Bug 824: Disable combining characters unless --enable-combining.

Label this as an experimental feature because it has so many bugs
and it is not clear how they can be fixed.

configure.in

@@ -305,7 +305,7 @@ AC_CHECK_FUNCS(setenv putenv, HAVE_SETENV_OR_PUTENV=yes)
 AC_CHECK_FUNCS(getuid, HAVE_GETUID=yes)
 AC_CHECK_FUNCS(geteuid, HAVE_GETEUID=yes)
 
-AC_CHECK_FUNCS(wcwidth)
+AC_CHECK_FUNCS(wcwidth, HAVE_WCWIDTH=yes)
 
 dnl These aren't probably needed now, as they are commented in links.h.
 dnl I've no idea about their historical background, but I keep them here
@@ -1338,6 +1338,9 @@ EL_ARG_ENABLE(CONFIG_SMALL, small, [Small binary],
 EL_ARG_ENABLE(CONFIG_UTF8, utf-8, [UTF-8],
 	      [  --disable-utf-8         disable UTF-8 support])
 
+EL_ARG_DEPEND(CONFIG_COMBINE, combining, [CONFIG_UTF8:yes HAVE_WCWIDTH:yes], [Combining characters],
+	      [  --enable-combining      support Unicode combining characters (experimental)])
+
 
 AC_ARG_ENABLE(weehoofooboomookerchoo,
               [

features.conf

@@ -621,7 +621,7 @@ CONFIG_SMALL=no
 # support for double-width characters (like Japanese, etc.).
 #
 # Some features of Unicode are not handled at all. Combining characters is
-# most visible absence.
+# most visible absence; but see CONFIG_COMBINE below.
 # Some features are partially supported. Like line breaking between
 # double-width characters. There is no other detection for determining when to
 # break or not.
@@ -633,6 +633,42 @@ CONFIG_SMALL=no
 CONFIG_UTF8=yes
 
 
+### Unicode combining characters support
+#
+# Extends CONFIG_UTF8 with spotty support for combining characters
+# such as U+0303 COMBINING TILDE.
+#
+# This feature is experimental and has been filed as enhancement 824.
+# Known bugs and weaknesses:
+#
+# - It assumes wcwidth(wc)==0 means wc is a combining character.
+#   However, wcwidth also returns 0 for various control characters
+#   (e.g. U+200E LEFT-TO-RIGHT MARK), and apparently returns -1 if
+#   LC_CTYPE does not support the wide character.  Besides, wchar_t
+#   might not be Unicode at all.  ELinks should instead use Unicode
+#   character properties, perhaps via ICU.
+#
+# - It assumes all combining characters are nonspacing.
+#
+# - It works only if the terminal is using the UTF-8 charset.
+#
+# - It allocates an internal code for each different combining
+#   character sequence.  A malicious web page could easily use up all
+#   the available codes, and the ELinks process would thenceforth be
+#   unable to display any new sequences.
+#
+# - It does not understand canonical equivalences.
+#
+# - Combining characters work only in HTML text.  They do not work in
+#   HTML forms, HTML links, HTML document titles, plain text, menus,
+#   dialog boxes, or keymaps.
+#
+# - Combining characters at the end of the document do not take effect.
+#
+# Default: disabled
+
+CONFIG_COMBINE=no
+
 
 ### Back-trace Printing
 #

src/config/options.inc

@@ -864,10 +864,12 @@ static struct option_info config_options_info[] = {
 		"only the subset of UTF-8 according to terminal codepage is used.\n"
 		"ELinks ignores this option if the terminal codepage is UTF-8.")),
 
+#ifdef CONFIG_COMBINE
 	INIT_OPT_BOOL("terminal._template_", N_("Combining characters"),
 		"combine", 0, 0,
 		N_("Enable combining characters. It works only with "
 		"the xterm in UTF-8 mode.")),
+#endif
 
 	INIT_OPT_BOOL("terminal._template_", N_("Restrict frames in cp850/852"),
 		"restrict_852", 0, 0,

src/dialogs/options.c

@@ -91,7 +91,9 @@ enum termopt {
 	TERM_OPT_UTF_8_IO,
 	TERM_OPT_TRANSPARENCY,
 	TERM_OPT_UNDERLINE,
+#ifdef CONFIG_COMBINE
 	TERM_OPT_COMBINE,
+#endif
 
 	TERM_OPTIONS,
 };
@@ -105,7 +107,9 @@ static struct option_resolver resolvers[] = {
 	{ TERM_OPT_TRANSPARENCY, "transparency"	},
 	{ TERM_OPT_UTF_8_IO,	 "utf_8_io"	},
 	{ TERM_OPT_UNDERLINE,	 "underline"	},
+#ifdef CONFIG_COMBINE
 	{ TERM_OPT_COMBINE,	 "combine"	},
+#endif
 };
 
 static widget_handler_status_T
@@ -151,7 +155,7 @@ push_save_button(struct dialog_data *dlg_data, struct widget_data *button)
 #define	RADIO_TRUE 0
 #endif
 
-#define TERMOPT_WIDGETS_COUNT (20 + RADIO_88 + RADIO_256 + RADIO_TRUE)
+#define TERMOPT_WIDGETS_COUNT (12 + TERM_OPTIONS + RADIO_88 + RADIO_256 + RADIO_TRUE)
 
 #define TERM_OPTION_VALUE_SIZE (sizeof(union option_value) * TERM_OPTIONS)
 
@@ -232,7 +236,9 @@ terminal_options(struct terminal *term, void *xxx, struct session *ses)
 	add_dlg_checkbox(dlg, _("Transparency", term), &values[TERM_OPT_TRANSPARENCY].number);
 	add_dlg_checkbox(dlg, _("Underline", term), &values[TERM_OPT_UNDERLINE].number);
 	add_dlg_checkbox(dlg, _("UTF-8 I/O", term), &values[TERM_OPT_UTF_8_IO].number);
+#ifdef CONFIG_COMBINE
 	add_dlg_checkbox(dlg, _("Combining characters", term), &values[TERM_OPT_COMBINE].number);
+#endif
 
 	add_dlg_button(dlg, _("~OK", term), B_ENTER, push_ok_button, NULL);
 	if (!anonymous)

src/document/document.c

@@ -57,7 +57,7 @@ init_document(struct cache_entry *cached, struct document_options *options)
 	init_list(document->onload_snippets);
 #endif
 
-#ifdef CONFIG_UTF8
+#ifdef CONFIG_COMBINE
 	document->comb_x = -1;
 	document->comb_y = -1;
 #endif

src/document/document.h

@@ -201,6 +201,8 @@ struct document {
 #ifdef CONFIG_UTF8
 	unsigned char buf[7];
 	unsigned char buf_length;
+#endif
+#ifdef CONFIG_COMBINE
 	/* base char + 5 combining chars = 6 */
 	unicode_val_T combi[UCS_MAX_LENGTH_COMBINED];
 	/* the number of combining characters. The base char is not counted. */

src/document/html/renderer.c

@@ -4,15 +4,18 @@
 #include "config.h"
 #endif
 
-#if defined(HAVE_WCHAR_H) && defined(HAVE_WCWIDTH)
-#define _XOPEN_SOURCE 500
+/* Our current implementation of combining characters requires
+ * wcwidth().  Therefore the configure script should have disabled
+ * CONFIG_COMBINE if wcwidth() doesn't exist.  */
+#ifdef CONFIG_COMBINE
+#define _XOPEN_SOURCE 500	/* for wcwidth */
 #endif
 
 #include <ctype.h>
 #include <stdarg.h>
 #include <string.h>
 
-#if defined(HAVE_WCHAR_H) && defined(HAVE_WCWIDTH)
+#ifdef HAVE_WCHAR_H
 #include <wchar.h>
 #endif
 
@@ -489,7 +492,8 @@ good_char:
 				if (data == UCS_NO_BREAK_SPACE
 				    && html_context->options->wrap_nbsp)
 					data = UCS_SPACE;
-#ifdef HAVE_WCWIDTH
+
+#ifdef CONFIG_COMBINE
 				if (wcwidth((wchar_t)data)) {
 					if (document->combi_length) {
 						if (document->comb_x != -1) {
@@ -521,7 +525,7 @@ good_char:
 					part->char_width[x] = unicode_to_cell(data);
 					schar->data = (unicode_val_T)data;
 				}
-#ifdef HAVE_WCWIDTH
+#ifdef CONFIG_COMBINE
 				document->comb_x = x;
 				document->comb_y = y;
 #endif

src/intl/charsets.c

@@ -770,8 +770,10 @@ cp_to_unicode(int codepage, unsigned char **string, unsigned char *end)
 	++*string;
 	return ret;
 }
+#endif	/* CONFIG_UTF8 */
 
 
+#ifdef CONFIG_COMBINE
 unicode_val_T last_combined = UCS_BEGIN_COMBINED - 1;
 unicode_val_T **combined;
 struct hash *combined_hash;
@@ -830,7 +832,8 @@ free_combined()
 		mem_free(combined[i]);
 	mem_free_if(combined);
 }
-#endif	/* CONFIG_UTF8 */
+#endif /* CONFIG_COMBINE */
+
 
 static void
 add_utf8(struct conv_table *ct, unicode_val_T u, const unsigned char *str)
@@ -1540,4 +1543,3 @@ is_cp_utf8(int cp_index)
 	cp_index &= ~SYSTEM_CHARSET_FLAG;
 	return is_cp_ptr_utf8(&codepages[cp_index]);
 }
-

src/intl/charsets.h

@@ -27,12 +27,14 @@ typedef uint32_t unicode_val_T;
  * for the second cell of a double-cell character.  */
 #define UCS_NO_CHAR ((unicode_val_T) 0xFFFFFFFD)
 
+#ifdef CONFIG_COMBINE
 #define UCS_END_COMBINED ((unicode_val_T) 0xFFFFFFFC)
 
 #define UCS_BEGIN_COMBINED ((unicode_val_T) (UCS_END_COMBINED - (unicode_val_T) 10000))
 
 /* Base character and up to 5 combining characters. */
 #define UCS_MAX_LENGTH_COMBINED 6
+#endif /* CONFIG_COMBINE */
 
 /* If ELinks should display a double-cell character but there is only
  * one cell available, it displays this character instead.  This must
@@ -154,15 +156,15 @@ unicode_val_T unicode_fold_label_case(unicode_val_T);
 inline int strlen_utf8(unsigned char **);
 inline unicode_val_T utf8_to_unicode(unsigned char **, const unsigned char *);
 unicode_val_T cp_to_unicode(int, unsigned char **, unsigned char *);
+#endif /* CONFIG_UTF8 */
 
-
+#ifdef CONFIG_COMBINE
 extern unicode_val_T last_combined;
 extern unicode_val_T **combined;
 extern struct hash *combined_hash;
 unicode_val_T get_combined(unicode_val_T *, int);
 void free_combined();
-
-#endif /* CONFIG_UTF8 */
+#endif /* CONFIG_COMBINE */
 
 unicode_val_T cp2u(int, unsigned char);
 const unsigned char *cp2utf8(int, int);

src/main/main.c

@@ -303,7 +303,7 @@ terminate_all_subsystems(void)
 	done_options();
 	done_event();
 	terminate_osdep();
-#ifdef CONFIG_UTF8
+#ifdef CONFIG_COMBINE
 	free_combined();
 #endif
 }

src/main/version.c

@@ -142,6 +142,9 @@ get_dyn_full_version(struct terminal *term, int more)
 #endif
 #ifdef CONFIG_UTF8
 		comma, "UTF-8",
+#endif
+#ifdef CONFIG_COMBINE
+		comma, _("Combining characters", term),
 #endif
 		comma,
 		(unsigned char *) NULL

src/terminal/screen.c

@@ -223,10 +223,12 @@ struct screen_driver {
 		 * is the same as is_cp_utf8(charsets[0]), except the
 		 * latter might crash if UTF-8 I/O is disabled.  */
 		unsigned int utf8_cp:1;
+#endif /* CONFIG_UTF8 */
 
+#ifdef CONFIG_COMBINE
 		/* Whether the terminal supports combining characters. */
 		unsigned int combine:1;
-#endif /* CONFIG_UTF8 */
+#endif /* CONFIG_COMBINE */
 	} opt;
 
 	/* The terminal._template_ name. */
@@ -243,8 +245,10 @@ static const struct screen_driver_opt dumb_screen_driver_opt = {
 	/* transparent: */	1,
 #ifdef CONFIG_UTF8
 	/* utf8_cp: */		0,
-	/* combine */		0,
 #endif /* CONFIG_UTF8 */
+#ifdef CONFIG_COMBINE
+	/* combine */		0,
+#endif /* CONFIG_COMBINE */
 };
 
 /** Default options for ::TERM_VT100.  */
@@ -257,8 +261,10 @@ static const struct screen_driver_opt vt100_screen_driver_opt = {
 	/* transparent: */	1,
 #ifdef CONFIG_UTF8
 	/* utf8_cp: */		0,
-	/* combine */		0,
 #endif /* CONFIG_UTF8 */
+#ifdef CONFIG_COMBINE
+	/* combine */		0,
+#endif /* CONFIG_COMBINE */
 };
 
 /** Default options for ::TERM_LINUX.  */
@@ -271,8 +277,10 @@ static const struct screen_driver_opt linux_screen_driver_opt = {
 	/* transparent: */	1,
 #ifdef CONFIG_UTF8
 	/* utf8_cp: */		0,
-	/* combine */		0,
 #endif /* CONFIG_UTF8 */
+#ifdef CONFIG_COMBINE
+	/* combine */		0,
+#endif /* CONFIG_COMBINE */
 };
 
 /** Default options for ::TERM_KOI8.  */
@@ -285,8 +293,10 @@ static const struct screen_driver_opt koi8_screen_driver_opt = {
 	/* transparent: */	1,
 #ifdef CONFIG_UTF8
 	/* utf8_cp: */		0,
-	/* combine */		0,
 #endif /* CONFIG_UTF8 */
+#ifdef CONFIG_COMBINE
+	/* combine */		0,
+#endif /* CONFIG_COMBINE */
 };
 
 /** Default options for ::TERM_FREEBSD.  */
@@ -299,8 +309,10 @@ static const struct screen_driver_opt freebsd_screen_driver_opt = {
 	/* transparent: */	1,
 #ifdef CONFIG_UTF8
 	/* utf8_cp: */		0,
-	/* combine */		0,
 #endif /* CONFIG_UTF8 */
+#ifdef CONFIG_COMBINE
+	/* combine */		0,
+#endif /* CONFIG_COMBINE */
 };
 
 /** Default options for all the different types of terminals.
@@ -332,8 +344,10 @@ set_screen_driver_opt(struct screen_driver *driver, struct option *term_spec)
 	 * function need not carefully restore options one by one.  */
 	copy_struct(&driver->opt, screen_driver_opts[driver->type]);
 
-#ifdef CONFIG_UTF8
+#ifdef CONFIG_COMBINE
 	driver->opt.combine = get_opt_bool_tree(term_spec, "combine", NULL);
+#endif /* CONFIG_COMBINE */
+#ifdef CONFIG_UTF8
 	/* Force UTF-8 I/O if the UTF-8 charset is selected.  Various
 	 * places assume that the terminal's charset is unibyte if
 	 * UTF-8 I/O is disabled.  (bug 827) */
@@ -648,6 +662,7 @@ add_char_data(struct string *screen, struct screen_driver *driver,
 		}
 		if (data == UCS_NO_CHAR)
 			return;
+#ifdef CONFIG_COMBINE
 		if (data >= UCS_BEGIN_COMBINED && data <= last_combined) {
 			unicode_val_T *text = combined[data - UCS_BEGIN_COMBINED];
 
@@ -663,6 +678,7 @@ add_char_data(struct string *screen, struct screen_driver *driver,
 				data = *text;
 			}
 		}
+#endif /* CONFIG_COMBINE */
 		if (!isscreensafe_ucs(data))
 			data = UCS_SPACE;
 		add_to_string(screen, encode_utf8(data));