string.c: use actual encodings

* string.c (get_actual_encoding): get actual encoding according to
  the BOM if exists.
* string.c (rb_str_inspect): use according encoding, instead of
  pseudo encodings, UTF-{16,32}.  [ruby-core:59757] [Bug #8940]

git-svn-id: svn+ssh://ci.ruby-lang.org/ruby/trunk@44605 b2dd03c8-39d4-4d8f-98ff-823fe69b080e

ChangeLog

@@ -1,3 +1,11 @@
+Wed Jan 15 14:03:47 2014  Nobuyoshi Nakada  <nobu@ruby-lang.org>
+
+	* string.c (get_actual_encoding): get actual encoding according to
+	  the BOM if exists.
+
+	* string.c (rb_str_inspect): use according encoding, instead of
+	  pseudo encodings, UTF-{16,32}.  [ruby-core:59757] [Bug #8940]
+
 Tue Jan 14 21:07:22 2014  Masaki Matsushita  <glass.saga@gmail.com>
 
 	* ext/thread/thread.c (rb_szqueue_clear): notify SZQUEUE_WAITERS

encoding.c

@@ -598,6 +598,12 @@ rb_enc_from_index(int index)
     return enc_table.list[index].enc;
 }
 
+rb_encoding *
+rb_enc_get_from_index(int index)
+{
+    return must_encindex(index);
+}
+
 int
 rb_enc_registered(const char *name)
 {

string.c

@@ -123,6 +123,38 @@ VALUE rb_cSymbol;
 
 #define STR_ENC_GET(str) rb_enc_from_index(ENCODING_GET(str))
 
+rb_encoding *rb_enc_get_from_index(int index);
+
+static rb_encoding *
+get_actual_encoding(const int encidx, VALUE str)
+{
+    const unsigned char *q;
+
+    switch (encidx) {
+      case ENCINDEX_UTF_16:
+	if (RSTRING_LEN(str) < 2) break;
+	q = (const unsigned char *)RSTRING_PTR(str);
+	if (q[0] == 0xFE && q[1] == 0xFF) {
+	    return rb_enc_get_from_index(ENCINDEX_UTF_16BE);
+	}
+	if (q[0] == 0xFF && q[1] == 0xFE) {
+	    return rb_enc_get_from_index(ENCINDEX_UTF_16LE);
+	}
+	return rb_ascii8bit_encoding();
+      case ENCINDEX_UTF_32:
+	if (RSTRING_LEN(str) < 4) break;
+	q = (const unsigned char *)RSTRING_PTR(str);
+	if (q[0] == 0 && q[1] == 0 && q[2] == 0xFE && q[3] == 0xFF) {
+	    return rb_enc_get_from_index(ENCINDEX_UTF_32BE);
+	}
+	if (q[3] == 0 && q[2] == 0 && q[1] == 0xFE && q[0] == 0xFF) {
+	    return rb_enc_get_from_index(ENCINDEX_UTF_32LE);
+	}
+	return rb_ascii8bit_encoding();
+    }
+    return rb_enc_from_index(encidx);
+}
+
 static int fstring_cmp(VALUE a, VALUE b);
 
 static st_table* frozen_strings;
@@ -4749,8 +4781,8 @@ rb_str_buf_cat_escaped_char(VALUE result, unsigned int c, int unicode_p)
 VALUE
 rb_str_inspect(VALUE str)
 {
-    rb_encoding *enc = STR_ENC_GET(str);
+    int encidx = ENCODING_GET(str);
-    int encidx = rb_enc_to_index(enc);
+    rb_encoding *enc = rb_enc_from_index(encidx), *actenc;
     const char *p, *pend, *prev;
     char buf[CHAR_ESC_LEN + 1];
     VALUE result = rb_str_buf_new(0);
@@ -4765,27 +4797,10 @@ rb_str_inspect(VALUE str)
 
     p = RSTRING_PTR(str); pend = RSTRING_END(str);
     prev = p;
-    if (encidx == ENCINDEX_UTF_16 && p + 2 <= pend) {
+    actenc = get_actual_encoding(encidx, str);
-	const unsigned char *q = (const unsigned char *)p;
+    if (actenc != enc) {
-	if (q[0] == 0xFE && q[1] == 0xFF)
+	enc = actenc;
-	    enc = rb_enc_from_index(ENCINDEX_UTF_16BE);
+	if (unicode_p) unicode_p = rb_enc_unicode_p(enc);
-	else if (q[0] == 0xFF && q[1] == 0xFE)
-	    enc = rb_enc_from_index(ENCINDEX_UTF_16LE);
-	else {
-	    enc = rb_ascii8bit_encoding();
-	    unicode_p = 0;
-	}
-    }
-    else if (encidx == ENCINDEX_UTF_32 && p + 4 <= pend) {
-	const unsigned char *q = (const unsigned char *)p;
-	if (q[0] == 0 && q[1] == 0 && q[2] == 0xFE && q[3] == 0xFF)
-	    enc = rb_enc_from_index(ENCINDEX_UTF_32BE);
-	else if (q[3] == 0 && q[2] == 0 && q[1] == 0xFE && q[0] == 0xFF)
-	    enc = rb_enc_from_index(ENCINDEX_UTF_32LE);
-	else {
-	    enc = rb_ascii8bit_encoding();
-	    unicode_p = 0;
-	}
     }
     while (p < pend) {
 	unsigned int c, cc;

test/ruby/test_m17n.rb

@@ -228,14 +228,18 @@ class TestM17N < Test::Unit::TestCase
 
   STR_WITHOUT_BOM = "\u3042".freeze
   STR_WITH_BOM = "\uFEFF\u3042".freeze
+  bug8940 = '[ruby-core:59757] [Bug #8940]'
   %w/UTF-16 UTF-32/.each do |enc|
     %w/BE LE/.each do |endian|
+      bom = "\uFEFF".encode("#{enc}#{endian}").force_encoding(enc)
+
       define_method("test_utf_16_32_inspect(#{enc}#{endian})") do
         s = STR_WITHOUT_BOM.encode(enc + endian)
         # When a UTF-16/32 string doesn't have a BOM,
         # inspect as a dummy encoding string.
         assert_equal(s.dup.force_encoding("ISO-2022-JP").inspect,
                      s.dup.force_encoding(enc).inspect)
+        assert_normal_exit("#{bom.b.dump}.force_encoding('#{enc}').inspect", bug8940)
       end
 
       define_method("test_utf_16_32_inspect(#{enc}#{endian}-BOM)") do