added test to test case converting with nonASCII

base/i18n/case_conversion_unittest.cc

Index: base/i18n/case_conversion_unittest.cc
diff --git a/base/i18n/case_conversion_unittest.cc b/base/i18n/case_conversion_unittest.cc
index 38e2c6878228de468124d80a7aabfddebf2eec0f..3bea75ef4b3bb55520bdcdc006354e09dd479de3 100644
--- a/base/i18n/case_conversion_unittest.cc
+++ b/base/i18n/case_conversion_unittest.cc
@@ -22,7 +22,17 @@ TEST(CaseConversionTest, UpperLower) {
   EXPECT_EQ(expected_upper, result);
 }
 
-// TODO(jshin): More tests are needed, especially with non-ASCII characters.
+TEST(CaseConversionTest, nonASCII) {
+  string16 mixed(UTF8ToUTF16(u8"ÄÖäö Ïï ÷%¤#*^`@£$‰‚~ ἇἏ Ḁḁ"));

[jungshik at Google, 2015/04/27 18:25:51]

MSVC does not support 'u8', yet, IIRC. It cannot even compile any source files
with non-ASCII byte sequences on CJK Windows if they happen to be valid in CJK
Windows code page. 

It's very unfortunate, but you have to use hex-escape here. 

+  const string16 expected_lower(UTF8ToUTF16(u8"äöäö ïï ÷%¤#*^`@£$‰‚~ ἇἇ ḁḁ"));

[jungshik at Google, 2015/04/27 18:25:51]

Add a test with Turkish dotless i and try case conversion in both Turkish locale
and non-Turkish locale. You can do that by setting the ICU's default locale.
There are other tests that do that. 

+  const string16 expected_upper(UTF8ToUTF16(u8"ÄÖÄÖ ÏÏ ÷%¤#*^`@£$‰‚~ ἏἏ ḀḀ"));
+
+  string16 result = base::i18n::ToLower(mixed);
+  EXPECT_EQ(expected_lower, result);
+
+  result = base::i18n::ToUpper(mixed);
+  EXPECT_EQ(expected_upper, result);
+}
 
 }  // namespace
 }  // namespace base