Skip to content

GitLab

Commit 0c5a6624 authored by vladlosev's avatar vladlosev
Browse files

Implement wide->UTF-8 string conversion more correctly

parent c6e674db
Hide whitespace changes
Inline Side-by-side
Showing with 301 additions and 66 deletions
include/gtest/internal/gtest-port.h
View file @ 0c5a6624
......@@ -225,6 +225,12 @@
#include <sys/mman.h>
#endif // GTEST_HAS_STD_STRING && defined(GTEST_OS_LINUX)
// Determines whether the system compiler uses UTF-16 for encoding wide strings.
#if defined(GTEST_OS_WINDOWS) || defined(GTEST_OS_CYGWIN) || \
defined(__SYMBIAN32__)
#define GTEST_WIDE_STRING_USES_UTF16_ 1
#endif
// Defines some utility macros.
// The GNU compiler emits a warning if nested "if" statements are followed by
......
src/gtest-internal-inl.h
View file @ 0c5a6624
......@@ -133,8 +133,30 @@ class GTestFlagSaver {
internal::Int32 repeat_;
} GTEST_ATTRIBUTE_UNUSED;
// Converts a Unicode code-point to its UTF-8 encoding.
String ToUtf8String(wchar_t wchar);
// Converts a Unicode code point to a narrow string in UTF-8 encoding.
// code_point parameter is of type UInt32 because wchar_t may not be
// wide enough to contain a code point.
// The output buffer str must containt at least 32 characters.
// The function returns the address of the output buffer.
// If the code_point is not a valid Unicode code point
// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be output
// as '(Invalid Unicode 0xXXXXXXXX)'.
char* CodePointToUtf8(UInt32 code_point, char* str);
// Converts a wide string to a narrow string in UTF-8 encoding.
// The wide string is assumed to have the following encoding:
// UTF-16 if sizeof(wchar_t) == 2 (on Windows, Cygwin, Symbian OS)
// UTF-32 if sizeof(wchar_t) == 4 (on Linux)
// Parameter str points to a null-terminated wide string.
// Parameter num_chars may additionally limit the number
// of wchar_t characters processed. -1 is used when the entire string
// should be processed.
// If the string contains code points that are not valid Unicode code points
// (i.e. outside of Unicode range U+0 to U+10FFFF) they will be output
// as '(Invalid Unicode 0xXXXXXXXX)'. If the string is in UTF16 encoding
// and contains invalid UTF-16 surrogate pairs, values in those pairs
// will be encoded as individual Unicode characters from Basic Normal Plane.
String WideStringToUtf8(const wchar_t* str, int num_chars);
// Returns the number of active threads, or 0 when there is an error.
size_t GetThreadCount();
......
src/gtest.cc
View file @ 0c5a6624
......@@ -784,16 +784,19 @@ bool String::CStringEquals(const char * lhs, const char * rhs) {
// encoding, and streams the result to the given Message object.
static void StreamWideCharsToMessage(const wchar_t* wstr, size_t len,
Message* msg) {
for (size_t i = 0; i != len; i++) {
// TODO(wan): consider allowing a testing::String object to
// contain '\0'. This will make it behave more like std::string,
// and will allow ToUtf8String() to return the correct encoding
// for '\0' s.t. we can get rid of the conditional here (and in
// several other places).
if (wstr[i]) {
*msg << internal::ToUtf8String(wstr[i]);
// TODO(wan): consider allowing a testing::String object to
// contain '\0'. This will make it behave more like std::string,
// and will allow ToUtf8String() to return the correct encoding
// for '\0' s.t. we can get rid of the conditional here (and in
// several other places).
for (size_t i = 0; i != len; ) { // NOLINT
if (wstr[i] != L'\0') {
*msg << WideStringToUtf8(wstr + i, len - i);
while (i != len && wstr[i] != L'\0')
i++;
} else {
*msg << '\0';
i++;
}
}
}
......@@ -852,8 +855,10 @@ String FormatForFailureMessage(wchar_t wchar) {
Message msg;
// A String object cannot contain '\0', so we print "\\0" when wchar is
// L'\0'.
msg << "L'" << (wchar ? ToUtf8String(wchar).c_str() : "\\0") << "' ("
<< wchar_as_uint64 << ", 0x" << ::std::setbase(16)
char buffer[32]; // CodePointToUtf8 requires a buffer that big.
msg << "L'"
<< (wchar ? CodePointToUtf8(static_cast<UInt32>(wchar), buffer) : "\\0")
<< "' (" << wchar_as_uint64 << ", 0x" << ::std::setbase(16)
<< wchar_as_uint64 << ")";
return msg.GetString();
}
......@@ -1317,31 +1322,118 @@ inline UInt32 ChopLowBits(UInt32* bits, int n) {
return low_bits;
}
// Converts a Unicode code-point to its UTF-8 encoding.
String ToUtf8String(wchar_t wchar) {
char str[5] = {}; // Initializes str to all '\0' characters.
UInt32 code = static_cast<UInt32>(wchar);
if (code <= kMaxCodePoint1) {
str[0] = static_cast<char>(code); // 0xxxxxxx
} else if (code <= kMaxCodePoint2) {
str[1] = static_cast<char>(0x80 | ChopLowBits(&code, 6)); // 10xxxxxx
str[0] = static_cast<char>(0xC0 | code); // 110xxxxx
} else if (code <= kMaxCodePoint3) {
str[2] = static_cast<char>(0x80 | ChopLowBits(&code, 6)); // 10xxxxxx
str[1] = static_cast<char>(0x80 | ChopLowBits(&code, 6)); // 10xxxxxx
str[0] = static_cast<char>(0xE0 | code); // 1110xxxx
} else if (code <= kMaxCodePoint4) {
str[3] = static_cast<char>(0x80 | ChopLowBits(&code, 6)); // 10xxxxxx
str[2] = static_cast<char>(0x80 | ChopLowBits(&code, 6)); // 10xxxxxx
str[1] = static_cast<char>(0x80 | ChopLowBits(&code, 6)); // 10xxxxxx
str[0] = static_cast<char>(0xF0 | code); // 11110xxx
// Converts a Unicode code point to a narrow string in UTF-8 encoding.
// code_point parameter is of type UInt32 because wchar_t may not be
// wide enough to contain a code point.
// The output buffer str must containt at least 32 characters.
// The function returns the address of the output buffer.
// If the code_point is not a valid Unicode code point
// (i.e. outside of Unicode range U+0 to U+10FFFF) it will be output
// as '(Invalid Unicode 0xXXXXXXXX)'.
char* CodePointToUtf8(UInt32 code_point, char* str) {
if (code_point <= kMaxCodePoint1) {
str[1] = '\0';
str[0] = static_cast<char>(code_point); // 0xxxxxxx
} else if (code_point <= kMaxCodePoint2) {
str[2] = '\0';
str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6)); // 10xxxxxx
str[0] = static_cast<char>(0xC0 | code_point); // 110xxxxx
} else if (code_point <= kMaxCodePoint3) {
str[3] = '\0';
str[2] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6)); // 10xxxxxx
str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6)); // 10xxxxxx
str[0] = static_cast<char>(0xE0 | code_point); // 1110xxxx
} else if (code_point <= kMaxCodePoint4) {
str[4] = '\0';
str[3] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6)); // 10xxxxxx
str[2] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6)); // 10xxxxxx
str[1] = static_cast<char>(0x80 | ChopLowBits(&code_point, 6)); // 10xxxxxx
str[0] = static_cast<char>(0xF0 | code_point); // 11110xxx
} else {
return String::Format("(Invalid Unicode 0x%llX)",
static_cast<UInt64>(wchar));
// The longest string String::Format can produce when invoked
// with these parameters is 28 character long (not including
// the terminating nul character). We are asking for 32 character
// buffer just in case. This is also enough for strncpy to
// null-terminate the destination string.
// MSVC 8 deprecates strncpy(), so we want to suppress warning
// 4996 (deprecated function) there.
#ifdef GTEST_OS_WINDOWS // We are on Windows.
#pragma warning(push) // Saves the current warning state.
#pragma warning(disable:4996) // Temporarily disables warning 4996.
#endif
strncpy(str, String::Format("(Invalid Unicode 0x%X)", code_point).c_str(),
32);
#ifdef GTEST_OS_WINDOWS // We are on Windows.
#pragma warning(pop) // Restores the warning state.
#endif
str[31] = '\0'; // Makes sure no change in the format to strncpy leaves
// the result unterminated.
}
return str;
}
// The following two functions only make sense if the the system
// uses UTF-16 for wide string encoding. All supported systems
// with 16 bit wchar_t (Windows, Cygwin, Symbian OS) do use UTF-16.
return String(str);
// Determines if the arguments constitute UTF-16 surrogate pair
// and thus should be combined into a single Unicode code point
// using CreateCodePointFromUtf16SurrogatePair.
inline bool IsUtf16SurrogatePair(wchar_t first, wchar_t second) {
if (sizeof(wchar_t) == 2)
return (first & 0xFC00) == 0xD800 && (second & 0xFC00) == 0xDC00;
else
return false;
}
// Creates a Unicode code point from UTF16 surrogate pair.
inline UInt32 CreateCodePointFromUtf16SurrogatePair(wchar_t first,
wchar_t second) {
if (sizeof(wchar_t) == 2) {
const UInt32 mask = (1 << 10) - 1;
return (((first & mask) << 10) | (second & mask)) + 0x10000;
} else {
// This should not be called, but we provide a sensible default
// in case it is.
return static_cast<UInt32>(first);
}
}
// Converts a wide string to a narrow string in UTF-8 encoding.
// The wide string is assumed to have the following encoding:
// UTF-16 if sizeof(wchar_t) == 2 (on Windows, Cygwin, Symbian OS)
// UTF-32 if sizeof(wchar_t) == 4 (on Linux)
// Parameter str points to a null-terminated wide string.
// Parameter num_chars may additionally limit the number
// of wchar_t characters processed. -1 is used when the entire string
// should be processed.
// If the string contains code points that are not valid Unicode code points
// (i.e. outside of Unicode range U+0 to U+10FFFF) they will be output
// as '(Invalid Unicode 0xXXXXXXXX)'. If the string is in UTF16 encoding
// and contains invalid UTF-16 surrogate pairs, values in those pairs
// will be encoded as individual Unicode characters from Basic Normal Plane.
String WideStringToUtf8(const wchar_t* str, int num_chars) {
if (num_chars == -1)
num_chars = wcslen(str);
StrStream stream;
for (int i = 0; i < num_chars; ++i) {
UInt32 unicode_code_point;
if (str[i] == L'\0') {
break;
} else if (i + 1 < num_chars && IsUtf16SurrogatePair(str[i], str[i + 1])) {
unicode_code_point = CreateCodePointFromUtf16SurrogatePair(str[i],
str[i + 1]);
i++;
} else {
unicode_code_point = static_cast<UInt32>(str[i]);
}
char buffer[32]; // CodePointToUtf8 requires a buffer this big.
stream << CodePointToUtf8(unicode_code_point, buffer);
}
return StrStreamToString(&stream);
}
// Converts a wide C string to a String using the UTF-8 encoding.
......@@ -1349,12 +1441,7 @@ String ToUtf8String(wchar_t wchar) {
String String::ShowWideCString(const wchar_t * wide_c_str) {
if (wide_c_str == NULL) return String("(null)");
StrStream ss;
while (*wide_c_str) {
ss << internal::ToUtf8String(*wide_c_str++);
}
return internal::StrStreamToString(&ss);
return String(internal::WideStringToUtf8(wide_c_str, -1).c_str());
}
// Similar to ShowWideCString(), except that this function encloses
......
test/gtest_unittest.cc
View file @ 0c5a6624
......@@ -101,6 +101,7 @@ using testing::TPRT_NONFATAL_FAILURE;
using testing::TPRT_SUCCESS;
using testing::UnitTest;
using testing::internal::AppendUserMessage;
using testing::internal::CodePointToUtf8;
using testing::internal::EqFailure;
using testing::internal::FloatingPoint;
using testing::internal::GTestFlagSaver;
......@@ -111,8 +112,8 @@ using testing::internal::StreamableToString;
using testing::internal::String;
using testing::internal::TestProperty;
using testing::internal::TestResult;
using testing::internal::ToUtf8String;
using testing::internal::UnitTestImpl;
using testing::internal::WideStringToUtf8;
// This line tests that we can define tests in an unnamed namespace.
namespace {
......@@ -142,65 +143,184 @@ TEST(NullLiteralTest, IsFalseForNonNullLiterals) {
}
#endif // __SYMBIAN32__
// Tests ToUtf8String().
//
// Tests CodePointToUtf8().
// Tests that the NUL character L'\0' is encoded correctly.
TEST(ToUtf8StringTest, CanEncodeNul) {
EXPECT_STREQ("", ToUtf8String(L'\0').c_str());
TEST(CodePointToUtf8Test, CanEncodeNul) {
char buffer[32];
EXPECT_STREQ("", CodePointToUtf8(L'\0', buffer));
}
// Tests that ASCII characters are encoded correctly.
TEST(ToUtf8StringTest, CanEncodeAscii) {
EXPECT_STREQ("a", ToUtf8String(L'a').c_str());
EXPECT_STREQ("Z", ToUtf8String(L'Z').c_str());
EXPECT_STREQ("&", ToUtf8String(L'&').c_str());
EXPECT_STREQ("\x7F", ToUtf8String(L'\x7F').c_str());
TEST(CodePointToUtf8Test, CanEncodeAscii) {
char buffer[32];
EXPECT_STREQ("a", CodePointToUtf8(L'a', buffer));
EXPECT_STREQ("Z", CodePointToUtf8(L'Z', buffer));
EXPECT_STREQ("&", CodePointToUtf8(L'&', buffer));
EXPECT_STREQ("\x7F", CodePointToUtf8(L'\x7F', buffer));
}
// Tests that Unicode code-points that have 8 to 11 bits are encoded
// as 110xxxxx 10xxxxxx.
TEST(ToUtf8StringTest, CanEncode8To11Bits) {
TEST(CodePointToUtf8Test, CanEncode8To11Bits) {
char buffer[32];
// 000 1101 0011 => 110-00011 10-010011
EXPECT_STREQ("\xC3\x93", ToUtf8String(L'\xD3').c_str());
EXPECT_STREQ("\xC3\x93", CodePointToUtf8(L'\xD3', buffer));
// 101 0111 0110 => 110-10101 10-110110
EXPECT_STREQ("\xD5\xB6", ToUtf8String(L'\x576').c_str());
EXPECT_STREQ("\xD5\xB6", CodePointToUtf8(L'\x576', buffer));
}
// Tests that Unicode code-points that have 12 to 16 bits are encoded
// as 1110xxxx 10xxxxxx 10xxxxxx.
TEST(ToUtf8StringTest, CanEncode12To16Bits) {
TEST(CodePointToUtf8Test, CanEncode12To16Bits) {
char buffer[32];
// 0000 1000 1101 0011 => 1110-0000 10-100011 10-010011
EXPECT_STREQ("\xE0\xA3\x93", ToUtf8String(L'\x8D3').c_str());
EXPECT_STREQ("\xE0\xA3\x93", CodePointToUtf8(L'\x8D3', buffer));
// 1100 0111 0100 1101 => 1110-1100 10-011101 10-001101
EXPECT_STREQ("\xEC\x9D\x8D", ToUtf8String(L'\xC74D').c_str());
EXPECT_STREQ("\xEC\x9D\x8D", CodePointToUtf8(L'\xC74D', buffer));
}
#if !defined(GTEST_OS_WINDOWS) && !defined(GTEST_OS_CYGWIN) && \
!defined(__SYMBIAN32__)
#ifndef GTEST_WIDE_STRING_USES_UTF16_
// Tests in this group require a wchar_t to hold > 16 bits, and thus
// are skipped on Windows, Cygwin, and Symbian, where a wchar_t is
// 16-bit wide.
// 16-bit wide. This code may not compile on those systems.
// Tests that Unicode code-points that have 17 to 21 bits are encoded
// as 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx.
TEST(ToUtf8StringTest, CanEncode17To21Bits) {
TEST(CodePointToUtf8Test, CanEncode17To21Bits) {
char buffer[32];
// 0 0001 0000 1000 1101 0011 => 11110-000 10-010000 10-100011 10-010011
EXPECT_STREQ("\xF0\x90\xA3\x93", ToUtf8String(L'\x108D3').c_str());
EXPECT_STREQ("\xF0\x90\xA3\x93", CodePointToUtf8(L'\x108D3', buffer));
// 0 0001 0000 0100 0000 0000 => 11110-000 10-010000 10-010000 10-000000
EXPECT_STREQ("\xF0\x90\x90\x80", CodePointToUtf8(L'\x10400', buffer));
// 1 0111 1000 0110 0011 0100 => 11110-101 10-111000 10-011000 10-110100
EXPECT_STREQ("\xF5\xB8\x98\xB4", ToUtf8String(L'\x178634').c_str());
// 1 0000 1000 0110 0011 0100 => 11110-100 10-001000 10-011000 10-110100
EXPECT_STREQ("\xF4\x88\x98\xB4", CodePointToUtf8(L'\x108634', buffer));
}
// Tests that encoding an invalid code-point generates the expected result.
TEST(ToUtf8StringTest, CanEncodeInvalidCodePoint) {
TEST(CodePointToUtf8Test, CanEncodeInvalidCodePoint) {
char buffer[32];
EXPECT_STREQ("(Invalid Unicode 0x1234ABCD)",
ToUtf8String(L'\x1234ABCD').c_str());
CodePointToUtf8(L'\x1234ABCD', buffer));
}
#endif // GTEST_WIDE_STRING_USES_UTF16_
// Tests WideStringToUtf8().
// Tests that the NUL character L'\0' is encoded correctly.
TEST(WideStringToUtf8Test, CanEncodeNul) {
EXPECT_STREQ("", WideStringToUtf8(L"", 0).c_str());
EXPECT_STREQ("", WideStringToUtf8(L"", -1).c_str());
}
// Tests that ASCII strings are encoded correctly.
TEST(WideStringToUtf8Test, CanEncodeAscii) {
EXPECT_STREQ("a", WideStringToUtf8(L"a", 1).c_str());
EXPECT_STREQ("ab", WideStringToUtf8(L"ab", 2).c_str());
EXPECT_STREQ("a", WideStringToUtf8(L"a", -1).c_str());
EXPECT_STREQ("ab", WideStringToUtf8(L"ab", -1).c_str());
}
// Tests that Unicode code-points that have 8 to 11 bits are encoded
// as 110xxxxx 10xxxxxx.
TEST(WideStringToUtf8Test, CanEncode8To11Bits) {
// 000 1101 0011 => 110-00011 10-010011
EXPECT_STREQ("\xC3\x93", WideStringToUtf8(L"\xD3", 1).c_str());
EXPECT_STREQ("\xC3\x93", WideStringToUtf8(L"\xD3", -1).c_str());
// 101 0111 0110 => 110-10101 10-110110
EXPECT_STREQ("\xD5\xB6", WideStringToUtf8(L"\x576", 1).c_str());
EXPECT_STREQ("\xD5\xB6", WideStringToUtf8(L"\x576", -1).c_str());
}
// Tests that Unicode code-points that have 12 to 16 bits are encoded
// as 1110xxxx 10xxxxxx 10xxxxxx.
TEST(WideStringToUtf8Test, CanEncode12To16Bits) {
// 0000 1000 1101 0011 => 1110-0000 10-100011 10-010011
EXPECT_STREQ("\xE0\xA3\x93", WideStringToUtf8(L"\x8D3", 1).c_str());
EXPECT_STREQ("\xE0\xA3\x93", WideStringToUtf8(L"\x8D3", -1).c_str());
// 1100 0111 0100 1101 => 1110-1100 10-011101 10-001101
EXPECT_STREQ("\xEC\x9D\x8D", WideStringToUtf8(L"\xC74D", 1).c_str());
EXPECT_STREQ("\xEC\x9D\x8D", WideStringToUtf8(L"\xC74D", -1).c_str());
}
#endif // Windows, Cygwin, or Symbian
// Tests that the conversion stops when the function encounters \0 character.
TEST(WideStringToUtf8Test, StopsOnNulCharacter) {
EXPECT_STREQ("ABC", WideStringToUtf8(L"ABC\0XYZ", 100).c_str());
}
// Tests that the conversion stops when the function reaches the limit
// specified by the 'length' parameter.
TEST(WideStringToUtf8Test, StopsWhenLengthLimitReached) {
EXPECT_STREQ("ABC", WideStringToUtf8(L"ABCDEF", 3).c_str());
}
#ifndef GTEST_WIDE_STRING_USES_UTF16_
// Tests that Unicode code-points that have 17 to 21 bits are encoded
// as 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx. This code may not compile
// on the systems using UTF-16 encoding.
TEST(WideStringToUtf8Test, CanEncode17To21Bits) {
// 0 0001 0000 1000 1101 0011 => 11110-000 10-010000 10-100011 10-010011
EXPECT_STREQ("\xF0\x90\xA3\x93", WideStringToUtf8(L"\x108D3", 1).c_str());
EXPECT_STREQ("\xF0\x90\xA3\x93", WideStringToUtf8(L"\x108D3", -1).c_str());
// 1 0000 1000 0110 0011 0100 => 11110-100 10-001000 10-011000 10-110100
EXPECT_STREQ("\xF4\x88\x98\xB4", WideStringToUtf8(L"\x108634", 1).c_str());
EXPECT_STREQ("\xF4\x88\x98\xB4", WideStringToUtf8(L"\x108634", -1).c_str());
}
// Tests that encoding an invalid code-point generates the expected result.
TEST(WideStringToUtf8Test, CanEncodeInvalidCodePoint) {
EXPECT_STREQ("(Invalid Unicode 0xABCDFF)",
WideStringToUtf8(L"\xABCDFF", -1).c_str());
}
#else
// Tests that surrogate pairs are encoded correctly on the systems using
// UTF-16 encoding in the wide strings.
TEST(WideStringToUtf8Test, CanEncodeValidUtf16SUrrogatePairs) {
EXPECT_STREQ("\xF0\x90\x90\x80",
WideStringToUtf8(L"\xD801\xDC00", -1).c_str());
}
// Tests that encoding an invalid UTF-16 surrogate pair
// generates the expected result.
TEST(WideStringToUtf8Test, CanEncodeInvalidUtf16SurrogatePair) {
// Leading surrogate is at the end of the string.
EXPECT_STREQ("\xED\xA0\x80", WideStringToUtf8(L"\xD800", -1).c_str());
// Leading surrogate is not followed by the trailing surrogate.
EXPECT_STREQ("\xED\xA0\x80$", WideStringToUtf8(L"\xD800$", -1).c_str());
// Trailing surrogate appearas without a leading surrogate.
EXPECT_STREQ("\xED\xB0\x80PQR", WideStringToUtf8(L"\xDC00PQR", -1).c_str());
}
#endif // GTEST_WIDE_STRING_USES_UTF16_
// Tests that codepoint concatenation works correctly.
#ifndef GTEST_WIDE_STRING_USES_UTF16_
TEST(WideStringToUtf8Test, ConcatenatesCodepointsCorrectly) {
EXPECT_STREQ(
"\xF4\x88\x98\xB4"
"\xEC\x9D\x8D"
"\n"
"\xD5\xB6"
"\xE0\xA3\x93"
"\xF4\x88\x98\xB4",
WideStringToUtf8(L"\x108634\xC74D\n\x576\x8D3\x108634", -1).c_str());
}
#else
TEST(WideStringToUtf8Test, ConcatenatesCodepointsCorrectly) {
EXPECT_STREQ(
"\xEC\x9D\x8D" "\n" "\xD5\xB6" "\xE0\xA3\x93",
WideStringToUtf8(L"\xC74D\n\x576\x8D3", -1).c_str());
}
#endif // GTEST_WIDE_STRING_USES_UTF16_
// Tests the List template class.
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment