- Generated by
1.8.11
|
proxygen
|
If you cannot find the answer to your question here, and you have read Primer and AdvancedGuide, send it to googletestframework@googlegroups.com.
First, let us say clearly that we don't want to get into the debate of which C++ testing framework is the best. There exist many fine frameworks for writing C++ tests, and we have tremendous respect for the developers and users of them. We don't think there is (or will be) a single best framework - you have to pick the right tool for the particular task you are tackling.
We created Google Test because we couldn't find the right combination of features and conveniences in an existing framework to satisfy our needs. The following is a list of things that we like about Google Test. We don't claim them to be unique to Google Test - rather, the combination of them makes Google Test the choice for us. We hope this list can help you decide whether it is for you too.
EXPECT_*) have proven to be great time savers, as they allow a test to report multiple failures in a single edit-compile-test cycle.ASSERT_EQ(5, Foo(i)) << " where i = " << i;. It doesn't require a new set of macros or special functions.SCOPED_TRACE helps you understand the context of an assertion failure when it comes from inside a sub-routine or loop.We strive to minimize compiler warnings Google Test generates. Before releasing a new version, we test to make sure that it doesn't generate warnings when compiled using its CMake script on Windows, Linux, and Mac OS.
Unfortunately, this doesn't mean you are guaranteed to see no warnings when compiling Google Test in your environment:
It is not always possible to make Google Test warning-free for everyone. Or, it may not be desirable if the warning is rarely enabled and fixing the violations makes the code more complex.
If you see warnings when compiling Google Test, we suggest that you use the -isystem flag (assuming your are using GCC) to mark Google Test headers as system headers. That'll suppress warnings from Google Test headers.
Underscore (_) is special, as C++ reserves the following to be used by the compiler and the standard library:
_ followed by an upper-case letter, and__) anywhere in its name.User code is prohibited from using such identifiers.
Now let's look at what this means for TEST and TEST_F.
Currently TEST(TestCaseName, TestName) generates a class named TestCaseName_TestName_Test. What happens if TestCaseName or TestName contains _?
TestCaseName starts with an _ followed by an upper-case letter (say, _Foo), we end up with _Foo_TestName_Test, which is reserved and thus invalid.TestCaseName ends with an _ (say, Foo_), we get Foo__TestName_Test, which is invalid.TestName starts with an _ (say, _Bar), we get TestCaseName__Bar_Test, which is invalid.TestName ends with an _ (say, Bar_), we get TestCaseName_Bar__Test, which is invalid.So clearly TestCaseName and TestName cannot start or end with _ (Actually, TestCaseName can start with _ – as long as the _ isn't followed by an upper-case letter. But that's getting complicated. So for simplicity we just say that it cannot start with _.).
It may seem fine for TestCaseName and TestName to contain _ in the middle. However, consider this:
Now, the two TESTs will both generate the same class (Time_Files_Like_An_Arrow_Test). That's not good.
So for simplicity, we just ask the users to avoid _ in TestCaseName and TestName. The rule is more constraining than necessary, but it's simple and easy to remember. It also gives Google Test some wiggle room in case its implementation needs to change in the future.
If you violate the rule, there may not be immediately consequences, but your test may (just may) break with a new compiler (or a new version of the compiler you are using) or with a new version of Google Test. Therefore it's best to follow the rule.
In the early days, we said that you could install compiled Google Test libraries on *nix systems using make install. Then every user of your machine can write tests without recompiling Google Test.
This seemed like a good idea, but it has a got-cha: every user needs to compile his tests using the same compiler flags used to compile the installed Google Test libraries; otherwise he may run into undefined behaviors (i.e. the tests can behave strangely and may even crash for no obvious reasons).
Why? Because C++ has this thing called the One-Definition Rule: if two C++ source files contain different definitions of the same class/function/variable, and you link them together, you violate the rule. The linker may or may not catch the error (in many cases it's not required by the C++ standard to catch the violation). If it doesn't, you get strange run-time behaviors that are unexpected and hard to debug.
If you compile Google Test and your test code using different compiler flags, they may see different definitions of the same class/function/variable (e.g. due to the use of if in Google Test). Therefore, for your sanity, we recommend to avoid installing pre-compiled Google Test libraries. Instead, each project should compile Google Test itself such that it can be sure that the same flags are used for both Google Test and the tests.
(Answered by Trevor Robinson)
Load the supplied Visual Studio solution file, either msvc\gtest-md.sln or msvc\gtest.sln. Go through the migration wizard to migrate the solution and project files to Visual Studio 2008. Select Configuration Manager... from the Build menu. Select <New...> from the Active solution platform dropdown. Select x64 from the new platform dropdown, leave Copy settings from set to Win32 and Create new project platforms checked, then click OK. You now have Win32 and x64 platform configurations, selectable from the Standard toolbar, which allow you to toggle between building 32-bit or 64-bit binaries (or both at once using Batch Build).
In order to prevent build output files from overwriting one another, you'll need to change the Intermediate Directory settings for the newly created platform configuration across all the projects. To do this, multi-select (e.g. using shift-click) all projects (but not the solution) in the Solution Explorer. Right-click one of them and select Properties. In the left pane, select Configuration Properties, and from the Configuration dropdown, select All Configurations. Make sure the selected platform is x64. For the Intermediate Directory setting, change the value from \ to \. Click OK and then build the solution. When the build is complete, the 64-bit binaries will be in the msvc\x64\Debug directory.
We haven't tested this ourselves, but Per Abrahamsen reported that he was able to compile and install Google Test successfully when using MinGW from Cygwin. You'll need to configure it with:
PATH/TO/configure CC="gcc -mno-cygwin" CXX="g++ -mno-cygwin"
You should be able to replace the -mno-cygwin option with direct links to the real MinGW binaries, but we haven't tried that.
Caveats:
make check will produce some errors as not all tests for Google Test itself are compatible with MinGW.We also have reports on successful cross compilation of Google Test MinGW binaries on Linux using these instructions on the WxWidgets site.
Please contact googletestframework@googlegroups.com if you are interested in improving the support for MinGW.
Due to some peculiarity of C++, it requires some non-trivial template meta programming tricks to support using NULL as an argument of the EXPECT_XX() and ASSERT_XX() macros. Therefore we only do it where it's most needed (otherwise we make the implementation of Google Test harder to maintain and more error-prone than necessary).
The EXPECT_EQ() macro takes the expected value as its first argument and the actual value as the second. It's reasonable that someone wants to write EXPECT_EQ(NULL, some_expression), and this indeed was requested several times. Therefore we implemented it.
The need for EXPECT_NE(NULL, ptr) isn't nearly as strong. When the assertion fails, you already know that ptr must be NULL, so it doesn't add any information to print ptr in this case. That means EXPECT_TRUE(ptr != NULL) works just as well.
If we were to support EXPECT_NE(NULL, ptr), for consistency we'll have to support EXPECT_NE(ptr, NULL) as well, as unlike EXPECT_EQ, we don't have a convention on the order of the two arguments for EXPECT_NE. This means using the template meta programming tricks twice in the implementation, making it even harder to understand and maintain. We believe the benefit doesn't justify the cost.
Finally, with the growth of Google Mock's matcher library, we are encouraging people to use the unified EXPECT_THAT(value, matcher) syntax more often in tests. One significant advantage of the matcher approach is that matchers can be easily combined to form new matchers, while the EXPECT_NE, etc, macros cannot be easily combined. Therefore we want to invest more in the matchers than in the EXPECT_XX() macros.
Test runners tend to be tightly coupled with the build/test environment, and Google Test doesn't try to solve the problem of running tests in parallel. Instead, we tried to make Google Test work nicely with test runners. For example, Google Test's XML report contains the time spent on each test, and its gtest_list_tests and gtest_filter flags can be used for splitting the execution of test methods into multiple processes. These functionalities can help the test runner run the tests in parallel.
It's difficult to write thread-safe code. Most tests are not written with thread-safety in mind, and thus may not work correctly in a multi-threaded setting.
If you think about it, it's already hard to make your code work when you know what other threads are doing. It's much harder, and sometimes even impossible, to make your code work when you don't know what other threads are doing (remember that test methods can be added, deleted, or modified after your test was written). If you want to run the tests in parallel, you'd better run them in different processes.
Our original motivation was to be able to use Google Test in projects that disable exceptions. Later we realized some additional benefits of this approach:
EXPECT_* family of macros will continue even after a failure, allowing multiple failures in a TEST to be reported in a single run. This is a popular feature, as in C++ the edit-compile-test cycle is usually quite long and being able to fixing more than one thing at a time is a blessing.ASSERT_TRUE throws. While it's unlikely for someone to write this in a test, it's possible to run into this pattern when you write assertions in callbacks that are called by the code under test.The downside of not using exceptions is that ASSERT_* (implemented using return) will only abort the current function, not the current TEST.
Unfortunately, C++'s macro system doesn't allow us to use the same macro for both cases. One possibility is to provide only one macro for tests with fixtures, and require the user to define an empty fixture sometimes:
or
Yet, many people think this is one line too many. :-) Our goal was to make it really easy to write tests, so we tried to make simple tests trivial to create. That means using a separate macro for such tests.
We think neither approach is ideal, yet either of them is reasonable. In the end, it probably doesn't matter much either way.
We like to use structs only when representing passive data. This distinction between structs and classes is good for documenting the intent of the code's author. Since test fixtures have logic like SetUp() and TearDown(), they are better defined as classes.
Our goal was to make death tests as convenient for a user as C++ possibly allows. In particular:
ASSERT_DEATH(statement, expected_message) specifies all necessary information in one place, in one language, without boilerplate code. It is very declarative.ASSERT_DEATH has a similar syntax and error-reporting semantics as other Google Test assertions, and thus is easy to learn.ASSERT_DEATH can be mixed with other assertions and other logic at your will. You are not limited to one death test per test method. For example, you can write something like: If you prefer one death test per test method, you can write your tests in that style too, but we don't want to impose that on the users. The fewer artificial limitations the better.ASSERT_DEATH can reference local variables in the current function, and you can decide how many death tests you want based on run-time information. For example, The runner-based approach tends to be more static and less flexible, or requires more user effort to get this kind of flexibility.Another interesting thing about ASSERT_DEATH is that it calls fork() to create a child process to run the death test. This is lightening fast, as fork() uses copy-on-write pages and incurs almost zero overhead, and the child process starts from the user-supplied statement directly, skipping all global and local initialization and any code leading to the given statement. If you launch the child process from scratch, it can take seconds just to load everything and start running if the test links to many libraries dynamically.
Death tests (EXPECT_DEATH, etc) are executed in a sub-process s.t. the expected crash won't kill the test program (i.e. the parent process). As a result, any in-memory side effects they incur are observable in their respective sub-processes, but not in the parent process. You can think of them as running in a parallel universe, more or less.
If your class has a static data member:
You also need to define it outside of the class body in foo.cc:
Otherwise your code is **invalid C++**, and may break in unexpected ways. In particular, using it in Google Test comparison assertions (EXPECT_EQ, etc) will generate an "undefined reference" linker error.
Google Test doesn't yet have good support for this kind of tests, or data-driven tests in general. We hope to be able to make improvements in this area soon.
Yes.
Each test fixture has a corresponding and same named test case. This means only one test case can use a particular fixture. Sometimes, however, multiple test cases may want to use the same or slightly different fixtures. For example, you may want to make sure that all of a GUI library's test cases don't leak important system resources like fonts and brushes.
In Google Test, you share a fixture among test cases by putting the shared logic in a base test fixture, then deriving from that base a separate fixture for each test case that wants to use this common logic. You then use TEST_F() to write tests using each derived fixture.
Typically, your code looks like this:
If necessary, you can continue to derive test fixtures from a derived fixture. Google Test has no limit on how deep the hierarchy can be.
For a complete example using derived test fixtures, see sample5.
You're probably using an ASSERT_*() in a function that doesn't return void. ASSERT_*() can only be used in void functions.
In Google Test, death tests are run in a child process and the way they work is delicate. To write death tests you really need to understand how they work. Please make sure you have read this.
In particular, death tests don't like having multiple threads in the parent process. So the first thing you can try is to eliminate creating threads outside of EXPECT_DEATH().
Sometimes this is impossible as some library you must use may be creating threads before main() is even reached. In this case, you can try to minimize the chance of conflicts by either moving as many activities as possible inside EXPECT_DEATH() (in the extreme case, you want to move everything inside), or leaving as few things as possible in it. Also, you can try to set the death test style to "threadsafe", which is safer but slower, and see if it helps.
If you go with thread-safe death tests, remember that they rerun the test program from the beginning in the child process. Therefore make sure your program can run side-by-side with itself and is deterministic.
In the end, this boils down to good concurrent programming. You have to make sure that there is no race conditions or dead locks in your program. No silver bullet - sorry!
The first thing to remember is that Google Test does not reuse the same test fixture object across multiple tests. For each TEST_F, Google Test will create a fresh test fixture object, immediately call SetUp(), run the test body, call TearDown(), and then immediately delete the test fixture object.
When you need to write per-test set-up and tear-down logic, you have the choice between using the test fixture constructor/destructor or SetUp()/TearDown(). The former is usually preferred, as it has the following benefits:
const, which helps prevent accidental changes to its value and makes the tests more obviously correct.SetUp()/TearDown(), a subclass may make the mistake of forgetting to call the base class' SetUp()/TearDown() or call them at the wrong moment.You may still want to use SetUp()/TearDown() in the following rare cases:
TearDown() as opposed to the destructor, as throwing in a destructor leads to undefined behavior and usually will kill your program right away. Note that many standard libraries (like STL) may throw when exceptions are enabled in the compiler. Therefore you should prefer TearDown() if you want to write portable tests that work with or without exceptions.--gtest_throw_on_failure is specified. Therefore, you shouldn't use Google Test assertions in a destructor if you plan to run your tests with this flag.SetUp()/TearDown().If the predicate function you use in ASSERT_PRED* or EXPECT_PRED* is overloaded or a template, the compiler will have trouble figuring out which overloaded version it should use. ASSERT_PRED_FORMAT* and EXPECT_PRED_FORMAT* don't have this problem.
If you see this error, you might want to switch to (ASSERT|EXPECT)_PRED_FORMAT*, which will also give you a better failure message. If, however, that is not an option, you can resolve the problem by explicitly telling the compiler which version to pick.
For example, suppose you have
you will get a compiler error if you write
However, this will work:
(The stuff inside the angled brackets for the static_cast operator is the type of the function pointer for the int-version of IsPositive().)
As another example, when you have a template function
you can use it in a predicate assertion like this:
Things are more interesting if your template has more than one parameters. The following won't compile:
as the C++ pre-processor thinks you are giving ASSERT_PRED2 4 arguments, which is one more than expected. The workaround is to wrap the predicate function in parentheses:
Some people had been ignoring the return value of RUN_ALL_TESTS(). That is, instead of
they write
This is wrong and dangerous. A test runner needs to see the return value of RUN_ALL_TESTS() in order to determine if a test has passed. If your main() function ignores it, your test will be considered successful even if it has a Google Test assertion failure. Very bad.
To help the users avoid this dangerous bug, the implementation of RUN_ALL_TESTS() causes gcc to raise this warning, when the return value is ignored. If you see this warning, the fix is simple: just make sure its value is used as the return value of main().
Due to a peculiarity of C++, in order to support the syntax for streaming messages to an ASSERT_*, e.g.
we had to give up using ASSERT* and FAIL* (but not EXPECT* and ADD_FAILURE*) in constructors and destructors. The workaround is to move the content of your constructor/destructor to a private void member function, or switch to EXPECT_*() if that works. This section in the user's guide explains it.
C++ is case-sensitive. It should be spelled as SetUp(). Did you spell it as Setup()?
Similarly, sometimes people spell SetUpTestCase() as SetupTestCase() and wonder why it's never called.
Google Test's failure message format is understood by Emacs and many other IDEs, like acme and XCode. If a Google Test message is in a compilation buffer in Emacs, then it's clickable. You can now hit enter on a message to jump to the corresponding source code, or use C-x` to jump to the next failure.
You don't have to. Instead of
you can simply typedef the test fixtures:
The Google Test output is meant to be a concise and human-friendly report. If your test generates textual output itself, it will mix with the Google Test output, making it hard to read. However, there is an easy solution to this problem.
Since most log messages go to stderr, we decided to let Google Test output go to stdout. This way, you can easily separate the two using redirection. For example:
There are several good reasons:
You should try to write testable code, which means classes should be easily tested from their public interface. One way to achieve this is the Pimpl idiom: you move all private members of a class into a helper class, and make all members of the helper class public.
You have several other options that don't require using FRIEND_TEST:
We find private static methods clutter the header file. They are implementation details and ideally should be kept out of a .h. So often I make them free functions instead.
Instead of:
You probably should better write:
No. You can use a feature called value-parameterized tests which lets you repeat your tests with different parameters, without defining it more than once.
To test a foo.cc file, you need to compile and link it into your unit test program. However, when the file contains a definition for the main() function, it will clash with the main() of your unit test, and will result in a build error.
The right solution is to split it into three files:
foo.h which contains the declarations,foo.cc which contains the definitions except main(), andfoo_main.cc which contains nothing but the definition of main().Then foo.cc can be easily tested.
If you are adding tests to an existing file and don't want an intrusive change like this, there is a hack: just include the entire foo.cc file in your unit test. For example:
However, please remember this is a hack and should only be used as the last resort.
ASSERT_DEATH(_statement_, _regex_) (or any death assertion macro) can be used wherever _statement_ is valid. So basically _statement_ can be any C++ statement that makes sense in the current context. In particular, it can reference global and/or local variables, and can be:
Some examples are shown here:
googletest_unittest.cc contains more examples if you are interested.
On POSIX systems, Google Test uses the POSIX Extended regular expression syntax (http://en.wikipedia.org/wiki/Regular_expression#POSIX_Extended_Regular_Expressions). On Windows, it uses a limited variant of regular expression syntax. For more details, see the regular expression syntax.
Google Test needs to be able to create objects of your test fixture class, so it must have a default constructor. Normally the compiler will define one for you. However, there are cases where you have to define your own:
Foo, then you need to define a default constructor, even if it would be empty.Foo has a const non-static data member, then you have to define the default constructor and initialize the const member in the initializer list of the constructor. (Early versions of gcc doesn't force you to initialize the const member. It's a bug that has been fixed in gcc 4.)With the Linux pthread library, there is no turning back once you cross the line from single thread to multiple threads. The first time you create a thread, a manager thread is created in addition, so you get 3, not 2, threads. Later when the thread you create joins the main thread, the thread count decrements by 1, but the manager thread will never be killed, so you still have 2 threads, which means you cannot safely run a death test.
The new NPTL thread library doesn't suffer from this problem, as it doesn't create a manager thread. However, if you don't control which machine your test runs on, you shouldn't depend on this.
Google Test does not interleave tests from different test cases. That is, it runs all tests in one test case first, and then runs all tests in the next test case, and so on. Google Test does this because it needs to set up a test case before the first test in it is run, and tear it down afterwords. Splitting up the test case would require multiple set-up and tear-down processes, which is inefficient and makes the semantics unclean.
If we were to determine the order of tests based on test name instead of test case name, then we would have a problem with the following situation:
Since FooTest.AbcDeathTest needs to run before BarTest.Xyz, and we don't interleave tests from different test cases, we need to run all tests in the FooTest case before running any test in the BarTest case. This contradicts with the requirement to run BarTest.DefDeathTest before FooTest.Uvw.
You don't have to, but if you like, you may split up the test case into FooTest and FooDeathTest, where the names make it clear that they are related:
If you use a user-defined type FooType in an assertion, you must make sure there is an std::ostream& operator<<(std::ostream&, const FooType&) function defined such that we can print a value of FooType.
In addition, if FooType is declared in a name space, the << operator also needs to be defined in the same name space.
Since the statically initialized Google Test singleton requires allocations on the heap, the Visual C++ memory leak detector will report memory leaks at the end of the program run. The easiest way to avoid this is to use the _CrtMemCheckpoint and _CrtMemDumpAllObjectsSince calls to not report any statically initialized heap objects. See MSDN for more details and additional heap check/debug routines.
You may get a number of the following linker error or warnings if you attempt to link your test project with the Google Test library when your project and the are not built using the same compiler settings.
The Google Test project (gtest.vcproj) has the Runtime Library option set to /MT (use multi-threaded static libraries, /MTd for debug). If your project uses something else, for example /MD (use multi-threaded DLLs, /MDd for debug), you need to change the setting in the Google Test project to match your project's.
To update this setting open the project properties in the Visual Studio IDE then select the branch Configuration Properties | C/C++ | Code Generation and change the option "Runtime Library". You may also try using gtest-md.vcproj instead of gtest.vcproj.
Have you read a warning on the Google Test Primer page?
Many people are in your position and one of the posted his solution to our mailing list.
Google Test uses parts of the standard C++ library that SunStudio does not support. Our users reported success using alternative implementations. Try running the build after runing this commad:
`export CC=cc CXX=CC CXXFLAGS='-library=stlport4'`
If you write code that sniffs whether it's running in a test and does different things accordingly, you are leaking test-only logic into production code and there is no easy way to ensure that the test-only code paths aren't run by mistake in production. Such cleverness also leads to Heisenbugs. Therefore we strongly advise against the practice, and Google Test doesn't provide a way to do it.
In general, the recommended way to cause the code to behave differently under test is dependency injection. You can inject different functionality from the test and from the production code. Since your production code doesn't link in the for-test logic at all, there is no danger in accidentally running it.
However, if you really, really, really have no choice, and if you follow the rule of ending your test program names with _test, you can use the horrible hack of sniffing your executable name (argv[0] in main()) to know whether the code is under test.
In C++, macros don't obey namespaces. Therefore two libraries that both define a macro of the same name will clash if you #include both definitions. In case a Google Test macro clashes with another library, you can force Google Test to rename its macro to avoid the conflict.
Specifically, if both Google Test and some other code define macro FOO, you can add
to the compiler flags to tell Google Test to change the macro's name from FOO to GTEST_FOO. For example, with -DGTEST_DONT_DEFINE_TEST=1, you'll need to write
instead of
in order to define a test.
Currently, the following TEST, FAIL, SUCCEED, and the basic comparison assertion macros can have alternative names. You can see the full list of covered macros here. More information can be found in the "Avoiding Macro Name Clashes" section of the README file.
TEST(Foo, Bar) test methods defined in different namespaces?Yes.
The rule is all test methods in the same test case must use the same fixture class. This means that the following is allowed because both tests use the same fixture class (testing::Test).
However, the following code is not allowed and will produce a runtime error from Google Test because the test methods are using different test fixture classes with the same test case name.
If you try to build Google Test's Xcode project with Xcode 4.0 or later, you may encounter an error message that looks like "Missing SDK in target gtest\_framework: /Developer/SDKs/MacOSX10.4u.sdk". That means that Xcode does not support the SDK the project is targeting. See the Xcode section in the ../README.md "README" file on how to resolve this.
If you cannot find the answer to your question in this FAQ, there are some other resources you can use:
Please note that creating an issue in the issue tracker is not a good way to get your answer, as it is monitored infrequently by a very small number of people.
When asking a question, it's helpful to provide as much of the following information as possible (people cannot help you if there's not enough information in your question):
1.8.11