From 58cae243738c77acf9b3722804ee96a0f68d7b9a Mon Sep 17 00:00:00 2001 From: Antony Woods <acron1@gmail.com> Date: Tue, 29 Oct 2013 08:36:33 +0000 Subject: Fixed a compiler error with std::chrono::duration --- gtest-1.6.0/test/gtest-port_test.cc | 2412 +++++++++++++++++------------------ 1 file changed, 1206 insertions(+), 1206 deletions(-) (limited to 'gtest-1.6.0/test/gtest-port_test.cc') diff --git a/gtest-1.6.0/test/gtest-port_test.cc b/gtest-1.6.0/test/gtest-port_test.cc index 1c6e2b0..3a3976c 100644 --- a/gtest-1.6.0/test/gtest-port_test.cc +++ b/gtest-1.6.0/test/gtest-port_test.cc @@ -1,1206 +1,1206 @@ -// Copyright 2008, Google Inc. -// All rights reserved. -// -// Redistribution and use in source and binary forms, with or without -// modification, are permitted provided that the following conditions are -// met: -// -// * Redistributions of source code must retain the above copyright -// notice, this list of conditions and the following disclaimer. -// * Redistributions in binary form must reproduce the above -// copyright notice, this list of conditions and the following disclaimer -// in the documentation and/or other materials provided with the -// distribution. -// * Neither the name of Google Inc. nor the names of its -// contributors may be used to endorse or promote products derived from -// this software without specific prior written permission. -// -// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS -// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT -// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR -// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT -// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, -// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT -// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, -// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY -// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT -// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE -// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -// -// Authors: vladl@google.com (Vlad Losev), wan@google.com (Zhanyong Wan) -// -// This file tests the internal cross-platform support utilities. - -#include "gtest/internal/gtest-port.h" - -#include <stdio.h> - -#if GTEST_OS_MAC -# include <time.h> -#endif // GTEST_OS_MAC - -#include <list> -#include <utility> // For std::pair and std::make_pair. -#include <vector> - -#include "gtest/gtest.h" -#include "gtest/gtest-spi.h" - -// Indicates that this translation unit is part of Google Test's -// implementation. It must come before gtest-internal-inl.h is -// included, or there will be a compiler error. This trick is to -// prevent a user from accidentally including gtest-internal-inl.h in -// his code. -#define GTEST_IMPLEMENTATION_ 1 -#include "src/gtest-internal-inl.h" -#undef GTEST_IMPLEMENTATION_ - -using std::make_pair; -using std::pair; - -namespace testing { -namespace internal { - -class Base { - public: - // Copy constructor and assignment operator do exactly what we need, so we - // use them. - Base() : member_(0) {} - explicit Base(int n) : member_(n) {} - virtual ~Base() {} - int member() { return member_; } - - private: - int member_; -}; - -class Derived : public Base { - public: - explicit Derived(int n) : Base(n) {} -}; - -TEST(ImplicitCastTest, ConvertsPointers) { - Derived derived(0); - EXPECT_TRUE(&derived == ::testing::internal::ImplicitCast_<Base*>(&derived)); -} - -TEST(ImplicitCastTest, CanUseInheritance) { - Derived derived(1); - Base base = ::testing::internal::ImplicitCast_<Base>(derived); - EXPECT_EQ(derived.member(), base.member()); -} - -class Castable { - public: - Castable(bool* converted) : converted_(converted) {} - operator Base() { - *converted_ = true; - return Base(); - } - - private: - bool* converted_; -}; - -TEST(ImplicitCastTest, CanUseNonConstCastOperator) { - bool converted = false; - Castable castable(&converted); - Base base = ::testing::internal::ImplicitCast_<Base>(castable); - EXPECT_TRUE(converted); -} - -class ConstCastable { - public: - ConstCastable(bool* converted) : converted_(converted) {} - operator Base() const { - *converted_ = true; - return Base(); - } - - private: - bool* converted_; -}; - -TEST(ImplicitCastTest, CanUseConstCastOperatorOnConstValues) { - bool converted = false; - const ConstCastable const_castable(&converted); - Base base = ::testing::internal::ImplicitCast_<Base>(const_castable); - EXPECT_TRUE(converted); -} - -class ConstAndNonConstCastable { - public: - ConstAndNonConstCastable(bool* converted, bool* const_converted) - : converted_(converted), const_converted_(const_converted) {} - operator Base() { - *converted_ = true; - return Base(); - } - operator Base() const { - *const_converted_ = true; - return Base(); - } - - private: - bool* converted_; - bool* const_converted_; -}; - -TEST(ImplicitCastTest, CanSelectBetweenConstAndNonConstCasrAppropriately) { - bool converted = false; - bool const_converted = false; - ConstAndNonConstCastable castable(&converted, &const_converted); - Base base = ::testing::internal::ImplicitCast_<Base>(castable); - EXPECT_TRUE(converted); - EXPECT_FALSE(const_converted); - - converted = false; - const_converted = false; - const ConstAndNonConstCastable const_castable(&converted, &const_converted); - base = ::testing::internal::ImplicitCast_<Base>(const_castable); - EXPECT_FALSE(converted); - EXPECT_TRUE(const_converted); -} - -class To { - public: - To(bool* converted) { *converted = true; } // NOLINT -}; - -TEST(ImplicitCastTest, CanUseImplicitConstructor) { - bool converted = false; - To to = ::testing::internal::ImplicitCast_<To>(&converted); - (void)to; - EXPECT_TRUE(converted); -} - -TEST(IteratorTraitsTest, WorksForSTLContainerIterators) { - StaticAssertTypeEq<int, - IteratorTraits< ::std::vector<int>::const_iterator>::value_type>(); - StaticAssertTypeEq<bool, - IteratorTraits< ::std::list<bool>::iterator>::value_type>(); -} - -TEST(IteratorTraitsTest, WorksForPointerToNonConst) { - StaticAssertTypeEq<char, IteratorTraits<char*>::value_type>(); - StaticAssertTypeEq<const void*, IteratorTraits<const void**>::value_type>(); -} - -TEST(IteratorTraitsTest, WorksForPointerToConst) { - StaticAssertTypeEq<char, IteratorTraits<const char*>::value_type>(); - StaticAssertTypeEq<const void*, - IteratorTraits<const void* const*>::value_type>(); -} - -// Tests that the element_type typedef is available in scoped_ptr and refers -// to the parameter type. -TEST(ScopedPtrTest, DefinesElementType) { - StaticAssertTypeEq<int, ::testing::internal::scoped_ptr<int>::element_type>(); -} - -// TODO(vladl@google.com): Implement THE REST of scoped_ptr tests. - -TEST(GtestCheckSyntaxTest, BehavesLikeASingleStatement) { - if (AlwaysFalse()) - GTEST_CHECK_(false) << "This should never be executed; " - "It's a compilation test only."; - - if (AlwaysTrue()) - GTEST_CHECK_(true); - else - ; // NOLINT - - if (AlwaysFalse()) - ; // NOLINT - else - GTEST_CHECK_(true) << ""; -} - -TEST(GtestCheckSyntaxTest, WorksWithSwitch) { - switch (0) { - case 1: - break; - default: - GTEST_CHECK_(true); - } - - switch(0) - case 0: - GTEST_CHECK_(true) << "Check failed in switch case"; -} - -// Verifies behavior of FormatFileLocation. -TEST(FormatFileLocationTest, FormatsFileLocation) { - EXPECT_PRED_FORMAT2(IsSubstring, "foo.cc", FormatFileLocation("foo.cc", 42)); - EXPECT_PRED_FORMAT2(IsSubstring, "42", FormatFileLocation("foo.cc", 42)); -} - -TEST(FormatFileLocationTest, FormatsUnknownFile) { - EXPECT_PRED_FORMAT2( - IsSubstring, "unknown file", FormatFileLocation(NULL, 42)); - EXPECT_PRED_FORMAT2(IsSubstring, "42", FormatFileLocation(NULL, 42)); -} - -TEST(FormatFileLocationTest, FormatsUknownLine) { - EXPECT_EQ("foo.cc:", FormatFileLocation("foo.cc", -1)); -} - -TEST(FormatFileLocationTest, FormatsUknownFileAndLine) { - EXPECT_EQ("unknown file:", FormatFileLocation(NULL, -1)); -} - -// Verifies behavior of FormatCompilerIndependentFileLocation. -TEST(FormatCompilerIndependentFileLocationTest, FormatsFileLocation) { - EXPECT_EQ("foo.cc:42", FormatCompilerIndependentFileLocation("foo.cc", 42)); -} - -TEST(FormatCompilerIndependentFileLocationTest, FormatsUknownFile) { - EXPECT_EQ("unknown file:42", - FormatCompilerIndependentFileLocation(NULL, 42)); -} - -TEST(FormatCompilerIndependentFileLocationTest, FormatsUknownLine) { - EXPECT_EQ("foo.cc", FormatCompilerIndependentFileLocation("foo.cc", -1)); -} - -TEST(FormatCompilerIndependentFileLocationTest, FormatsUknownFileAndLine) { - EXPECT_EQ("unknown file", FormatCompilerIndependentFileLocation(NULL, -1)); -} - -#if GTEST_OS_MAC -void* ThreadFunc(void* data) { - pthread_mutex_t* mutex = static_cast<pthread_mutex_t*>(data); - pthread_mutex_lock(mutex); - pthread_mutex_unlock(mutex); - return NULL; -} - -TEST(GetThreadCountTest, ReturnsCorrectValue) { - EXPECT_EQ(1U, GetThreadCount()); - pthread_mutex_t mutex; - pthread_attr_t attr; - pthread_t thread_id; - - // TODO(vladl@google.com): turn mutex into internal::Mutex for automatic - // destruction. - pthread_mutex_init(&mutex, NULL); - pthread_mutex_lock(&mutex); - ASSERT_EQ(0, pthread_attr_init(&attr)); - ASSERT_EQ(0, pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE)); - - const int status = pthread_create(&thread_id, &attr, &ThreadFunc, &mutex); - ASSERT_EQ(0, pthread_attr_destroy(&attr)); - ASSERT_EQ(0, status); - EXPECT_EQ(2U, GetThreadCount()); - pthread_mutex_unlock(&mutex); - - void* dummy; - ASSERT_EQ(0, pthread_join(thread_id, &dummy)); - - // MacOS X may not immediately report the updated thread count after - // joining a thread, causing flakiness in this test. To counter that, we - // wait for up to .5 seconds for the OS to report the correct value. - for (int i = 0; i < 5; ++i) { - if (GetThreadCount() == 1) - break; - - SleepMilliseconds(100); - } - EXPECT_EQ(1U, GetThreadCount()); - pthread_mutex_destroy(&mutex); -} -#else -TEST(GetThreadCountTest, ReturnsZeroWhenUnableToCountThreads) { - EXPECT_EQ(0U, GetThreadCount()); -} -#endif // GTEST_OS_MAC - -TEST(GtestCheckDeathTest, DiesWithCorrectOutputOnFailure) { - const bool a_false_condition = false; - const char regex[] = -#ifdef _MSC_VER - "gtest-port_test\\.cc\\(\\d+\\):" -#elif GTEST_USES_POSIX_RE - "gtest-port_test\\.cc:[0-9]+" -#else - "gtest-port_test\\.cc:\\d+" -#endif // _MSC_VER - ".*a_false_condition.*Extra info.*"; - - EXPECT_DEATH_IF_SUPPORTED(GTEST_CHECK_(a_false_condition) << "Extra info", - regex); -} - -#if GTEST_HAS_DEATH_TEST - -TEST(GtestCheckDeathTest, LivesSilentlyOnSuccess) { - EXPECT_EXIT({ - GTEST_CHECK_(true) << "Extra info"; - ::std::cerr << "Success\n"; - exit(0); }, - ::testing::ExitedWithCode(0), "Success"); -} - -#endif // GTEST_HAS_DEATH_TEST - -// Verifies that Google Test choose regular expression engine appropriate to -// the platform. The test will produce compiler errors in case of failure. -// For simplicity, we only cover the most important platforms here. -TEST(RegexEngineSelectionTest, SelectsCorrectRegexEngine) { -#if GTEST_HAS_POSIX_RE - - EXPECT_TRUE(GTEST_USES_POSIX_RE); - -#else - - EXPECT_TRUE(GTEST_USES_SIMPLE_RE); - -#endif -} - -#if GTEST_USES_POSIX_RE - -# if GTEST_HAS_TYPED_TEST - -template <typename Str> -class RETest : public ::testing::Test {}; - -// Defines StringTypes as the list of all string types that class RE -// supports. -typedef testing::Types< - ::std::string, -# if GTEST_HAS_GLOBAL_STRING - ::string, -# endif // GTEST_HAS_GLOBAL_STRING - const char*> StringTypes; - -TYPED_TEST_CASE(RETest, StringTypes); - -// Tests RE's implicit constructors. -TYPED_TEST(RETest, ImplicitConstructorWorks) { - const RE empty(TypeParam("")); - EXPECT_STREQ("", empty.pattern()); - - const RE simple(TypeParam("hello")); - EXPECT_STREQ("hello", simple.pattern()); - - const RE normal(TypeParam(".*(\\w+)")); - EXPECT_STREQ(".*(\\w+)", normal.pattern()); -} - -// Tests that RE's constructors reject invalid regular expressions. -TYPED_TEST(RETest, RejectsInvalidRegex) { - EXPECT_NONFATAL_FAILURE({ - const RE invalid(TypeParam("?")); - }, "\"?\" is not a valid POSIX Extended regular expression."); -} - -// Tests RE::FullMatch(). -TYPED_TEST(RETest, FullMatchWorks) { - const RE empty(TypeParam("")); - EXPECT_TRUE(RE::FullMatch(TypeParam(""), empty)); - EXPECT_FALSE(RE::FullMatch(TypeParam("a"), empty)); - - const RE re(TypeParam("a.*z")); - EXPECT_TRUE(RE::FullMatch(TypeParam("az"), re)); - EXPECT_TRUE(RE::FullMatch(TypeParam("axyz"), re)); - EXPECT_FALSE(RE::FullMatch(TypeParam("baz"), re)); - EXPECT_FALSE(RE::FullMatch(TypeParam("azy"), re)); -} - -// Tests RE::PartialMatch(). -TYPED_TEST(RETest, PartialMatchWorks) { - const RE empty(TypeParam("")); - EXPECT_TRUE(RE::PartialMatch(TypeParam(""), empty)); - EXPECT_TRUE(RE::PartialMatch(TypeParam("a"), empty)); - - const RE re(TypeParam("a.*z")); - EXPECT_TRUE(RE::PartialMatch(TypeParam("az"), re)); - EXPECT_TRUE(RE::PartialMatch(TypeParam("axyz"), re)); - EXPECT_TRUE(RE::PartialMatch(TypeParam("baz"), re)); - EXPECT_TRUE(RE::PartialMatch(TypeParam("azy"), re)); - EXPECT_FALSE(RE::PartialMatch(TypeParam("zza"), re)); -} - -# endif // GTEST_HAS_TYPED_TEST - -#elif GTEST_USES_SIMPLE_RE - -TEST(IsInSetTest, NulCharIsNotInAnySet) { - EXPECT_FALSE(IsInSet('\0', "")); - EXPECT_FALSE(IsInSet('\0', "\0")); - EXPECT_FALSE(IsInSet('\0', "a")); -} - -TEST(IsInSetTest, WorksForNonNulChars) { - EXPECT_FALSE(IsInSet('a', "Ab")); - EXPECT_FALSE(IsInSet('c', "")); - - EXPECT_TRUE(IsInSet('b', "bcd")); - EXPECT_TRUE(IsInSet('b', "ab")); -} - -TEST(IsAsciiDigitTest, IsFalseForNonDigit) { - EXPECT_FALSE(IsAsciiDigit('\0')); - EXPECT_FALSE(IsAsciiDigit(' ')); - EXPECT_FALSE(IsAsciiDigit('+')); - EXPECT_FALSE(IsAsciiDigit('-')); - EXPECT_FALSE(IsAsciiDigit('.')); - EXPECT_FALSE(IsAsciiDigit('a')); -} - -TEST(IsAsciiDigitTest, IsTrueForDigit) { - EXPECT_TRUE(IsAsciiDigit('0')); - EXPECT_TRUE(IsAsciiDigit('1')); - EXPECT_TRUE(IsAsciiDigit('5')); - EXPECT_TRUE(IsAsciiDigit('9')); -} - -TEST(IsAsciiPunctTest, IsFalseForNonPunct) { - EXPECT_FALSE(IsAsciiPunct('\0')); - EXPECT_FALSE(IsAsciiPunct(' ')); - EXPECT_FALSE(IsAsciiPunct('\n')); - EXPECT_FALSE(IsAsciiPunct('a')); - EXPECT_FALSE(IsAsciiPunct('0')); -} - -TEST(IsAsciiPunctTest, IsTrueForPunct) { - for (const char* p = "^-!\"#$%&'()*+,./:;<=>?@[\\]_`{|}~"; *p; p++) { - EXPECT_PRED1(IsAsciiPunct, *p); - } -} - -TEST(IsRepeatTest, IsFalseForNonRepeatChar) { - EXPECT_FALSE(IsRepeat('\0')); - EXPECT_FALSE(IsRepeat(' ')); - EXPECT_FALSE(IsRepeat('a')); - EXPECT_FALSE(IsRepeat('1')); - EXPECT_FALSE(IsRepeat('-')); -} - -TEST(IsRepeatTest, IsTrueForRepeatChar) { - EXPECT_TRUE(IsRepeat('?')); - EXPECT_TRUE(IsRepeat('*')); - EXPECT_TRUE(IsRepeat('+')); -} - -TEST(IsAsciiWhiteSpaceTest, IsFalseForNonWhiteSpace) { - EXPECT_FALSE(IsAsciiWhiteSpace('\0')); - EXPECT_FALSE(IsAsciiWhiteSpace('a')); - EXPECT_FALSE(IsAsciiWhiteSpace('1')); - EXPECT_FALSE(IsAsciiWhiteSpace('+')); - EXPECT_FALSE(IsAsciiWhiteSpace('_')); -} - -TEST(IsAsciiWhiteSpaceTest, IsTrueForWhiteSpace) { - EXPECT_TRUE(IsAsciiWhiteSpace(' ')); - EXPECT_TRUE(IsAsciiWhiteSpace('\n')); - EXPECT_TRUE(IsAsciiWhiteSpace('\r')); - EXPECT_TRUE(IsAsciiWhiteSpace('\t')); - EXPECT_TRUE(IsAsciiWhiteSpace('\v')); - EXPECT_TRUE(IsAsciiWhiteSpace('\f')); -} - -TEST(IsAsciiWordCharTest, IsFalseForNonWordChar) { - EXPECT_FALSE(IsAsciiWordChar('\0')); - EXPECT_FALSE(IsAsciiWordChar('+')); - EXPECT_FALSE(IsAsciiWordChar('.')); - EXPECT_FALSE(IsAsciiWordChar(' ')); - EXPECT_FALSE(IsAsciiWordChar('\n')); -} - -TEST(IsAsciiWordCharTest, IsTrueForLetter) { - EXPECT_TRUE(IsAsciiWordChar('a')); - EXPECT_TRUE(IsAsciiWordChar('b')); - EXPECT_TRUE(IsAsciiWordChar('A')); - EXPECT_TRUE(IsAsciiWordChar('Z')); -} - -TEST(IsAsciiWordCharTest, IsTrueForDigit) { - EXPECT_TRUE(IsAsciiWordChar('0')); - EXPECT_TRUE(IsAsciiWordChar('1')); - EXPECT_TRUE(IsAsciiWordChar('7')); - EXPECT_TRUE(IsAsciiWordChar('9')); -} - -TEST(IsAsciiWordCharTest, IsTrueForUnderscore) { - EXPECT_TRUE(IsAsciiWordChar('_')); -} - -TEST(IsValidEscapeTest, IsFalseForNonPrintable) { - EXPECT_FALSE(IsValidEscape('\0')); - EXPECT_FALSE(IsValidEscape('\007')); -} - -TEST(IsValidEscapeTest, IsFalseForDigit) { - EXPECT_FALSE(IsValidEscape('0')); - EXPECT_FALSE(IsValidEscape('9')); -} - -TEST(IsValidEscapeTest, IsFalseForWhiteSpace) { - EXPECT_FALSE(IsValidEscape(' ')); - EXPECT_FALSE(IsValidEscape('\n')); -} - -TEST(IsValidEscapeTest, IsFalseForSomeLetter) { - EXPECT_FALSE(IsValidEscape('a')); - EXPECT_FALSE(IsValidEscape('Z')); -} - -TEST(IsValidEscapeTest, IsTrueForPunct) { - EXPECT_TRUE(IsValidEscape('.')); - EXPECT_TRUE(IsValidEscape('-')); - EXPECT_TRUE(IsValidEscape('^')); - EXPECT_TRUE(IsValidEscape('$')); - EXPECT_TRUE(IsValidEscape('(')); - EXPECT_TRUE(IsValidEscape(']')); - EXPECT_TRUE(IsValidEscape('{')); - EXPECT_TRUE(IsValidEscape('|')); -} - -TEST(IsValidEscapeTest, IsTrueForSomeLetter) { - EXPECT_TRUE(IsValidEscape('d')); - EXPECT_TRUE(IsValidEscape('D')); - EXPECT_TRUE(IsValidEscape('s')); - EXPECT_TRUE(IsValidEscape('S')); - EXPECT_TRUE(IsValidEscape('w')); - EXPECT_TRUE(IsValidEscape('W')); -} - -TEST(AtomMatchesCharTest, EscapedPunct) { - EXPECT_FALSE(AtomMatchesChar(true, '\\', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, '\\', ' ')); - EXPECT_FALSE(AtomMatchesChar(true, '_', '.')); - EXPECT_FALSE(AtomMatchesChar(true, '.', 'a')); - - EXPECT_TRUE(AtomMatchesChar(true, '\\', '\\')); - EXPECT_TRUE(AtomMatchesChar(true, '_', '_')); - EXPECT_TRUE(AtomMatchesChar(true, '+', '+')); - EXPECT_TRUE(AtomMatchesChar(true, '.', '.')); -} - -TEST(AtomMatchesCharTest, Escaped_d) { - EXPECT_FALSE(AtomMatchesChar(true, 'd', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 'd', 'a')); - EXPECT_FALSE(AtomMatchesChar(true, 'd', '.')); - - EXPECT_TRUE(AtomMatchesChar(true, 'd', '0')); - EXPECT_TRUE(AtomMatchesChar(true, 'd', '9')); -} - -TEST(AtomMatchesCharTest, Escaped_D) { - EXPECT_FALSE(AtomMatchesChar(true, 'D', '0')); - EXPECT_FALSE(AtomMatchesChar(true, 'D', '9')); - - EXPECT_TRUE(AtomMatchesChar(true, 'D', '\0')); - EXPECT_TRUE(AtomMatchesChar(true, 'D', 'a')); - EXPECT_TRUE(AtomMatchesChar(true, 'D', '-')); -} - -TEST(AtomMatchesCharTest, Escaped_s) { - EXPECT_FALSE(AtomMatchesChar(true, 's', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 's', 'a')); - EXPECT_FALSE(AtomMatchesChar(true, 's', '.')); - EXPECT_FALSE(AtomMatchesChar(true, 's', '9')); - - EXPECT_TRUE(AtomMatchesChar(true, 's', ' ')); - EXPECT_TRUE(AtomMatchesChar(true, 's', '\n')); - EXPECT_TRUE(AtomMatchesChar(true, 's', '\t')); -} - -TEST(AtomMatchesCharTest, Escaped_S) { - EXPECT_FALSE(AtomMatchesChar(true, 'S', ' ')); - EXPECT_FALSE(AtomMatchesChar(true, 'S', '\r')); - - EXPECT_TRUE(AtomMatchesChar(true, 'S', '\0')); - EXPECT_TRUE(AtomMatchesChar(true, 'S', 'a')); - EXPECT_TRUE(AtomMatchesChar(true, 'S', '9')); -} - -TEST(AtomMatchesCharTest, Escaped_w) { - EXPECT_FALSE(AtomMatchesChar(true, 'w', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 'w', '+')); - EXPECT_FALSE(AtomMatchesChar(true, 'w', ' ')); - EXPECT_FALSE(AtomMatchesChar(true, 'w', '\n')); - - EXPECT_TRUE(AtomMatchesChar(true, 'w', '0')); - EXPECT_TRUE(AtomMatchesChar(true, 'w', 'b')); - EXPECT_TRUE(AtomMatchesChar(true, 'w', 'C')); - EXPECT_TRUE(AtomMatchesChar(true, 'w', '_')); -} - -TEST(AtomMatchesCharTest, Escaped_W) { - EXPECT_FALSE(AtomMatchesChar(true, 'W', 'A')); - EXPECT_FALSE(AtomMatchesChar(true, 'W', 'b')); - EXPECT_FALSE(AtomMatchesChar(true, 'W', '9')); - EXPECT_FALSE(AtomMatchesChar(true, 'W', '_')); - - EXPECT_TRUE(AtomMatchesChar(true, 'W', '\0')); - EXPECT_TRUE(AtomMatchesChar(true, 'W', '*')); - EXPECT_TRUE(AtomMatchesChar(true, 'W', '\n')); -} - -TEST(AtomMatchesCharTest, EscapedWhiteSpace) { - EXPECT_FALSE(AtomMatchesChar(true, 'f', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 'f', '\n')); - EXPECT_FALSE(AtomMatchesChar(true, 'n', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 'n', '\r')); - EXPECT_FALSE(AtomMatchesChar(true, 'r', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 'r', 'a')); - EXPECT_FALSE(AtomMatchesChar(true, 't', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 't', 't')); - EXPECT_FALSE(AtomMatchesChar(true, 'v', '\0')); - EXPECT_FALSE(AtomMatchesChar(true, 'v', '\f')); - - EXPECT_TRUE(AtomMatchesChar(true, 'f', '\f')); - EXPECT_TRUE(AtomMatchesChar(true, 'n', '\n')); - EXPECT_TRUE(AtomMatchesChar(true, 'r', '\r')); - EXPECT_TRUE(AtomMatchesChar(true, 't', '\t')); - EXPECT_TRUE(AtomMatchesChar(true, 'v', '\v')); -} - -TEST(AtomMatchesCharTest, UnescapedDot) { - EXPECT_FALSE(AtomMatchesChar(false, '.', '\n')); - - EXPECT_TRUE(AtomMatchesChar(false, '.', '\0')); - EXPECT_TRUE(AtomMatchesChar(false, '.', '.')); - EXPECT_TRUE(AtomMatchesChar(false, '.', 'a')); - EXPECT_TRUE(AtomMatchesChar(false, '.', ' ')); -} - -TEST(AtomMatchesCharTest, UnescapedChar) { - EXPECT_FALSE(AtomMatchesChar(false, 'a', '\0')); - EXPECT_FALSE(AtomMatchesChar(false, 'a', 'b')); - EXPECT_FALSE(AtomMatchesChar(false, '$', 'a')); - - EXPECT_TRUE(AtomMatchesChar(false, '$', '$')); - EXPECT_TRUE(AtomMatchesChar(false, '5', '5')); - EXPECT_TRUE(AtomMatchesChar(false, 'Z', 'Z')); -} - -TEST(ValidateRegexTest, GeneratesFailureAndReturnsFalseForInvalid) { - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex(NULL)), - "NULL is not a valid simple regular expression"); - EXPECT_NONFATAL_FAILURE( - ASSERT_FALSE(ValidateRegex("a\\")), - "Syntax error at index 1 in simple regular expression \"a\\\": "); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("a\\")), - "'\\' cannot appear at the end"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("\\n\\")), - "'\\' cannot appear at the end"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("\\s\\hb")), - "invalid escape sequence \"\\h\""); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("^^")), - "'^' can only appear at the beginning"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex(".*^b")), - "'^' can only appear at the beginning"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("$$")), - "'$' can only appear at the end"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("^$a")), - "'$' can only appear at the end"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("a(b")), - "'(' is unsupported"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("ab)")), - "')' is unsupported"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("[ab")), - "'[' is unsupported"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("a{2")), - "'{' is unsupported"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("?")), - "'?' can only follow a repeatable token"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("^*")), - "'*' can only follow a repeatable token"); - EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("5*+")), - "'+' can only follow a repeatable token"); -} - -TEST(ValidateRegexTest, ReturnsTrueForValid) { - EXPECT_TRUE(ValidateRegex("")); - EXPECT_TRUE(ValidateRegex("a")); - EXPECT_TRUE(ValidateRegex(".*")); - EXPECT_TRUE(ValidateRegex("^a_+")); - EXPECT_TRUE(ValidateRegex("^a\\t\\&?")); - EXPECT_TRUE(ValidateRegex("09*$")); - EXPECT_TRUE(ValidateRegex("^Z$")); - EXPECT_TRUE(ValidateRegex("a\\^Z\\$\\(\\)\\|\\[\\]\\{\\}")); -} - -TEST(MatchRepetitionAndRegexAtHeadTest, WorksForZeroOrOne) { - EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "a", "ba")); - // Repeating more than once. - EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "b", "aab")); - - // Repeating zero times. - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "b", "ba")); - // Repeating once. - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "b", "ab")); - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '#', '?', ".", "##")); -} - -TEST(MatchRepetitionAndRegexAtHeadTest, WorksForZeroOrMany) { - EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, '.', '*', "a$", "baab")); - - // Repeating zero times. - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '.', '*', "b", "bc")); - // Repeating once. - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '.', '*', "b", "abc")); - // Repeating more than once. - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(true, 'w', '*', "-", "ab_1-g")); -} - -TEST(MatchRepetitionAndRegexAtHeadTest, WorksForOneOrMany) { - EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, '.', '+', "a$", "baab")); - // Repeating zero times. - EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, '.', '+', "b", "bc")); - - // Repeating once. - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '.', '+', "b", "abc")); - // Repeating more than once. - EXPECT_TRUE(MatchRepetitionAndRegexAtHead(true, 'w', '+', "-", "ab_1-g")); -} - -TEST(MatchRegexAtHeadTest, ReturnsTrueForEmptyRegex) { - EXPECT_TRUE(MatchRegexAtHead("", "")); - EXPECT_TRUE(MatchRegexAtHead("", "ab")); -} - -TEST(MatchRegexAtHeadTest, WorksWhenDollarIsInRegex) { - EXPECT_FALSE(MatchRegexAtHead("$", "a")); - - EXPECT_TRUE(MatchRegexAtHead("$", "")); - EXPECT_TRUE(MatchRegexAtHead("a$", "a")); -} - -TEST(MatchRegexAtHeadTest, WorksWhenRegexStartsWithEscapeSequence) { - EXPECT_FALSE(MatchRegexAtHead("\\w", "+")); - EXPECT_FALSE(MatchRegexAtHead("\\W", "ab")); - - EXPECT_TRUE(MatchRegexAtHead("\\sa", "\nab")); - EXPECT_TRUE(MatchRegexAtHead("\\d", "1a")); -} - -TEST(MatchRegexAtHeadTest, WorksWhenRegexStartsWithRepetition) { - EXPECT_FALSE(MatchRegexAtHead(".+a", "abc")); - EXPECT_FALSE(MatchRegexAtHead("a?b", "aab")); - - EXPECT_TRUE(MatchRegexAtHead(".*a", "bc12-ab")); - EXPECT_TRUE(MatchRegexAtHead("a?b", "b")); - EXPECT_TRUE(MatchRegexAtHead("a?b", "ab")); -} - -TEST(MatchRegexAtHeadTest, - WorksWhenRegexStartsWithRepetionOfEscapeSequence) { - EXPECT_FALSE(MatchRegexAtHead("\\.+a", "abc")); - EXPECT_FALSE(MatchRegexAtHead("\\s?b", " b")); - - EXPECT_TRUE(MatchRegexAtHead("\\(*a", "((((ab")); - EXPECT_TRUE(MatchRegexAtHead("\\^?b", "^b")); - EXPECT_TRUE(MatchRegexAtHead("\\\\?b", "b")); - EXPECT_TRUE(MatchRegexAtHead("\\\\?b", "\\b")); -} - -TEST(MatchRegexAtHeadTest, MatchesSequentially) { - EXPECT_FALSE(MatchRegexAtHead("ab.*c", "acabc")); - - EXPECT_TRUE(MatchRegexAtHead("ab.*c", "ab-fsc")); -} - -TEST(MatchRegexAnywhereTest, ReturnsFalseWhenStringIsNull) { - EXPECT_FALSE(MatchRegexAnywhere("", NULL)); -} - -TEST(MatchRegexAnywhereTest, WorksWhenRegexStartsWithCaret) { - EXPECT_FALSE(MatchRegexAnywhere("^a", "ba")); - EXPECT_FALSE(MatchRegexAnywhere("^$", "a")); - - EXPECT_TRUE(MatchRegexAnywhere("^a", "ab")); - EXPECT_TRUE(MatchRegexAnywhere("^", "ab")); - EXPECT_TRUE(MatchRegexAnywhere("^$", "")); -} - -TEST(MatchRegexAnywhereTest, ReturnsFalseWhenNoMatch) { - EXPECT_FALSE(MatchRegexAnywhere("a", "bcde123")); - EXPECT_FALSE(MatchRegexAnywhere("a.+a", "--aa88888888")); -} - -TEST(MatchRegexAnywhereTest, ReturnsTrueWhenMatchingPrefix) { - EXPECT_TRUE(MatchRegexAnywhere("\\w+", "ab1_ - 5")); - EXPECT_TRUE(MatchRegexAnywhere(".*=", "=")); - EXPECT_TRUE(MatchRegexAnywhere("x.*ab?.*bc", "xaaabc")); -} - -TEST(MatchRegexAnywhereTest, ReturnsTrueWhenMatchingNonPrefix) { - EXPECT_TRUE(MatchRegexAnywhere("\\w+", "$$$ ab1_ - 5")); - EXPECT_TRUE(MatchRegexAnywhere("\\.+=", "= ...=")); -} - -// Tests RE's implicit constructors. -TEST(RETest, ImplicitConstructorWorks) { - const RE empty(""); - EXPECT_STREQ("", empty.pattern()); - - const RE simple("hello"); - EXPECT_STREQ("hello", simple.pattern()); -} - -// Tests that RE's constructors reject invalid regular expressions. -TEST(RETest, RejectsInvalidRegex) { - EXPECT_NONFATAL_FAILURE({ - const RE normal(NULL); - }, "NULL is not a valid simple regular expression"); - - EXPECT_NONFATAL_FAILURE({ - const RE normal(".*(\\w+"); - }, "'(' is unsupported"); - - EXPECT_NONFATAL_FAILURE({ - const RE invalid("^?"); - }, "'?' can only follow a repeatable token"); -} - -// Tests RE::FullMatch(). -TEST(RETest, FullMatchWorks) { - const RE empty(""); - EXPECT_TRUE(RE::FullMatch("", empty)); - EXPECT_FALSE(RE::FullMatch("a", empty)); - - const RE re1("a"); - EXPECT_TRUE(RE::FullMatch("a", re1)); - - const RE re("a.*z"); - EXPECT_TRUE(RE::FullMatch("az", re)); - EXPECT_TRUE(RE::FullMatch("axyz", re)); - EXPECT_FALSE(RE::FullMatch("baz", re)); - EXPECT_FALSE(RE::FullMatch("azy", re)); -} - -// Tests RE::PartialMatch(). -TEST(RETest, PartialMatchWorks) { - const RE empty(""); - EXPECT_TRUE(RE::PartialMatch("", empty)); - EXPECT_TRUE(RE::PartialMatch("a", empty)); - - const RE re("a.*z"); - EXPECT_TRUE(RE::PartialMatch("az", re)); - EXPECT_TRUE(RE::PartialMatch("axyz", re)); - EXPECT_TRUE(RE::PartialMatch("baz", re)); - EXPECT_TRUE(RE::PartialMatch("azy", re)); - EXPECT_FALSE(RE::PartialMatch("zza", re)); -} - -#endif // GTEST_USES_POSIX_RE - -#if !GTEST_OS_WINDOWS_MOBILE - -TEST(CaptureTest, CapturesStdout) { - CaptureStdout(); - fprintf(stdout, "abc"); - EXPECT_STREQ("abc", GetCapturedStdout().c_str()); - - CaptureStdout(); - fprintf(stdout, "def%cghi", '\0'); - EXPECT_EQ(::std::string("def\0ghi", 7), ::std::string(GetCapturedStdout())); -} - -TEST(CaptureTest, CapturesStderr) { - CaptureStderr(); - fprintf(stderr, "jkl"); - EXPECT_STREQ("jkl", GetCapturedStderr().c_str()); - - CaptureStderr(); - fprintf(stderr, "jkl%cmno", '\0'); - EXPECT_EQ(::std::string("jkl\0mno", 7), ::std::string(GetCapturedStderr())); -} - -// Tests that stdout and stderr capture don't interfere with each other. -TEST(CaptureTest, CapturesStdoutAndStderr) { - CaptureStdout(); - CaptureStderr(); - fprintf(stdout, "pqr"); - fprintf(stderr, "stu"); - EXPECT_STREQ("pqr", GetCapturedStdout().c_str()); - EXPECT_STREQ("stu", GetCapturedStderr().c_str()); -} - -TEST(CaptureDeathTest, CannotReenterStdoutCapture) { - CaptureStdout(); - EXPECT_DEATH_IF_SUPPORTED(CaptureStdout();, - "Only one stdout capturer can exist at a time"); - GetCapturedStdout(); - - // We cannot test stderr capturing using death tests as they use it - // themselves. -} - -#endif // !GTEST_OS_WINDOWS_MOBILE - -TEST(ThreadLocalTest, DefaultConstructorInitializesToDefaultValues) { - ThreadLocal<int> t1; - EXPECT_EQ(0, t1.get()); - - ThreadLocal<void*> t2; - EXPECT_TRUE(t2.get() == NULL); -} - -TEST(ThreadLocalTest, SingleParamConstructorInitializesToParam) { - ThreadLocal<int> t1(123); - EXPECT_EQ(123, t1.get()); - - int i = 0; - ThreadLocal<int*> t2(&i); - EXPECT_EQ(&i, t2.get()); -} - -class NoDefaultContructor { - public: - explicit NoDefaultContructor(const char*) {} - NoDefaultContructor(const NoDefaultContructor&) {} -}; - -TEST(ThreadLocalTest, ValueDefaultContructorIsNotRequiredForParamVersion) { - ThreadLocal<NoDefaultContructor> bar(NoDefaultContructor("foo")); - bar.pointer(); -} - -TEST(ThreadLocalTest, GetAndPointerReturnSameValue) { - ThreadLocal<String> thread_local; - - EXPECT_EQ(thread_local.pointer(), &(thread_local.get())); - - // Verifies the condition still holds after calling set. - thread_local.set("foo"); - EXPECT_EQ(thread_local.pointer(), &(thread_local.get())); -} - -TEST(ThreadLocalTest, PointerAndConstPointerReturnSameValue) { - ThreadLocal<String> thread_local; - const ThreadLocal<String>& const_thread_local = thread_local; - - EXPECT_EQ(thread_local.pointer(), const_thread_local.pointer()); - - thread_local.set("foo"); - EXPECT_EQ(thread_local.pointer(), const_thread_local.pointer()); -} - -#if GTEST_IS_THREADSAFE - -void AddTwo(int* param) { *param += 2; } - -TEST(ThreadWithParamTest, ConstructorExecutesThreadFunc) { - int i = 40; - ThreadWithParam<int*> thread(&AddTwo, &i, NULL); - thread.Join(); - EXPECT_EQ(42, i); -} - -TEST(MutexDeathTest, AssertHeldShouldAssertWhenNotLocked) { - // AssertHeld() is flaky only in the presence of multiple threads accessing - // the lock. In this case, the test is robust. - EXPECT_DEATH_IF_SUPPORTED({ - Mutex m; - { MutexLock lock(&m); } - m.AssertHeld(); - }, - "thread .*hold"); -} - -TEST(MutexTest, AssertHeldShouldNotAssertWhenLocked) { - Mutex m; - MutexLock lock(&m); - m.AssertHeld(); -} - -class AtomicCounterWithMutex { - public: - explicit AtomicCounterWithMutex(Mutex* mutex) : - value_(0), mutex_(mutex), random_(42) {} - - void Increment() { - MutexLock lock(mutex_); - int temp = value_; - { - // Locking a mutex puts up a memory barrier, preventing reads and - // writes to value_ rearranged when observed from other threads. - // - // We cannot use Mutex and MutexLock here or rely on their memory - // barrier functionality as we are testing them here. - pthread_mutex_t memory_barrier_mutex; - GTEST_CHECK_POSIX_SUCCESS_( - pthread_mutex_init(&memory_barrier_mutex, NULL)); - GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_lock(&memory_barrier_mutex)); - - SleepMilliseconds(random_.Generate(30)); - - GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_unlock(&memory_barrier_mutex)); - } - value_ = temp + 1; - } - int value() const { return value_; } - - private: - volatile int value_; - Mutex* const mutex_; // Protects value_. - Random random_; -}; - -void CountingThreadFunc(pair<AtomicCounterWithMutex*, int> param) { - for (int i = 0; i < param.second; ++i) - param.first->Increment(); -} - -// Tests that the mutex only lets one thread at a time to lock it. -TEST(MutexTest, OnlyOneThreadCanLockAtATime) { - Mutex mutex; - AtomicCounterWithMutex locked_counter(&mutex); - - typedef ThreadWithParam<pair<AtomicCounterWithMutex*, int> > ThreadType; - const int kCycleCount = 20; - const int kThreadCount = 7; - scoped_ptr<ThreadType> counting_threads[kThreadCount]; - Notification threads_can_start; - // Creates and runs kThreadCount threads that increment locked_counter - // kCycleCount times each. - for (int i = 0; i < kThreadCount; ++i) { - counting_threads[i].reset(new ThreadType(&CountingThreadFunc, - make_pair(&locked_counter, - kCycleCount), - &threads_can_start)); - } - threads_can_start.Notify(); - for (int i = 0; i < kThreadCount; ++i) - counting_threads[i]->Join(); - - // If the mutex lets more than one thread to increment the counter at a - // time, they are likely to encounter a race condition and have some - // increments overwritten, resulting in the lower then expected counter - // value. - EXPECT_EQ(kCycleCount * kThreadCount, locked_counter.value()); -} - -template <typename T> -void RunFromThread(void (func)(T), T param) { - ThreadWithParam<T> thread(func, param, NULL); - thread.Join(); -} - -void RetrieveThreadLocalValue(pair<ThreadLocal<String>*, String*> param) { - *param.second = param.first->get(); -} - -TEST(ThreadLocalTest, ParameterizedConstructorSetsDefault) { - ThreadLocal<String> thread_local("foo"); - EXPECT_STREQ("foo", thread_local.get().c_str()); - - thread_local.set("bar"); - EXPECT_STREQ("bar", thread_local.get().c_str()); - - String result; - RunFromThread(&RetrieveThreadLocalValue, make_pair(&thread_local, &result)); - EXPECT_STREQ("foo", result.c_str()); -} - -// DestructorTracker keeps track of whether its instances have been -// destroyed. -static std::vector<bool> g_destroyed; - -class DestructorTracker { - public: - DestructorTracker() : index_(GetNewIndex()) {} - DestructorTracker(const DestructorTracker& /* rhs */) - : index_(GetNewIndex()) {} - ~DestructorTracker() { - // We never access g_destroyed concurrently, so we don't need to - // protect the write operation under a mutex. - g_destroyed[index_] = true; - } - - private: - static int GetNewIndex() { - g_destroyed.push_back(false); - return g_destroyed.size() - 1; - } - const int index_; -}; - -typedef ThreadLocal<DestructorTracker>* ThreadParam; - -void CallThreadLocalGet(ThreadParam thread_local) { - thread_local->get(); -} - -// Tests that when a ThreadLocal object dies in a thread, it destroys -// the managed object for that thread. -TEST(ThreadLocalTest, DestroysManagedObjectForOwnThreadWhenDying) { - g_destroyed.clear(); - - { - // The next line default constructs a DestructorTracker object as - // the default value of objects managed by thread_local. - ThreadLocal<DestructorTracker> thread_local; - ASSERT_EQ(1U, g_destroyed.size()); - ASSERT_FALSE(g_destroyed[0]); - - // This creates another DestructorTracker object for the main thread. - thread_local.get(); - ASSERT_EQ(2U, g_destroyed.size()); - ASSERT_FALSE(g_destroyed[0]); - ASSERT_FALSE(g_destroyed[1]); - } - - // Now thread_local has died. It should have destroyed both the - // default value shared by all threads and the value for the main - // thread. - ASSERT_EQ(2U, g_destroyed.size()); - EXPECT_TRUE(g_destroyed[0]); - EXPECT_TRUE(g_destroyed[1]); - - g_destroyed.clear(); -} - -// Tests that when a thread exits, the thread-local object for that -// thread is destroyed. -TEST(ThreadLocalTest, DestroysManagedObjectAtThreadExit) { - g_destroyed.clear(); - - { - // The next line default constructs a DestructorTracker object as - // the default value of objects managed by thread_local. - ThreadLocal<DestructorTracker> thread_local; - ASSERT_EQ(1U, g_destroyed.size()); - ASSERT_FALSE(g_destroyed[0]); - - // This creates another DestructorTracker object in the new thread. - ThreadWithParam<ThreadParam> thread( - &CallThreadLocalGet, &thread_local, NULL); - thread.Join(); - - // Now the new thread has exited. The per-thread object for it - // should have been destroyed. - ASSERT_EQ(2U, g_destroyed.size()); - ASSERT_FALSE(g_destroyed[0]); - ASSERT_TRUE(g_destroyed[1]); - } - - // Now thread_local has died. The default value should have been - // destroyed too. - ASSERT_EQ(2U, g_destroyed.size()); - EXPECT_TRUE(g_destroyed[0]); - EXPECT_TRUE(g_destroyed[1]); - - g_destroyed.clear(); -} - -TEST(ThreadLocalTest, ThreadLocalMutationsAffectOnlyCurrentThread) { - ThreadLocal<String> thread_local; - thread_local.set("Foo"); - EXPECT_STREQ("Foo", thread_local.get().c_str()); - - String result; - RunFromThread(&RetrieveThreadLocalValue, make_pair(&thread_local, &result)); - EXPECT_TRUE(result.c_str() == NULL); -} - -#endif // GTEST_IS_THREADSAFE - -} // namespace internal -} // namespace testing +// Copyright 2008, Google Inc. +// All rights reserved. +// +// Redistribution and use in source and binary forms, with or without +// modification, are permitted provided that the following conditions are +// met: +// +// * Redistributions of source code must retain the above copyright +// notice, this list of conditions and the following disclaimer. +// * Redistributions in binary form must reproduce the above +// copyright notice, this list of conditions and the following disclaimer +// in the documentation and/or other materials provided with the +// distribution. +// * Neither the name of Google Inc. nor the names of its +// contributors may be used to endorse or promote products derived from +// this software without specific prior written permission. +// +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT +// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT +// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, +// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT +// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY +// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT +// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE +// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +// +// Authors: vladl@google.com (Vlad Losev), wan@google.com (Zhanyong Wan) +// +// This file tests the internal cross-platform support utilities. + +#include "gtest/internal/gtest-port.h" + +#include <stdio.h> + +#if GTEST_OS_MAC +# include <time.h> +#endif // GTEST_OS_MAC + +#include <list> +#include <utility> // For std::pair and std::make_pair. +#include <vector> + +#include "gtest/gtest.h" +#include "gtest/gtest-spi.h" + +// Indicates that this translation unit is part of Google Test's +// implementation. It must come before gtest-internal-inl.h is +// included, or there will be a compiler error. This trick is to +// prevent a user from accidentally including gtest-internal-inl.h in +// his code. +#define GTEST_IMPLEMENTATION_ 1 +#include "src/gtest-internal-inl.h" +#undef GTEST_IMPLEMENTATION_ + +using std::make_pair; +using std::pair; + +namespace testing { +namespace internal { + +class Base { + public: + // Copy constructor and assignment operator do exactly what we need, so we + // use them. + Base() : member_(0) {} + explicit Base(int n) : member_(n) {} + virtual ~Base() {} + int member() { return member_; } + + private: + int member_; +}; + +class Derived : public Base { + public: + explicit Derived(int n) : Base(n) {} +}; + +TEST(ImplicitCastTest, ConvertsPointers) { + Derived derived(0); + EXPECT_TRUE(&derived == ::testing::internal::ImplicitCast_<Base*>(&derived)); +} + +TEST(ImplicitCastTest, CanUseInheritance) { + Derived derived(1); + Base base = ::testing::internal::ImplicitCast_<Base>(derived); + EXPECT_EQ(derived.member(), base.member()); +} + +class Castable { + public: + Castable(bool* converted) : converted_(converted) {} + operator Base() { + *converted_ = true; + return Base(); + } + + private: + bool* converted_; +}; + +TEST(ImplicitCastTest, CanUseNonConstCastOperator) { + bool converted = false; + Castable castable(&converted); + Base base = ::testing::internal::ImplicitCast_<Base>(castable); + EXPECT_TRUE(converted); +} + +class ConstCastable { + public: + ConstCastable(bool* converted) : converted_(converted) {} + operator Base() const { + *converted_ = true; + return Base(); + } + + private: + bool* converted_; +}; + +TEST(ImplicitCastTest, CanUseConstCastOperatorOnConstValues) { + bool converted = false; + const ConstCastable const_castable(&converted); + Base base = ::testing::internal::ImplicitCast_<Base>(const_castable); + EXPECT_TRUE(converted); +} + +class ConstAndNonConstCastable { + public: + ConstAndNonConstCastable(bool* converted, bool* const_converted) + : converted_(converted), const_converted_(const_converted) {} + operator Base() { + *converted_ = true; + return Base(); + } + operator Base() const { + *const_converted_ = true; + return Base(); + } + + private: + bool* converted_; + bool* const_converted_; +}; + +TEST(ImplicitCastTest, CanSelectBetweenConstAndNonConstCasrAppropriately) { + bool converted = false; + bool const_converted = false; + ConstAndNonConstCastable castable(&converted, &const_converted); + Base base = ::testing::internal::ImplicitCast_<Base>(castable); + EXPECT_TRUE(converted); + EXPECT_FALSE(const_converted); + + converted = false; + const_converted = false; + const ConstAndNonConstCastable const_castable(&converted, &const_converted); + base = ::testing::internal::ImplicitCast_<Base>(const_castable); + EXPECT_FALSE(converted); + EXPECT_TRUE(const_converted); +} + +class To { + public: + To(bool* converted) { *converted = true; } // NOLINT +}; + +TEST(ImplicitCastTest, CanUseImplicitConstructor) { + bool converted = false; + To to = ::testing::internal::ImplicitCast_<To>(&converted); + (void)to; + EXPECT_TRUE(converted); +} + +TEST(IteratorTraitsTest, WorksForSTLContainerIterators) { + StaticAssertTypeEq<int, + IteratorTraits< ::std::vector<int>::const_iterator>::value_type>(); + StaticAssertTypeEq<bool, + IteratorTraits< ::std::list<bool>::iterator>::value_type>(); +} + +TEST(IteratorTraitsTest, WorksForPointerToNonConst) { + StaticAssertTypeEq<char, IteratorTraits<char*>::value_type>(); + StaticAssertTypeEq<const void*, IteratorTraits<const void**>::value_type>(); +} + +TEST(IteratorTraitsTest, WorksForPointerToConst) { + StaticAssertTypeEq<char, IteratorTraits<const char*>::value_type>(); + StaticAssertTypeEq<const void*, + IteratorTraits<const void* const*>::value_type>(); +} + +// Tests that the element_type typedef is available in scoped_ptr and refers +// to the parameter type. +TEST(ScopedPtrTest, DefinesElementType) { + StaticAssertTypeEq<int, ::testing::internal::scoped_ptr<int>::element_type>(); +} + +// TODO(vladl@google.com): Implement THE REST of scoped_ptr tests. + +TEST(GtestCheckSyntaxTest, BehavesLikeASingleStatement) { + if (AlwaysFalse()) + GTEST_CHECK_(false) << "This should never be executed; " + "It's a compilation test only."; + + if (AlwaysTrue()) + GTEST_CHECK_(true); + else + ; // NOLINT + + if (AlwaysFalse()) + ; // NOLINT + else + GTEST_CHECK_(true) << ""; +} + +TEST(GtestCheckSyntaxTest, WorksWithSwitch) { + switch (0) { + case 1: + break; + default: + GTEST_CHECK_(true); + } + + switch(0) + case 0: + GTEST_CHECK_(true) << "Check failed in switch case"; +} + +// Verifies behavior of FormatFileLocation. +TEST(FormatFileLocationTest, FormatsFileLocation) { + EXPECT_PRED_FORMAT2(IsSubstring, "foo.cc", FormatFileLocation("foo.cc", 42)); + EXPECT_PRED_FORMAT2(IsSubstring, "42", FormatFileLocation("foo.cc", 42)); +} + +TEST(FormatFileLocationTest, FormatsUnknownFile) { + EXPECT_PRED_FORMAT2( + IsSubstring, "unknown file", FormatFileLocation(NULL, 42)); + EXPECT_PRED_FORMAT2(IsSubstring, "42", FormatFileLocation(NULL, 42)); +} + +TEST(FormatFileLocationTest, FormatsUknownLine) { + EXPECT_EQ("foo.cc:", FormatFileLocation("foo.cc", -1)); +} + +TEST(FormatFileLocationTest, FormatsUknownFileAndLine) { + EXPECT_EQ("unknown file:", FormatFileLocation(NULL, -1)); +} + +// Verifies behavior of FormatCompilerIndependentFileLocation. +TEST(FormatCompilerIndependentFileLocationTest, FormatsFileLocation) { + EXPECT_EQ("foo.cc:42", FormatCompilerIndependentFileLocation("foo.cc", 42)); +} + +TEST(FormatCompilerIndependentFileLocationTest, FormatsUknownFile) { + EXPECT_EQ("unknown file:42", + FormatCompilerIndependentFileLocation(NULL, 42)); +} + +TEST(FormatCompilerIndependentFileLocationTest, FormatsUknownLine) { + EXPECT_EQ("foo.cc", FormatCompilerIndependentFileLocation("foo.cc", -1)); +} + +TEST(FormatCompilerIndependentFileLocationTest, FormatsUknownFileAndLine) { + EXPECT_EQ("unknown file", FormatCompilerIndependentFileLocation(NULL, -1)); +} + +#if GTEST_OS_MAC +void* ThreadFunc(void* data) { + pthread_mutex_t* mutex = static_cast<pthread_mutex_t*>(data); + pthread_mutex_lock(mutex); + pthread_mutex_unlock(mutex); + return NULL; +} + +TEST(GetThreadCountTest, ReturnsCorrectValue) { + EXPECT_EQ(1U, GetThreadCount()); + pthread_mutex_t mutex; + pthread_attr_t attr; + pthread_t thread_id; + + // TODO(vladl@google.com): turn mutex into internal::Mutex for automatic + // destruction. + pthread_mutex_init(&mutex, NULL); + pthread_mutex_lock(&mutex); + ASSERT_EQ(0, pthread_attr_init(&attr)); + ASSERT_EQ(0, pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_JOINABLE)); + + const int status = pthread_create(&thread_id, &attr, &ThreadFunc, &mutex); + ASSERT_EQ(0, pthread_attr_destroy(&attr)); + ASSERT_EQ(0, status); + EXPECT_EQ(2U, GetThreadCount()); + pthread_mutex_unlock(&mutex); + + void* dummy; + ASSERT_EQ(0, pthread_join(thread_id, &dummy)); + + // MacOS X may not immediately report the updated thread count after + // joining a thread, causing flakiness in this test. To counter that, we + // wait for up to .5 seconds for the OS to report the correct value. + for (int i = 0; i < 5; ++i) { + if (GetThreadCount() == 1) + break; + + SleepMilliseconds(100); + } + EXPECT_EQ(1U, GetThreadCount()); + pthread_mutex_destroy(&mutex); +} +#else +TEST(GetThreadCountTest, ReturnsZeroWhenUnableToCountThreads) { + EXPECT_EQ(0U, GetThreadCount()); +} +#endif // GTEST_OS_MAC + +TEST(GtestCheckDeathTest, DiesWithCorrectOutputOnFailure) { + const bool a_false_condition = false; + const char regex[] = +#ifdef _MSC_VER + "gtest-port_test\\.cc\\(\\d+\\):" +#elif GTEST_USES_POSIX_RE + "gtest-port_test\\.cc:[0-9]+" +#else + "gtest-port_test\\.cc:\\d+" +#endif // _MSC_VER + ".*a_false_condition.*Extra info.*"; + + EXPECT_DEATH_IF_SUPPORTED(GTEST_CHECK_(a_false_condition) << "Extra info", + regex); +} + +#if GTEST_HAS_DEATH_TEST + +TEST(GtestCheckDeathTest, LivesSilentlyOnSuccess) { + EXPECT_EXIT({ + GTEST_CHECK_(true) << "Extra info"; + ::std::cerr << "Success\n"; + exit(0); }, + ::testing::ExitedWithCode(0), "Success"); +} + +#endif // GTEST_HAS_DEATH_TEST + +// Verifies that Google Test choose regular expression engine appropriate to +// the platform. The test will produce compiler errors in case of failure. +// For simplicity, we only cover the most important platforms here. +TEST(RegexEngineSelectionTest, SelectsCorrectRegexEngine) { +#if GTEST_HAS_POSIX_RE + + EXPECT_TRUE(GTEST_USES_POSIX_RE); + +#else + + EXPECT_TRUE(GTEST_USES_SIMPLE_RE); + +#endif +} + +#if GTEST_USES_POSIX_RE + +# if GTEST_HAS_TYPED_TEST + +template <typename Str> +class RETest : public ::testing::Test {}; + +// Defines StringTypes as the list of all string types that class RE +// supports. +typedef testing::Types< + ::std::string, +# if GTEST_HAS_GLOBAL_STRING + ::string, +# endif // GTEST_HAS_GLOBAL_STRING + const char*> StringTypes; + +TYPED_TEST_CASE(RETest, StringTypes); + +// Tests RE's implicit constructors. +TYPED_TEST(RETest, ImplicitConstructorWorks) { + const RE empty(TypeParam("")); + EXPECT_STREQ("", empty.pattern()); + + const RE simple(TypeParam("hello")); + EXPECT_STREQ("hello", simple.pattern()); + + const RE normal(TypeParam(".*(\\w+)")); + EXPECT_STREQ(".*(\\w+)", normal.pattern()); +} + +// Tests that RE's constructors reject invalid regular expressions. +TYPED_TEST(RETest, RejectsInvalidRegex) { + EXPECT_NONFATAL_FAILURE({ + const RE invalid(TypeParam("?")); + }, "\"?\" is not a valid POSIX Extended regular expression."); +} + +// Tests RE::FullMatch(). +TYPED_TEST(RETest, FullMatchWorks) { + const RE empty(TypeParam("")); + EXPECT_TRUE(RE::FullMatch(TypeParam(""), empty)); + EXPECT_FALSE(RE::FullMatch(TypeParam("a"), empty)); + + const RE re(TypeParam("a.*z")); + EXPECT_TRUE(RE::FullMatch(TypeParam("az"), re)); + EXPECT_TRUE(RE::FullMatch(TypeParam("axyz"), re)); + EXPECT_FALSE(RE::FullMatch(TypeParam("baz"), re)); + EXPECT_FALSE(RE::FullMatch(TypeParam("azy"), re)); +} + +// Tests RE::PartialMatch(). +TYPED_TEST(RETest, PartialMatchWorks) { + const RE empty(TypeParam("")); + EXPECT_TRUE(RE::PartialMatch(TypeParam(""), empty)); + EXPECT_TRUE(RE::PartialMatch(TypeParam("a"), empty)); + + const RE re(TypeParam("a.*z")); + EXPECT_TRUE(RE::PartialMatch(TypeParam("az"), re)); + EXPECT_TRUE(RE::PartialMatch(TypeParam("axyz"), re)); + EXPECT_TRUE(RE::PartialMatch(TypeParam("baz"), re)); + EXPECT_TRUE(RE::PartialMatch(TypeParam("azy"), re)); + EXPECT_FALSE(RE::PartialMatch(TypeParam("zza"), re)); +} + +# endif // GTEST_HAS_TYPED_TEST + +#elif GTEST_USES_SIMPLE_RE + +TEST(IsInSetTest, NulCharIsNotInAnySet) { + EXPECT_FALSE(IsInSet('\0', "")); + EXPECT_FALSE(IsInSet('\0', "\0")); + EXPECT_FALSE(IsInSet('\0', "a")); +} + +TEST(IsInSetTest, WorksForNonNulChars) { + EXPECT_FALSE(IsInSet('a', "Ab")); + EXPECT_FALSE(IsInSet('c', "")); + + EXPECT_TRUE(IsInSet('b', "bcd")); + EXPECT_TRUE(IsInSet('b', "ab")); +} + +TEST(IsAsciiDigitTest, IsFalseForNonDigit) { + EXPECT_FALSE(IsAsciiDigit('\0')); + EXPECT_FALSE(IsAsciiDigit(' ')); + EXPECT_FALSE(IsAsciiDigit('+')); + EXPECT_FALSE(IsAsciiDigit('-')); + EXPECT_FALSE(IsAsciiDigit('.')); + EXPECT_FALSE(IsAsciiDigit('a')); +} + +TEST(IsAsciiDigitTest, IsTrueForDigit) { + EXPECT_TRUE(IsAsciiDigit('0')); + EXPECT_TRUE(IsAsciiDigit('1')); + EXPECT_TRUE(IsAsciiDigit('5')); + EXPECT_TRUE(IsAsciiDigit('9')); +} + +TEST(IsAsciiPunctTest, IsFalseForNonPunct) { + EXPECT_FALSE(IsAsciiPunct('\0')); + EXPECT_FALSE(IsAsciiPunct(' ')); + EXPECT_FALSE(IsAsciiPunct('\n')); + EXPECT_FALSE(IsAsciiPunct('a')); + EXPECT_FALSE(IsAsciiPunct('0')); +} + +TEST(IsAsciiPunctTest, IsTrueForPunct) { + for (const char* p = "^-!\"#$%&'()*+,./:;<=>?@[\\]_`{|}~"; *p; p++) { + EXPECT_PRED1(IsAsciiPunct, *p); + } +} + +TEST(IsRepeatTest, IsFalseForNonRepeatChar) { + EXPECT_FALSE(IsRepeat('\0')); + EXPECT_FALSE(IsRepeat(' ')); + EXPECT_FALSE(IsRepeat('a')); + EXPECT_FALSE(IsRepeat('1')); + EXPECT_FALSE(IsRepeat('-')); +} + +TEST(IsRepeatTest, IsTrueForRepeatChar) { + EXPECT_TRUE(IsRepeat('?')); + EXPECT_TRUE(IsRepeat('*')); + EXPECT_TRUE(IsRepeat('+')); +} + +TEST(IsAsciiWhiteSpaceTest, IsFalseForNonWhiteSpace) { + EXPECT_FALSE(IsAsciiWhiteSpace('\0')); + EXPECT_FALSE(IsAsciiWhiteSpace('a')); + EXPECT_FALSE(IsAsciiWhiteSpace('1')); + EXPECT_FALSE(IsAsciiWhiteSpace('+')); + EXPECT_FALSE(IsAsciiWhiteSpace('_')); +} + +TEST(IsAsciiWhiteSpaceTest, IsTrueForWhiteSpace) { + EXPECT_TRUE(IsAsciiWhiteSpace(' ')); + EXPECT_TRUE(IsAsciiWhiteSpace('\n')); + EXPECT_TRUE(IsAsciiWhiteSpace('\r')); + EXPECT_TRUE(IsAsciiWhiteSpace('\t')); + EXPECT_TRUE(IsAsciiWhiteSpace('\v')); + EXPECT_TRUE(IsAsciiWhiteSpace('\f')); +} + +TEST(IsAsciiWordCharTest, IsFalseForNonWordChar) { + EXPECT_FALSE(IsAsciiWordChar('\0')); + EXPECT_FALSE(IsAsciiWordChar('+')); + EXPECT_FALSE(IsAsciiWordChar('.')); + EXPECT_FALSE(IsAsciiWordChar(' ')); + EXPECT_FALSE(IsAsciiWordChar('\n')); +} + +TEST(IsAsciiWordCharTest, IsTrueForLetter) { + EXPECT_TRUE(IsAsciiWordChar('a')); + EXPECT_TRUE(IsAsciiWordChar('b')); + EXPECT_TRUE(IsAsciiWordChar('A')); + EXPECT_TRUE(IsAsciiWordChar('Z')); +} + +TEST(IsAsciiWordCharTest, IsTrueForDigit) { + EXPECT_TRUE(IsAsciiWordChar('0')); + EXPECT_TRUE(IsAsciiWordChar('1')); + EXPECT_TRUE(IsAsciiWordChar('7')); + EXPECT_TRUE(IsAsciiWordChar('9')); +} + +TEST(IsAsciiWordCharTest, IsTrueForUnderscore) { + EXPECT_TRUE(IsAsciiWordChar('_')); +} + +TEST(IsValidEscapeTest, IsFalseForNonPrintable) { + EXPECT_FALSE(IsValidEscape('\0')); + EXPECT_FALSE(IsValidEscape('\007')); +} + +TEST(IsValidEscapeTest, IsFalseForDigit) { + EXPECT_FALSE(IsValidEscape('0')); + EXPECT_FALSE(IsValidEscape('9')); +} + +TEST(IsValidEscapeTest, IsFalseForWhiteSpace) { + EXPECT_FALSE(IsValidEscape(' ')); + EXPECT_FALSE(IsValidEscape('\n')); +} + +TEST(IsValidEscapeTest, IsFalseForSomeLetter) { + EXPECT_FALSE(IsValidEscape('a')); + EXPECT_FALSE(IsValidEscape('Z')); +} + +TEST(IsValidEscapeTest, IsTrueForPunct) { + EXPECT_TRUE(IsValidEscape('.')); + EXPECT_TRUE(IsValidEscape('-')); + EXPECT_TRUE(IsValidEscape('^')); + EXPECT_TRUE(IsValidEscape('$')); + EXPECT_TRUE(IsValidEscape('(')); + EXPECT_TRUE(IsValidEscape(']')); + EXPECT_TRUE(IsValidEscape('{')); + EXPECT_TRUE(IsValidEscape('|')); +} + +TEST(IsValidEscapeTest, IsTrueForSomeLetter) { + EXPECT_TRUE(IsValidEscape('d')); + EXPECT_TRUE(IsValidEscape('D')); + EXPECT_TRUE(IsValidEscape('s')); + EXPECT_TRUE(IsValidEscape('S')); + EXPECT_TRUE(IsValidEscape('w')); + EXPECT_TRUE(IsValidEscape('W')); +} + +TEST(AtomMatchesCharTest, EscapedPunct) { + EXPECT_FALSE(AtomMatchesChar(true, '\\', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, '\\', ' ')); + EXPECT_FALSE(AtomMatchesChar(true, '_', '.')); + EXPECT_FALSE(AtomMatchesChar(true, '.', 'a')); + + EXPECT_TRUE(AtomMatchesChar(true, '\\', '\\')); + EXPECT_TRUE(AtomMatchesChar(true, '_', '_')); + EXPECT_TRUE(AtomMatchesChar(true, '+', '+')); + EXPECT_TRUE(AtomMatchesChar(true, '.', '.')); +} + +TEST(AtomMatchesCharTest, Escaped_d) { + EXPECT_FALSE(AtomMatchesChar(true, 'd', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 'd', 'a')); + EXPECT_FALSE(AtomMatchesChar(true, 'd', '.')); + + EXPECT_TRUE(AtomMatchesChar(true, 'd', '0')); + EXPECT_TRUE(AtomMatchesChar(true, 'd', '9')); +} + +TEST(AtomMatchesCharTest, Escaped_D) { + EXPECT_FALSE(AtomMatchesChar(true, 'D', '0')); + EXPECT_FALSE(AtomMatchesChar(true, 'D', '9')); + + EXPECT_TRUE(AtomMatchesChar(true, 'D', '\0')); + EXPECT_TRUE(AtomMatchesChar(true, 'D', 'a')); + EXPECT_TRUE(AtomMatchesChar(true, 'D', '-')); +} + +TEST(AtomMatchesCharTest, Escaped_s) { + EXPECT_FALSE(AtomMatchesChar(true, 's', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 's', 'a')); + EXPECT_FALSE(AtomMatchesChar(true, 's', '.')); + EXPECT_FALSE(AtomMatchesChar(true, 's', '9')); + + EXPECT_TRUE(AtomMatchesChar(true, 's', ' ')); + EXPECT_TRUE(AtomMatchesChar(true, 's', '\n')); + EXPECT_TRUE(AtomMatchesChar(true, 's', '\t')); +} + +TEST(AtomMatchesCharTest, Escaped_S) { + EXPECT_FALSE(AtomMatchesChar(true, 'S', ' ')); + EXPECT_FALSE(AtomMatchesChar(true, 'S', '\r')); + + EXPECT_TRUE(AtomMatchesChar(true, 'S', '\0')); + EXPECT_TRUE(AtomMatchesChar(true, 'S', 'a')); + EXPECT_TRUE(AtomMatchesChar(true, 'S', '9')); +} + +TEST(AtomMatchesCharTest, Escaped_w) { + EXPECT_FALSE(AtomMatchesChar(true, 'w', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 'w', '+')); + EXPECT_FALSE(AtomMatchesChar(true, 'w', ' ')); + EXPECT_FALSE(AtomMatchesChar(true, 'w', '\n')); + + EXPECT_TRUE(AtomMatchesChar(true, 'w', '0')); + EXPECT_TRUE(AtomMatchesChar(true, 'w', 'b')); + EXPECT_TRUE(AtomMatchesChar(true, 'w', 'C')); + EXPECT_TRUE(AtomMatchesChar(true, 'w', '_')); +} + +TEST(AtomMatchesCharTest, Escaped_W) { + EXPECT_FALSE(AtomMatchesChar(true, 'W', 'A')); + EXPECT_FALSE(AtomMatchesChar(true, 'W', 'b')); + EXPECT_FALSE(AtomMatchesChar(true, 'W', '9')); + EXPECT_FALSE(AtomMatchesChar(true, 'W', '_')); + + EXPECT_TRUE(AtomMatchesChar(true, 'W', '\0')); + EXPECT_TRUE(AtomMatchesChar(true, 'W', '*')); + EXPECT_TRUE(AtomMatchesChar(true, 'W', '\n')); +} + +TEST(AtomMatchesCharTest, EscapedWhiteSpace) { + EXPECT_FALSE(AtomMatchesChar(true, 'f', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 'f', '\n')); + EXPECT_FALSE(AtomMatchesChar(true, 'n', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 'n', '\r')); + EXPECT_FALSE(AtomMatchesChar(true, 'r', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 'r', 'a')); + EXPECT_FALSE(AtomMatchesChar(true, 't', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 't', 't')); + EXPECT_FALSE(AtomMatchesChar(true, 'v', '\0')); + EXPECT_FALSE(AtomMatchesChar(true, 'v', '\f')); + + EXPECT_TRUE(AtomMatchesChar(true, 'f', '\f')); + EXPECT_TRUE(AtomMatchesChar(true, 'n', '\n')); + EXPECT_TRUE(AtomMatchesChar(true, 'r', '\r')); + EXPECT_TRUE(AtomMatchesChar(true, 't', '\t')); + EXPECT_TRUE(AtomMatchesChar(true, 'v', '\v')); +} + +TEST(AtomMatchesCharTest, UnescapedDot) { + EXPECT_FALSE(AtomMatchesChar(false, '.', '\n')); + + EXPECT_TRUE(AtomMatchesChar(false, '.', '\0')); + EXPECT_TRUE(AtomMatchesChar(false, '.', '.')); + EXPECT_TRUE(AtomMatchesChar(false, '.', 'a')); + EXPECT_TRUE(AtomMatchesChar(false, '.', ' ')); +} + +TEST(AtomMatchesCharTest, UnescapedChar) { + EXPECT_FALSE(AtomMatchesChar(false, 'a', '\0')); + EXPECT_FALSE(AtomMatchesChar(false, 'a', 'b')); + EXPECT_FALSE(AtomMatchesChar(false, '$', 'a')); + + EXPECT_TRUE(AtomMatchesChar(false, '$', '$')); + EXPECT_TRUE(AtomMatchesChar(false, '5', '5')); + EXPECT_TRUE(AtomMatchesChar(false, 'Z', 'Z')); +} + +TEST(ValidateRegexTest, GeneratesFailureAndReturnsFalseForInvalid) { + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex(NULL)), + "NULL is not a valid simple regular expression"); + EXPECT_NONFATAL_FAILURE( + ASSERT_FALSE(ValidateRegex("a\\")), + "Syntax error at index 1 in simple regular expression \"a\\\": "); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("a\\")), + "'\\' cannot appear at the end"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("\\n\\")), + "'\\' cannot appear at the end"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("\\s\\hb")), + "invalid escape sequence \"\\h\""); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("^^")), + "'^' can only appear at the beginning"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex(".*^b")), + "'^' can only appear at the beginning"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("$$")), + "'$' can only appear at the end"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("^$a")), + "'$' can only appear at the end"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("a(b")), + "'(' is unsupported"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("ab)")), + "')' is unsupported"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("[ab")), + "'[' is unsupported"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("a{2")), + "'{' is unsupported"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("?")), + "'?' can only follow a repeatable token"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("^*")), + "'*' can only follow a repeatable token"); + EXPECT_NONFATAL_FAILURE(ASSERT_FALSE(ValidateRegex("5*+")), + "'+' can only follow a repeatable token"); +} + +TEST(ValidateRegexTest, ReturnsTrueForValid) { + EXPECT_TRUE(ValidateRegex("")); + EXPECT_TRUE(ValidateRegex("a")); + EXPECT_TRUE(ValidateRegex(".*")); + EXPECT_TRUE(ValidateRegex("^a_+")); + EXPECT_TRUE(ValidateRegex("^a\\t\\&?")); + EXPECT_TRUE(ValidateRegex("09*$")); + EXPECT_TRUE(ValidateRegex("^Z$")); + EXPECT_TRUE(ValidateRegex("a\\^Z\\$\\(\\)\\|\\[\\]\\{\\}")); +} + +TEST(MatchRepetitionAndRegexAtHeadTest, WorksForZeroOrOne) { + EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "a", "ba")); + // Repeating more than once. + EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "b", "aab")); + + // Repeating zero times. + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "b", "ba")); + // Repeating once. + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, 'a', '?', "b", "ab")); + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '#', '?', ".", "##")); +} + +TEST(MatchRepetitionAndRegexAtHeadTest, WorksForZeroOrMany) { + EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, '.', '*', "a$", "baab")); + + // Repeating zero times. + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '.', '*', "b", "bc")); + // Repeating once. + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '.', '*', "b", "abc")); + // Repeating more than once. + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(true, 'w', '*', "-", "ab_1-g")); +} + +TEST(MatchRepetitionAndRegexAtHeadTest, WorksForOneOrMany) { + EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, '.', '+', "a$", "baab")); + // Repeating zero times. + EXPECT_FALSE(MatchRepetitionAndRegexAtHead(false, '.', '+', "b", "bc")); + + // Repeating once. + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(false, '.', '+', "b", "abc")); + // Repeating more than once. + EXPECT_TRUE(MatchRepetitionAndRegexAtHead(true, 'w', '+', "-", "ab_1-g")); +} + +TEST(MatchRegexAtHeadTest, ReturnsTrueForEmptyRegex) { + EXPECT_TRUE(MatchRegexAtHead("", "")); + EXPECT_TRUE(MatchRegexAtHead("", "ab")); +} + +TEST(MatchRegexAtHeadTest, WorksWhenDollarIsInRegex) { + EXPECT_FALSE(MatchRegexAtHead("$", "a")); + + EXPECT_TRUE(MatchRegexAtHead("$", "")); + EXPECT_TRUE(MatchRegexAtHead("a$", "a")); +} + +TEST(MatchRegexAtHeadTest, WorksWhenRegexStartsWithEscapeSequence) { + EXPECT_FALSE(MatchRegexAtHead("\\w", "+")); + EXPECT_FALSE(MatchRegexAtHead("\\W", "ab")); + + EXPECT_TRUE(MatchRegexAtHead("\\sa", "\nab")); + EXPECT_TRUE(MatchRegexAtHead("\\d", "1a")); +} + +TEST(MatchRegexAtHeadTest, WorksWhenRegexStartsWithRepetition) { + EXPECT_FALSE(MatchRegexAtHead(".+a", "abc")); + EXPECT_FALSE(MatchRegexAtHead("a?b", "aab")); + + EXPECT_TRUE(MatchRegexAtHead(".*a", "bc12-ab")); + EXPECT_TRUE(MatchRegexAtHead("a?b", "b")); + EXPECT_TRUE(MatchRegexAtHead("a?b", "ab")); +} + +TEST(MatchRegexAtHeadTest, + WorksWhenRegexStartsWithRepetionOfEscapeSequence) { + EXPECT_FALSE(MatchRegexAtHead("\\.+a", "abc")); + EXPECT_FALSE(MatchRegexAtHead("\\s?b", " b")); + + EXPECT_TRUE(MatchRegexAtHead("\\(*a", "((((ab")); + EXPECT_TRUE(MatchRegexAtHead("\\^?b", "^b")); + EXPECT_TRUE(MatchRegexAtHead("\\\\?b", "b")); + EXPECT_TRUE(MatchRegexAtHead("\\\\?b", "\\b")); +} + +TEST(MatchRegexAtHeadTest, MatchesSequentially) { + EXPECT_FALSE(MatchRegexAtHead("ab.*c", "acabc")); + + EXPECT_TRUE(MatchRegexAtHead("ab.*c", "ab-fsc")); +} + +TEST(MatchRegexAnywhereTest, ReturnsFalseWhenStringIsNull) { + EXPECT_FALSE(MatchRegexAnywhere("", NULL)); +} + +TEST(MatchRegexAnywhereTest, WorksWhenRegexStartsWithCaret) { + EXPECT_FALSE(MatchRegexAnywhere("^a", "ba")); + EXPECT_FALSE(MatchRegexAnywhere("^$", "a")); + + EXPECT_TRUE(MatchRegexAnywhere("^a", "ab")); + EXPECT_TRUE(MatchRegexAnywhere("^", "ab")); + EXPECT_TRUE(MatchRegexAnywhere("^$", "")); +} + +TEST(MatchRegexAnywhereTest, ReturnsFalseWhenNoMatch) { + EXPECT_FALSE(MatchRegexAnywhere("a", "bcde123")); + EXPECT_FALSE(MatchRegexAnywhere("a.+a", "--aa88888888")); +} + +TEST(MatchRegexAnywhereTest, ReturnsTrueWhenMatchingPrefix) { + EXPECT_TRUE(MatchRegexAnywhere("\\w+", "ab1_ - 5")); + EXPECT_TRUE(MatchRegexAnywhere(".*=", "=")); + EXPECT_TRUE(MatchRegexAnywhere("x.*ab?.*bc", "xaaabc")); +} + +TEST(MatchRegexAnywhereTest, ReturnsTrueWhenMatchingNonPrefix) { + EXPECT_TRUE(MatchRegexAnywhere("\\w+", "$$$ ab1_ - 5")); + EXPECT_TRUE(MatchRegexAnywhere("\\.+=", "= ...=")); +} + +// Tests RE's implicit constructors. +TEST(RETest, ImplicitConstructorWorks) { + const RE empty(""); + EXPECT_STREQ("", empty.pattern()); + + const RE simple("hello"); + EXPECT_STREQ("hello", simple.pattern()); +} + +// Tests that RE's constructors reject invalid regular expressions. +TEST(RETest, RejectsInvalidRegex) { + EXPECT_NONFATAL_FAILURE({ + const RE normal(NULL); + }, "NULL is not a valid simple regular expression"); + + EXPECT_NONFATAL_FAILURE({ + const RE normal(".*(\\w+"); + }, "'(' is unsupported"); + + EXPECT_NONFATAL_FAILURE({ + const RE invalid("^?"); + }, "'?' can only follow a repeatable token"); +} + +// Tests RE::FullMatch(). +TEST(RETest, FullMatchWorks) { + const RE empty(""); + EXPECT_TRUE(RE::FullMatch("", empty)); + EXPECT_FALSE(RE::FullMatch("a", empty)); + + const RE re1("a"); + EXPECT_TRUE(RE::FullMatch("a", re1)); + + const RE re("a.*z"); + EXPECT_TRUE(RE::FullMatch("az", re)); + EXPECT_TRUE(RE::FullMatch("axyz", re)); + EXPECT_FALSE(RE::FullMatch("baz", re)); + EXPECT_FALSE(RE::FullMatch("azy", re)); +} + +// Tests RE::PartialMatch(). +TEST(RETest, PartialMatchWorks) { + const RE empty(""); + EXPECT_TRUE(RE::PartialMatch("", empty)); + EXPECT_TRUE(RE::PartialMatch("a", empty)); + + const RE re("a.*z"); + EXPECT_TRUE(RE::PartialMatch("az", re)); + EXPECT_TRUE(RE::PartialMatch("axyz", re)); + EXPECT_TRUE(RE::PartialMatch("baz", re)); + EXPECT_TRUE(RE::PartialMatch("azy", re)); + EXPECT_FALSE(RE::PartialMatch("zza", re)); +} + +#endif // GTEST_USES_POSIX_RE + +#if !GTEST_OS_WINDOWS_MOBILE + +TEST(CaptureTest, CapturesStdout) { + CaptureStdout(); + fprintf(stdout, "abc"); + EXPECT_STREQ("abc", GetCapturedStdout().c_str()); + + CaptureStdout(); + fprintf(stdout, "def%cghi", '\0'); + EXPECT_EQ(::std::string("def\0ghi", 7), ::std::string(GetCapturedStdout())); +} + +TEST(CaptureTest, CapturesStderr) { + CaptureStderr(); + fprintf(stderr, "jkl"); + EXPECT_STREQ("jkl", GetCapturedStderr().c_str()); + + CaptureStderr(); + fprintf(stderr, "jkl%cmno", '\0'); + EXPECT_EQ(::std::string("jkl\0mno", 7), ::std::string(GetCapturedStderr())); +} + +// Tests that stdout and stderr capture don't interfere with each other. +TEST(CaptureTest, CapturesStdoutAndStderr) { + CaptureStdout(); + CaptureStderr(); + fprintf(stdout, "pqr"); + fprintf(stderr, "stu"); + EXPECT_STREQ("pqr", GetCapturedStdout().c_str()); + EXPECT_STREQ("stu", GetCapturedStderr().c_str()); +} + +TEST(CaptureDeathTest, CannotReenterStdoutCapture) { + CaptureStdout(); + EXPECT_DEATH_IF_SUPPORTED(CaptureStdout();, + "Only one stdout capturer can exist at a time"); + GetCapturedStdout(); + + // We cannot test stderr capturing using death tests as they use it + // themselves. +} + +#endif // !GTEST_OS_WINDOWS_MOBILE + +TEST(ThreadLocalTest, DefaultConstructorInitializesToDefaultValues) { + ThreadLocal<int> t1; + EXPECT_EQ(0, t1.get()); + + ThreadLocal<void*> t2; + EXPECT_TRUE(t2.get() == NULL); +} + +TEST(ThreadLocalTest, SingleParamConstructorInitializesToParam) { + ThreadLocal<int> t1(123); + EXPECT_EQ(123, t1.get()); + + int i = 0; + ThreadLocal<int*> t2(&i); + EXPECT_EQ(&i, t2.get()); +} + +class NoDefaultContructor { + public: + explicit NoDefaultContructor(const char*) {} + NoDefaultContructor(const NoDefaultContructor&) {} +}; + +TEST(ThreadLocalTest, ValueDefaultContructorIsNotRequiredForParamVersion) { + ThreadLocal<NoDefaultContructor> bar(NoDefaultContructor("foo")); + bar.pointer(); +} + +TEST(ThreadLocalTest, GetAndPointerReturnSameValue) { + ThreadLocal<String> thread_local; + + EXPECT_EQ(thread_local.pointer(), &(thread_local.get())); + + // Verifies the condition still holds after calling set. + thread_local.set("foo"); + EXPECT_EQ(thread_local.pointer(), &(thread_local.get())); +} + +TEST(ThreadLocalTest, PointerAndConstPointerReturnSameValue) { + ThreadLocal<String> thread_local; + const ThreadLocal<String>& const_thread_local = thread_local; + + EXPECT_EQ(thread_local.pointer(), const_thread_local.pointer()); + + thread_local.set("foo"); + EXPECT_EQ(thread_local.pointer(), const_thread_local.pointer()); +} + +#if GTEST_IS_THREADSAFE + +void AddTwo(int* param) { *param += 2; } + +TEST(ThreadWithParamTest, ConstructorExecutesThreadFunc) { + int i = 40; + ThreadWithParam<int*> thread(&AddTwo, &i, NULL); + thread.Join(); + EXPECT_EQ(42, i); +} + +TEST(MutexDeathTest, AssertHeldShouldAssertWhenNotLocked) { + // AssertHeld() is flaky only in the presence of multiple threads accessing + // the lock. In this case, the test is robust. + EXPECT_DEATH_IF_SUPPORTED({ + Mutex m; + { MutexLock lock(&m); } + m.AssertHeld(); + }, + "thread .*hold"); +} + +TEST(MutexTest, AssertHeldShouldNotAssertWhenLocked) { + Mutex m; + MutexLock lock(&m); + m.AssertHeld(); +} + +class AtomicCounterWithMutex { + public: + explicit AtomicCounterWithMutex(Mutex* mutex) : + value_(0), mutex_(mutex), random_(42) {} + + void Increment() { + MutexLock lock(mutex_); + int temp = value_; + { + // Locking a mutex puts up a memory barrier, preventing reads and + // writes to value_ rearranged when observed from other threads. + // + // We cannot use Mutex and MutexLock here or rely on their memory + // barrier functionality as we are testing them here. + pthread_mutex_t memory_barrier_mutex; + GTEST_CHECK_POSIX_SUCCESS_( + pthread_mutex_init(&memory_barrier_mutex, NULL)); + GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_lock(&memory_barrier_mutex)); + + SleepMilliseconds(random_.Generate(30)); + + GTEST_CHECK_POSIX_SUCCESS_(pthread_mutex_unlock(&memory_barrier_mutex)); + } + value_ = temp + 1; + } + int value() const { return value_; } + + private: + volatile int value_; + Mutex* const mutex_; // Protects value_. + Random random_; +}; + +void CountingThreadFunc(pair<AtomicCounterWithMutex*, int> param) { + for (int i = 0; i < param.second; ++i) + param.first->Increment(); +} + +// Tests that the mutex only lets one thread at a time to lock it. +TEST(MutexTest, OnlyOneThreadCanLockAtATime) { + Mutex mutex; + AtomicCounterWithMutex locked_counter(&mutex); + + typedef ThreadWithParam<pair<AtomicCounterWithMutex*, int> > ThreadType; + const int kCycleCount = 20; + const int kThreadCount = 7; + scoped_ptr<ThreadType> counting_threads[kThreadCount]; + Notification threads_can_start; + // Creates and runs kThreadCount threads that increment locked_counter + // kCycleCount times each. + for (int i = 0; i < kThreadCount; ++i) { + counting_threads[i].reset(new ThreadType(&CountingThreadFunc, + make_pair(&locked_counter, + kCycleCount), + &threads_can_start)); + } + threads_can_start.Notify(); + for (int i = 0; i < kThreadCount; ++i) + counting_threads[i]->Join(); + + // If the mutex lets more than one thread to increment the counter at a + // time, they are likely to encounter a race condition and have some + // increments overwritten, resulting in the lower then expected counter + // value. + EXPECT_EQ(kCycleCount * kThreadCount, locked_counter.value()); +} + +template <typename T> +void RunFromThread(void (func)(T), T param) { + ThreadWithParam<T> thread(func, param, NULL); + thread.Join(); +} + +void RetrieveThreadLocalValue(pair<ThreadLocal<String>*, String*> param) { + *param.second = param.first->get(); +} + +TEST(ThreadLocalTest, ParameterizedConstructorSetsDefault) { + ThreadLocal<String> thread_local("foo"); + EXPECT_STREQ("foo", thread_local.get().c_str()); + + thread_local.set("bar"); + EXPECT_STREQ("bar", thread_local.get().c_str()); + + String result; + RunFromThread(&RetrieveThreadLocalValue, make_pair(&thread_local, &result)); + EXPECT_STREQ("foo", result.c_str()); +} + +// DestructorTracker keeps track of whether its instances have been +// destroyed. +static std::vector<bool> g_destroyed; + +class DestructorTracker { + public: + DestructorTracker() : index_(GetNewIndex()) {} + DestructorTracker(const DestructorTracker& /* rhs */) + : index_(GetNewIndex()) {} + ~DestructorTracker() { + // We never access g_destroyed concurrently, so we don't need to + // protect the write operation under a mutex. + g_destroyed[index_] = true; + } + + private: + static int GetNewIndex() { + g_destroyed.push_back(false); + return g_destroyed.size() - 1; + } + const int index_; +}; + +typedef ThreadLocal<DestructorTracker>* ThreadParam; + +void CallThreadLocalGet(ThreadParam thread_local) { + thread_local->get(); +} + +// Tests that when a ThreadLocal object dies in a thread, it destroys +// the managed object for that thread. +TEST(ThreadLocalTest, DestroysManagedObjectForOwnThreadWhenDying) { + g_destroyed.clear(); + + { + // The next line default constructs a DestructorTracker object as + // the default value of objects managed by thread_local. + ThreadLocal<DestructorTracker> thread_local; + ASSERT_EQ(1U, g_destroyed.size()); + ASSERT_FALSE(g_destroyed[0]); + + // This creates another DestructorTracker object for the main thread. + thread_local.get(); + ASSERT_EQ(2U, g_destroyed.size()); + ASSERT_FALSE(g_destroyed[0]); + ASSERT_FALSE(g_destroyed[1]); + } + + // Now thread_local has died. It should have destroyed both the + // default value shared by all threads and the value for the main + // thread. + ASSERT_EQ(2U, g_destroyed.size()); + EXPECT_TRUE(g_destroyed[0]); + EXPECT_TRUE(g_destroyed[1]); + + g_destroyed.clear(); +} + +// Tests that when a thread exits, the thread-local object for that +// thread is destroyed. +TEST(ThreadLocalTest, DestroysManagedObjectAtThreadExit) { + g_destroyed.clear(); + + { + // The next line default constructs a DestructorTracker object as + // the default value of objects managed by thread_local. + ThreadLocal<DestructorTracker> thread_local; + ASSERT_EQ(1U, g_destroyed.size()); + ASSERT_FALSE(g_destroyed[0]); + + // This creates another DestructorTracker object in the new thread. + ThreadWithParam<ThreadParam> thread( + &CallThreadLocalGet, &thread_local, NULL); + thread.Join(); + + // Now the new thread has exited. The per-thread object for it + // should have been destroyed. + ASSERT_EQ(2U, g_destroyed.size()); + ASSERT_FALSE(g_destroyed[0]); + ASSERT_TRUE(g_destroyed[1]); + } + + // Now thread_local has died. The default value should have been + // destroyed too. + ASSERT_EQ(2U, g_destroyed.size()); + EXPECT_TRUE(g_destroyed[0]); + EXPECT_TRUE(g_destroyed[1]); + + g_destroyed.clear(); +} + +TEST(ThreadLocalTest, ThreadLocalMutationsAffectOnlyCurrentThread) { + ThreadLocal<String> thread_local; + thread_local.set("Foo"); + EXPECT_STREQ("Foo", thread_local.get().c_str()); + + String result; + RunFromThread(&RetrieveThreadLocalValue, make_pair(&thread_local, &result)); + EXPECT_TRUE(result.c_str() == NULL); +} + +#endif // GTEST_IS_THREADSAFE + +} // namespace internal +} // namespace testing -- cgit v1.2.3