1# Advanced googletest Topics
2
3## Introduction
4
5Now that you have read the [googletest Primer](primer.md) and learned how to
6write tests using googletest, it's time to learn some new tricks. This document
7will show you more assertions as well as how to construct complex failure
8messages, propagate fatal failures, reuse and speed up your test fixtures, and
9use various flags with your tests.
10
11## More Assertions
12
13This section covers some less frequently used, but still significant,
14assertions.
15
16### Explicit Success and Failure
17
18These three assertions do not actually test a value or expression. Instead, they
19generate a success or failure directly. Like the macros that actually perform a
20test, you may stream a custom failure message into them.
21
22```c++
23SUCCEED();
24```
25
26Generates a success. This does **NOT** make the overall test succeed. A test is
27considered successful only if none of its assertions fail during its execution.
28
29{: .callout .note}
30NOTE: `SUCCEED()` is purely documentary and currently doesn't generate any
31user-visible output. However, we may add `SUCCEED()` messages to googletest's
32output in the future.
33
34```c++
35FAIL();
36ADD_FAILURE();
37ADD_FAILURE_AT("file_path", line_number);
38```
39
40`FAIL()` generates a fatal failure, while `ADD_FAILURE()` and `ADD_FAILURE_AT()`
41generate a nonfatal failure. These are useful when control flow, rather than a
42Boolean expression, determines the test's success or failure. For example, you
43might want to write something like:
44
45```c++
46switch(expression) {
47  case 1:
48     ... some checks ...
49  case 2:
50     ... some other checks ...
51  default:
52     FAIL() << "We shouldn't get here.";
53}
54```
55
56{: .callout .note}
57NOTE: you can only use `FAIL()` in functions that return `void`. See the
58[Assertion Placement section](#assertion-placement) for more information.
59
60### Exception Assertions
61
62These are for verifying that a piece of code throws (or does not throw) an
63exception of the given type:
64
65Fatal assertion                            | Nonfatal assertion                         | Verifies
66------------------------------------------ | ------------------------------------------ | --------
67`ASSERT_THROW(statement, exception_type);` | `EXPECT_THROW(statement, exception_type);` | `statement` throws an exception of the given type
68`ASSERT_ANY_THROW(statement);`             | `EXPECT_ANY_THROW(statement);`             | `statement` throws an exception of any type
69`ASSERT_NO_THROW(statement);`              | `EXPECT_NO_THROW(statement);`              | `statement` doesn't throw any exception
70
71Examples:
72
73```c++
74ASSERT_THROW(Foo(5), bar_exception);
75
76EXPECT_NO_THROW({
77  int n = 5;
78  Bar(&n);
79});
80```
81
82**Availability**: requires exceptions to be enabled in the build environment
83
84### Predicate Assertions for Better Error Messages
85
86Even though googletest has a rich set of assertions, they can never be complete,
87as it's impossible (nor a good idea) to anticipate all scenarios a user might
88run into. Therefore, sometimes a user has to use `EXPECT_TRUE()` to check a
89complex expression, for lack of a better macro. This has the problem of not
90showing you the values of the parts of the expression, making it hard to
91understand what went wrong. As a workaround, some users choose to construct the
92failure message by themselves, streaming it into `EXPECT_TRUE()`. However, this
93is awkward especially when the expression has side-effects or is expensive to
94evaluate.
95
96googletest gives you three different options to solve this problem:
97
98#### Using an Existing Boolean Function
99
100If you already have a function or functor that returns `bool` (or a type that
101can be implicitly converted to `bool`), you can use it in a *predicate
102assertion* to get the function arguments printed for free:
103
104
105| Fatal assertion                   | Nonfatal assertion                | Verifies                    |
106| --------------------------------- | --------------------------------- | --------------------------- |
107| `ASSERT_PRED1(pred1, val1)`       | `EXPECT_PRED1(pred1, val1)`       | `pred1(val1)` is true       |
108| `ASSERT_PRED2(pred2, val1, val2)` | `EXPECT_PRED2(pred2, val1, val2)` | `pred2(val1, val2)` is true |
109| `...`                             | `...`                             | `...`                       |
110
111In the above, `predn` is an `n`-ary predicate function or functor, where `val1`,
112`val2`, ..., and `valn` are its arguments. The assertion succeeds if the
113predicate returns `true` when applied to the given arguments, and fails
114otherwise. When the assertion fails, it prints the value of each argument. In
115either case, the arguments are evaluated exactly once.
116
117Here's an example. Given
118
119```c++
120// Returns true if m and n have no common divisors except 1.
121bool MutuallyPrime(int m, int n) { ... }
122
123const int a = 3;
124const int b = 4;
125const int c = 10;
126```
127
128the assertion
129
130```c++
131  EXPECT_PRED2(MutuallyPrime, a, b);
132```
133
134will succeed, while the assertion
135
136```c++
137  EXPECT_PRED2(MutuallyPrime, b, c);
138```
139
140will fail with the message
141
142```none
143MutuallyPrime(b, c) is false, where
144b is 4
145c is 10
146```
147
148{: .callout .note}
149> NOTE:
150>
151> 1.  If you see a compiler error "no matching function to call" when using
152>     `ASSERT_PRED*` or `EXPECT_PRED*`, please see
153>     [this](faq.md#the-compiler-complains-no-matching-function-to-call-when-i-use-assert_pred-how-do-i-fix-it)
154>     for how to resolve it.
155
156#### Using a Function That Returns an AssertionResult
157
158While `EXPECT_PRED*()` and friends are handy for a quick job, the syntax is not
159satisfactory: you have to use different macros for different arities, and it
160feels more like Lisp than C++. The `::testing::AssertionResult` class solves
161this problem.
162
163An `AssertionResult` object represents the result of an assertion (whether it's
164a success or a failure, and an associated message). You can create an
165`AssertionResult` using one of these factory functions:
166
167```c++
168namespace testing {
169
170// Returns an AssertionResult object to indicate that an assertion has
171// succeeded.
172AssertionResult AssertionSuccess();
173
174// Returns an AssertionResult object to indicate that an assertion has
175// failed.
176AssertionResult AssertionFailure();
177
178}
179```
180
181You can then use the `<<` operator to stream messages to the `AssertionResult`
182object.
183
184To provide more readable messages in Boolean assertions (e.g. `EXPECT_TRUE()`),
185write a predicate function that returns `AssertionResult` instead of `bool`. For
186example, if you define `IsEven()` as:
187
188```c++
189testing::AssertionResult IsEven(int n) {
190  if ((n % 2) == 0)
191    return testing::AssertionSuccess();
192  else
193    return testing::AssertionFailure() << n << " is odd";
194}
195```
196
197instead of:
198
199```c++
200bool IsEven(int n) {
201  return (n % 2) == 0;
202}
203```
204
205the failed assertion `EXPECT_TRUE(IsEven(Fib(4)))` will print:
206
207```none
208Value of: IsEven(Fib(4))
209  Actual: false (3 is odd)
210Expected: true
211```
212
213instead of a more opaque
214
215```none
216Value of: IsEven(Fib(4))
217  Actual: false
218Expected: true
219```
220
221If you want informative messages in `EXPECT_FALSE` and `ASSERT_FALSE` as well
222(one third of Boolean assertions in the Google code base are negative ones), and
223are fine with making the predicate slower in the success case, you can supply a
224success message:
225
226```c++
227testing::AssertionResult IsEven(int n) {
228  if ((n % 2) == 0)
229    return testing::AssertionSuccess() << n << " is even";
230  else
231    return testing::AssertionFailure() << n << " is odd";
232}
233```
234
235Then the statement `EXPECT_FALSE(IsEven(Fib(6)))` will print
236
237```none
238  Value of: IsEven(Fib(6))
239     Actual: true (8 is even)
240  Expected: false
241```
242
243#### Using a Predicate-Formatter
244
245If you find the default message generated by `(ASSERT|EXPECT)_PRED*` and
246`(ASSERT|EXPECT)_(TRUE|FALSE)` unsatisfactory, or some arguments to your
247predicate do not support streaming to `ostream`, you can instead use the
248following *predicate-formatter assertions* to *fully* customize how the message
249is formatted:
250
251Fatal assertion                                  | Nonfatal assertion                               | Verifies
252------------------------------------------------ | ------------------------------------------------ | --------
253`ASSERT_PRED_FORMAT1(pred_format1, val1);`       | `EXPECT_PRED_FORMAT1(pred_format1, val1);`       | `pred_format1(val1)` is successful
254`ASSERT_PRED_FORMAT2(pred_format2, val1, val2);` | `EXPECT_PRED_FORMAT2(pred_format2, val1, val2);` | `pred_format2(val1, val2)` is successful
255`...`                                            | `...`                                            | ...
256
257The difference between this and the previous group of macros is that instead of
258a predicate, `(ASSERT|EXPECT)_PRED_FORMAT*` take a *predicate-formatter*
259(`pred_formatn`), which is a function or functor with the signature:
260
261```c++
262testing::AssertionResult PredicateFormattern(const char* expr1,
263                                             const char* expr2,
264                                             ...
265                                             const char* exprn,
266                                             T1 val1,
267                                             T2 val2,
268                                             ...
269                                             Tn valn);
270```
271
272where `val1`, `val2`, ..., and `valn` are the values of the predicate arguments,
273and `expr1`, `expr2`, ..., and `exprn` are the corresponding expressions as they
274appear in the source code. The types `T1`, `T2`, ..., and `Tn` can be either
275value types or reference types. For example, if an argument has type `Foo`, you
276can declare it as either `Foo` or `const Foo&`, whichever is appropriate.
277
278As an example, let's improve the failure message in `MutuallyPrime()`, which was
279used with `EXPECT_PRED2()`:
280
281```c++
282// Returns the smallest prime common divisor of m and n,
283// or 1 when m and n are mutually prime.
284int SmallestPrimeCommonDivisor(int m, int n) { ... }
285
286// A predicate-formatter for asserting that two integers are mutually prime.
287testing::AssertionResult AssertMutuallyPrime(const char* m_expr,
288                                             const char* n_expr,
289                                             int m,
290                                             int n) {
291  if (MutuallyPrime(m, n)) return testing::AssertionSuccess();
292
293  return testing::AssertionFailure() << m_expr << " and " << n_expr
294      << " (" << m << " and " << n << ") are not mutually prime, "
295      << "as they have a common divisor " << SmallestPrimeCommonDivisor(m, n);
296}
297```
298
299With this predicate-formatter, we can use
300
301```c++
302  EXPECT_PRED_FORMAT2(AssertMutuallyPrime, b, c);
303```
304
305to generate the message
306
307```none
308b and c (4 and 10) are not mutually prime, as they have a common divisor 2.
309```
310
311As you may have realized, many of the built-in assertions we introduced earlier
312are special cases of `(EXPECT|ASSERT)_PRED_FORMAT*`. In fact, most of them are
313indeed defined using `(EXPECT|ASSERT)_PRED_FORMAT*`.
314
315### Floating-Point Comparison
316
317Comparing floating-point numbers is tricky. Due to round-off errors, it is very
318unlikely that two floating-points will match exactly. Therefore, `ASSERT_EQ` 's
319naive comparison usually doesn't work. And since floating-points can have a wide
320value range, no single fixed error bound works. It's better to compare by a
321fixed relative error bound, except for values close to 0 due to the loss of
322precision there.
323
324In general, for floating-point comparison to make sense, the user needs to
325carefully choose the error bound. If they don't want or care to, comparing in
326terms of Units in the Last Place (ULPs) is a good default, and googletest
327provides assertions to do this. Full details about ULPs are quite long; if you
328want to learn more, see
329[here](https://randomascii.wordpress.com/2012/02/25/comparing-floating-point-numbers-2012-edition/).
330
331#### Floating-Point Macros
332
333
334| Fatal assertion                 | Nonfatal assertion              | Verifies                                 |
335| ------------------------------- | ------------------------------- | ---------------------------------------- |
336| `ASSERT_FLOAT_EQ(val1, val2);`  | `EXPECT_FLOAT_EQ(val1, val2);`  | the two `float` values are almost equal  |
337| `ASSERT_DOUBLE_EQ(val1, val2);` | `EXPECT_DOUBLE_EQ(val1, val2);` | the two `double` values are almost equal |
338
339
340By "almost equal" we mean the values are within 4 ULP's from each other.
341
342The following assertions allow you to choose the acceptable error bound:
343
344
345| Fatal assertion                       | Nonfatal assertion                    | Verifies                                                                         |
346| ------------------------------------- | ------------------------------------- | -------------------------------------------------------------------------------- |
347| `ASSERT_NEAR(val1, val2, abs_error);` | `EXPECT_NEAR(val1, val2, abs_error);` | the difference between `val1` and `val2` doesn't exceed the given absolute error |
348
349
350#### Floating-Point Predicate-Format Functions
351
352Some floating-point operations are useful, but not that often used. In order to
353avoid an explosion of new macros, we provide them as predicate-format functions
354that can be used in predicate assertion macros (e.g. `EXPECT_PRED_FORMAT2`,
355etc).
356
357```c++
358EXPECT_PRED_FORMAT2(testing::FloatLE, val1, val2);
359EXPECT_PRED_FORMAT2(testing::DoubleLE, val1, val2);
360```
361
362Verifies that `val1` is less than, or almost equal to, `val2`. You can replace
363`EXPECT_PRED_FORMAT2` in the above table with `ASSERT_PRED_FORMAT2`.
364
365### Asserting Using gMock Matchers
366
367gMock comes with a library of *matchers* for validating arguments passed to mock
368objects. A gMock matcher is basically a predicate that knows how to describe
369itself. It can be used in these assertion macros:
370
371
372| Fatal assertion                | Nonfatal assertion             | Verifies              |
373| ------------------------------ | ------------------------------ | --------------------- |
374| `ASSERT_THAT(value, matcher);` | `EXPECT_THAT(value, matcher);` | value matches matcher |
375
376
377For example, `StartsWith(prefix)` is a matcher that matches a string starting
378with `prefix`, and you can write:
379
380```c++
381using ::testing::StartsWith;
382...
383    // Verifies that Foo() returns a string starting with "Hello".
384    EXPECT_THAT(Foo(), StartsWith("Hello"));
385```
386
387See
388[Using Matchers in googletest Assertions](gmock_cook_book.md#using-matchers-in-googletest-assertions)
389in the gMock Cookbook for more details. For a list of built-in matchers, see the
390[Matchers Reference](reference/matchers.md). You can also write your own
391matchers—see [Writing New Matchers Quickly](gmock_cook_book.md#NewMatchers).
392
393gMock is bundled with googletest, so you don't need to add any build dependency
394in order to take advantage of this. Just include `"gmock/gmock.h"`
395and you're ready to go.
396
397### More String Assertions
398
399(Please read the [previous](#asserting-using-gmock-matchers) section first if
400you haven't.)
401
402You can use the gMock [string matchers](reference/matchers.md#string-matchers)
403with `EXPECT_THAT()` or `ASSERT_THAT()` to do more string comparison tricks
404(sub-string, prefix, suffix, regular expression, and etc). For example,
405
406```c++
407using ::testing::HasSubstr;
408using ::testing::MatchesRegex;
409...
410  ASSERT_THAT(foo_string, HasSubstr("needle"));
411  EXPECT_THAT(bar_string, MatchesRegex("\\w*\\d+"));
412```
413
414If the string contains a well-formed HTML or XML document, you can check whether
415its DOM tree matches an
416[XPath expression](http://www.w3.org/TR/xpath/#contents):
417
418```c++
419// Currently still in //template/prototemplate/testing:xpath_matcher
420#include "template/prototemplate/testing/xpath_matcher.h"
421using ::prototemplate::testing::MatchesXPath;
422EXPECT_THAT(html_string, MatchesXPath("//a[text()='click here']"));
423```
424
425### Windows HRESULT assertions
426
427These assertions test for `HRESULT` success or failure.
428
429Fatal assertion                        | Nonfatal assertion                     | Verifies
430-------------------------------------- | -------------------------------------- | --------
431`ASSERT_HRESULT_SUCCEEDED(expression)` | `EXPECT_HRESULT_SUCCEEDED(expression)` | `expression` is a success `HRESULT`
432`ASSERT_HRESULT_FAILED(expression)`    | `EXPECT_HRESULT_FAILED(expression)`    | `expression` is a failure `HRESULT`
433
434The generated output contains the human-readable error message associated with
435the `HRESULT` code returned by `expression`.
436
437You might use them like this:
438
439```c++
440CComPtr<IShellDispatch2> shell;
441ASSERT_HRESULT_SUCCEEDED(shell.CoCreateInstance(L"Shell.Application"));
442CComVariant empty;
443ASSERT_HRESULT_SUCCEEDED(shell->ShellExecute(CComBSTR(url), empty, empty, empty, empty));
444```
445
446### Type Assertions
447
448You can call the function
449
450```c++
451::testing::StaticAssertTypeEq<T1, T2>();
452```
453
454to assert that types `T1` and `T2` are the same. The function does nothing if
455the assertion is satisfied. If the types are different, the function call will
456fail to compile, the compiler error message will say that
457`T1 and T2 are not the same type` and most likely (depending on the compiler)
458show you the actual values of `T1` and `T2`. This is mainly useful inside
459template code.
460
461**Caveat**: When used inside a member function of a class template or a function
462template, `StaticAssertTypeEq<T1, T2>()` is effective only if the function is
463instantiated. For example, given:
464
465```c++
466template <typename T> class Foo {
467 public:
468  void Bar() { testing::StaticAssertTypeEq<int, T>(); }
469};
470```
471
472the code:
473
474```c++
475void Test1() { Foo<bool> foo; }
476```
477
478will not generate a compiler error, as `Foo<bool>::Bar()` is never actually
479instantiated. Instead, you need:
480
481```c++
482void Test2() { Foo<bool> foo; foo.Bar(); }
483```
484
485to cause a compiler error.
486
487### Assertion Placement
488
489You can use assertions in any C++ function. In particular, it doesn't have to be
490a method of the test fixture class. The one constraint is that assertions that
491generate a fatal failure (`FAIL*` and `ASSERT_*`) can only be used in
492void-returning functions. This is a consequence of Google's not using
493exceptions. By placing it in a non-void function you'll get a confusing compile
494error like `"error: void value not ignored as it ought to be"` or `"cannot
495initialize return object of type 'bool' with an rvalue of type 'void'"` or
496`"error: no viable conversion from 'void' to 'string'"`.
497
498If you need to use fatal assertions in a function that returns non-void, one
499option is to make the function return the value in an out parameter instead. For
500example, you can rewrite `T2 Foo(T1 x)` to `void Foo(T1 x, T2* result)`. You
501need to make sure that `*result` contains some sensible value even when the
502function returns prematurely. As the function now returns `void`, you can use
503any assertion inside of it.
504
505If changing the function's type is not an option, you should just use assertions
506that generate non-fatal failures, such as `ADD_FAILURE*` and `EXPECT_*`.
507
508{: .callout .note}
509NOTE: Constructors and destructors are not considered void-returning functions,
510according to the C++ language specification, and so you may not use fatal
511assertions in them; you'll get a compilation error if you try. Instead, either
512call `abort` and crash the entire test executable, or put the fatal assertion in
513a `SetUp`/`TearDown` function; see
514[constructor/destructor vs. `SetUp`/`TearDown`](faq.md#CtorVsSetUp)
515
516{: .callout .warning}
517WARNING: A fatal assertion in a helper function (private void-returning method)
518called from a constructor or destructor does not terminate the current test, as
519your intuition might suggest: it merely returns from the constructor or
520destructor early, possibly leaving your object in a partially-constructed or
521partially-destructed state! You almost certainly want to `abort` or use
522`SetUp`/`TearDown` instead.
523
524## Skipping test execution
525
526Related to the assertions `SUCCEED()` and `FAIL()`, you can prevent further test
527execution at runtime with the `GTEST_SKIP()` macro. This is useful when you need
528to check for preconditions of the system under test during runtime and skip
529tests in a meaningful way.
530
531`GTEST_SKIP()` can be used in individual test cases or in the `SetUp()` methods
532of classes derived from either `::testing::Environment` or `::testing::Test`.
533For example:
534
535```c++
536TEST(SkipTest, DoesSkip) {
537  GTEST_SKIP() << "Skipping single test";
538  EXPECT_EQ(0, 1);  // Won't fail; it won't be executed
539}
540
541class SkipFixture : public ::testing::Test {
542 protected:
543  void SetUp() override {
544    GTEST_SKIP() << "Skipping all tests for this fixture";
545  }
546};
547
548// Tests for SkipFixture won't be executed.
549TEST_F(SkipFixture, SkipsOneTest) {
550  EXPECT_EQ(5, 7);  // Won't fail
551}
552```
553
554As with assertion macros, you can stream a custom message into `GTEST_SKIP()`.
555
556## Teaching googletest How to Print Your Values
557
558When a test assertion such as `EXPECT_EQ` fails, googletest prints the argument
559values to help you debug. It does this using a user-extensible value printer.
560
561This printer knows how to print built-in C++ types, native arrays, STL
562containers, and any type that supports the `<<` operator. For other types, it
563prints the raw bytes in the value and hopes that you the user can figure it out.
564
565As mentioned earlier, the printer is *extensible*. That means you can teach it
566to do a better job at printing your particular type than to dump the bytes. To
567do that, define `<<` for your type:
568
569```c++
570#include <ostream>
571
572namespace foo {
573
574class Bar {  // We want googletest to be able to print instances of this.
575...
576  // Create a free inline friend function.
577  friend std::ostream& operator<<(std::ostream& os, const Bar& bar) {
578    return os << bar.DebugString();  // whatever needed to print bar to os
579  }
580};
581
582// If you can't declare the function in the class it's important that the
583// << operator is defined in the SAME namespace that defines Bar.  C++'s look-up
584// rules rely on that.
585std::ostream& operator<<(std::ostream& os, const Bar& bar) {
586  return os << bar.DebugString();  // whatever needed to print bar to os
587}
588
589}  // namespace foo
590```
591
592Sometimes, this might not be an option: your team may consider it bad style to
593have a `<<` operator for `Bar`, or `Bar` may already have a `<<` operator that
594doesn't do what you want (and you cannot change it). If so, you can instead
595define a `PrintTo()` function like this:
596
597```c++
598#include <ostream>
599
600namespace foo {
601
602class Bar {
603  ...
604  friend void PrintTo(const Bar& bar, std::ostream* os) {
605    *os << bar.DebugString();  // whatever needed to print bar to os
606  }
607};
608
609// If you can't declare the function in the class it's important that PrintTo()
610// is defined in the SAME namespace that defines Bar.  C++'s look-up rules rely
611// on that.
612void PrintTo(const Bar& bar, std::ostream* os) {
613  *os << bar.DebugString();  // whatever needed to print bar to os
614}
615
616}  // namespace foo
617```
618
619If you have defined both `<<` and `PrintTo()`, the latter will be used when
620googletest is concerned. This allows you to customize how the value appears in
621googletest's output without affecting code that relies on the behavior of its
622`<<` operator.
623
624If you want to print a value `x` using googletest's value printer yourself, just
625call `::testing::PrintToString(x)`, which returns an `std::string`:
626
627```c++
628vector<pair<Bar, int> > bar_ints = GetBarIntVector();
629
630EXPECT_TRUE(IsCorrectBarIntVector(bar_ints))
631    << "bar_ints = " << testing::PrintToString(bar_ints);
632```
633
634## Death Tests
635
636In many applications, there are assertions that can cause application failure if
637a condition is not met. These sanity checks, which ensure that the program is in
638a known good state, are there to fail at the earliest possible time after some
639program state is corrupted. If the assertion checks the wrong condition, then
640the program may proceed in an erroneous state, which could lead to memory
641corruption, security holes, or worse. Hence it is vitally important to test that
642such assertion statements work as expected.
643
644Since these precondition checks cause the processes to die, we call such tests
645_death tests_. More generally, any test that checks that a program terminates
646(except by throwing an exception) in an expected fashion is also a death test.
647
648Note that if a piece of code throws an exception, we don't consider it "death"
649for the purpose of death tests, as the caller of the code could catch the
650exception and avoid the crash. If you want to verify exceptions thrown by your
651code, see [Exception Assertions](#ExceptionAssertions).
652
653If you want to test `EXPECT_*()/ASSERT_*()` failures in your test code, see
654Catching Failures
655
656### How to Write a Death Test
657
658googletest has the following macros to support death tests:
659
660Fatal assertion                                  | Nonfatal assertion                               | Verifies
661------------------------------------------------ | ------------------------------------------------ | --------
662`ASSERT_DEATH(statement, matcher);`              | `EXPECT_DEATH(statement, matcher);`              | `statement` crashes with the given error
663`ASSERT_DEATH_IF_SUPPORTED(statement, matcher);` | `EXPECT_DEATH_IF_SUPPORTED(statement, matcher);` | if death tests are supported, verifies that `statement` crashes with the given error; otherwise verifies nothing
664`ASSERT_DEBUG_DEATH(statement, matcher);`        | `EXPECT_DEBUG_DEATH(statement, matcher);`        | `statement` crashes with the given error **in debug mode**. When not in debug (i.e. `NDEBUG` is defined), this just executes `statement`
665`ASSERT_EXIT(statement, predicate, matcher);`    | `EXPECT_EXIT(statement, predicate, matcher);`    | `statement` exits with the given error, and its exit code matches `predicate`
666
667where `statement` is a statement that is expected to cause the process to die,
668`predicate` is a function or function object that evaluates an integer exit
669status, and `matcher` is either a gMock matcher matching a `const std::string&`
670or a (Perl) regular expression - either of which is matched against the stderr
671output of `statement`. For legacy reasons, a bare string (i.e. with no matcher)
672is interpreted as `ContainsRegex(str)`, **not** `Eq(str)`. Note that `statement`
673can be *any valid statement* (including *compound statement*) and doesn't have
674to be an expression.
675
676As usual, the `ASSERT` variants abort the current test function, while the
677`EXPECT` variants do not.
678
679{: .callout .note}
680> NOTE: We use the word "crash" here to mean that the process terminates with a
681> *non-zero* exit status code. There are two possibilities: either the process
682> has called `exit()` or `_exit()` with a non-zero value, or it may be killed by
683> a signal.
684>
685> This means that if *`statement`* terminates the process with a 0 exit code, it
686> is *not* considered a crash by `EXPECT_DEATH`. Use `EXPECT_EXIT` instead if
687> this is the case, or if you want to restrict the exit code more precisely.
688
689A predicate here must accept an `int` and return a `bool`. The death test
690succeeds only if the predicate returns `true`. googletest defines a few
691predicates that handle the most common cases:
692
693```c++
694::testing::ExitedWithCode(exit_code)
695```
696
697This expression is `true` if the program exited normally with the given exit
698code.
699
700```c++
701testing::KilledBySignal(signal_number)  // Not available on Windows.
702```
703
704This expression is `true` if the program was killed by the given signal.
705
706The `*_DEATH` macros are convenient wrappers for `*_EXIT` that use a predicate
707that verifies the process' exit code is non-zero.
708
709Note that a death test only cares about three things:
710
7111.  does `statement` abort or exit the process?
7122.  (in the case of `ASSERT_EXIT` and `EXPECT_EXIT`) does the exit status
713    satisfy `predicate`? Or (in the case of `ASSERT_DEATH` and `EXPECT_DEATH`)
714    is the exit status non-zero? And
7153.  does the stderr output match `matcher`?
716
717In particular, if `statement` generates an `ASSERT_*` or `EXPECT_*` failure, it
718will **not** cause the death test to fail, as googletest assertions don't abort
719the process.
720
721To write a death test, simply use one of the above macros inside your test
722function. For example,
723
724```c++
725TEST(MyDeathTest, Foo) {
726  // This death test uses a compound statement.
727  ASSERT_DEATH({
728    int n = 5;
729    Foo(&n);
730  }, "Error on line .* of Foo()");
731}
732
733TEST(MyDeathTest, NormalExit) {
734  EXPECT_EXIT(NormalExit(), testing::ExitedWithCode(0), "Success");
735}
736
737TEST(MyDeathTest, KillMyself) {
738  EXPECT_EXIT(KillMyself(), testing::KilledBySignal(SIGKILL),
739              "Sending myself unblockable signal");
740}
741```
742
743verifies that:
744
745*   calling `Foo(5)` causes the process to die with the given error message,
746*   calling `NormalExit()` causes the process to print `"Success"` to stderr and
747    exit with exit code 0, and
748*   calling `KillMyself()` kills the process with signal `SIGKILL`.
749
750The test function body may contain other assertions and statements as well, if
751necessary.
752
753### Death Test Naming
754
755{: .callout .important}
756IMPORTANT: We strongly recommend you to follow the convention of naming your
757**test suite** (not test) `*DeathTest` when it contains a death test, as
758demonstrated in the above example. The
759[Death Tests And Threads](#death-tests-and-threads) section below explains why.
760
761If a test fixture class is shared by normal tests and death tests, you can use
762`using` or `typedef` to introduce an alias for the fixture class and avoid
763duplicating its code:
764
765```c++
766class FooTest : public testing::Test { ... };
767
768using FooDeathTest = FooTest;
769
770TEST_F(FooTest, DoesThis) {
771  // normal test
772}
773
774TEST_F(FooDeathTest, DoesThat) {
775  // death test
776}
777```
778
779### Regular Expression Syntax
780
781On POSIX systems (e.g. Linux, Cygwin, and Mac), googletest uses the
782[POSIX extended regular expression](http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap09.html#tag_09_04)
783syntax. To learn about this syntax, you may want to read this
784[Wikipedia entry](http://en.wikipedia.org/wiki/Regular_expression#POSIX_Extended_Regular_Expressions).
785
786On Windows, googletest uses its own simple regular expression implementation. It
787lacks many features. For example, we don't support union (`"x|y"`), grouping
788(`"(xy)"`), brackets (`"[xy]"`), and repetition count (`"x{5,7}"`), among
789others. Below is what we do support (`A` denotes a literal character, period
790(`.`), or a single `\\ ` escape sequence; `x` and `y` denote regular
791expressions.):
792
793Expression | Meaning
794---------- | --------------------------------------------------------------
795`c`        | matches any literal character `c`
796`\\d`      | matches any decimal digit
797`\\D`      | matches any character that's not a decimal digit
798`\\f`      | matches `\f`
799`\\n`      | matches `\n`
800`\\r`      | matches `\r`
801`\\s`      | matches any ASCII whitespace, including `\n`
802`\\S`      | matches any character that's not a whitespace
803`\\t`      | matches `\t`
804`\\v`      | matches `\v`
805`\\w`      | matches any letter, `_`, or decimal digit
806`\\W`      | matches any character that `\\w` doesn't match
807`\\c`      | matches any literal character `c`, which must be a punctuation
808`.`        | matches any single character except `\n`
809`A?`       | matches 0 or 1 occurrences of `A`
810`A*`       | matches 0 or many occurrences of `A`
811`A+`       | matches 1 or many occurrences of `A`
812`^`        | matches the beginning of a string (not that of each line)
813`$`        | matches the end of a string (not that of each line)
814`xy`       | matches `x` followed by `y`
815
816To help you determine which capability is available on your system, googletest
817defines macros to govern which regular expression it is using. The macros are:
818`GTEST_USES_SIMPLE_RE=1` or `GTEST_USES_POSIX_RE=1`. If you want your death
819tests to work in all cases, you can either `#if` on these macros or use the more
820limited syntax only.
821
822### How It Works
823
824Under the hood, `ASSERT_EXIT()` spawns a new process and executes the death test
825statement in that process. The details of how precisely that happens depend on
826the platform and the variable `::testing::GTEST_FLAG(death_test_style)` (which is
827initialized from the command-line flag `--gtest_death_test_style`).
828
829*   On POSIX systems, `fork()` (or `clone()` on Linux) is used to spawn the
830    child, after which:
831    *   If the variable's value is `"fast"`, the death test statement is
832        immediately executed.
833    *   If the variable's value is `"threadsafe"`, the child process re-executes
834        the unit test binary just as it was originally invoked, but with some
835        extra flags to cause just the single death test under consideration to
836        be run.
837*   On Windows, the child is spawned using the `CreateProcess()` API, and
838    re-executes the binary to cause just the single death test under
839    consideration to be run - much like the `threadsafe` mode on POSIX.
840
841Other values for the variable are illegal and will cause the death test to fail.
842Currently, the flag's default value is
843**`"fast"`**.
844
8451.  the child's exit status satisfies the predicate, and
8462.  the child's stderr matches the regular expression.
847
848If the death test statement runs to completion without dying, the child process
849will nonetheless terminate, and the assertion fails.
850
851### Death Tests And Threads
852
853The reason for the two death test styles has to do with thread safety. Due to
854well-known problems with forking in the presence of threads, death tests should
855be run in a single-threaded context. Sometimes, however, it isn't feasible to
856arrange that kind of environment. For example, statically-initialized modules
857may start threads before main is ever reached. Once threads have been created,
858it may be difficult or impossible to clean them up.
859
860googletest has three features intended to raise awareness of threading issues.
861
8621.  A warning is emitted if multiple threads are running when a death test is
863    encountered.
8642.  Test suites with a name ending in "DeathTest" are run before all other
865    tests.
8663.  It uses `clone()` instead of `fork()` to spawn the child process on Linux
867    (`clone()` is not available on Cygwin and Mac), as `fork()` is more likely
868    to cause the child to hang when the parent process has multiple threads.
869
870It's perfectly fine to create threads inside a death test statement; they are
871executed in a separate process and cannot affect the parent.
872
873### Death Test Styles
874
875The "threadsafe" death test style was introduced in order to help mitigate the
876risks of testing in a possibly multithreaded environment. It trades increased
877test execution time (potentially dramatically so) for improved thread safety.
878
879The automated testing framework does not set the style flag. You can choose a
880particular style of death tests by setting the flag programmatically:
881
882```c++
883testing::FLAGS_gtest_death_test_style="threadsafe"
884```
885
886You can do this in `main()` to set the style for all death tests in the binary,
887or in individual tests. Recall that flags are saved before running each test and
888restored afterwards, so you need not do that yourself. For example:
889
890```c++
891int main(int argc, char** argv) {
892  testing::InitGoogleTest(&argc, argv);
893  testing::FLAGS_gtest_death_test_style = "fast";
894  return RUN_ALL_TESTS();
895}
896
897TEST(MyDeathTest, TestOne) {
898  testing::FLAGS_gtest_death_test_style = "threadsafe";
899  // This test is run in the "threadsafe" style:
900  ASSERT_DEATH(ThisShouldDie(), "");
901}
902
903TEST(MyDeathTest, TestTwo) {
904  // This test is run in the "fast" style:
905  ASSERT_DEATH(ThisShouldDie(), "");
906}
907```
908
909### Caveats
910
911The `statement` argument of `ASSERT_EXIT()` can be any valid C++ statement. If
912it leaves the current function via a `return` statement or by throwing an
913exception, the death test is considered to have failed. Some googletest macros
914may return from the current function (e.g. `ASSERT_TRUE()`), so be sure to avoid
915them in `statement`.
916
917Since `statement` runs in the child process, any in-memory side effect (e.g.
918modifying a variable, releasing memory, etc) it causes will *not* be observable
919in the parent process. In particular, if you release memory in a death test,
920your program will fail the heap check as the parent process will never see the
921memory reclaimed. To solve this problem, you can
922
9231.  try not to free memory in a death test;
9242.  free the memory again in the parent process; or
9253.  do not use the heap checker in your program.
926
927Due to an implementation detail, you cannot place multiple death test assertions
928on the same line; otherwise, compilation will fail with an unobvious error
929message.
930
931Despite the improved thread safety afforded by the "threadsafe" style of death
932test, thread problems such as deadlock are still possible in the presence of
933handlers registered with `pthread_atfork(3)`.
934
935
936## Using Assertions in Sub-routines
937
938{: .callout .note}
939Note: If you want to put a series of test assertions in a subroutine to check
940for a complex condition, consider using
941[a custom GMock matcher](gmock_cook_book.md#NewMatchers)
942instead. This lets you provide a more readable error message in case of failure
943and avoid all of the issues described below.
944
945### Adding Traces to Assertions
946
947If a test sub-routine is called from several places, when an assertion inside it
948fails, it can be hard to tell which invocation of the sub-routine the failure is
949from. You can alleviate this problem using extra logging or custom failure
950messages, but that usually clutters up your tests. A better solution is to use
951the `SCOPED_TRACE` macro or the `ScopedTrace` utility:
952
953```c++
954SCOPED_TRACE(message);
955```
956```c++
957ScopedTrace trace("file_path", line_number, message);
958```
959
960where `message` can be anything streamable to `std::ostream`. `SCOPED_TRACE`
961macro will cause the current file name, line number, and the given message to be
962added in every failure message. `ScopedTrace` accepts explicit file name and
963line number in arguments, which is useful for writing test helpers. The effect
964will be undone when the control leaves the current lexical scope.
965
966For example,
967
968```c++
96910: void Sub1(int n) {
97011:   EXPECT_EQ(Bar(n), 1);
97112:   EXPECT_EQ(Bar(n + 1), 2);
97213: }
97314:
97415: TEST(FooTest, Bar) {
97516:   {
97617:     SCOPED_TRACE("A");  // This trace point will be included in
97718:                         // every failure in this scope.
97819:     Sub1(1);
97920:   }
98021:   // Now it won't.
98122:   Sub1(9);
98223: }
983```
984
985could result in messages like these:
986
987```none
988path/to/foo_test.cc:11: Failure
989Value of: Bar(n)
990Expected: 1
991  Actual: 2
992Google Test trace:
993path/to/foo_test.cc:17: A
994
995path/to/foo_test.cc:12: Failure
996Value of: Bar(n + 1)
997Expected: 2
998  Actual: 3
999```
1000
1001Without the trace, it would've been difficult to know which invocation of
1002`Sub1()` the two failures come from respectively. (You could add an extra
1003message to each assertion in `Sub1()` to indicate the value of `n`, but that's
1004tedious.)
1005
1006Some tips on using `SCOPED_TRACE`:
1007
10081.  With a suitable message, it's often enough to use `SCOPED_TRACE` at the
1009    beginning of a sub-routine, instead of at each call site.
10102.  When calling sub-routines inside a loop, make the loop iterator part of the
1011    message in `SCOPED_TRACE` such that you can know which iteration the failure
1012    is from.
10133.  Sometimes the line number of the trace point is enough for identifying the
1014    particular invocation of a sub-routine. In this case, you don't have to
1015    choose a unique message for `SCOPED_TRACE`. You can simply use `""`.
10164.  You can use `SCOPED_TRACE` in an inner scope when there is one in the outer
1017    scope. In this case, all active trace points will be included in the failure
1018    messages, in reverse order they are encountered.
10195.  The trace dump is clickable in Emacs - hit `return` on a line number and
1020    you'll be taken to that line in the source file!
1021
1022### Propagating Fatal Failures
1023
1024A common pitfall when using `ASSERT_*` and `FAIL*` is not understanding that
1025when they fail they only abort the _current function_, not the entire test. For
1026example, the following test will segfault:
1027
1028```c++
1029void Subroutine() {
1030  // Generates a fatal failure and aborts the current function.
1031  ASSERT_EQ(1, 2);
1032
1033  // The following won't be executed.
1034  ...
1035}
1036
1037TEST(FooTest, Bar) {
1038  Subroutine();  // The intended behavior is for the fatal failure
1039                 // in Subroutine() to abort the entire test.
1040
1041  // The actual behavior: the function goes on after Subroutine() returns.
1042  int* p = nullptr;
1043  *p = 3;  // Segfault!
1044}
1045```
1046
1047To alleviate this, googletest provides three different solutions. You could use
1048either exceptions, the `(ASSERT|EXPECT)_NO_FATAL_FAILURE` assertions or the
1049`HasFatalFailure()` function. They are described in the following two
1050subsections.
1051
1052#### Asserting on Subroutines with an exception
1053
1054The following code can turn ASSERT-failure into an exception:
1055
1056```c++
1057class ThrowListener : public testing::EmptyTestEventListener {
1058  void OnTestPartResult(const testing::TestPartResult& result) override {
1059    if (result.type() == testing::TestPartResult::kFatalFailure) {
1060      throw testing::AssertionException(result);
1061    }
1062  }
1063};
1064int main(int argc, char** argv) {
1065  ...
1066  testing::UnitTest::GetInstance()->listeners().Append(new ThrowListener);
1067  return RUN_ALL_TESTS();
1068}
1069```
1070
1071This listener should be added after other listeners if you have any, otherwise
1072they won't see failed `OnTestPartResult`.
1073
1074#### Asserting on Subroutines
1075
1076As shown above, if your test calls a subroutine that has an `ASSERT_*` failure
1077in it, the test will continue after the subroutine returns. This may not be what
1078you want.
1079
1080Often people want fatal failures to propagate like exceptions. For that
1081googletest offers the following macros:
1082
1083Fatal assertion                       | Nonfatal assertion                    | Verifies
1084------------------------------------- | ------------------------------------- | --------
1085`ASSERT_NO_FATAL_FAILURE(statement);` | `EXPECT_NO_FATAL_FAILURE(statement);` | `statement` doesn't generate any new fatal failures in the current thread.
1086
1087Only failures in the thread that executes the assertion are checked to determine
1088the result of this type of assertions. If `statement` creates new threads,
1089failures in these threads are ignored.
1090
1091Examples:
1092
1093```c++
1094ASSERT_NO_FATAL_FAILURE(Foo());
1095
1096int i;
1097EXPECT_NO_FATAL_FAILURE({
1098  i = Bar();
1099});
1100```
1101
1102Assertions from multiple threads are currently not supported on Windows.
1103
1104#### Checking for Failures in the Current Test
1105
1106`HasFatalFailure()` in the `::testing::Test` class returns `true` if an
1107assertion in the current test has suffered a fatal failure. This allows
1108functions to catch fatal failures in a sub-routine and return early.
1109
1110```c++
1111class Test {
1112 public:
1113  ...
1114  static bool HasFatalFailure();
1115};
1116```
1117
1118The typical usage, which basically simulates the behavior of a thrown exception,
1119is:
1120
1121```c++
1122TEST(FooTest, Bar) {
1123  Subroutine();
1124  // Aborts if Subroutine() had a fatal failure.
1125  if (HasFatalFailure()) return;
1126
1127  // The following won't be executed.
1128  ...
1129}
1130```
1131
1132If `HasFatalFailure()` is used outside of `TEST()` , `TEST_F()` , or a test
1133fixture, you must add the `::testing::Test::` prefix, as in:
1134
1135```c++
1136if (testing::Test::HasFatalFailure()) return;
1137```
1138
1139Similarly, `HasNonfatalFailure()` returns `true` if the current test has at
1140least one non-fatal failure, and `HasFailure()` returns `true` if the current
1141test has at least one failure of either kind.
1142
1143## Logging Additional Information
1144
1145In your test code, you can call `RecordProperty("key", value)` to log additional
1146information, where `value` can be either a string or an `int`. The *last* value
1147recorded for a key will be emitted to the
1148[XML output](#generating-an-xml-report) if you specify one. For example, the
1149test
1150
1151```c++
1152TEST_F(WidgetUsageTest, MinAndMaxWidgets) {
1153  RecordProperty("MaximumWidgets", ComputeMaxUsage());
1154  RecordProperty("MinimumWidgets", ComputeMinUsage());
1155}
1156```
1157
1158will output XML like this:
1159
1160```xml
1161  ...
1162    <testcase name="MinAndMaxWidgets" status="run" time="0.006" classname="WidgetUsageTest" MaximumWidgets="12" MinimumWidgets="9" />
1163  ...
1164```
1165
1166{: .callout .note}
1167> NOTE:
1168>
1169> *   `RecordProperty()` is a static member of the `Test` class. Therefore it
1170>     needs to be prefixed with `::testing::Test::` if used outside of the
1171>     `TEST` body and the test fixture class.
1172> *   *`key`* must be a valid XML attribute name, and cannot conflict with the
1173>     ones already used by googletest (`name`, `status`, `time`, `classname`,
1174>     `type_param`, and `value_param`).
1175> *   Calling `RecordProperty()` outside of the lifespan of a test is allowed.
1176>     If it's called outside of a test but between a test suite's
1177>     `SetUpTestSuite()` and `TearDownTestSuite()` methods, it will be
1178>     attributed to the XML element for the test suite. If it's called outside
1179>     of all test suites (e.g. in a test environment), it will be attributed to
1180>     the top-level XML element.
1181
1182## Sharing Resources Between Tests in the Same Test Suite
1183
1184googletest creates a new test fixture object for each test in order to make
1185tests independent and easier to debug. However, sometimes tests use resources
1186that are expensive to set up, making the one-copy-per-test model prohibitively
1187expensive.
1188
1189If the tests don't change the resource, there's no harm in their sharing a
1190single resource copy. So, in addition to per-test set-up/tear-down, googletest
1191also supports per-test-suite set-up/tear-down. To use it:
1192
11931.  In your test fixture class (say `FooTest` ), declare as `static` some member
1194    variables to hold the shared resources.
11952.  Outside your test fixture class (typically just below it), define those
1196    member variables, optionally giving them initial values.
11973.  In the same test fixture class, define a `static void SetUpTestSuite()`
1198    function (remember not to spell it as **`SetupTestSuite`** with a small
1199    `u`!) to set up the shared resources and a `static void TearDownTestSuite()`
1200    function to tear them down.
1201
1202That's it! googletest automatically calls `SetUpTestSuite()` before running the
1203*first test* in the `FooTest` test suite (i.e. before creating the first
1204`FooTest` object), and calls `TearDownTestSuite()` after running the *last test*
1205in it (i.e. after deleting the last `FooTest` object). In between, the tests can
1206use the shared resources.
1207
1208Remember that the test order is undefined, so your code can't depend on a test
1209preceding or following another. Also, the tests must either not modify the state
1210of any shared resource, or, if they do modify the state, they must restore the
1211state to its original value before passing control to the next test.
1212
1213Here's an example of per-test-suite set-up and tear-down:
1214
1215```c++
1216class FooTest : public testing::Test {
1217 protected:
1218  // Per-test-suite set-up.
1219  // Called before the first test in this test suite.
1220  // Can be omitted if not needed.
1221  static void SetUpTestSuite() {
1222    shared_resource_ = new ...;
1223  }
1224
1225  // Per-test-suite tear-down.
1226  // Called after the last test in this test suite.
1227  // Can be omitted if not needed.
1228  static void TearDownTestSuite() {
1229    delete shared_resource_;
1230    shared_resource_ = nullptr;
1231  }
1232
1233  // You can define per-test set-up logic as usual.
1234  void SetUp() override { ... }
1235
1236  // You can define per-test tear-down logic as usual.
1237  void TearDown() override { ... }
1238
1239  // Some expensive resource shared by all tests.
1240  static T* shared_resource_;
1241};
1242
1243T* FooTest::shared_resource_ = nullptr;
1244
1245TEST_F(FooTest, Test1) {
1246  ... you can refer to shared_resource_ here ...
1247}
1248
1249TEST_F(FooTest, Test2) {
1250  ... you can refer to shared_resource_ here ...
1251}
1252```
1253
1254{: .callout .note}
1255NOTE: Though the above code declares `SetUpTestSuite()` protected, it may
1256sometimes be necessary to declare it public, such as when using it with
1257`TEST_P`.
1258
1259## Global Set-Up and Tear-Down
1260
1261Just as you can do set-up and tear-down at the test level and the test suite
1262level, you can also do it at the test program level. Here's how.
1263
1264First, you subclass the `::testing::Environment` class to define a test
1265environment, which knows how to set-up and tear-down:
1266
1267```c++
1268class Environment : public ::testing::Environment {
1269 public:
1270  ~Environment() override {}
1271
1272  // Override this to define how to set up the environment.
1273  void SetUp() override {}
1274
1275  // Override this to define how to tear down the environment.
1276  void TearDown() override {}
1277};
1278```
1279
1280Then, you register an instance of your environment class with googletest by
1281calling the `::testing::AddGlobalTestEnvironment()` function:
1282
1283```c++
1284Environment* AddGlobalTestEnvironment(Environment* env);
1285```
1286
1287Now, when `RUN_ALL_TESTS()` is called, it first calls the `SetUp()` method of
1288each environment object, then runs the tests if none of the environments
1289reported fatal failures and `GTEST_SKIP()` was not called. `RUN_ALL_TESTS()`
1290always calls `TearDown()` with each environment object, regardless of whether or
1291not the tests were run.
1292
1293It's OK to register multiple environment objects. In this suite, their `SetUp()`
1294will be called in the order they are registered, and their `TearDown()` will be
1295called in the reverse order.
1296
1297Note that googletest takes ownership of the registered environment objects.
1298Therefore **do not delete them** by yourself.
1299
1300You should call `AddGlobalTestEnvironment()` before `RUN_ALL_TESTS()` is called,
1301probably in `main()`. If you use `gtest_main`, you need to call this before
1302`main()` starts for it to take effect. One way to do this is to define a global
1303variable like this:
1304
1305```c++
1306testing::Environment* const foo_env =
1307    testing::AddGlobalTestEnvironment(new FooEnvironment);
1308```
1309
1310However, we strongly recommend you to write your own `main()` and call
1311`AddGlobalTestEnvironment()` there, as relying on initialization of global
1312variables makes the code harder to read and may cause problems when you register
1313multiple environments from different translation units and the environments have
1314dependencies among them (remember that the compiler doesn't guarantee the order
1315in which global variables from different translation units are initialized).
1316
1317## Value-Parameterized Tests
1318
1319*Value-parameterized tests* allow you to test your code with different
1320parameters without writing multiple copies of the same test. This is useful in a
1321number of situations, for example:
1322
1323*   You have a piece of code whose behavior is affected by one or more
1324    command-line flags. You want to make sure your code performs correctly for
1325    various values of those flags.
1326*   You want to test different implementations of an OO interface.
1327*   You want to test your code over various inputs (a.k.a. data-driven testing).
1328    This feature is easy to abuse, so please exercise your good sense when doing
1329    it!
1330
1331### How to Write Value-Parameterized Tests
1332
1333To write value-parameterized tests, first you should define a fixture class. It
1334must be derived from both `testing::Test` and `testing::WithParamInterface<T>`
1335(the latter is a pure interface), where `T` is the type of your parameter
1336values. For convenience, you can just derive the fixture class from
1337`testing::TestWithParam<T>`, which itself is derived from both `testing::Test`
1338and `testing::WithParamInterface<T>`. `T` can be any copyable type. If it's a
1339raw pointer, you are responsible for managing the lifespan of the pointed
1340values.
1341
1342{: .callout .note}
1343NOTE: If your test fixture defines `SetUpTestSuite()` or `TearDownTestSuite()`
1344they must be declared **public** rather than **protected** in order to use
1345`TEST_P`.
1346
1347```c++
1348class FooTest :
1349    public testing::TestWithParam<const char*> {
1350  // You can implement all the usual fixture class members here.
1351  // To access the test parameter, call GetParam() from class
1352  // TestWithParam<T>.
1353};
1354
1355// Or, when you want to add parameters to a pre-existing fixture class:
1356class BaseTest : public testing::Test {
1357  ...
1358};
1359class BarTest : public BaseTest,
1360                public testing::WithParamInterface<const char*> {
1361  ...
1362};
1363```
1364
1365Then, use the `TEST_P` macro to define as many test patterns using this fixture
1366as you want. The `_P` suffix is for "parameterized" or "pattern", whichever you
1367prefer to think.
1368
1369```c++
1370TEST_P(FooTest, DoesBlah) {
1371  // Inside a test, access the test parameter with the GetParam() method
1372  // of the TestWithParam<T> class:
1373  EXPECT_TRUE(foo.Blah(GetParam()));
1374  ...
1375}
1376
1377TEST_P(FooTest, HasBlahBlah) {
1378  ...
1379}
1380```
1381
1382Finally, you can use `INSTANTIATE_TEST_SUITE_P` to instantiate the test suite
1383with any set of parameters you want. googletest defines a number of functions
1384for generating test parameters. They return what we call (surprise!) *parameter
1385generators*. Here is a summary of them, which are all in the `testing`
1386namespace:
1387
1388
1389| Parameter Generator                                                                       | Behavior                                                                                                          |
1390| ----------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------- |
1391| `Range(begin, end [, step])`                                                              | Yields values `{begin, begin+step, begin+step+step, ...}`. The values do not include `end`. `step` defaults to 1. |
1392| `Values(v1, v2, ..., vN)`                                                                 | Yields values `{v1, v2, ..., vN}`.                                                                                |
1393| `ValuesIn(container)` and  `ValuesIn(begin,end)`                                          | Yields values from a C-style array, an  STL-style container, or an iterator range `[begin, end)`                  |
1394| `Bool()`                                                                                  | Yields sequence `{false, true}`.                                                                                  |
1395| `Combine(g1, g2, ..., gN)`                                                                | Yields all combinations (Cartesian product) as std\:\:tuples of the values generated by the `N` generators.       |
1396
1397
1398For more details, see the comments at the definitions of these functions.
1399
1400The following statement will instantiate tests from the `FooTest` test suite
1401each with parameter values `"meeny"`, `"miny"`, and `"moe"`.
1402
1403```c++
1404INSTANTIATE_TEST_SUITE_P(MeenyMinyMoe,
1405                         FooTest,
1406                         testing::Values("meeny", "miny", "moe"));
1407```
1408
1409{: .callout .note}
1410NOTE: The code above must be placed at global or namespace scope, not at
1411function scope.
1412
1413The first argument to `INSTANTIATE_TEST_SUITE_P` is a unique name for the
1414instantiation of the test suite. The next argument is the name of the test
1415pattern, and the last is the parameter generator.
1416
1417You can instantiate a test pattern more than once, so to distinguish different
1418instances of the pattern, the instantiation name is added as a prefix to the
1419actual test suite name. Remember to pick unique prefixes for different
1420instantiations. The tests from the instantiation above will have these names:
1421
1422*   `MeenyMinyMoe/FooTest.DoesBlah/0` for `"meeny"`
1423*   `MeenyMinyMoe/FooTest.DoesBlah/1` for `"miny"`
1424*   `MeenyMinyMoe/FooTest.DoesBlah/2` for `"moe"`
1425*   `MeenyMinyMoe/FooTest.HasBlahBlah/0` for `"meeny"`
1426*   `MeenyMinyMoe/FooTest.HasBlahBlah/1` for `"miny"`
1427*   `MeenyMinyMoe/FooTest.HasBlahBlah/2` for `"moe"`
1428
1429You can use these names in [`--gtest_filter`](#running-a-subset-of-the-tests).
1430
1431The following statement will instantiate all tests from `FooTest` again, each
1432with parameter values `"cat"` and `"dog"`:
1433
1434```c++
1435const char* pets[] = {"cat", "dog"};
1436INSTANTIATE_TEST_SUITE_P(Pets, FooTest, testing::ValuesIn(pets));
1437```
1438
1439The tests from the instantiation above will have these names:
1440
1441*   `Pets/FooTest.DoesBlah/0` for `"cat"`
1442*   `Pets/FooTest.DoesBlah/1` for `"dog"`
1443*   `Pets/FooTest.HasBlahBlah/0` for `"cat"`
1444*   `Pets/FooTest.HasBlahBlah/1` for `"dog"`
1445
1446Please note that `INSTANTIATE_TEST_SUITE_P` will instantiate *all* tests in the
1447given test suite, whether their definitions come before or *after* the
1448`INSTANTIATE_TEST_SUITE_P` statement.
1449
1450Additionally, by default, every `TEST_P` without a corresponding
1451`INSTANTIATE_TEST_SUITE_P` causes a failing test in test suite
1452`GoogleTestVerification`. If you have a test suite where that omission is not an
1453error, for example it is in a library that may be linked in for other reasons or
1454where the list of test cases is dynamic and may be empty, then this check can be
1455suppressed by tagging the test suite:
1456
1457```c++
1458GTEST_ALLOW_UNINSTANTIATED_PARAMETERIZED_TEST(FooTest);
1459```
1460
1461You can see [sample7_unittest.cc] and [sample8_unittest.cc] for more examples.
1462
1463[sample7_unittest.cc]: https://github.com/google/googletest/blob/master/googletest/samples/sample7_unittest.cc "Parameterized Test example"
1464[sample8_unittest.cc]: https://github.com/google/googletest/blob/master/googletest/samples/sample8_unittest.cc "Parameterized Test example with multiple parameters"
1465
1466### Creating Value-Parameterized Abstract Tests
1467
1468In the above, we define and instantiate `FooTest` in the *same* source file.
1469Sometimes you may want to define value-parameterized tests in a library and let
1470other people instantiate them later. This pattern is known as *abstract tests*.
1471As an example of its application, when you are designing an interface you can
1472write a standard suite of abstract tests (perhaps using a factory function as
1473the test parameter) that all implementations of the interface are expected to
1474pass. When someone implements the interface, they can instantiate your suite to
1475get all the interface-conformance tests for free.
1476
1477To define abstract tests, you should organize your code like this:
1478
14791.  Put the definition of the parameterized test fixture class (e.g. `FooTest`)
1480    in a header file, say `foo_param_test.h`. Think of this as *declaring* your
1481    abstract tests.
14822.  Put the `TEST_P` definitions in `foo_param_test.cc`, which includes
1483    `foo_param_test.h`. Think of this as *implementing* your abstract tests.
1484
1485Once they are defined, you can instantiate them by including `foo_param_test.h`,
1486invoking `INSTANTIATE_TEST_SUITE_P()`, and depending on the library target that
1487contains `foo_param_test.cc`. You can instantiate the same abstract test suite
1488multiple times, possibly in different source files.
1489
1490### Specifying Names for Value-Parameterized Test Parameters
1491
1492The optional last argument to `INSTANTIATE_TEST_SUITE_P()` allows the user to
1493specify a function or functor that generates custom test name suffixes based on
1494the test parameters. The function should accept one argument of type
1495`testing::TestParamInfo<class ParamType>`, and return `std::string`.
1496
1497`testing::PrintToStringParamName` is a builtin test suffix generator that
1498returns the value of `testing::PrintToString(GetParam())`. It does not work for
1499`std::string` or C strings.
1500
1501{: .callout .note}
1502NOTE: test names must be non-empty, unique, and may only contain ASCII
1503alphanumeric characters. In particular, they
1504[should not contain underscores](faq.md#why-should-test-suite-names-and-test-names-not-contain-underscore)
1505
1506```c++
1507class MyTestSuite : public testing::TestWithParam<int> {};
1508
1509TEST_P(MyTestSuite, MyTest)
1510{
1511  std::cout << "Example Test Param: " << GetParam() << std::endl;
1512}
1513
1514INSTANTIATE_TEST_SUITE_P(MyGroup, MyTestSuite, testing::Range(0, 10),
1515                         testing::PrintToStringParamName());
1516```
1517
1518Providing a custom functor allows for more control over test parameter name
1519generation, especially for types where the automatic conversion does not
1520generate helpful parameter names (e.g. strings as demonstrated above). The
1521following example illustrates this for multiple parameters, an enumeration type
1522and a string, and also demonstrates how to combine generators. It uses a lambda
1523for conciseness:
1524
1525```c++
1526enum class MyType { MY_FOO = 0, MY_BAR = 1 };
1527
1528class MyTestSuite : public testing::TestWithParam<std::tuple<MyType, std::string>> {
1529};
1530
1531INSTANTIATE_TEST_SUITE_P(
1532    MyGroup, MyTestSuite,
1533    testing::Combine(
1534        testing::Values(MyType::MY_FOO, MyType::MY_BAR),
1535        testing::Values("A", "B")),
1536    [](const testing::TestParamInfo<MyTestSuite::ParamType>& info) {
1537      std::string name = absl::StrCat(
1538          std::get<0>(info.param) == MyType::MY_FOO ? "Foo" : "Bar",
1539          std::get<1>(info.param));
1540      absl::c_replace_if(name, [](char c) { return !std::isalnum(c); }, '');
1541      return name;
1542    });
1543```
1544
1545## Typed Tests
1546
1547Suppose you have multiple implementations of the same interface and want to make
1548sure that all of them satisfy some common requirements. Or, you may have defined
1549several types that are supposed to conform to the same "concept" and you want to
1550verify it. In both cases, you want the same test logic repeated for different
1551types.
1552
1553While you can write one `TEST` or `TEST_F` for each type you want to test (and
1554you may even factor the test logic into a function template that you invoke from
1555the `TEST`), it's tedious and doesn't scale: if you want `m` tests over `n`
1556types, you'll end up writing `m*n` `TEST`s.
1557
1558*Typed tests* allow you to repeat the same test logic over a list of types. You
1559only need to write the test logic once, although you must know the type list
1560when writing typed tests. Here's how you do it:
1561
1562First, define a fixture class template. It should be parameterized by a type.
1563Remember to derive it from `::testing::Test`:
1564
1565```c++
1566template <typename T>
1567class FooTest : public testing::Test {
1568 public:
1569  ...
1570  using List = std::list<T>;
1571  static T shared_;
1572  T value_;
1573};
1574```
1575
1576Next, associate a list of types with the test suite, which will be repeated for
1577each type in the list:
1578
1579```c++
1580using MyTypes = ::testing::Types<char, int, unsigned int>;
1581TYPED_TEST_SUITE(FooTest, MyTypes);
1582```
1583
1584The type alias (`using` or `typedef`) is necessary for the `TYPED_TEST_SUITE`
1585macro to parse correctly. Otherwise the compiler will think that each comma in
1586the type list introduces a new macro argument.
1587
1588Then, use `TYPED_TEST()` instead of `TEST_F()` to define a typed test for this
1589test suite. You can repeat this as many times as you want:
1590
1591```c++
1592TYPED_TEST(FooTest, DoesBlah) {
1593  // Inside a test, refer to the special name TypeParam to get the type
1594  // parameter.  Since we are inside a derived class template, C++ requires
1595  // us to visit the members of FooTest via 'this'.
1596  TypeParam n = this->value_;
1597
1598  // To visit static members of the fixture, add the 'TestFixture::'
1599  // prefix.
1600  n += TestFixture::shared_;
1601
1602  // To refer to typedefs in the fixture, add the 'typename TestFixture::'
1603  // prefix.  The 'typename' is required to satisfy the compiler.
1604  typename TestFixture::List values;
1605
1606  values.push_back(n);
1607  ...
1608}
1609
1610TYPED_TEST(FooTest, HasPropertyA) { ... }
1611```
1612
1613You can see [sample6_unittest.cc] for a complete example.
1614
1615[sample6_unittest.cc]: https://github.com/google/googletest/blob/master/googletest/samples/sample6_unittest.cc "Typed Test example"
1616
1617## Type-Parameterized Tests
1618
1619*Type-parameterized tests* are like typed tests, except that they don't require
1620you to know the list of types ahead of time. Instead, you can define the test
1621logic first and instantiate it with different type lists later. You can even
1622instantiate it more than once in the same program.
1623
1624If you are designing an interface or concept, you can define a suite of
1625type-parameterized tests to verify properties that any valid implementation of
1626the interface/concept should have. Then, the author of each implementation can
1627just instantiate the test suite with their type to verify that it conforms to
1628the requirements, without having to write similar tests repeatedly. Here's an
1629example:
1630
1631First, define a fixture class template, as we did with typed tests:
1632
1633```c++
1634template <typename T>
1635class FooTest : public testing::Test {
1636  ...
1637};
1638```
1639
1640Next, declare that you will define a type-parameterized test suite:
1641
1642```c++
1643TYPED_TEST_SUITE_P(FooTest);
1644```
1645
1646Then, use `TYPED_TEST_P()` to define a type-parameterized test. You can repeat
1647this as many times as you want:
1648
1649```c++
1650TYPED_TEST_P(FooTest, DoesBlah) {
1651  // Inside a test, refer to TypeParam to get the type parameter.
1652  TypeParam n = 0;
1653  ...
1654}
1655
1656TYPED_TEST_P(FooTest, HasPropertyA) { ... }
1657```
1658
1659Now the tricky part: you need to register all test patterns using the
1660`REGISTER_TYPED_TEST_SUITE_P` macro before you can instantiate them. The first
1661argument of the macro is the test suite name; the rest are the names of the
1662tests in this test suite:
1663
1664```c++
1665REGISTER_TYPED_TEST_SUITE_P(FooTest,
1666                            DoesBlah, HasPropertyA);
1667```
1668
1669Finally, you are free to instantiate the pattern with the types you want. If you
1670put the above code in a header file, you can `#include` it in multiple C++
1671source files and instantiate it multiple times.
1672
1673```c++
1674using MyTypes = ::testing::Types<char, int, unsigned int>;
1675INSTANTIATE_TYPED_TEST_SUITE_P(My, FooTest, MyTypes);
1676```
1677
1678To distinguish different instances of the pattern, the first argument to the
1679`INSTANTIATE_TYPED_TEST_SUITE_P` macro is a prefix that will be added to the
1680actual test suite name. Remember to pick unique prefixes for different
1681instances.
1682
1683In the special case where the type list contains only one type, you can write
1684that type directly without `::testing::Types<...>`, like this:
1685
1686```c++
1687INSTANTIATE_TYPED_TEST_SUITE_P(My, FooTest, int);
1688```
1689
1690You can see [sample6_unittest.cc] for a complete example.
1691
1692## Testing Private Code
1693
1694If you change your software's internal implementation, your tests should not
1695break as long as the change is not observable by users. Therefore, **per the
1696black-box testing principle, most of the time you should test your code through
1697its public interfaces.**
1698
1699**If you still find yourself needing to test internal implementation code,
1700consider if there's a better design.** The desire to test internal
1701implementation is often a sign that the class is doing too much. Consider
1702extracting an implementation class, and testing it. Then use that implementation
1703class in the original class.
1704
1705If you absolutely have to test non-public interface code though, you can. There
1706are two cases to consider:
1707
1708*   Static functions ( *not* the same as static member functions!) or unnamed
1709    namespaces, and
1710*   Private or protected class members
1711
1712To test them, we use the following special techniques:
1713
1714*   Both static functions and definitions/declarations in an unnamed namespace
1715    are only visible within the same translation unit. To test them, you can
1716    `#include` the entire `.cc` file being tested in your `*_test.cc` file.
1717    (#including `.cc` files is not a good way to reuse code - you should not do
1718    this in production code!)
1719
1720    However, a better approach is to move the private code into the
1721    `foo::internal` namespace, where `foo` is the namespace your project
1722    normally uses, and put the private declarations in a `*-internal.h` file.
1723    Your production `.cc` files and your tests are allowed to include this
1724    internal header, but your clients are not. This way, you can fully test your
1725    internal implementation without leaking it to your clients.
1726
1727*   Private class members are only accessible from within the class or by
1728    friends. To access a class' private members, you can declare your test
1729    fixture as a friend to the class and define accessors in your fixture. Tests
1730    using the fixture can then access the private members of your production
1731    class via the accessors in the fixture. Note that even though your fixture
1732    is a friend to your production class, your tests are not automatically
1733    friends to it, as they are technically defined in sub-classes of the
1734    fixture.
1735
1736    Another way to test private members is to refactor them into an
1737    implementation class, which is then declared in a `*-internal.h` file. Your
1738    clients aren't allowed to include this header but your tests can. Such is
1739    called the
1740    [Pimpl](https://www.gamedev.net/articles/programming/general-and-gameplay-programming/the-c-pimpl-r1794/)
1741    (Private Implementation) idiom.
1742
1743    Or, you can declare an individual test as a friend of your class by adding
1744    this line in the class body:
1745
1746    ```c++
1747        FRIEND_TEST(TestSuiteName, TestName);
1748    ```
1749
1750    For example,
1751
1752    ```c++
1753    // foo.h
1754    class Foo {
1755      ...
1756     private:
1757      FRIEND_TEST(FooTest, BarReturnsZeroOnNull);
1758
1759      int Bar(void* x);
1760    };
1761
1762    // foo_test.cc
1763    ...
1764    TEST(FooTest, BarReturnsZeroOnNull) {
1765      Foo foo;
1766      EXPECT_EQ(foo.Bar(NULL), 0);  // Uses Foo's private member Bar().
1767    }
1768    ```
1769
1770    Pay special attention when your class is defined in a namespace. If you want
1771    your test fixtures and tests to be friends of your class, then they must be
1772    defined in the exact same namespace (no anonymous or inline namespaces).
1773
1774    For example, if the code to be tested looks like:
1775
1776    ```c++
1777    namespace my_namespace {
1778
1779    class Foo {
1780      friend class FooTest;
1781      FRIEND_TEST(FooTest, Bar);
1782      FRIEND_TEST(FooTest, Baz);
1783      ... definition of the class Foo ...
1784    };
1785
1786    }  // namespace my_namespace
1787    ```
1788
1789    Your test code should be something like:
1790
1791    ```c++
1792    namespace my_namespace {
1793
1794    class FooTest : public testing::Test {
1795     protected:
1796      ...
1797    };
1798
1799    TEST_F(FooTest, Bar) { ... }
1800    TEST_F(FooTest, Baz) { ... }
1801
1802    }  // namespace my_namespace
1803    ```
1804
1805## "Catching" Failures
1806
1807If you are building a testing utility on top of googletest, you'll want to test
1808your utility. What framework would you use to test it? googletest, of course.
1809
1810The challenge is to verify that your testing utility reports failures correctly.
1811In frameworks that report a failure by throwing an exception, you could catch
1812the exception and assert on it. But googletest doesn't use exceptions, so how do
1813we test that a piece of code generates an expected failure?
1814
1815`"gtest/gtest-spi.h"` contains some constructs to do this. After #including this header,
1816you can use
1817
1818```c++
1819  EXPECT_FATAL_FAILURE(statement, substring);
1820```
1821
1822to assert that `statement` generates a fatal (e.g. `ASSERT_*`) failure in the
1823current thread whose message contains the given `substring`, or use
1824
1825```c++
1826  EXPECT_NONFATAL_FAILURE(statement, substring);
1827```
1828
1829if you are expecting a non-fatal (e.g. `EXPECT_*`) failure.
1830
1831Only failures in the current thread are checked to determine the result of this
1832type of expectations. If `statement` creates new threads, failures in these
1833threads are also ignored. If you want to catch failures in other threads as
1834well, use one of the following macros instead:
1835
1836```c++
1837  EXPECT_FATAL_FAILURE_ON_ALL_THREADS(statement, substring);
1838  EXPECT_NONFATAL_FAILURE_ON_ALL_THREADS(statement, substring);
1839```
1840
1841{: .callout .note}
1842NOTE: Assertions from multiple threads are currently not supported on Windows.
1843
1844For technical reasons, there are some caveats:
1845
18461.  You cannot stream a failure message to either macro.
1847
18482.  `statement` in `EXPECT_FATAL_FAILURE{_ON_ALL_THREADS}()` cannot reference
1849    local non-static variables or non-static members of `this` object.
1850
18513.  `statement` in `EXPECT_FATAL_FAILURE{_ON_ALL_THREADS}()` cannot return a
1852    value.
1853
1854## Registering tests programmatically
1855
1856The `TEST` macros handle the vast majority of all use cases, but there are few
1857where runtime registration logic is required. For those cases, the framework
1858provides the `::testing::RegisterTest` that allows callers to register arbitrary
1859tests dynamically.
1860
1861This is an advanced API only to be used when the `TEST` macros are insufficient.
1862The macros should be preferred when possible, as they avoid most of the
1863complexity of calling this function.
1864
1865It provides the following signature:
1866
1867```c++
1868template <typename Factory>
1869TestInfo* RegisterTest(const char* test_suite_name, const char* test_name,
1870                       const char* type_param, const char* value_param,
1871                       const char* file, int line, Factory factory);
1872```
1873
1874The `factory` argument is a factory callable (move-constructible) object or
1875function pointer that creates a new instance of the Test object. It handles
1876ownership to the caller. The signature of the callable is `Fixture*()`, where
1877`Fixture` is the test fixture class for the test. All tests registered with the
1878same `test_suite_name` must return the same fixture type. This is checked at
1879runtime.
1880
1881The framework will infer the fixture class from the factory and will call the
1882`SetUpTestSuite` and `TearDownTestSuite` for it.
1883
1884Must be called before `RUN_ALL_TESTS()` is invoked, otherwise behavior is
1885undefined.
1886
1887Use case example:
1888
1889```c++
1890class MyFixture : public testing::Test {
1891 public:
1892  // All of these optional, just like in regular macro usage.
1893  static void SetUpTestSuite() { ... }
1894  static void TearDownTestSuite() { ... }
1895  void SetUp() override { ... }
1896  void TearDown() override { ... }
1897};
1898
1899class MyTest : public MyFixture {
1900 public:
1901  explicit MyTest(int data) : data_(data) {}
1902  void TestBody() override { ... }
1903
1904 private:
1905  int data_;
1906};
1907
1908void RegisterMyTests(const std::vector<int>& values) {
1909  for (int v : values) {
1910    testing::RegisterTest(
1911        "MyFixture", ("Test" + std::to_string(v)).c_str(), nullptr,
1912        std::to_string(v).c_str(),
1913        __FILE__, __LINE__,
1914        // Important to use the fixture type as the return type here.
1915        [=]() -> MyFixture* { return new MyTest(v); });
1916  }
1917}
1918...
1919int main(int argc, char** argv) {
1920  std::vector<int> values_to_test = LoadValuesFromConfig();
1921  RegisterMyTests(values_to_test);
1922  ...
1923  return RUN_ALL_TESTS();
1924}
1925```
1926## Getting the Current Test's Name
1927
1928Sometimes a function may need to know the name of the currently running test.
1929For example, you may be using the `SetUp()` method of your test fixture to set
1930the golden file name based on which test is running. The `::testing::TestInfo`
1931class has this information:
1932
1933```c++
1934namespace testing {
1935
1936class TestInfo {
1937 public:
1938  // Returns the test suite name and the test name, respectively.
1939  //
1940  // Do NOT delete or free the return value - it's managed by the
1941  // TestInfo class.
1942  const char* test_suite_name() const;
1943  const char* name() const;
1944};
1945
1946}
1947```
1948
1949To obtain a `TestInfo` object for the currently running test, call
1950`current_test_info()` on the `UnitTest` singleton object:
1951
1952```c++
1953  // Gets information about the currently running test.
1954  // Do NOT delete the returned object - it's managed by the UnitTest class.
1955  const testing::TestInfo* const test_info =
1956      testing::UnitTest::GetInstance()->current_test_info();
1957
1958  printf("We are in test %s of test suite %s.\n",
1959         test_info->name(),
1960         test_info->test_suite_name());
1961```
1962
1963`current_test_info()` returns a null pointer if no test is running. In
1964particular, you cannot find the test suite name in `SetUpTestSuite()`,
1965`TearDownTestSuite()` (where you know the test suite name implicitly), or
1966functions called from them.
1967
1968## Extending googletest by Handling Test Events
1969
1970googletest provides an **event listener API** to let you receive notifications
1971about the progress of a test program and test failures. The events you can
1972listen to include the start and end of the test program, a test suite, or a test
1973method, among others. You may use this API to augment or replace the standard
1974console output, replace the XML output, or provide a completely different form
1975of output, such as a GUI or a database. You can also use test events as
1976checkpoints to implement a resource leak checker, for example.
1977
1978### Defining Event Listeners
1979
1980To define a event listener, you subclass either testing::TestEventListener or
1981testing::EmptyTestEventListener The former is an (abstract) interface, where
1982*each pure virtual method can be overridden to handle a test event* (For
1983example, when a test starts, the `OnTestStart()` method will be called.). The
1984latter provides an empty implementation of all methods in the interface, such
1985that a subclass only needs to override the methods it cares about.
1986
1987When an event is fired, its context is passed to the handler function as an
1988argument. The following argument types are used:
1989
1990*   UnitTest reflects the state of the entire test program,
1991*   TestSuite has information about a test suite, which can contain one or more
1992    tests,
1993*   TestInfo contains the state of a test, and
1994*   TestPartResult represents the result of a test assertion.
1995
1996An event handler function can examine the argument it receives to find out
1997interesting information about the event and the test program's state.
1998
1999Here's an example:
2000
2001```c++
2002  class MinimalistPrinter : public testing::EmptyTestEventListener {
2003    // Called before a test starts.
2004    void OnTestStart(const testing::TestInfo& test_info) override {
2005      printf("*** Test %s.%s starting.\n",
2006             test_info.test_suite_name(), test_info.name());
2007    }
2008
2009    // Called after a failed assertion or a SUCCESS().
2010    void OnTestPartResult(const testing::TestPartResult& test_part_result) override {
2011      printf("%s in %s:%d\n%s\n",
2012             test_part_result.failed() ? "*** Failure" : "Success",
2013             test_part_result.file_name(),
2014             test_part_result.line_number(),
2015             test_part_result.summary());
2016    }
2017
2018    // Called after a test ends.
2019    void OnTestEnd(const testing::TestInfo& test_info) override {
2020      printf("*** Test %s.%s ending.\n",
2021             test_info.test_suite_name(), test_info.name());
2022    }
2023  };
2024```
2025
2026### Using Event Listeners
2027
2028To use the event listener you have defined, add an instance of it to the
2029googletest event listener list (represented by class TestEventListeners - note
2030the "s" at the end of the name) in your `main()` function, before calling
2031`RUN_ALL_TESTS()`:
2032
2033```c++
2034int main(int argc, char** argv) {
2035  testing::InitGoogleTest(&argc, argv);
2036  // Gets hold of the event listener list.
2037  testing::TestEventListeners& listeners =
2038      testing::UnitTest::GetInstance()->listeners();
2039  // Adds a listener to the end.  googletest takes the ownership.
2040  listeners.Append(new MinimalistPrinter);
2041  return RUN_ALL_TESTS();
2042}
2043```
2044
2045There's only one problem: the default test result printer is still in effect, so
2046its output will mingle with the output from your minimalist printer. To suppress
2047the default printer, just release it from the event listener list and delete it.
2048You can do so by adding one line:
2049
2050```c++
2051  ...
2052  delete listeners.Release(listeners.default_result_printer());
2053  listeners.Append(new MinimalistPrinter);
2054  return RUN_ALL_TESTS();
2055```
2056
2057Now, sit back and enjoy a completely different output from your tests. For more
2058details, see [sample9_unittest.cc].
2059
2060[sample9_unittest.cc]: https://github.com/google/googletest/blob/master/googletest/samples/sample9_unittest.cc "Event listener example"
2061
2062You may append more than one listener to the list. When an `On*Start()` or
2063`OnTestPartResult()` event is fired, the listeners will receive it in the order
2064they appear in the list (since new listeners are added to the end of the list,
2065the default text printer and the default XML generator will receive the event
2066first). An `On*End()` event will be received by the listeners in the *reverse*
2067order. This allows output by listeners added later to be framed by output from
2068listeners added earlier.
2069
2070### Generating Failures in Listeners
2071
2072You may use failure-raising macros (`EXPECT_*()`, `ASSERT_*()`, `FAIL()`, etc)
2073when processing an event. There are some restrictions:
2074
20751.  You cannot generate any failure in `OnTestPartResult()` (otherwise it will
2076    cause `OnTestPartResult()` to be called recursively).
20772.  A listener that handles `OnTestPartResult()` is not allowed to generate any
2078    failure.
2079
2080When you add listeners to the listener list, you should put listeners that
2081handle `OnTestPartResult()` *before* listeners that can generate failures. This
2082ensures that failures generated by the latter are attributed to the right test
2083by the former.
2084
2085See [sample10_unittest.cc] for an example of a failure-raising listener.
2086
2087[sample10_unittest.cc]: https://github.com/google/googletest/blob/master/googletest/samples/sample10_unittest.cc "Failure-raising listener example"
2088
2089## Running Test Programs: Advanced Options
2090
2091googletest test programs are ordinary executables. Once built, you can run them
2092directly and affect their behavior via the following environment variables
2093and/or command line flags. For the flags to work, your programs must call
2094`::testing::InitGoogleTest()` before calling `RUN_ALL_TESTS()`.
2095
2096To see a list of supported flags and their usage, please run your test program
2097with the `--help` flag. You can also use `-h`, `-?`, or `/?` for short.
2098
2099If an option is specified both by an environment variable and by a flag, the
2100latter takes precedence.
2101
2102### Selecting Tests
2103
2104#### Listing Test Names
2105
2106Sometimes it is necessary to list the available tests in a program before
2107running them so that a filter may be applied if needed. Including the flag
2108`--gtest_list_tests` overrides all other flags and lists tests in the following
2109format:
2110
2111```none
2112TestSuite1.
2113  TestName1
2114  TestName2
2115TestSuite2.
2116  TestName
2117```
2118
2119None of the tests listed are actually run if the flag is provided. There is no
2120corresponding environment variable for this flag.
2121
2122#### Running a Subset of the Tests
2123
2124By default, a googletest program runs all tests the user has defined. Sometimes,
2125you want to run only a subset of the tests (e.g. for debugging or quickly
2126verifying a change). If you set the `GTEST_FILTER` environment variable or the
2127`--gtest_filter` flag to a filter string, googletest will only run the tests
2128whose full names (in the form of `TestSuiteName.TestName`) match the filter.
2129
2130The format of a filter is a '`:`'-separated list of wildcard patterns (called
2131the *positive patterns*) optionally followed by a '`-`' and another
2132'`:`'-separated pattern list (called the *negative patterns*). A test matches
2133the filter if and only if it matches any of the positive patterns but does not
2134match any of the negative patterns.
2135
2136A pattern may contain `'*'` (matches any string) or `'?'` (matches any single
2137character). For convenience, the filter `'*-NegativePatterns'` can be also
2138written as `'-NegativePatterns'`.
2139
2140For example:
2141
2142*   `./foo_test` Has no flag, and thus runs all its tests.
2143*   `./foo_test --gtest_filter=*` Also runs everything, due to the single
2144    match-everything `*` value.
2145*   `./foo_test --gtest_filter=FooTest.*` Runs everything in test suite
2146    `FooTest` .
2147*   `./foo_test --gtest_filter=*Null*:*Constructor*` Runs any test whose full
2148    name contains either `"Null"` or `"Constructor"` .
2149*   `./foo_test --gtest_filter=-*DeathTest.*` Runs all non-death tests.
2150*   `./foo_test --gtest_filter=FooTest.*-FooTest.Bar` Runs everything in test
2151    suite `FooTest` except `FooTest.Bar`.
2152*   `./foo_test --gtest_filter=FooTest.*:BarTest.*-FooTest.Bar:BarTest.Foo` Runs
2153    everything in test suite `FooTest` except `FooTest.Bar` and everything in
2154    test suite `BarTest` except `BarTest.Foo`.
2155
2156#### Stop test execution upon first failure
2157
2158By default, a googletest program runs all tests the user has defined. In some
2159cases (e.g. iterative test development & execution) it may be desirable stop
2160test execution upon first failure (trading improved latency for completeness).
2161If `GTEST_FAIL_FAST` environment variable or `--gtest_fail_fast` flag is set,
2162the test runner will stop execution as soon as the first test failure is
2163found.
2164
2165#### Temporarily Disabling Tests
2166
2167If you have a broken test that you cannot fix right away, you can add the
2168`DISABLED_` prefix to its name. This will exclude it from execution. This is
2169better than commenting out the code or using `#if 0`, as disabled tests are
2170still compiled (and thus won't rot).
2171
2172If you need to disable all tests in a test suite, you can either add `DISABLED_`
2173to the front of the name of each test, or alternatively add it to the front of
2174the test suite name.
2175
2176For example, the following tests won't be run by googletest, even though they
2177will still be compiled:
2178
2179```c++
2180// Tests that Foo does Abc.
2181TEST(FooTest, DISABLED_DoesAbc) { ... }
2182
2183class DISABLED_BarTest : public testing::Test { ... };
2184
2185// Tests that Bar does Xyz.
2186TEST_F(DISABLED_BarTest, DoesXyz) { ... }
2187```
2188
2189{: .callout .note}
2190NOTE: This feature should only be used for temporary pain-relief. You still have
2191to fix the disabled tests at a later date. As a reminder, googletest will print
2192a banner warning you if a test program contains any disabled tests.
2193
2194{: .callout .tip}
2195TIP: You can easily count the number of disabled tests you have using
2196`grep`. This number can be used as a metric for
2197improving your test quality.
2198
2199#### Temporarily Enabling Disabled Tests
2200
2201To include disabled tests in test execution, just invoke the test program with
2202the `--gtest_also_run_disabled_tests` flag or set the
2203`GTEST_ALSO_RUN_DISABLED_TESTS` environment variable to a value other than `0`.
2204You can combine this with the `--gtest_filter` flag to further select which
2205disabled tests to run.
2206
2207### Repeating the Tests
2208
2209Once in a while you'll run into a test whose result is hit-or-miss. Perhaps it
2210will fail only 1% of the time, making it rather hard to reproduce the bug under
2211a debugger. This can be a major source of frustration.
2212
2213The `--gtest_repeat` flag allows you to repeat all (or selected) test methods in
2214a program many times. Hopefully, a flaky test will eventually fail and give you
2215a chance to debug. Here's how to use it:
2216
2217```none
2218$ foo_test --gtest_repeat=1000
2219Repeat foo_test 1000 times and don't stop at failures.
2220
2221$ foo_test --gtest_repeat=-1
2222A negative count means repeating forever.
2223
2224$ foo_test --gtest_repeat=1000 --gtest_break_on_failure
2225Repeat foo_test 1000 times, stopping at the first failure.  This
2226is especially useful when running under a debugger: when the test
2227fails, it will drop into the debugger and you can then inspect
2228variables and stacks.
2229
2230$ foo_test --gtest_repeat=1000 --gtest_filter=FooBar.*
2231Repeat the tests whose name matches the filter 1000 times.
2232```
2233
2234If your test program contains
2235[global set-up/tear-down](#global-set-up-and-tear-down) code, it will be
2236repeated in each iteration as well, as the flakiness may be in it. You can also
2237specify the repeat count by setting the `GTEST_REPEAT` environment variable.
2238
2239### Shuffling the Tests
2240
2241You can specify the `--gtest_shuffle` flag (or set the `GTEST_SHUFFLE`
2242environment variable to `1`) to run the tests in a program in a random order.
2243This helps to reveal bad dependencies between tests.
2244
2245By default, googletest uses a random seed calculated from the current time.
2246Therefore you'll get a different order every time. The console output includes
2247the random seed value, such that you can reproduce an order-related test failure
2248later. To specify the random seed explicitly, use the `--gtest_random_seed=SEED`
2249flag (or set the `GTEST_RANDOM_SEED` environment variable), where `SEED` is an
2250integer in the range [0, 99999]. The seed value 0 is special: it tells
2251googletest to do the default behavior of calculating the seed from the current
2252time.
2253
2254If you combine this with `--gtest_repeat=N`, googletest will pick a different
2255random seed and re-shuffle the tests in each iteration.
2256
2257### Controlling Test Output
2258
2259#### Colored Terminal Output
2260
2261googletest can use colors in its terminal output to make it easier to spot the
2262important information:
2263
2264<pre>...
2265<font color="green">[----------]</font> 1 test from FooTest
2266<font color="green">[ RUN      ]</font> FooTest.DoesAbc
2267<font color="green">[       OK ]</font> FooTest.DoesAbc
2268<font color="green">[----------]</font> 2 tests from BarTest
2269<font color="green">[ RUN      ]</font> BarTest.HasXyzProperty
2270<font color="green">[       OK ]</font> BarTest.HasXyzProperty
2271<font color="green">[ RUN      ]</font> BarTest.ReturnsTrueOnSuccess
2272... some error messages ...
2273<font color="red">[   FAILED ]</font> BarTest.ReturnsTrueOnSuccess
2274...
2275<font color="green">[==========]</font> 30 tests from 14 test suites ran.
2276<font color="green">[   PASSED ]</font> 28 tests.
2277<font color="red">[   FAILED ]</font> 2 tests, listed below:
2278<font color="red">[   FAILED ]</font> BarTest.ReturnsTrueOnSuccess
2279<font color="red">[   FAILED ]</font> AnotherTest.DoesXyz
2280
2281 2 FAILED TESTS
2282</pre>
2283
2284You can set the `GTEST_COLOR` environment variable or the `--gtest_color`
2285command line flag to `yes`, `no`, or `auto` (the default) to enable colors,
2286disable colors, or let googletest decide. When the value is `auto`, googletest
2287will use colors if and only if the output goes to a terminal and (on non-Windows
2288platforms) the `TERM` environment variable is set to `xterm` or `xterm-color`.
2289
2290#### Suppressing test passes
2291
2292By default, googletest prints 1 line of output for each test, indicating if it
2293passed or failed. To show only test failures, run the test program with
2294`--gtest_brief=1`, or set the GTEST_BRIEF environment variable to `1`.
2295
2296#### Suppressing the Elapsed Time
2297
2298By default, googletest prints the time it takes to run each test. To disable
2299that, run the test program with the `--gtest_print_time=0` command line flag, or
2300set the GTEST_PRINT_TIME environment variable to `0`.
2301
2302#### Suppressing UTF-8 Text Output
2303
2304In case of assertion failures, googletest prints expected and actual values of
2305type `string` both as hex-encoded strings as well as in readable UTF-8 text if
2306they contain valid non-ASCII UTF-8 characters. If you want to suppress the UTF-8
2307text because, for example, you don't have an UTF-8 compatible output medium, run
2308the test program with `--gtest_print_utf8=0` or set the `GTEST_PRINT_UTF8`
2309environment variable to `0`.
2310
2311
2312
2313#### Generating an XML Report
2314
2315googletest can emit a detailed XML report to a file in addition to its normal
2316textual output. The report contains the duration of each test, and thus can help
2317you identify slow tests.
2318
2319To generate the XML report, set the `GTEST_OUTPUT` environment variable or the
2320`--gtest_output` flag to the string `"xml:path_to_output_file"`, which will
2321create the file at the given location. You can also just use the string `"xml"`,
2322in which case the output can be found in the `test_detail.xml` file in the
2323current directory.
2324
2325If you specify a directory (for example, `"xml:output/directory/"` on Linux or
2326`"xml:output\directory\"` on Windows), googletest will create the XML file in
2327that directory, named after the test executable (e.g. `foo_test.xml` for test
2328program `foo_test` or `foo_test.exe`). If the file already exists (perhaps left
2329over from a previous run), googletest will pick a different name (e.g.
2330`foo_test_1.xml`) to avoid overwriting it.
2331
2332The report is based on the `junitreport` Ant task. Since that format was
2333originally intended for Java, a little interpretation is required to make it
2334apply to googletest tests, as shown here:
2335
2336```xml
2337<testsuites name="AllTests" ...>
2338  <testsuite name="test_case_name" ...>
2339    <testcase    name="test_name" ...>
2340      <failure message="..."/>
2341      <failure message="..."/>
2342      <failure message="..."/>
2343    </testcase>
2344  </testsuite>
2345</testsuites>
2346```
2347
2348*   The root `<testsuites>` element corresponds to the entire test program.
2349*   `<testsuite>` elements correspond to googletest test suites.
2350*   `<testcase>` elements correspond to googletest test functions.
2351
2352For instance, the following program
2353
2354```c++
2355TEST(MathTest, Addition) { ... }
2356TEST(MathTest, Subtraction) { ... }
2357TEST(LogicTest, NonContradiction) { ... }
2358```
2359
2360could generate this report:
2361
2362```xml
2363<?xml version="1.0" encoding="UTF-8"?>
2364<testsuites tests="3" failures="1" errors="0" time="0.035" timestamp="2011-10-31T18:52:42" name="AllTests">
2365  <testsuite name="MathTest" tests="2" failures="1" errors="0" time="0.015">
2366    <testcase name="Addition" status="run" time="0.007" classname="">
2367      <failure message="Value of: add(1, 1)&#x0A;  Actual: 3&#x0A;Expected: 2" type="">...</failure>
2368      <failure message="Value of: add(1, -1)&#x0A;  Actual: 1&#x0A;Expected: 0" type="">...</failure>
2369    </testcase>
2370    <testcase name="Subtraction" status="run" time="0.005" classname="">
2371    </testcase>
2372  </testsuite>
2373  <testsuite name="LogicTest" tests="1" failures="0" errors="0" time="0.005">
2374    <testcase name="NonContradiction" status="run" time="0.005" classname="">
2375    </testcase>
2376  </testsuite>
2377</testsuites>
2378```
2379
2380Things to note:
2381
2382*   The `tests` attribute of a `<testsuites>` or `<testsuite>` element tells how
2383    many test functions the googletest program or test suite contains, while the
2384    `failures` attribute tells how many of them failed.
2385
2386*   The `time` attribute expresses the duration of the test, test suite, or
2387    entire test program in seconds.
2388
2389*   The `timestamp` attribute records the local date and time of the test
2390    execution.
2391
2392*   Each `<failure>` element corresponds to a single failed googletest
2393    assertion.
2394
2395#### Generating a JSON Report
2396
2397googletest can also emit a JSON report as an alternative format to XML. To
2398generate the JSON report, set the `GTEST_OUTPUT` environment variable or the
2399`--gtest_output` flag to the string `"json:path_to_output_file"`, which will
2400create the file at the given location. You can also just use the string
2401`"json"`, in which case the output can be found in the `test_detail.json` file
2402in the current directory.
2403
2404The report format conforms to the following JSON Schema:
2405
2406```json
2407{
2408  "$schema": "http://json-schema.org/schema#",
2409  "type": "object",
2410  "definitions": {
2411    "TestCase": {
2412      "type": "object",
2413      "properties": {
2414        "name": { "type": "string" },
2415        "tests": { "type": "integer" },
2416        "failures": { "type": "integer" },
2417        "disabled": { "type": "integer" },
2418        "time": { "type": "string" },
2419        "testsuite": {
2420          "type": "array",
2421          "items": {
2422            "$ref": "#/definitions/TestInfo"
2423          }
2424        }
2425      }
2426    },
2427    "TestInfo": {
2428      "type": "object",
2429      "properties": {
2430        "name": { "type": "string" },
2431        "status": {
2432          "type": "string",
2433          "enum": ["RUN", "NOTRUN"]
2434        },
2435        "time": { "type": "string" },
2436        "classname": { "type": "string" },
2437        "failures": {
2438          "type": "array",
2439          "items": {
2440            "$ref": "#/definitions/Failure"
2441          }
2442        }
2443      }
2444    },
2445    "Failure": {
2446      "type": "object",
2447      "properties": {
2448        "failures": { "type": "string" },
2449        "type": { "type": "string" }
2450      }
2451    }
2452  },
2453  "properties": {
2454    "tests": { "type": "integer" },
2455    "failures": { "type": "integer" },
2456    "disabled": { "type": "integer" },
2457    "errors": { "type": "integer" },
2458    "timestamp": {
2459      "type": "string",
2460      "format": "date-time"
2461    },
2462    "time": { "type": "string" },
2463    "name": { "type": "string" },
2464    "testsuites": {
2465      "type": "array",
2466      "items": {
2467        "$ref": "#/definitions/TestCase"
2468      }
2469    }
2470  }
2471}
2472```
2473
2474The report uses the format that conforms to the following Proto3 using the
2475[JSON encoding](https://developers.google.com/protocol-buffers/docs/proto3#json):
2476
2477```proto
2478syntax = "proto3";
2479
2480package googletest;
2481
2482import "google/protobuf/timestamp.proto";
2483import "google/protobuf/duration.proto";
2484
2485message UnitTest {
2486  int32 tests = 1;
2487  int32 failures = 2;
2488  int32 disabled = 3;
2489  int32 errors = 4;
2490  google.protobuf.Timestamp timestamp = 5;
2491  google.protobuf.Duration time = 6;
2492  string name = 7;
2493  repeated TestCase testsuites = 8;
2494}
2495
2496message TestCase {
2497  string name = 1;
2498  int32 tests = 2;
2499  int32 failures = 3;
2500  int32 disabled = 4;
2501  int32 errors = 5;
2502  google.protobuf.Duration time = 6;
2503  repeated TestInfo testsuite = 7;
2504}
2505
2506message TestInfo {
2507  string name = 1;
2508  enum Status {
2509    RUN = 0;
2510    NOTRUN = 1;
2511  }
2512  Status status = 2;
2513  google.protobuf.Duration time = 3;
2514  string classname = 4;
2515  message Failure {
2516    string failures = 1;
2517    string type = 2;
2518  }
2519  repeated Failure failures = 5;
2520}
2521```
2522
2523For instance, the following program
2524
2525```c++
2526TEST(MathTest, Addition) { ... }
2527TEST(MathTest, Subtraction) { ... }
2528TEST(LogicTest, NonContradiction) { ... }
2529```
2530
2531could generate this report:
2532
2533```json
2534{
2535  "tests": 3,
2536  "failures": 1,
2537  "errors": 0,
2538  "time": "0.035s",
2539  "timestamp": "2011-10-31T18:52:42Z",
2540  "name": "AllTests",
2541  "testsuites": [
2542    {
2543      "name": "MathTest",
2544      "tests": 2,
2545      "failures": 1,
2546      "errors": 0,
2547      "time": "0.015s",
2548      "testsuite": [
2549        {
2550          "name": "Addition",
2551          "status": "RUN",
2552          "time": "0.007s",
2553          "classname": "",
2554          "failures": [
2555            {
2556              "message": "Value of: add(1, 1)\n  Actual: 3\nExpected: 2",
2557              "type": ""
2558            },
2559            {
2560              "message": "Value of: add(1, -1)\n  Actual: 1\nExpected: 0",
2561              "type": ""
2562            }
2563          ]
2564        },
2565        {
2566          "name": "Subtraction",
2567          "status": "RUN",
2568          "time": "0.005s",
2569          "classname": ""
2570        }
2571      ]
2572    },
2573    {
2574      "name": "LogicTest",
2575      "tests": 1,
2576      "failures": 0,
2577      "errors": 0,
2578      "time": "0.005s",
2579      "testsuite": [
2580        {
2581          "name": "NonContradiction",
2582          "status": "RUN",
2583          "time": "0.005s",
2584          "classname": ""
2585        }
2586      ]
2587    }
2588  ]
2589}
2590```
2591
2592{: .callout .important}
2593IMPORTANT: The exact format of the JSON document is subject to change.
2594
2595### Controlling How Failures Are Reported
2596
2597#### Detecting Test Premature Exit
2598
2599Google Test implements the _premature-exit-file_ protocol for test runners
2600to catch any kind of unexpected exits of test programs. Upon start,
2601Google Test creates the file which will be automatically deleted after
2602all work has been finished. Then, the test runner can check if this file
2603exists. In case the file remains undeleted, the inspected test has exited
2604prematurely.
2605
2606This feature is enabled only if the `TEST_PREMATURE_EXIT_FILE` environment
2607variable has been set.
2608
2609#### Turning Assertion Failures into Break-Points
2610
2611When running test programs under a debugger, it's very convenient if the
2612debugger can catch an assertion failure and automatically drop into interactive
2613mode. googletest's *break-on-failure* mode supports this behavior.
2614
2615To enable it, set the `GTEST_BREAK_ON_FAILURE` environment variable to a value
2616other than `0`. Alternatively, you can use the `--gtest_break_on_failure`
2617command line flag.
2618
2619#### Disabling Catching Test-Thrown Exceptions
2620
2621googletest can be used either with or without exceptions enabled. If a test
2622throws a C++ exception or (on Windows) a structured exception (SEH), by default
2623googletest catches it, reports it as a test failure, and continues with the next
2624test method. This maximizes the coverage of a test run. Also, on Windows an
2625uncaught exception will cause a pop-up window, so catching the exceptions allows
2626you to run the tests automatically.
2627
2628When debugging the test failures, however, you may instead want the exceptions
2629to be handled by the debugger, such that you can examine the call stack when an
2630exception is thrown. To achieve that, set the `GTEST_CATCH_EXCEPTIONS`
2631environment variable to `0`, or use the `--gtest_catch_exceptions=0` flag when
2632running the tests.
2633
2634### Sanitizer Integration
2635
2636The
2637[Undefined Behavior Sanitizer](https://clang.llvm.org/docs/UndefinedBehaviorSanitizer.html),
2638[Address Sanitizer](https://github.com/google/sanitizers/wiki/AddressSanitizer),
2639and
2640[Thread Sanitizer](https://github.com/google/sanitizers/wiki/ThreadSanitizerCppManual)
2641all provide weak functions that you can override to trigger explicit failures
2642when they detect sanitizer errors, such as creating a reference from `nullptr`.
2643To override these functions, place definitions for them in a source file that
2644you compile as part of your main binary:
2645
2646```
2647extern "C" {
2648void __ubsan_on_report() {
2649  FAIL() << "Encountered an undefined behavior sanitizer error";
2650}
2651void __asan_on_error() {
2652  FAIL() << "Encountered an address sanitizer error";
2653}
2654void __tsan_on_report() {
2655  FAIL() << "Encountered a thread sanitizer error";
2656}
2657}  // extern "C"
2658```
2659
2660After compiling your project with one of the sanitizers enabled, if a particular
2661test triggers a sanitizer error, googletest will report that it failed.
2662