Merge pull request #5046 from UditDewan:fix/print-invalid-char-code-units

PiperOrigin-RevId: 968678073
Change-Id: I669221b694231e385acb9a0bf6640e6f1c319e1e

googletest/include/gtest/gtest-printers.h

@@ -522,17 +522,14 @@ inline void PrintTo(bool x, ::std::ostream* os) {
 GTEST_API_ void PrintTo(wchar_t wc, ::std::ostream* os);
 
 GTEST_API_ void PrintTo(char32_t c, ::std::ostream* os);
-inline void PrintTo(char16_t c, ::std::ostream* os) {
+
-  // TODO(b/418738869): Incorrect for values not representing valid codepoints.
+// Overloads for the UTF-8 and UTF-16 code unit types.  A code unit that
-  // Also see https://github.com/google/googletest/issues/4762.
+// encodes a code point all by itself is printed with the U+XXXX notation;
-  PrintTo(static_cast<char32_t>(c), os);
+// one that does not (any non-ASCII char8_t, and the UTF-16 surrogates) is
-}
+// printed as a code unit instead, the way wchar_t is.
+GTEST_API_ void PrintTo(char16_t c, ::std::ostream* os);
 #ifdef __cpp_lib_char8_t
-inline void PrintTo(char8_t c, ::std::ostream* os) {
+GTEST_API_ void PrintTo(char8_t c, ::std::ostream* os);
-  // TODO(b/418738869): Incorrect for values not representing valid codepoints.
-  // Also see https://github.com/google/googletest/issues/4762.
-  PrintTo(static_cast<char32_t>(c), os);
-}
 #endif
 
 // gcc/clang __{u,}int128_t

googletest/src/gtest-printers.cc

@@ -286,6 +286,28 @@ void PrintTo(char32_t c, ::std::ostream* os) {
       << static_cast<uint32_t>(c);
 }
 
+// char8_t and char16_t hold code units, not code points, so the U+XXXX
+// notation only applies to the values that encode a code point on their own.
+// The rest -- non-ASCII char8_t, which are always part of a multi-unit UTF-8
+// sequence, and the UTF-16 surrogates -- are printed as code units.
+void PrintTo(char16_t c, ::std::ostream* os) {
+  if (c >= 0xD800 && c <= 0xDFFF) {
+    PrintCharAndCodeTo(c, os);
+  } else {
+    PrintTo(static_cast<char32_t>(c), os);
+  }
+}
+
+#ifdef __cpp_lib_char8_t
+void PrintTo(char8_t c, ::std::ostream* os) {
+  if (c >= 0x80) {
+    PrintCharAndCodeTo(c, os);
+  } else {
+    PrintTo(static_cast<char32_t>(c), os);
+  }
+}
+#endif
+
 // gcc/clang __{u,}int128_t
 #if defined(__SIZEOF_INT128__)
 void PrintTo(__uint128_t v, ::std::ostream* os) {

googletest/test/googletest-printers-test.cc

@@ -396,12 +396,24 @@ TEST(PrintCharTest, UnsignedChar) {
   EXPECT_EQ("'b' (98, 0x62)", Print(static_cast<unsigned char>('b')));
 }
 
-TEST(PrintCharTest, Char16) { EXPECT_EQ("U+0041", Print(u'A')); }
+TEST(PrintCharTest, Char16) {
+  EXPECT_EQ("U+0041", Print(u'A'));
+  EXPECT_EQ("U+754C", Print(u'界'));
+  // Surrogates are not code points, so they are printed as code units.
+  EXPECT_EQ("u'\\xD800' (55296)", Print(static_cast<char16_t>(0xD800)));
+  EXPECT_EQ("u'\\xDFFF' (57343)", Print(static_cast<char16_t>(0xDFFF)));
+}
 
 TEST(PrintCharTest, Char32) { EXPECT_EQ("U+0041", Print(U'A')); }
 
 #ifdef __cpp_lib_char8_t
-TEST(PrintCharTest, Char8) { EXPECT_EQ("U+0041", Print(u8'A')); }
+TEST(PrintCharTest, Char8) {
+  EXPECT_EQ("U+0041", Print(u8'A'));
+  // Only ASCII code units encode a code point on their own; the rest are
+  // printed as code units.
+  EXPECT_EQ("u8'\\x80' (128)", Print(static_cast<char8_t>(0x80)));
+  EXPECT_EQ("u8'\\xFF' (255)", Print(static_cast<char8_t>(0xFF)));
+}
 #endif
 
 // Tests printing other simple, built-in types.
@@ -456,7 +468,7 @@ TEST(PrintBuiltInTypeTest, Integer) {
 #ifdef __cpp_lib_char8_t
   EXPECT_EQ("U+0000",
             Print(std::numeric_limits<char8_t>::min()));  // char8_t
-  EXPECT_EQ("U+00FF",
+  EXPECT_EQ("u8'\\xFF' (255)",
             Print(std::numeric_limits<char8_t>::max()));  // char8_t
 #endif
   EXPECT_EQ("U+0000",