possible next states: AFTER_OPEN_ANGLE */
+          do
+            {
+              if (*context->iter == '<')
+                break;
+            }
+          while (advance_char (context));
+
+          /* The text hasn't necessarily ended. Merge with
+           * partial chunk, leave state unchanged.
+           */
+
+          add_to_partial (context, context->start, context->iter);
+
+          if (context->iter != context->current_text_end)
+            {
+              bool is_ascii;
+
+              /* The text has ended at the open angle. Call the text
+               * callback.
+               */
+              if (unescape_string_inplace (context, context->partial_chunk,
+                                           &is_ascii)
+                  && (is_ascii
+                      || text_validate (context,
+                                        context->partial_chunk->buffer,
+                                        context->partial_chunk->buflen)))
+                {
+                  if (context->parser->text)
+                    (*context->parser->text) (context,
+                                              context->partial_chunk->buffer,
+                                              context->partial_chunk->buflen,
+                                              context->user_data);
+
+                  /* advance past open angle and set state. */
+                  advance_char (context);
+                  context->state = STATE_AFTER_OPEN_ANGLE;
+                  /* could begin a passthrough */
+                  context->start = context->iter;
+                }
+
+              truncate_partial (context);
+            }
+          break;
+
+        case STATE_AFTER_CLOSE_TAG_SLASH:
+          /* Possible next state: INSIDE_CLOSE_TAG_NAME */
+          if (!IS_COMMON_NAME_END_CHAR (*(context->iter)))
+            {
+              context->state = STATE_INSIDE_CLOSE_TAG_NAME;
+
+              /* start of tag name */
+              context->start = context->iter;
+            }
+          else
+            {
+              emit_error (context, _("invalid character after '</'"));
+            }
+          break;
+
+        case STATE_INSIDE_CLOSE_TAG_NAME:
+          /* Possible next state: AFTER_CLOSE_TAG_NAME */
+          advance_to_name_end (context);
+          add_to_partial (context, context->start, context->iter);
+
+          if (context->iter != context->current_text_end)
+            context->state = STATE_AFTER_CLOSE_TAG_NAME;
+          break;
+
+        case STATE_AFTER_CLOSE_TAG_NAME:
+          /* Possible next state: AFTER_CLOSE_TAG_SLASH */
+
+          skip_spaces (context);
+
+          if (context->iter != context->current_text_end)
+            {
+              markup_string_ty *close_name;
+
+              close_name = context->partial_chunk;
+              context->partial_chunk = NULL;
+
+              if (*context->iter != '>')
+                {
+                  emit_error (context,
+                              _("invalid character after a close element name"));
+                }
+              else if (gl_list_size (context->tag_stack) == 0)
+                {
+                  emit_error (context, _("element is closed"));
+                }
+              else if (strcmp (close_name->buffer, current_element (context)) != 0)
+                {
+                  emit_error (context, _("element is closed"));
+                }
+              else
+                {
+                  advance_char (context);
+                  context->state = STATE_AFTER_CLOSE_ANGLE;
+                  context->start = NULL;
+
+                  emit_end_element (context);
+                }
+              context->partial_chunk = close_name;
+              truncate_partial (context);
+            }
+          break;
+
+        case STATE_INSIDE_PASSTHROUGH:
+          /* Possible next state: AFTER_CLOSE_ANGLE */
+          do
+            {
+              if (*context->iter == '<')
+                context->balance++;
+              if (*context->iter == '>')
+                {
+                  char *str;
+                  size_t len;
+
+                  context->balance--;
+                  add_to_partial (context, context->start, context->iter);
+                  context->start = context->iter;
+
+                  str = context->partial_chunk->buffer;
+                  len = context->partial_chunk->buflen;
+
+                  if (str[1] == '?' && str[len - 1] == '?')
+                    break;
+                  if (strncmp (str, "<!--", 4) == 0 &&
+                      strcmp (str + len - 2, "--") == 0)
+                    break;
+                  if (strncmp (str, "<![CDATA[", 9) == 0 &&
+                      strcmp (str + len - 2, "]]") == 0)
+                    break;
+                  if (strncmp (str, "<!DOCTYPE", 9) == 0 &&
+                      context->balance == 0)
+                    break;
+                }
+            }
+          while (advance_char (context));
+
+          if (context->iter == context->current_text_end)
+            {
+              /* The passthrough hasn't necessarily ended. Merge with
+               * partial chunk, leave state unchanged.
+               */
+               add_to_partial (context, context->start, context->iter);
+            }
+          else
+            {
+              /* The passthrough has ended at the close angle. Combine
+               * it with the partial chunk if any. Call the passthrough
+               * callback. Note that the open/close angles are
+               * included in the text of the passthrough.
+               */
+              advance_char (context); /* advance past close angle */
+              add_to_partial (context, context->start, context->iter);
+
+              if (context->flags & MARKUP_TREAT_CDATA_AS_TEXT &&
+                  strncmp (context->partial_chunk->buffer, "<![CDATA[", 9) == 0)
+                {
+                  if (context->parser->text &&
+                      text_validate (context,
+                                     context->partial_chunk->buffer + 9,
+                                     context->partial_chunk->buflen - 12))
+                    (*context->parser->text) (context,
+                                              context->partial_chunk->buffer + 9,
+                                              context->partial_chunk->buflen - 12,
+                                              context->user_data);
+                }
+              else if (context->parser->passthrough &&
+                       text_validate (context,
+                                      context->partial_chunk->buffer,
+                                      context->partial_chunk->buflen))
+                (*context->parser->passthrough) (context,
+                                                 context->partial_chunk->buffer,
+                                                 context->partial_chunk->buflen,
+                                                 context->user_data);
+
+              truncate_partial (context);
+
+              context->state = STATE_AFTER_CLOSE_ANGLE;
+              context->start = context->iter; /* could begin text */
+            }
+          break;
+
+        case STATE_ERROR:
+          goto finished;
+          break;
+
+        default:
+          abort ();
+          break;
+        }
+    }
+
+ finished:
+  context->parsing = false;
+
+  return context->state != STATE_ERROR;
+}
+
+/* Signals to the parse context that all data has been fed into the
+ * parse context with markup_parse_context_parse.
+ *
+ * This function reports an error if the document isn't complete,
+ * for example if elements are still open.  */
+bool
+markup_parse_context_end_parse (markup_parse_context_ty *context)
+{
+  assert (context != NULL);
+  assert (!context->parsing);
+  assert (context->state != STATE_ERROR);
+
+  if (context->partial_chunk != NULL)
+    {
+      markup_string_free (context->partial_chunk, true);
+      context->partial_chunk = NULL;
+    }
+
+  if (context->document_empty)
+    {
+      emit_error (context, _("empty document"));
+      return false;
+    }
+
+  context->parsing = true;
+
+  switch (context->state)
+    {
+    case STATE_START:
+      /* Nothing to do */
+      break;
+
+    case STATE_AFTER_OPEN_ANGLE:
+      emit_error (context,
+                  _("document ended unexpectedly just after '<'"));
+      break;
+
+    case STATE_AFTER_CLOSE_ANGLE:
+      if (gl_list_size (context->tag_stack) > 0)
+        {
+          /* Error message the same as for INSIDE_TEXT */
+          emit_error (context,
+                      _("document ended unexpectedly with elements still open"));
+        }
+      break;
+
+    case STATE_AFTER_ELISION_SLASH:
+      emit_error (context, _("document ended unexpectedly without '>'"));
+      break;
+
+    case STATE_INSIDE_OPEN_TAG_NAME:
+      emit_error (context,
+                  _("document ended unexpectedly inside an element name"));
+      break;
+
+    case STATE_INSIDE_ATTRIBUTE_NAME:
+    case STATE_AFTER_ATTRIBUTE_NAME:
+      emit_error (context,
+                  _("document ended unexpectedly inside an attribute name"));
+      break;
+
+    case STATE_BETWEEN_ATTRIBUTES:
+      emit_error (context,
+                  _("document ended unexpectedly inside an open tag"));
+      break;
+
+    case STATE_AFTER_ATTRIBUTE_EQUALS_SIGN:
+      emit_error (context, _("document ended unexpectedly after '='"));
+      break;
+
+    case STATE_INSIDE_ATTRIBUTE_VALUE_SQ:
+    case STATE_INSIDE_ATTRIBUTE_VALUE_DQ:
+      emit_error (context,
+                  _("document ended unexpectedly inside an attribute value"));
+      break;
+
+    case STATE_INSIDE_TEXT:
+      assert (gl_list_size (context->tag_stack) > 0);
+      emit_error (context,
+                  _("document ended unexpectedly with elements still open"));
+      break;
+
+    case STATE_AFTER_CLOSE_TAG_SLASH:
+    case STATE_INSIDE_CLOSE_TAG_NAME:
+    case STATE_AFTER_CLOSE_TAG_NAME:
+      emit_error (context,
+                  _("document ended unexpectedly inside the close tag"));
+      break;
+
+    case STATE_INSIDE_PASSTHROUGH:
+      emit_error (context,
+                  _("document ended unexpectedly inside a comment or "
+                    "processing instruction"));
+      break;
+
+    case STATE_ERROR:
+    default:
+      abort ();
+      break;
+    }
+
+  context->parsing = false;
+
+  return context->state != STATE_ERROR;
+}
+
+const char *
+markup_parse_context_get_error (markup_parse_context_ty *context)
+{
+  return context->error_text;
+}
diff --git a/gnulib-local/lib/markup.h b/gnulib-local/lib/markup.h
new file mode 100644
index 0000000..61e5b0e
--- /dev/null
+++ b/gnulib-local/lib/markup.h
@@ -0,0 +1,164 @@
+/* markup.h -- simple XML-like string parser
+   Copyright (C) 2015 Free Software Foundation, Inc.
+
+   This file is not part of the GNU gettext program, but is used with
+   GNU gettext.
+
+   This is a stripped down version of GLib's gmarkup.h.  The original
+   copyright notice is as follows:
+ */
+
+/* gmarkup.h - Simple XML-like string parser/writer
+ *
+ *  Copyright 2000 Red Hat, Inc.
+ *
+ * GLib is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 3 of the
+ * License, or (at your option) any later version.
+ *
+ * GLib is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with GLib; see the file COPYING.LIB.  If not,
+ * see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __MARKUP_H__
+#define __MARKUP_H__ 1
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#include <stdbool.h>
+#include <stddef.h>
+#include <sys/types.h>
+
+/**
+ * markup_parse_flags_ty:
+ * @MARKUP_DO_NOT_USE_THIS_UNSUPPORTED_FLAG: flag you should not use
+ * @MARKUP_TREAT_CDATA_AS_TEXT: When this flag is set, CDATA marked
+ *     sections are not passed literally to the @passthrough function of
+ *     the parser. Instead, the content of the section (without the
+ *     `<![CDATA[` and `]]>`) is
+ *     passed to the @text function. This flag was added in GLib 2.12
+ * @MARKUP_PREFIX_ERROR_POSITION: Normally errors caught by GMarkup
+ *     itself have line/column information prefixed to them to let the
+ *     caller know the location of the error. When this flag is set the
+ *     location information is also prefixed to errors generated by the
+ *     #GMarkupParser implementation functions
+ * @MARKUP_IGNORE_QUALIFIED: Ignore (don't report) qualified
+ *     attributes and tags, along with their contents.  A qualified
+ *     attribute or tag is one that contains ':' in its name (ie: is in
+ *     another namespace).  Since: 2.40.
+ *
+ * Flags that affect the behaviour of the parser.
+ */
+typedef enum
+  {
+    MARKUP_DO_NOT_USE_THIS_UNSUPPORTED_FLAG = 1 << 0,
+    MARKUP_TREAT_CDATA_AS_TEXT = 1 << 1,
+    MARKUP_PREFIX_ERROR_POSITION = 1 << 2,
+    MARKUP_IGNORE_QUALIFIED = 1 << 3
+  } markup_parse_flags_ty;
+
+/**
+ * markup_parse_context_ty:
+ *
+ * A parse context is used to parse a stream of bytes that
+ * you expect to contain marked-up text.
+ *
+ * See markup_parse_context_new(), #markup_parser_ty, and so
+ * on for more details.
+ */
+typedef struct _markup_parse_context_ty markup_parse_context_ty;
+typedef struct _markup_parser_ty markup_parser_ty;
+
+/**
+ * markup_parser_ty:
+ * @start_element: Callback to invoke when the opening tag of an element
+ *     is seen. The callback's @attribute_names and @attribute_values parameters
+ *     are %NULL-terminated.
+ * @end_element: Callback to invoke when the closing tag of an element
+ *     is seen. Note that this is also called for empty tags like
+ *     `<empty/>`.
+ * @text: Callback to invoke when some text is seen (text is always
+ *     inside an element). Note that the text of an element may be spread
+ *     over multiple calls of this function. If the
+ *     %MARKUP_TREAT_CDATA_AS_TEXT flag is set, this function is also
+ *     called for the content of CDATA marked sections.
+ * @passthrough: Callback to invoke for comments, processing instructions
+ *     and doctype declarations; if you're re-writing the parsed document,
+ *     write the passthrough text back out in the same position. If the
+ *     %MARKUP_TREAT_CDATA_AS_TEXT flag is not set, this function is also
+ *     called for CDATA marked sections.
+ * @error: Callback to invoke when an error occurs.
+ *
+ * Any of the fields in #markup_parser_ty can be %NULL, in which case they
+ * will be ignored. Except for the @error function, any of these callbacks
+ * can set an error; in particular the %MARKUP_ERROR_UNKNOWN_ELEMENT,
+ * %MARKUP_ERROR_UNKNOWN_ATTRIBUTE, and %MARKUP_ERROR_INVALID_CONTENT
+ * errors are intended to be set from these callbacks. If you set an error
+ * from a callback, markup_parse_context_parse() will report that error
+ * back to its caller.
+ */
+struct _markup_parser_ty
+{
+  /* Called for open tags <foo bar="baz"> */
+  bool (*start_element) (markup_parse_context_ty *context,
+                         const char *element_name,
+                         const char **attribute_names,
+                         const char **attribute_values,
+                         void *user_data);
+
+  /* Called for close tags </foo> */
+  bool (*end_element) (markup_parse_context_ty *context,
+                       const char *element_name,
+                       void *user_data);
+
+  /* Called for character data */
+  /* text is not nul-terminated */
+  bool (*text) (markup_parse_context_ty *context,
+                const char *text,
+                size_t text_len,
+                void *user_data);
+
+  /* Called for strings that should be re-saved verbatim in this same
+   * position, but are not otherwise interpretable.  At the moment
+   * this includes comments and processing instructions.
+   */
+  /* text is not nul-terminated. */
+  bool (*passthrough) (markup_parse_context_ty *context,
+                       const char *passthrough_text,
+                       size_t text_len,
+                       void *user_data);
+
+  /* Called on error, including one set by other
+   * methods in the vtable. The GError should not be freed.
+   */
+  void (*error) (markup_parse_context_ty *context,
+                 const char *error_text,
+                 void *user_data);
+};
+
+extern markup_parse_context_ty *
+       markup_parse_context_new (const markup_parser_ty *parser,
+                                 markup_parse_flags_ty flags,
+                                 void *user_data);
+extern void markup_parse_context_free (markup_parse_context_ty *context);
+extern bool markup_parse_context_parse (markup_parse_context_ty *context,
+                                        const char *text,
+                                        ssize_t text_len);
+extern bool markup_parse_context_end_parse (markup_parse_context_ty *context);
+extern const char *
+       markup_parse_context_get_error (markup_parse_context_ty *context);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* __MARKUP_H__ */
diff --git a/gnulib-local/modules/markup b/gnulib-local/modules/markup
new file mode 100644
index 0000000..8d969d4
--- /dev/null
+++ b/gnulib-local/modules/markup
@@ -0,0 +1,31 @@
+Description:
+Simple XML-like parser
+
+Files:
+lib/markup.h
+lib/markup.c
+
+Depends-on:
+c-ctype
+linked-list
+unistr/u8-mbtouc
+unistr/u8-next
+unictype/ctype-alpha
+xalloc
+xlist
+xvasprintf
+
+configure.ac:
+
+Makefile.am:
+lib_SOURCES += markup.h markup.c
+
+Include:
+"markup.h"
+
+License:
+LGPL
+
+Maintainer:
+Daiki Ueno
+