Re: Included libraries in the Windows binary distribution

[Eli Zaretskii <eliz@gnu.org>]

> Date: Wed, 04 Jan 2012 00:09:51 -0500
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: emacs-devel@gnu.org
> 
> > From: Lars Magne Ingebrigtsen <larsi@gnus.org>
> > Date: Wed, 04 Jan 2012 00:55:08 +0100
> > 
> > There's been a lot of discussion about including libgnutls with the
> > Emacs binaries, and I'm obviously all for that.  But nobody has
> > mentioned libxml2.  I'd rather like to have that included, too, so that
> > people can read HTML in Emacs.
> 
> Making it possible to build Emacs on Windows with libxml2 is a
> standing todo item that I hope will be done soon.

The changes to support libxml2 in the Windows build are below.  Would
people who build Emacs on Windows please try them?  Precompiled
Windows binaries of libxml2 can be found here:

    http://sourceforge.net/projects/ezwinports/files/

Note that if you install the binaries from the above site, you need to
use "--cflags -I/path/to/include/libxml2" when invoking configure.bat,
because libxml2 headers are installed into the include/libxml2
directory.  Other distributions may use other directories.  (I could
have used pkg-config to figure that out automatically, but I'm
reluctant to request users to install pkg-config as prerequisite for
building Emacs.  Besides, at least one binary distro of libxml2 for
Windows I saw doesn't include the libxml-2.0.pc file that pkg-config
needs to do its job.)

Stefan and Chong: assuming that these changes are reported to work
well, is it safe to install them now, or do we want to postpone them
until after 24.1?


=== modified file 'lisp/ChangeLog'
--- a/lisp/ChangeLog	2012-02-04 13:25:57 +0000
+++ b/lisp/ChangeLog	2012-02-04 13:27:51 +0000
@@ -1,3 +1,7 @@
+2012-02-04  Eli Zaretskii  <eliz@gnu.org>
+
+	* term/w32-win.el (dynamic-library-alist): Add libxml2 DLLs.
+
 2012-02-04  Lars Ljung  <lars@matholka.se>  (tiny change)
 
 	* eshell/esh-ext.el (eshell-windows-shell-file): Match "cmdproxy"

=== modified file 'lisp/term/w32-win.el'
--- a/lisp/term/w32-win.el	2012-02-04 13:25:57 +0000
+++ b/lisp/term/w32-win.el	2012-02-04 13:27:51 +0000
@@ -210,7 +210,8 @@
        '(gdk-pixbuf "libgdk_pixbuf-2.0-0.dll")
        '(glib "libglib-2.0-0.dll")
        '(gobject "libgobject-2.0-0.dll")
-       '(gnutls "libgnutls-28.dll" "libgnutls-26.dll")))
+       '(gnutls "libgnutls-28.dll" "libgnutls-26.dll")
+       '(libxml2 "libxml2-2.dll" "libxml2.dll")))
 
 ;;; multi-tty support
 (defvar w32-initialized nil

=== modified file 'nt/ChangeLog'
--- a/nt/ChangeLog	2012-02-04 13:25:57 +0000
+++ b/nt/ChangeLog	2012-02-04 13:27:51 +0000
@@ -1,5 +1,10 @@
 2012-02-04  Eli Zaretskii  <eliz@gnu.org>
 
+	* configure.bat: Support building with libxml2.
+
+	* INSTALL:
+	* README.W32: Add information about libxml2.
+
 	* inc/sys/stat.h (_STAT_DEFINED): Define, to prevent redefinitions
 	by other headers.
 

=== modified file 'nt/INSTALL'
--- a/nt/INSTALL	2012-02-04 13:25:57 +0000
+++ b/nt/INSTALL	2012-02-04 13:27:51 +0000
@@ -367,6 +367,20 @@
   You can get pre-built binaries (including any required DLL and the
   header files) at http://sourceforge.net/projects/ezwinports/files/.
 
+* Optional libxml2 support
+
+  If configure.bat finds the libxml/HTMLparser.h file in the include path,
+  Emacs is built with libxml2 support by default; to avoid that you can
+  pass the argument --without-libxml2.
+
+  In order to support libxml2 at runtime, a libxml2-enabled Emacs must
+  be able to find the relevant DLLs during startup; failure to do so
+  is not an error, but libxml2 features won't be available to the
+  running session.
+
+  You can get pre-built binaries (including any required DLL and the
+  header files) at http://sourceforge.net/projects/ezwinports/files/.
+
 * Experimental SVG support
 
   SVG support is currently experimental, and not built by default.

=== modified file 'nt/README.W32'
--- a/nt/README.W32	2012-02-04 13:25:57 +0000
+++ b/nt/README.W32	2012-02-04 13:27:51 +0000
@@ -169,6 +169,16 @@
   You can get pre-built binaries (including any required DLL and the
   header files) at http://sourceforge.net/projects/ezwinports/files/.
 
+* libxml2 support
+
+  In order to support libxml2 at runtime, a libxml2-enabled Emacs must
+  be able to find the relevant DLLs during startup; failure to do so
+  is not an error, but libxml2 features won't be available to the
+  running session.
+
+  You can get pre-built binaries (including any required DLL and the
+  header files) at http://sourceforge.net/projects/ezwinports/files/.
+
 * Uninstalling Emacs
 
   If you should need to uninstall Emacs, simply delete all the files

=== modified file 'nt/configure.bat'
--- a/nt/configure.bat	2012-02-04 13:25:57 +0000
+++ b/nt/configure.bat	2012-02-04 13:27:51 +0000
@@ -131,6 +131,7 @@
 if "%1" == "--without-gif" goto withoutgif
 if "%1" == "--without-tiff" goto withouttiff
 if "%1" == "--without-gnutls" goto withoutgnutls
+if "%1" == "--without-libxml2" goto withoutlibxml2
 if "%1" == "--without-xpm" goto withoutxpm
 if "%1" == "--with-svg" goto withsvg
 if "%1" == "--distfiles" goto distfiles
@@ -156,6 +157,7 @@
 echo.   --without-tiff          do not use TIFF library even if it is installed
 echo.   --without-xpm           do not use XPM library even if it is installed
 echo.   --without-gnutls        do not use GnuTLS library even if it is installed
+echo.   --without-libxml2       do not use libxml2 library even if it is installed
 echo.   --with-svg              use the RSVG library (experimental)
 echo.   --distfiles             path to files for make dist, e.g. libXpm.dll
 if "%use_extensions%" == "0" goto end
@@ -317,6 +319,14 @@
 
 rem ----------------------------------------------------------------------
 
+:withoutlibxml2
+set libxml2support=N
+set HAVE_LIBXML2=
+shift
+goto again
+
+rem ----------------------------------------------------------------------
+
 :withouttiff
 set tiffsupport=N
 set HAVE_TIFF=
@@ -569,6 +579,28 @@
 :tlsDone
 rm -f junk.c junk.obj
 
+if (%libxml2support%) == (N) goto xml2Done
+
+echo Checking for libxml2....
+echo #include "libxml/HTMLparser.h" >junk.c
+echo main(){} >>junk.c
+echo %COMPILER% %usercflags% %mingwflag% -c junk.c -o junk.obj >>config.log
+%COMPILER% %usercflags% %mingwflag% -c junk.c -o junk.obj >junk.out 2>>config.log
+if exist junk.obj goto havelibxml2
+
+echo ...libxml/HTMLparser.h not found, building without libxml2 support
+echo The failed program was: >>config.log
+type junk.c >>config.log
+set HAVE_LIBXML2=
+goto xml2Done
+
+:havelibxml2
+echo ...libxml2 header available, building with libxml2 support
+set HAVE_LIBXML2=1
+
+:xml2Done
+rm -f junk.c junk.obj
+
 if (%jpegsupport%) == (N) goto jpegDone
 
 echo Checking for jpeg-6b...
@@ -757,6 +789,7 @@
 if (%profile%) == (Y) echo #define PROFILING 1 >>config.tmp
 if not "(%HAVE_PNG%)" == "()" echo #define HAVE_PNG 1 >>config.tmp
 if not "(%HAVE_GNUTLS%)" == "()" echo #define HAVE_GNUTLS 1 >>config.tmp
+if not "(%HAVE_LIBXML2%)" == "()" echo #define HAVE_LIBXML2 1 >>config.tmp
 if not "(%HAVE_JPEG%)" == "()" echo #define HAVE_JPEG 1 >>config.tmp
 if not "(%HAVE_GIF%)" == "()" echo #define HAVE_GIF 1 >>config.tmp
 if not "(%HAVE_TIFF%)" == "()" echo #define HAVE_TIFF 1 >>config.tmp
@@ -896,6 +929,7 @@
 set distFilesOk=
 set pngsupport=
 set tlssupport=
+set libxml2support=
 set jpegsupport=
 set gifsupport=
 set tiffsupport=

=== modified file 'src/ChangeLog'
--- a/src/ChangeLog	2012-02-04 13:25:57 +0000
+++ b/src/ChangeLog	2012-02-04 13:27:51 +0000
@@ -1,3 +1,29 @@
+2012-02-04  Eli Zaretskii  <eliz@gnu.org>
+
+	Support building on MS-Windows with libxml2.
+
+	* makefile.w32-in (OBJ2): Add xml.$(O).
+	(GLOBAL_SOURCES): Add xml.c.
+	($(BLD)/xml.$(O)): New dependency list.
+
+	* xml.c (DEF_XML2_FN, LOAD_XML2_FN) [WINDOWSNT]: New macros.
+	(fn_htmlReadMemory, fn_xmlReadMemory, fn_xmlDocGetRootElement)
+	(fn_xmlFreeDoc, fn_xmlCleanupParser, fn_xmlCheckVersion)
+	[!WINDOWSNT]: New macros.
+	(init_libxml2_functions): New function.
+	(parse_region): Call fn_xmlCheckVersion instead of using the macro
+	LIBXML_TEST_VERSION.  Call libxml2 functions via the fn_* macros.
+	(xml_cleanup_parser): New function, export for fn_xmlCleanupParser.
+	(Flibxml_parse_html_region, Flibxml_parse_xml_region): Call
+	init_libxml2_functions before calling libxml2 functions.
+	(syms_of_xml) <Qlibxml2_dll>: DEFSYM it.
+
+	* emacs.c: Don't include libxml/parser.h.
+	(shut_down_emacs): Call xml_cleanup_parser, instead of calling
+	xmlCleanupParser directly.
+
+	* lisp.h [HAVE_LIBXML2]: Add prototype for xml_cleanup_parser.
+
 2012-02-04  Martin Rudalics  <rudalics@gmx.at>
 
 	* dispnew.c (change_frame_size_1): Calculate new_frame_total_cols

=== modified file 'src/emacs.c'
--- a/src/emacs.c	2012-02-04 13:25:57 +0000
+++ b/src/emacs.c	2012-02-04 13:27:51 +0000
@@ -82,10 +82,6 @@
 #include <sys/personality.h>
 #endif
 
-#ifdef HAVE_LIBXML2
-#include <libxml/parser.h>
-#endif
-
 #ifndef O_RDWR
 #define O_RDWR 2
 #endif
@@ -2124,7 +2120,7 @@
 #endif
 
 #ifdef HAVE_LIBXML2
-  xmlCleanupParser ();
+  xml_cleanup_parser ();
 #endif
 }
 

=== modified file 'src/lisp.h'
--- a/src/lisp.h	2012-02-04 13:25:57 +0000
+++ b/src/lisp.h	2012-02-04 13:27:51 +0000
@@ -3542,6 +3542,7 @@
 #ifdef HAVE_LIBXML2
 /* Defined in xml.c */
 extern void syms_of_xml (void);
+extern void xml_cleanup_parser (void);
 #endif
 
 #ifdef HAVE_MENUS

=== modified file 'src/makefile.w32-in'
--- a/src/makefile.w32-in	2012-02-04 13:25:57 +0000
+++ b/src/makefile.w32-in	2012-02-04 13:27:51 +0000
@@ -128,6 +128,7 @@
 	$(BLD)/image.$(O)		\
 	$(BLD)/terminal.$(O)            \
 	$(BLD)/menu.$(O)		\
+	$(BLD)/xml.$(O)			\
 	$(BLD)/w32term.$(O)		\
 	$(BLD)/w32xfns.$(O)		\
 	$(BLD)/w32fns.$(O)		\
@@ -225,7 +226,7 @@
 	process.c callproc.c unexw32.c \
 	region-cache.c sound.c atimer.c \
 	doprnt.c intervals.c textprop.c composite.c \
-	gnutls.c
+	gnutls.c xml.c
 SOME_MACHINE_OBJECTS = dosfns.o msdos.o \
 	xterm.o xfns.o xmenu.o xselect.o xrdb.o xsmfns.o dbusbind.o
 obj = $(GLOBAL_SOURCES:.c=.o)
@@ -940,6 +941,13 @@
 	$(LISP_H) \
 	$(PROCESS_H)
 
+$(BLD)/xml.$(O) : \
+	$(SRC)/xml.c \
+	$(SRC)/w32.h \
+	$(CONFIG_H) \
+	$(LISP_H) \
+	$(SRC)/buffer.h
+
 $(BLD)/image.$(O) : \
 	$(SRC)/image.c \
 	$(SRC)/epaths.h \

=== modified file 'src/xml.c'
--- a/src/xml.c	2012-02-04 13:25:57 +0000
+++ b/src/xml.c	2012-02-04 13:27:51 +0000
@@ -28,6 +28,83 @@
 #include "lisp.h"
 #include "buffer.h"
 
+\f
+#ifdef WINDOWSNT
+
+#include <windows.h>
+#include "w32.h"
+
+/* Macro for defining functions that will be loaded from the libxml2 DLL.  */
+#define DEF_XML2_FN(rettype,func,args) static rettype (FAR CDECL *fn_##func)args
+
+/* Macro for loading libxml2 functions from the library.  */
+#define LOAD_XML2_FN(lib,func) {					\
+    fn_##func = (void *) GetProcAddress (lib, #func);			\
+    if (!fn_##func) goto bad_library;					\
+  }
+
+DEF_XML2_FN (htmlDocPtr, htmlReadMemory,
+	     (const char *, int, const char *, const char *, int));
+DEF_XML2_FN (xmlDocPtr, xmlReadMemory,
+	     (const char *, int, const char *, const char *, int));
+DEF_XML2_FN (xmlNodePtr, xmlDocGetRootElement, (xmlDocPtr));
+DEF_XML2_FN (void, xmlFreeDoc, (xmlDocPtr));
+DEF_XML2_FN (void, xmlCleanupParser, (void));
+DEF_XML2_FN (void, xmlCheckVersion, (int));
+
+#else  /* !WINDOWSNT */
+
+#define fn_htmlReadMemory       htmlReadMemory
+#define fn_xmlReadMemory        xmlReadMemory
+#define fn_xmlDocGetRootElement xmlDocGetRootElement
+#define fn_xmlFreeDoc           xmlFreeDoc
+#define fn_xmlCleanupParser     xmlCleanupParser
+#define fn_xmlCheckVersion      xmlCheckVersion
+
+#endif	/* !WINDOWSNT */
+
+static Lisp_Object Qlibxml2_dll;
+
+static int
+init_libxml2_functions (Lisp_Object libraries)
+{
+#ifdef WINDOWSNT
+  Lisp_Object found = Fassq (Qlibxml2_dll, Vlibrary_cache);
+
+  if (CONSP (found))
+    return EQ (XCDR (found), Qt) ? 1 : 0;
+  else
+    {
+      HMODULE library;
+
+      if (!(library = w32_delayed_load (libraries, Qlibxml2_dll)))
+	{
+	  message ("%s", "libxml2 library not found");
+	  return 0;
+	}
+
+      /* LOAD_XML2_FN jumps to bad_library if it fails to find the
+	 named function.  */
+      LOAD_XML2_FN (library, htmlReadMemory);
+      LOAD_XML2_FN (library, xmlReadMemory);
+      LOAD_XML2_FN (library, xmlDocGetRootElement);
+      LOAD_XML2_FN (library, xmlFreeDoc);
+      LOAD_XML2_FN (library, xmlCleanupParser);
+      LOAD_XML2_FN (library, xmlCheckVersion);
+
+      Vlibrary_cache = Fcons (Fcons (Qlibxml2_dll, Qt), Vlibrary_cache);
+      return 1;
+    }
+
+ bad_library:
+  Vlibrary_cache = Fcons (Fcons (Qlibxml2_dll, Qnil), Vlibrary_cache);
+
+  return 0;
+#else  /* !WINDOWSNT */
+  return 1;
+#endif	/* !WINDOWSNT */
+}
+
 static Lisp_Object
 make_dom (xmlNode *node)
 {
@@ -92,7 +169,7 @@
   EMACS_INT bytes;
   EMACS_INT istart, iend;
 
-  LIBXML_TEST_VERSION;
+  fn_xmlCheckVersion (LIBXML_VERSION);
 
   validate_region (&start, &end);
 
@@ -111,16 +188,16 @@
   bytes = CHAR_TO_BYTE (iend) - CHAR_TO_BYTE (istart);
 
   if (htmlp)
-    doc = htmlReadMemory ((char *) BYTE_POS_ADDR (CHAR_TO_BYTE (istart)),
-			  bytes, burl, "utf-8",
-			  HTML_PARSE_RECOVER|HTML_PARSE_NONET|
-			  HTML_PARSE_NOWARNING|HTML_PARSE_NOERROR|
-			  HTML_PARSE_NOBLANKS);
+    doc = fn_htmlReadMemory ((char *) BYTE_POS_ADDR (CHAR_TO_BYTE (istart)),
+			     bytes, burl, "utf-8",
+			     HTML_PARSE_RECOVER|HTML_PARSE_NONET|
+			     HTML_PARSE_NOWARNING|HTML_PARSE_NOERROR|
+			     HTML_PARSE_NOBLANKS);
   else
-    doc = xmlReadMemory ((char *) BYTE_POS_ADDR (CHAR_TO_BYTE (istart)),
-			 bytes, burl, "utf-8",
-			 XML_PARSE_NONET|XML_PARSE_NOWARNING|
-			 XML_PARSE_NOBLANKS |XML_PARSE_NOERROR);
+    doc = fn_xmlReadMemory ((char *) BYTE_POS_ADDR (CHAR_TO_BYTE (istart)),
+			    bytes, burl, "utf-8",
+			    XML_PARSE_NONET|XML_PARSE_NOWARNING|
+			    XML_PARSE_NOBLANKS |XML_PARSE_NOERROR);
 
   if (doc != NULL)
     {
@@ -139,19 +216,26 @@
       if (NILP (result)) {
 	/* The document isn't just comments, so get the tree the
 	   proper way. */
-	xmlNode *node = xmlDocGetRootElement (doc);
+	xmlNode *node = fn_xmlDocGetRootElement (doc);
 	if (node != NULL)
 	  result = make_dom (node);
       } else
 	result = Fcons (intern ("top"),
 			Fcons (Qnil, Fnreverse (Fcons (r, result))));
 
-      xmlFreeDoc (doc);
+      fn_xmlFreeDoc (doc);
     }
 
   return result;
 }
 
+void
+xml_cleanup_parser (void)
+{
+  if (init_libxml2_functions (Vdynamic_library_alist))
+    fn_xmlCleanupParser ();
+}
+
 DEFUN ("libxml-parse-html-region", Flibxml_parse_html_region,
        Slibxml_parse_html_region,
        2, 3, 0,
@@ -159,7 +243,9 @@
 If BASE-URL is non-nil, it is used to expand relative URLs.  */)
   (Lisp_Object start, Lisp_Object end, Lisp_Object base_url)
 {
-  return parse_region (start, end, base_url, 1);
+  if (init_libxml2_functions (Vdynamic_library_alist))
+    return parse_region (start, end, base_url, 1);
+  return Qnil;
 }
 
 DEFUN ("libxml-parse-xml-region", Flibxml_parse_xml_region,
@@ -169,7 +255,9 @@
 If BASE-URL is non-nil, it is used to expand relative URLs.  */)
   (Lisp_Object start, Lisp_Object end, Lisp_Object base_url)
 {
-  return parse_region (start, end, base_url, 0);
+  if (init_libxml2_functions (Vdynamic_library_alist))
+    return parse_region (start, end, base_url, 0);
+  return Qnil;
 }
 
 \f
@@ -181,6 +269,8 @@
 {
   defsubr (&Slibxml_parse_html_region);
   defsubr (&Slibxml_parse_xml_region);
+
+  DEFSYM (Qlibxml2_dll, "libxml2");
 }
 
 #endif /* HAVE_LIBXML2 */