From: "naruse (Yui NARUSE)" <naruse@...>
Date: 2013-04-09T18:24:45+09:00
Subject: [ruby-dev:47241] [ruby-trunk - Feature #6752] Replacing ill-formed subsequencce


Issue #6752 has been updated by naruse (Yui NARUSE).


I wrote a updated patch which include String#scrub and String#encode with extension.
String#scrub allows replacement as both argument or block.

diff --git a/string.c b/string.c
index 8b85739..7131ac5 100644
--- a/string.c
+++ b/string.c
@@ -7741,6 +7741,272 @@ rb_str_ellipsize(VALUE str, long len)
     return ret;
 }
 
+static VALUE
+str_compat_and_valid(VALUE str, rb_encoding *enc)
+{
+    int cr;
+    str = StringValue(str);
+    cr = rb_enc_str_coderange(str);
+    if (cr == ENC_CODERANGE_BROKEN) {
+	rb_raise(rb_eArgError, "replacement must be valid byte sequence '%+"PRIsVALUE"'", str);
+    }
+    else if (cr == ENC_CODERANGE_7BIT) {
+	rb_encoding *e = STR_ENC_GET(str);
+	if (!rb_enc_asciicompat(enc)) {
+	    rb_raise(rb_eEncCompatError, "incompatible character encodings: %s and %s",
+		    rb_enc_name(enc), rb_enc_name(e));
+	}
+    }
+    else { /* ENC_CODERANGE_VALID */
+	rb_encoding *e = STR_ENC_GET(str);
+	if (enc != e) {
+	    rb_raise(rb_eEncCompatError, "incompatible character encodings: %s and %s",
+		    rb_enc_name(enc), rb_enc_name(e));
+	}
+    }
+    return str;
+}
+
+/*
+ *  call-seq:
+ *    str.scrub -> new_str
+ *    str.scrub(repl) -> new_str
+ *    str.scrub{|bytes|} -> new_str
+ *
+ *  If the string is invalid byte sequence then replace invalid bytes with given replacement
+ *  character, else returns self.
+ */
+VALUE
+rb_str_scrub(int argc, VALUE *argv, VALUE str)
+{
+    int cr = ENC_CODERANGE(str);
+    rb_encoding *enc;
+    VALUE repl;
+
+    if (cr == ENC_CODERANGE_7BIT || cr == ENC_CODERANGE_VALID)
+	return rb_str_dup(str);
+
+    enc = STR_ENC_GET(str);
+    rb_scan_args(argc, argv, "01", &repl);
+    if (argc != 0) {
+	repl = str_compat_and_valid(repl, enc);
+    }
+
+    if (rb_enc_dummy_p(enc)) {
+	return rb_str_dup(str);
+    }
+
+    if (rb_enc_asciicompat(enc)) {
+	const char *p = RSTRING_PTR(str);
+	const char *e = RSTRING_END(str);
+	const char *p1 = p;
+	const char *rep;
+	long replen;
+	int rep7bit_p;
+	VALUE buf = rb_str_buf_new(RSTRING_LEN(str));
+	if (rb_block_given_p()) {
+	    rep = NULL;
+	}
+	else if (!NIL_P(repl)) {
+	    rep = RSTRING_PTR(repl);
+	    replen = RSTRING_LEN(repl);
+	    rep7bit_p = (ENC_CODERANGE(repl) == ENC_CODERANGE_7BIT);
+	}
+	else if (enc == rb_utf8_encoding()) {
+	    rep = "\xEF\xBF\xBD";
+	    replen = strlen(rep);
+	    rep7bit_p = FALSE;
+	}
+	else {
+	    rep = "?";
+	    replen = strlen(rep);
+	    rep7bit_p = TRUE;
+	}
+	cr = ENC_CODERANGE_7BIT;
+
+	p = search_nonascii(p, e);
+	if (!p) {
+	    p = e;
+	}
+	while (p < e) {
+	    int ret = rb_enc_precise_mbclen(p, e, enc);
+	    if (MBCLEN_NEEDMORE_P(ret)) {
+		break;
+	    }
+	    else if (MBCLEN_CHARFOUND_P(ret)) {
+		cr = ENC_CODERANGE_VALID;
+		p += MBCLEN_CHARFOUND_LEN(ret);
+	    }
+	    else if (MBCLEN_INVALID_P(ret)) {
+		/*
+		 * p1~p: valid ascii/multibyte chars
+		 * p ~e: invalid bytes + unknown bytes
+		 */
+		long clen = rb_enc_mbmaxlen(enc);
+		if (p > p1) {
+		    rb_str_buf_cat(buf, p1, p - p1);
+		}
+
+		if (e - p < clen) clen = e - p;
+		if (clen <= 2) {
+		    clen = 1;
+		}
+		else {
+		    const char *q = p;
+		    clen--;
+		    for (; clen > 1; clen--) {
+			ret = rb_enc_precise_mbclen(q, q + clen, enc);
+			if (MBCLEN_NEEDMORE_P(ret)) break;
+			else if (MBCLEN_INVALID_P(ret)) continue;
+			else UNREACHABLE;
+		    }
+		}
+		if (rep) {
+		    rb_str_buf_cat(buf, rep, replen);
+		    if (!rep7bit_p) cr = ENC_CODERANGE_VALID;
+		}
+		else {
+		    repl = rb_yield(rb_enc_str_new(p1, clen, enc));
+		    repl = str_compat_and_valid(repl, enc);
+		    rb_str_buf_cat(buf, RSTRING_PTR(repl), RSTRING_LEN(repl));
+		    if (ENC_CODERANGE(repl) == ENC_CODERANGE_VALID)
+			cr = ENC_CODERANGE_VALID;
+		}
+		p += clen;
+		p1 = p;
+		p = search_nonascii(p, e);
+		if (!p) {
+		    p = e;
+		    break;
+		}
+	    }
+	    else {
+		UNREACHABLE;
+	    }
+	}
+	if (p1 < p) {
+	    rb_str_buf_cat(buf, p1, p - p1);
+	}
+	if (p < e) {
+	    if (rep) {
+		rb_str_buf_cat(buf, rep, replen);
+		if (!rep7bit_p) cr = ENC_CODERANGE_VALID;
+	    }
+	    else {
+		repl = rb_yield(rb_enc_str_new(p, e-p, enc));
+		repl = str_compat_and_valid(repl, enc);
+		rb_str_buf_cat(buf, RSTRING_PTR(repl), RSTRING_LEN(repl));
+		if (ENC_CODERANGE(repl) == ENC_CODERANGE_VALID)
+		    cr = ENC_CODERANGE_VALID;
+	    }
+	}
+	ENCODING_CODERANGE_SET(buf, rb_enc_to_index(enc), cr);
+	return buf;
+    }
+    else {
+	/* ASCII incompatible */
+	const char *p = RSTRING_PTR(str);
+	const char *e = RSTRING_END(str);
+	const char *p1 = p;
+	VALUE buf = rb_str_buf_new(RSTRING_LEN(str));
+	const char *rep;
+	long replen;
+	long mbminlen = rb_enc_mbminlen(enc);
+	static rb_encoding *utf16be;
+	static rb_encoding *utf16le;
+	static rb_encoding *utf32be;
+	static rb_encoding *utf32le;
+	if (!utf16be) {
+	    utf16be = rb_enc_find("UTF-16BE");
+	    utf16le = rb_enc_find("UTF-16LE");
+	    utf32be = rb_enc_find("UTF-32BE");
+	    utf32le = rb_enc_find("UTF-32LE");
+	}
+	if (!NIL_P(repl)) {
+	    rep = RSTRING_PTR(repl);
+	    replen = RSTRING_LEN(repl);
+	}
+	else if (enc == utf16be) {
+	    rep = "\xFF\xFD";
+	    replen = strlen(rep);
+	}
+	else if (enc == utf16le) {
+	    rep = "\xFD\xFF";
+	    replen = strlen(rep);
+	}
+	else if (enc == utf32be) {
+	    rep = "\x00\x00\xFF\xFD";
+	    replen = strlen(rep);
+	}
+	else if (enc == utf32le) {
+	    rep = "\xFD\xFF\x00\x00";
+	    replen = strlen(rep);
+	}
+	else {
+	    rep = "?";
+	    replen = strlen(rep);
+	}
+
+	while (p < e) {
+	    int ret = rb_enc_precise_mbclen(p, e, enc);
+	    if (MBCLEN_NEEDMORE_P(ret)) {
+		break;
+	    }
+	    else if (MBCLEN_CHARFOUND_P(ret)) {
+		p += MBCLEN_CHARFOUND_LEN(ret);
+	    }
+	    else if (MBCLEN_INVALID_P(ret)) {
+		const char *q = p;
+		long clen = rb_enc_mbmaxlen(enc);
+		if (p > p1) rb_str_buf_cat(buf, p1, p - p1);
+
+		if (e - p < clen) clen = e - p;
+		if (clen <= mbminlen * 2) {
+		    clen = mbminlen;
+		}
+		else {
+		    clen -= mbminlen;
+		    for (; clen > mbminlen; clen-=mbminlen) {
+			ret = rb_enc_precise_mbclen(q, q + clen, enc);
+			if (MBCLEN_NEEDMORE_P(ret)) break;
+			else if (MBCLEN_INVALID_P(ret)) continue;
+			else UNREACHABLE;
+		    }
+		    rb_str_set_len(buf, len);
+		}
+		if (rep) {
+		    rb_str_buf_cat(buf, rep, replen);
+		}
+		else {
+		    repl = rb_yield(rb_enc_str_new(p, e-p, enc));
+		    repl = str_compat_and_valid(repl, enc);
+		    rb_str_buf_cat(buf, RSTRING_PTR(repl), RSTRING_LEN(repl));
+		}
+		p += clen;
+		p1 = p;
+	    }
+	    else {
+		UNREACHABLE;
+	    }
+	}
+	if (p1 < p) {
+	    rb_str_buf_cat(buf, p1, p - p1);
+	}
+	if (p < e) {
+	    if (rep) {
+		rb_str_buf_cat(buf, rep, replen);
+	    }
+	    else {
+		repl = rb_yield(rb_enc_str_new(p, e-p, enc));
+		repl = str_compat_and_valid(repl, enc);
+		rb_str_buf_cat(buf, RSTRING_PTR(repl), RSTRING_LEN(repl));
+	    }
+	}
+	ENCODING_CODERANGE_SET(buf, rb_enc_to_index(enc), ENC_CODERANGE_VALID);
+	return buf;
+    }
+}
+
 /**********************************************************************
  * Document-class: Symbol
  *
@@ -8222,6 +8488,7 @@ Init_String(void)
     rb_define_method(rb_cString, "getbyte", rb_str_getbyte, 1);
     rb_define_method(rb_cString, "setbyte", rb_str_setbyte, 2);
     rb_define_method(rb_cString, "byteslice", rb_str_byteslice, -1);
+    rb_define_method(rb_cString, "scrub", rb_str_scrub, -1);
 
     rb_define_method(rb_cString, "to_i", rb_str_to_i, -1);
     rb_define_method(rb_cString, "to_f", rb_str_to_f, 0);
diff --git a/test/ruby/test_m17n.rb b/test/ruby/test_m17n.rb
index a8d56a4..60834bb 100644
--- a/test/ruby/test_m17n.rb
+++ b/test/ruby/test_m17n.rb
@@ -1489,4 +1489,38 @@ class TestM17N < Test::Unit::TestCase
     s.untrust
     assert_equal(true, s.b.untrusted?)
   end
+
+  def test_scrub
+    assert_equal("\uFFFD\uFFFD\uFFFD", u("\x80\x80\x80").scrub)
+    assert_equal("\uFFFDA", u("\xF4\x80\x80A").scrub)
+
+    # exapmles in Unicode 6.1.0 D93b
+    assert_equal("\x41\uFFFD\uFFFD\x41\uFFFD\x41",
+                 u("\x41\xC0\xAF\x41\xF4\x80\x80\x41").scrub)
+    assert_equal("\x41\uFFFD\uFFFD\uFFFD\x41",
+                 u("\x41\xE0\x9F\x80\x41").scrub)
+    assert_equal("\u0061\uFFFD\uFFFD\uFFFD\u0062\uFFFD\u0063\uFFFD\uFFFD\u0064",
+                 u("\x61\xF1\x80\x80\xE1\x80\xC2\x62\x80\x63\x80\xBF\x64").scrub)
+    assert_equal("abcdefghijklmnopqrstuvwxyz\u0061\uFFFD\uFFFD\uFFFD\u0062\uFFFD\u0063\uFFFD\uFFFD\u0064",
+                 u("abcdefghijklmnopqrstuvwxyz\x61\xF1\x80\x80\xE1\x80\xC2\x62\x80\x63\x80\xBF\x64").scrub)
+
+    assert_equal("\u3042\u3013", u("\xE3\x81\x82\xE3\x81").scrub("\u3013"))
+    assert_raise(Encoding::CompatibilityError){ u("\xE3\x81\x82\xE3\x81").scrub(e("\xA4\xA2")) }
+    assert_raise(TypeError){ u("\xE3\x81\x82\xE3\x81").scrub(1) }
+    assert_raise(ArgumentError){ u("\xE3\x81\x82\xE3\x81\x82\xE3\x81").scrub(u("\x81")) }
+    assert_equal(e("\xA4\xA2\xA2\xAE"), e("\xA4\xA2\xA4").scrub(e("\xA2\xAE")))
+
+    assert_equal("\u3042<e381>", u("\xE3\x81\x82\xE3\x81").scrub{|x|'<'+x.unpack('H*')[0]+'>'})
+    assert_raise(Encoding::CompatibilityError){ u("\xE3\x81\x82\xE3\x81").scrub{e("\xA4\xA2")} }
+    assert_raise(TypeError){ u("\xE3\x81\x82\xE3\x81").scrub{1} }
+    assert_raise(ArgumentError){ u("\xE3\x81\x82\xE3\x81\x82\xE3\x81").scrub{u("\x81")} }
+    assert_equal(e("\xA4\xA2\xA2\xAE"), e("\xA4\xA2\xA4").scrub{e("\xA2\xAE")})
+
+    assert_equal("\uFFFD\u3042".encode("UTF-16BE"),
+                 "\xD8\x00\x30\x42".force_encoding(Encoding::UTF_16BE).
+                 scrub)
+    assert_equal("\uFFFD\u3042".encode("UTF-16LE"),
+                 "\x00\xD8\x42\x30".force_encoding(Encoding::UTF_16LE).
+                 scrub)
+  end
 end
diff --git a/transcode.c b/transcode.c
index de12c04..9c940ed 100644
--- a/transcode.c
+++ b/transcode.c
@@ -2652,6 +2652,8 @@ str_transcode_enc_args(VALUE str, volatile VALUE *arg1, volatile VALUE *arg2,
     return dencidx;
 }
 
+VALUE rb_str_scrub(int argc, VALUE *argv, VALUE str);
+
 static int
 str_transcode0(int argc, VALUE *argv, VALUE *self, int ecflags, VALUE ecopts)
 {
@@ -2686,6 +2688,17 @@ str_transcode0(int argc, VALUE *argv, VALUE *self, int ecflags, VALUE ecopts)
                     ECONV_XML_ATTR_CONTENT_DECORATOR|
                     ECONV_XML_ATTR_QUOTE_DECORATOR)) == 0) {
         if (senc && senc == denc) {
+	    if (ecflags & ECONV_INVALID_MASK) {
+		if (!NIL_P(ecopts)) {
+		    VALUE rep = rb_hash_aref(ecopts, sym_replace);
+		    dest = rb_str_scrub(1, &rep, str);
+		}
+		else {
+		    dest = rb_str_scrub(0, NULL, str);
+		}
+		*self = dest;
+		return dencidx;
+	    }
             return NIL_P(arg2) ? -1 : dencidx;
         }
         if (senc && denc && rb_enc_asciicompat(senc) && rb_enc_asciicompat(denc)) {
----------------------------------------
Feature #6752: Replacing ill-formed subsequencce
https://bugs.ruby-lang.org/issues/6752#change-38389

Author: naruse (Yui NARUSE)
Status: Assigned
Priority: Normal
Assignee: matz (Yukihiro Matsumoto)
Category: core
Target version: next minor


=begin
== 概要
Stringになんらかの理由で不正なバイト列が含まれている時に、それを置換文字で置き換えたい。

== ユースケース
実際に確認されているユースケースは以下の通りです。
* twitterのtitle
* IRCのログ
* ニコニコ動画の API
* Webクローリング
これらの不正なバイト列の生成過程は、おそらく、バイト単位で文字列を切り詰めた時に末尾が切れて、
末尾がおかしい不正な文字列が作られます。（前二者）
これをコンテナに入れたり結合することによって、途中にも混ざった文字列が作られます。（後二者）

* https://twitter.com/takahashim/status/18974040397
* https://twitter.com/n0kada/status/215674740705210368
* https://twitter.com/n0kada/status/215686490070585346
* https://twitter.com/hajimehoshi/status/215671146769682432
* http://po-ru.com/diary/fixing-invalid-utf-8-in-ruby-revisited/
* http://stackoverflow.com/questions/2982677/ruby-1-9-invalid-byte-sequence-in-utf-8

== 必要な引数: 置換文字
省略可能、String。
デフォルトは、Unicode系ならU+FFFD、それ以外では「?」。
デフォルトが空文字でない理由は、削除してしまうことで、従来は存在しなかったトークンを作れてしまい、
上位のレイヤーの脆弱性に繋がるからです。
http://unicode.org/reports/tr36/#UTF-8_Exploit

== API
--- str.encode(str.encoding, invalid: replace, [replace: "〓"])
* CSI的じゃなくて気持ち悪い
* iconv でできるのは glibc iconv か GNU libiconv に //IGNORE つけた時で他はできない
* 実装上のメリットは後述の通り、直感に反してあまりない(と思う)

== 別メソッド
* 新しいメソッドである
* fix/repair invalid/illegal bytes/sequence あたりの名前か

== 実装
=== 鬼車ベース
int ret = rb_enc_precise_mbclen(p, e, enc); して、
MBCLEN_INVALID_P(ret) が真な時、何バイト目が不正なのかわからないのが微妙。
ONIGENC_CONSTRUCT_MBCLEN_INVALID() がバイト数を取らないのが原因なので、
鬼車のエンコーディングモジュール全てに影響してしまうため、修正困難。
不正なバイトはほとんど存在しないと仮定して、効率を犠牲にすれば回避は可能。

=== transcodeベース
UCS正規化なglibc iconv, GNU libiconv, Perl Encodeなどと違って、
CSIなtranscodeでは、自分自身に変換する場合、
エンコーディングごとに「何もしない」変換モジュールを用意しないといけない。


とりあえず鬼車ベースのコンセプト実装とテストを添付しておきます。

 diff --git a/string.c b/string.c
 index d038835..4808f15 100644
 --- a/string.c
 +++ b/string.c
 @@ -7426,6 +7426,199 @@ rb_str_ellipsize(VALUE str, long len)
      return ret;
  }
  
 +/*
 + *  call-seq:
 + *    str.fix_invalid -> new_str
 + *
 + *  If the string is well-formed, it returns self.
 + *  If the string has invalid byte sequence, repair it with given replacement
 + *  character.
 + */
 +VALUE
 +rb_str_fix_invalid(VALUE str)
 +{
 +    int cr = ENC_CODERANGE(str);
 +    rb_encoding *enc;
 +    if (cr == ENC_CODERANGE_7BIT || cr == ENC_CODERANGE_VALID)
 +	return rb_str_dup(str);
 +
 +    enc = STR_ENC_GET(str);
 +    if (rb_enc_asciicompat(enc)) {
 +	const char *p = RSTRING_PTR(str);
 +	const char *e = RSTRING_END(str);
 +	const char *p1 = p;
 +	/* 10 should be enough for the usual use case,
 +	 * fixing a wrongly chopped character at the end of the string
 +	 */
 +	long room = 10;
 +	VALUE buf = rb_str_buf_new(RSTRING_LEN(str) + room);
 +	const char *rep;
 +	if (enc == rb_utf8_encoding())
 +	    rep = "\xEF\xBF\xBD";
 +	else
 +	    rep = "?";
 +	cr = ENC_CODERANGE_7BIT;
 +
 +	p = search_nonascii(p, e);
 +	if (!p) {
 +	    p = e;
 +	}
 +	while (p < e) {
 +	    int ret = rb_enc_precise_mbclen(p, e, enc);
 +	    if (MBCLEN_CHARFOUND_P(ret)) {
 +		if ((unsigned char)*p > 127) cr = ENC_CODERANGE_VALID;
 +		p += MBCLEN_CHARFOUND_LEN(ret);
 +	    }
 +	    else if (MBCLEN_INVALID_P(ret)) {
 +		const char *q;
 +		long clen = rb_enc_mbmaxlen(enc);
 +		if (p > p1) rb_str_buf_cat(buf, p1, p - p1);
 +		q = RSTRING_END(buf);
 +
 +		if (e - p < clen) clen = e - p;
 +		if (clen < 3) {
 +		    clen = 1;
 +		}
 +		else {
 +		    long len = RSTRING_LEN(buf);
 +		    clen--;
 +		    rb_str_buf_cat(buf, p, clen);
 +		    for (; clen > 1; clen--) {
 +			ret = rb_enc_precise_mbclen(q, q + clen, enc);
 +			if (MBCLEN_NEEDMORE_P(ret)) {
 +			    break;
 +			}
 +			else if (MBCLEN_INVALID_P(ret)) {
 +			    continue;
 +			}
 +			else {
 +			    rb_bug("shouldn't reach here '%s'", q);
 +			}
 +		    }
 +		    rb_str_set_len(buf, len);
 +		}
 +		p += clen;
 +		p1 = p;
 +		rb_str_buf_cat2(buf, rep);
 +		p = search_nonascii(p, e);
 +		if (!p) {
 +		    p = e;
 +		    break;
 +		}
 +	    }
 +	    else if (MBCLEN_NEEDMORE_P(ret)) {
 +		break;
 +	    }
 +	    else {
 +		rb_bug("shouldn't reach here");
 +	    }
 +	}
 +	if (p1 < p) {
 +	    rb_str_buf_cat(buf, p1, p - p1);
 +	}
 +	if (p < e) {
 +	    rb_str_buf_cat2(buf, rep);
 +	    cr = ENC_CODERANGE_VALID;
 +	}
 +	ENCODING_CODERANGE_SET(buf, rb_enc_to_index(enc), cr);
 +	return buf;
 +    }
 +    else if (rb_enc_dummy_p(enc)) {
 +	return rb_str_dup(str);
 +    }
 +    else {
 +	/* ASCII incompatible */
 +	const char *p = RSTRING_PTR(str);
 +	const char *e = RSTRING_END(str);
 +	const char *p1 = p;
 +	/* 10 should be enough for the usual use case,
 +	 * fixing a wrongly chopped character at the end of the string
 +	 */
 +	long room = 10;
 +	VALUE buf = rb_str_buf_new(RSTRING_LEN(str) + room);
 +	const char *rep;
 +	long mbminlen = rb_enc_mbminlen(enc);
 +	static rb_encoding *utf16be;
 +	static rb_encoding *utf16le;
 +	static rb_encoding *utf32be;
 +	static rb_encoding *utf32le;
 +	if (!utf16be) {
 +	    utf16be = rb_enc_find("UTF-16BE");
 +	    utf16le = rb_enc_find("UTF-16LE");
 +	    utf32be = rb_enc_find("UTF-32BE");
 +	    utf32le = rb_enc_find("UTF-32LE");
 +	}
 +	if (enc == utf16be) {
 +	    rep = "\xFF\xFD";
 +	}
 +	else if (enc == utf16le) {
 +	    rep = "\xFD\xFF";
 +	}
 +	else if (enc == utf32be) {
 +	    rep = "\x00\x00\xFF\xFD";
 +	}
 +	else if (enc == utf32le) {
 +	    rep = "\xFD\xFF\x00\x00";
 +	}
 +	else {
 +	    rep = "?";
 +	}
 +
 +	while (p < e) {
 +	    int ret = rb_enc_precise_mbclen(p, e, enc);
 +	    if (MBCLEN_CHARFOUND_P(ret)) {
 +		p += MBCLEN_CHARFOUND_LEN(ret);
 +	    }
 +	    else if (MBCLEN_INVALID_P(ret)) {
 +		const char *q;
 +		long clen = rb_enc_mbmaxlen(enc);
 +		if (p > p1) rb_str_buf_cat(buf, p1, p - p1);
 +		q = RSTRING_END(buf);
 +
 +		if (e - p < clen) clen = e - p;
 +		if (clen < mbminlen * 3) {
 +		    clen = mbminlen;
 +		}
 +		else {
 +		    long len = RSTRING_LEN(buf);
 +		    clen -= mbminlen;
 +		    rb_str_buf_cat(buf, p, clen);
 +		    for (; clen > mbminlen; clen-=mbminlen) {
 +			ret = rb_enc_precise_mbclen(q, q + clen, enc);
 +			if (MBCLEN_NEEDMORE_P(ret)) {
 +			    break;
 +			}
 +			else if (MBCLEN_INVALID_P(ret)) {
 +			    continue;
 +			}
 +			else {
 +			    rb_bug("shouldn't reach here '%s'", q);
 +			}
 +		    }
 +		    rb_str_set_len(buf, len);
 +		}
 +		p += clen;
 +		p1 = p;
 +		rb_str_buf_cat2(buf, rep);
 +	    }
 +	    else if (MBCLEN_NEEDMORE_P(ret)) {
 +		break;
 +	    }
 +	    else {
 +		rb_bug("shouldn't reach here");
 +	    }
 +	}
 +	if (p1 < p) {
 +	    rb_str_buf_cat(buf, p1, p - p1);
 +	}
 +	if (p < e) {
 +	    rb_str_buf_cat2(buf, rep);
 +	}
 +	ENCODING_CODERANGE_SET(buf, rb_enc_to_index(enc), ENC_CODERANGE_VALID);
 +	return buf;
 +    }
 +}
 +
  /**********************************************************************
   * Document-class: Symbol
   *
 @@ -7882,6 +8075,7 @@ Init_String(void)
      rb_define_method(rb_cString, "getbyte", rb_str_getbyte, 1);
      rb_define_method(rb_cString, "setbyte", rb_str_setbyte, 2);
      rb_define_method(rb_cString, "byteslice", rb_str_byteslice, -1);
 +    rb_define_method(rb_cString, "fix_invalid", rb_str_fix_invalid, 0);
  
      rb_define_method(rb_cString, "to_i", rb_str_to_i, -1);
      rb_define_method(rb_cString, "to_f", rb_str_to_f, 0);
 diff --git a/test/ruby/test_string.rb b/test/ruby/test_string.rb
 index 47f349c..2b0cfeb 100644
 --- a/test/ruby/test_string.rb
 +++ b/test/ruby/test_string.rb
 @@ -2031,6 +2031,29 @@ class TestString < Test::Unit::TestCase
  
      assert_equal(u("\x82")+("\u3042"*9), ("\u3042"*10).byteslice(2, 28))
    end
 +
 +  def test_fix_invalid
 +    assert_equal("\uFFFD\uFFFD\uFFFD", "\x80\x80\x80".fix_invalid)
 +    assert_equal("\uFFFDA", "\xF4\x80\x80A".fix_invalid)
 +
 +    # exapmles in Unicode 6.1.0 D93b
 +    assert_equal("\x41\uFFFD\uFFFD\x41\uFFFD\x41",
 +                 "\x41\xC0\xAF\x41\xF4\x80\x80\x41".fix_invalid)
 +    assert_equal("\x41\uFFFD\uFFFD\uFFFD\x41",
 +                 "\x41\xE0\x9F\x80\x41".fix_invalid)
 +    assert_equal("\u0061\uFFFD\uFFFD\uFFFD\u0062\uFFFD\u0063\uFFFD\uFFFD\u0064",
 +                 "\x61\xF1\x80\x80\xE1\x80\xC2\x62\x80\x63\x80\xBF\x64".fix_invalid)
 +
 +    assert_equal("abcdefghijklmnopqrstuvwxyz\u0061\uFFFD\uFFFD\uFFFD\u0062\uFFFD\u0063\uFFFD\uFFFD\u0064",
 +                 "abcdefghijklmnopqrstuvwxyz\x61\xF1\x80\x80\xE1\x80\xC2\x62\x80\x63\x80\xBF\x64".fix_invalid)
 +
 +    assert_equal("\uFFFD\u3042".encode("UTF-16BE"),
 +                 "\xD8\x00\x30\x42".force_encoding(Encoding::UTF_16BE).
 +                 fix_invalid)
 +    assert_equal("\uFFFD\u3042".encode("UTF-16LE"),
 +                 "\x00\xD8\x42\x30".force_encoding(Encoding::UTF_16LE).
 +                 fix_invalid)
 +  end
  end
  
  class TestString2 < TestString
=end



-- 
http://bugs.ruby-lang.org/