Revert "Revert "adb/base: fix adb push of Unicode filenames on Win32""

This reverts commit cc8cd59456ca485a51cd6fd388c8bcb1af4a8f9b.

With the dependency on libcutils (for gettid for non-bionic) removed,
this no longer breaks the build.

Change-Id: I645bd6876e2502ddc1535b69af1e645c0df9d178
diff --git a/base/utf8_test.cpp b/base/utf8_test.cpp
new file mode 100755
index 0000000..bbb54b1
--- /dev/null
+++ b/base/utf8_test.cpp
@@ -0,0 +1,402 @@
+/*
+* Copyright (C) 2015 The Android Open Source Project
+*
+* Licensed under the Apache License, Version 2.0 (the "License");
+* you may not use this file except in compliance with the License.
+* You may obtain a copy of the License at
+*
+*      http://www.apache.org/licenses/LICENSE-2.0
+*
+* Unless required by applicable law or agreed to in writing, software
+* distributed under the License is distributed on an "AS IS" BASIS,
+* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+* See the License for the specific language governing permissions and
+* limitations under the License.
+*/
+
+#include "base/utf8.h"
+
+#include <gtest/gtest.h>
+
+#include "base/macros.h"
+
+namespace android {
+namespace base {
+
+TEST(UTFStringConversionsTest, ConvertInvalidUTF8) {
+  std::wstring wide;
+
+  // Standalone \xa2 is an invalid UTF-8 sequence, so this should return an
+  // error. Concatenate two C/C++ literal string constants to prevent the
+  // compiler from giving an error about "\xa2af" containing a "hex escape
+  // sequence out of range".
+  EXPECT_FALSE(android::base::UTF8ToWide("before\xa2" "after", &wide));
+
+  // Even if an invalid character is encountered, UTF8ToWide() should still do
+  // its best to convert the rest of the string. sysdeps_win32.cpp:
+  // _console_write_utf8() depends on this behavior.
+  //
+  // Thus, we verify that the valid characters are converted, but we ignore the
+  // specific replacement character that UTF8ToWide() may replace the invalid
+  // UTF-8 characters with because we want to allow that to change if the
+  // implementation changes.
+  EXPECT_EQ(0, wide.find(L"before"));
+  const wchar_t after_wide[] = L"after";
+  EXPECT_EQ(wide.length() - (arraysize(after_wide) - 1), wide.find(after_wide));
+}
+
+// Below is adapted from https://chromium.googlesource.com/chromium/src/+/master/base/strings/utf_string_conversions_unittest.cc
+
+// Copyright (c) 2010 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+// The tests below from utf_string_conversions_unittest.cc check for this
+// preprocessor symbol, so define it, as it is appropriate for Windows.
+#define WCHAR_T_IS_UTF16
+static_assert(sizeof(wchar_t) == 2, "wchar_t is not 2 bytes");
+
+// The tests below from utf_string_conversions_unittest.cc call versions of
+// UTF8ToWide() and WideToUTF8() that don't return success/failure, so these are
+// stub implementations with that signature. These are just for testing and
+// should not be moved to base because they assert/expect no errors which is
+// probably not a good idea (or at least it is something that should be left
+// up to the caller, not a base library).
+
+static std::wstring UTF8ToWide(const std::string& utf8) {
+  std::wstring utf16;
+  EXPECT_TRUE(UTF8ToWide(utf8, &utf16));
+  return utf16;
+}
+
+static std::string WideToUTF8(const std::wstring& utf16) {
+  std::string utf8;
+  EXPECT_TRUE(WideToUTF8(utf16, &utf8));
+  return utf8;
+}
+
+namespace {
+
+const wchar_t* const kConvertRoundtripCases[] = {
+  L"Google Video",
+  // "网页 图片 资讯更多 »"
+  L"\x7f51\x9875\x0020\x56fe\x7247\x0020\x8d44\x8baf\x66f4\x591a\x0020\x00bb",
+  //  "Παγκόσμιος Ιστός"
+  L"\x03a0\x03b1\x03b3\x03ba\x03cc\x03c3\x03bc\x03b9"
+  L"\x03bf\x03c2\x0020\x0399\x03c3\x03c4\x03cc\x03c2",
+  // "Поиск страниц на русском"
+  L"\x041f\x043e\x0438\x0441\x043a\x0020\x0441\x0442"
+  L"\x0440\x0430\x043d\x0438\x0446\x0020\x043d\x0430"
+  L"\x0020\x0440\x0443\x0441\x0441\x043a\x043e\x043c",
+  // "전체서비스"
+  L"\xc804\xccb4\xc11c\xbe44\xc2a4",
+
+  // Test characters that take more than 16 bits. This will depend on whether
+  // wchar_t is 16 or 32 bits.
+#if defined(WCHAR_T_IS_UTF16)
+  L"\xd800\xdf00",
+  // ?????  (Mathematical Alphanumeric Symbols (U+011d40 - U+011d44 : A,B,C,D,E)
+  L"\xd807\xdd40\xd807\xdd41\xd807\xdd42\xd807\xdd43\xd807\xdd44",
+#elif defined(WCHAR_T_IS_UTF32)
+  L"\x10300",
+  // ?????  (Mathematical Alphanumeric Symbols (U+011d40 - U+011d44 : A,B,C,D,E)
+  L"\x11d40\x11d41\x11d42\x11d43\x11d44",
+#endif
+};
+
+}  // namespace
+
+TEST(UTFStringConversionsTest, ConvertUTF8AndWide) {
+  // we round-trip all the wide strings through UTF-8 to make sure everything
+  // agrees on the conversion. This uses the stream operators to test them
+  // simultaneously.
+  for (size_t i = 0; i < arraysize(kConvertRoundtripCases); ++i) {
+    std::ostringstream utf8;
+    utf8 << WideToUTF8(kConvertRoundtripCases[i]);
+    std::wostringstream wide;
+    wide << UTF8ToWide(utf8.str());
+
+    EXPECT_EQ(kConvertRoundtripCases[i], wide.str());
+  }
+}
+
+TEST(UTFStringConversionsTest, ConvertUTF8AndWideEmptyString) {
+  // An empty std::wstring should be converted to an empty std::string,
+  // and vice versa.
+  std::wstring wempty;
+  std::string empty;
+  EXPECT_EQ(empty, WideToUTF8(wempty));
+  EXPECT_EQ(wempty, UTF8ToWide(empty));
+}
+
+TEST(UTFStringConversionsTest, ConvertUTF8ToWide) {
+  struct UTF8ToWideCase {
+    const char* utf8;
+    const wchar_t* wide;
+    bool success;
+  } convert_cases[] = {
+    // Regular UTF-8 input.
+    {"\xe4\xbd\xa0\xe5\xa5\xbd", L"\x4f60\x597d", true},
+    // Non-character is passed through.
+    {"\xef\xbf\xbfHello", L"\xffffHello", true},
+    // Truncated UTF-8 sequence.
+    {"\xe4\xa0\xe5\xa5\xbd", L"\xfffd\x597d", false},
+    // Truncated off the end.
+    {"\xe5\xa5\xbd\xe4\xa0", L"\x597d\xfffd", false},
+    // Non-shortest-form UTF-8.
+    {"\xf0\x84\xbd\xa0\xe5\xa5\xbd", L"\xfffd\x597d", false},
+    // This UTF-8 character decodes to a UTF-16 surrogate, which is illegal.
+    // Note that for whatever reason, this test fails on Windows XP.
+    {"\xed\xb0\x80", L"\xfffd", false},
+    // Non-BMP characters. The second is a non-character regarded as valid.
+    // The result will either be in UTF-16 or UTF-32.
+#if defined(WCHAR_T_IS_UTF16)
+    {"A\xF0\x90\x8C\x80z", L"A\xd800\xdf00z", true},
+    {"A\xF4\x8F\xBF\xBEz", L"A\xdbff\xdffez", true},
+#elif defined(WCHAR_T_IS_UTF32)
+    {"A\xF0\x90\x8C\x80z", L"A\x10300z", true},
+    {"A\xF4\x8F\xBF\xBEz", L"A\x10fffez", true},
+#endif
+  };
+
+  for (size_t i = 0; i < arraysize(convert_cases); i++) {
+    std::wstring converted;
+    const bool success = UTF8ToWide(convert_cases[i].utf8,
+                                    strlen(convert_cases[i].utf8),
+                                    &converted);
+    EXPECT_EQ(convert_cases[i].success, success);
+    // The original test always compared expected and converted, but don't do
+    // that because our implementation of UTF8ToWide() does not guarantee to
+    // produce the same output in error situations.
+    if (success) {
+      std::wstring expected(convert_cases[i].wide);
+      EXPECT_EQ(expected, converted);
+    }
+  }
+
+  // Manually test an embedded NULL.
+  std::wstring converted;
+  EXPECT_TRUE(UTF8ToWide("\00Z\t", 3, &converted));
+  ASSERT_EQ(3U, converted.length());
+  EXPECT_EQ(static_cast<wchar_t>(0), converted[0]);
+  EXPECT_EQ('Z', converted[1]);
+  EXPECT_EQ('\t', converted[2]);
+
+  // Make sure that conversion replaces, not appends.
+  EXPECT_TRUE(UTF8ToWide("B", 1, &converted));
+  ASSERT_EQ(1U, converted.length());
+  EXPECT_EQ('B', converted[0]);
+}
+
+#if defined(WCHAR_T_IS_UTF16)
+// This test is only valid when wchar_t == UTF-16.
+TEST(UTFStringConversionsTest, ConvertUTF16ToUTF8) {
+  struct WideToUTF8Case {
+    const wchar_t* utf16;
+    const char* utf8;
+    bool success;
+  } convert_cases[] = {
+    // Regular UTF-16 input.
+    {L"\x4f60\x597d", "\xe4\xbd\xa0\xe5\xa5\xbd", true},
+    // Test a non-BMP character.
+    {L"\xd800\xdf00", "\xF0\x90\x8C\x80", true},
+    // Non-characters are passed through.
+    {L"\xffffHello", "\xEF\xBF\xBFHello", true},
+    {L"\xdbff\xdffeHello", "\xF4\x8F\xBF\xBEHello", true},
+    // The first character is a truncated UTF-16 character.
+    // Note that for whatever reason, this test fails on Windows XP.
+    {L"\xd800\x597d", "\xef\xbf\xbd\xe5\xa5\xbd",
+#if (WINVER >= 0x0600)
+    // Only Vista and later has a new API/flag that correctly returns false.
+    false
+#else
+    true
+#endif
+    },
+    // Truncated at the end.
+    // Note that for whatever reason, this test fails on Windows XP.
+    {L"\x597d\xd800", "\xe5\xa5\xbd\xef\xbf\xbd",
+#if (WINVER >= 0x0600)
+    // Only Vista and later has a new API/flag that correctly returns false.
+    false
+#else
+    true
+#endif
+    },
+  };
+
+  for (size_t i = 0; i < arraysize(convert_cases); i++) {
+    std::string converted;
+    const bool success = WideToUTF8(convert_cases[i].utf16,
+                                    wcslen(convert_cases[i].utf16),
+                                    &converted);
+    EXPECT_EQ(convert_cases[i].success, success);
+    // The original test always compared expected and converted, but don't do
+    // that because our implementation of WideToUTF8() does not guarantee to
+    // produce the same output in error situations.
+    if (success) {
+      std::string expected(convert_cases[i].utf8);
+      EXPECT_EQ(expected, converted);
+    }
+  }
+}
+
+#elif defined(WCHAR_T_IS_UTF32)
+// This test is only valid when wchar_t == UTF-32.
+TEST(UTFStringConversionsTest, ConvertUTF32ToUTF8) {
+  struct WideToUTF8Case {
+    const wchar_t* utf32;
+    const char* utf8;
+    bool success;
+  } convert_cases[] = {
+    // Regular 16-bit input.
+    {L"\x4f60\x597d", "\xe4\xbd\xa0\xe5\xa5\xbd", true},
+    // Test a non-BMP character.
+    {L"A\x10300z", "A\xF0\x90\x8C\x80z", true},
+    // Non-characters are passed through.
+    {L"\xffffHello", "\xEF\xBF\xBFHello", true},
+    {L"\x10fffeHello", "\xF4\x8F\xBF\xBEHello", true},
+    // Invalid Unicode code points.
+    {L"\xfffffffHello", "\xEF\xBF\xBDHello", false},
+    // The first character is a truncated UTF-16 character.
+    {L"\xd800\x597d", "\xef\xbf\xbd\xe5\xa5\xbd", false},
+    {L"\xdc01Hello", "\xef\xbf\xbdHello", false},
+  };
+
+  for (size_t i = 0; i < arraysize(convert_cases); i++) {
+    std::string converted;
+    EXPECT_EQ(convert_cases[i].success,
+              WideToUTF8(convert_cases[i].utf32,
+                         wcslen(convert_cases[i].utf32),
+                         &converted));
+    std::string expected(convert_cases[i].utf8);
+    EXPECT_EQ(expected, converted);
+  }
+}
+#endif  // defined(WCHAR_T_IS_UTF32)
+
+// The test below uses these types and functions, so just do enough to get the
+// test running.
+typedef wchar_t char16;
+typedef std::wstring string16;
+
+template<typename T>
+static void* WriteInto(T* t, size_t size) {
+  // std::(w)string::resize() already includes space for a NULL terminator.
+  t->resize(size - 1);
+  return &(*t)[0];
+}
+
+// A stub implementation that calls a helper from above, just to get the test
+// below working. This is just for testing and should not be moved to base
+// because this ignores errors which is probably not a good idea, plus it takes
+// a string16 type which we don't really have.
+static std::string UTF16ToUTF8(const string16& utf16) {
+  return WideToUTF8(utf16);
+}
+
+TEST(UTFStringConversionsTest, ConvertMultiString) {
+  static char16 multi16[] = {
+    'f', 'o', 'o', '\0',
+    'b', 'a', 'r', '\0',
+    'b', 'a', 'z', '\0',
+    '\0'
+  };
+  static char multi[] = {
+    'f', 'o', 'o', '\0',
+    'b', 'a', 'r', '\0',
+    'b', 'a', 'z', '\0',
+    '\0'
+  };
+  string16 multistring16;
+  memcpy(WriteInto(&multistring16, arraysize(multi16)), multi16,
+                   sizeof(multi16));
+  EXPECT_EQ(arraysize(multi16) - 1, multistring16.length());
+  std::string expected;
+  memcpy(WriteInto(&expected, arraysize(multi)), multi, sizeof(multi));
+  EXPECT_EQ(arraysize(multi) - 1, expected.length());
+  const std::string& converted = UTF16ToUTF8(multistring16);
+  EXPECT_EQ(arraysize(multi) - 1, converted.length());
+  EXPECT_EQ(expected, converted);
+}
+
+// The tests below from sys_string_conversions_unittest.cc call SysWideToUTF8()
+// and SysUTF8ToWide(), so these are stub implementations that call the helpers
+// above. These are just for testing and should not be moved to base because
+// they ignore errors which is probably not a good idea.
+
+static std::string SysWideToUTF8(const std::wstring& utf16) {
+  return WideToUTF8(utf16);
+}
+
+static std::wstring SysUTF8ToWide(const std::string& utf8) {
+  return UTF8ToWide(utf8);
+}
+
+// Below is adapted from https://chromium.googlesource.com/chromium/src/+/master/base/strings/sys_string_conversions_unittest.cc
+
+// Copyright (c) 2011 The Chromium Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifdef WCHAR_T_IS_UTF32
+static const std::wstring kSysWideOldItalicLetterA = L"\x10300";
+#else
+static const std::wstring kSysWideOldItalicLetterA = L"\xd800\xdf00";
+#endif
+
+TEST(SysStrings, SysWideToUTF8) {
+  EXPECT_EQ("Hello, world", SysWideToUTF8(L"Hello, world"));
+  EXPECT_EQ("\xe4\xbd\xa0\xe5\xa5\xbd", SysWideToUTF8(L"\x4f60\x597d"));
+
+  // >16 bits
+  EXPECT_EQ("\xF0\x90\x8C\x80", SysWideToUTF8(kSysWideOldItalicLetterA));
+
+  // Error case. When Windows finds a UTF-16 character going off the end of
+  // a string, it just converts that literal value to UTF-8, even though this
+  // is invalid.
+  //
+  // This is what XP does, but Vista has different behavior, so we don't bother
+  // verifying it:
+  // EXPECT_EQ("\xE4\xBD\xA0\xED\xA0\x80zyxw",
+  //           SysWideToUTF8(L"\x4f60\xd800zyxw"));
+
+  // Test embedded NULLs.
+  std::wstring wide_null(L"a");
+  wide_null.push_back(0);
+  wide_null.push_back('b');
+
+  std::string expected_null("a");
+  expected_null.push_back(0);
+  expected_null.push_back('b');
+
+  EXPECT_EQ(expected_null, SysWideToUTF8(wide_null));
+}
+
+TEST(SysStrings, SysUTF8ToWide) {
+  EXPECT_EQ(L"Hello, world", SysUTF8ToWide("Hello, world"));
+  EXPECT_EQ(L"\x4f60\x597d", SysUTF8ToWide("\xe4\xbd\xa0\xe5\xa5\xbd"));
+  // >16 bits
+  EXPECT_EQ(kSysWideOldItalicLetterA, SysUTF8ToWide("\xF0\x90\x8C\x80"));
+
+  // Error case. When Windows finds an invalid UTF-8 character, it just skips
+  // it. This seems weird because it's inconsistent with the reverse conversion.
+  //
+  // This is what XP does, but Vista has different behavior, so we don't bother
+  // verifying it:
+  // EXPECT_EQ(L"\x4f60zyxw", SysUTF8ToWide("\xe4\xbd\xa0\xe5\xa5zyxw"));
+
+  // Test embedded NULLs.
+  std::string utf8_null("a");
+  utf8_null.push_back(0);
+  utf8_null.push_back('b');
+
+  std::wstring expected_null(L"a");
+  expected_null.push_back(0);
+  expected_null.push_back('b');
+
+  EXPECT_EQ(expected_null, SysUTF8ToWide(utf8_null));
+}
+
+}  // namespace base
+}  // namespace android