1 2 3Now that you have read [Primer](V1_5_Primer.md) and learned how to write tests 4using Google Test, it's time to learn some new tricks. This document 5will show you more assertions as well as how to construct complex 6failure messages, propagate fatal failures, reuse and speed up your 7test fixtures, and use various flags with your tests. 8 9# More Assertions # 10 11This section covers some less frequently used, but still significant, 12assertions. 13 14## Explicit Success and Failure ## 15 16These three assertions do not actually test a value or expression. Instead, 17they generate a success or failure directly. Like the macros that actually 18perform a test, you may stream a custom failure message into the them. 19 20| `SUCCEED();` | 21|:-------------| 22 23Generates a success. This does NOT make the overall test succeed. A test is 24considered successful only if none of its assertions fail during its execution. 25 26Note: `SUCCEED()` is purely documentary and currently doesn't generate any 27user-visible output. However, we may add `SUCCEED()` messages to Google Test's 28output in the future. 29 30| `FAIL();` | `ADD_FAILURE();` | 31|:-----------|:-----------------| 32 33`FAIL*` generates a fatal failure while `ADD_FAILURE*` generates a nonfatal 34failure. These are useful when control flow, rather than a Boolean expression, 35deteremines the test's success or failure. For example, you might want to write 36something like: 37 38``` 39switch(expression) { 40 case 1: ... some checks ... 41 case 2: ... some other checks 42 ... 43 default: FAIL() << "We shouldn't get here."; 44} 45``` 46 47_Availability_: Linux, Windows, Mac. 48 49## Exception Assertions ## 50 51These are for verifying that a piece of code throws (or does not 52throw) an exception of the given type: 53 54| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 55|:--------------------|:-----------------------|:-------------| 56| `ASSERT_THROW(`_statement_, _exception\_type_`);` | `EXPECT_THROW(`_statement_, _exception\_type_`);` | _statement_ throws an exception of the given type | 57| `ASSERT_ANY_THROW(`_statement_`);` | `EXPECT_ANY_THROW(`_statement_`);` | _statement_ throws an exception of any type | 58| `ASSERT_NO_THROW(`_statement_`);` | `EXPECT_NO_THROW(`_statement_`);` | _statement_ doesn't throw any exception | 59 60Examples: 61 62``` 63ASSERT_THROW(Foo(5), bar_exception); 64 65EXPECT_NO_THROW({ 66 int n = 5; 67 Bar(&n); 68}); 69``` 70 71_Availability_: Linux, Windows, Mac; since version 1.1.0. 72 73## Predicate Assertions for Better Error Messages ## 74 75Even though Google Test has a rich set of assertions, they can never be 76complete, as it's impossible (nor a good idea) to anticipate all the scenarios 77a user might run into. Therefore, sometimes a user has to use `EXPECT_TRUE()` 78to check a complex expression, for lack of a better macro. This has the problem 79of not showing you the values of the parts of the expression, making it hard to 80understand what went wrong. As a workaround, some users choose to construct the 81failure message by themselves, streaming it into `EXPECT_TRUE()`. However, this 82is awkward especially when the expression has side-effects or is expensive to 83evaluate. 84 85Google Test gives you three different options to solve this problem: 86 87### Using an Existing Boolean Function ### 88 89If you already have a function or a functor that returns `bool` (or a type 90that can be implicitly converted to `bool`), you can use it in a _predicate 91assertion_ to get the function arguments printed for free: 92 93| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 94|:--------------------|:-----------------------|:-------------| 95| `ASSERT_PRED1(`_pred1, val1_`);` | `EXPECT_PRED1(`_pred1, val1_`);` | _pred1(val1)_ returns true | 96| `ASSERT_PRED2(`_pred2, val1, val2_`);` | `EXPECT_PRED2(`_pred2, val1, val2_`);` | _pred2(val1, val2)_ returns true | 97| ... | ... | ... | 98 99In the above, _predn_ is an _n_-ary predicate function or functor, where 100_val1_, _val2_, ..., and _valn_ are its arguments. The assertion succeeds 101if the predicate returns `true` when applied to the given arguments, and fails 102otherwise. When the assertion fails, it prints the value of each argument. In 103either case, the arguments are evaluated exactly once. 104 105Here's an example. Given 106 107``` 108// Returns true iff m and n have no common divisors except 1. 109bool MutuallyPrime(int m, int n) { ... } 110const int a = 3; 111const int b = 4; 112const int c = 10; 113``` 114 115the assertion `EXPECT_PRED2(MutuallyPrime, a, b);` will succeed, while the 116assertion `EXPECT_PRED2(MutuallyPrime, b, c);` will fail with the message 117 118<pre> 119!MutuallyPrime(b, c) is false, where<br> 120b is 4<br> 121c is 10<br> 122</pre> 123 124**Notes:** 125 126 1. If you see a compiler error "no matching function to call" when using `ASSERT_PRED*` or `EXPECT_PRED*`, please see [this](http://code.google.com/p/googletest/wiki/V1_5_FAQ#The_compiler_complains_%22no_matching_function_to_call%22) for how to resolve it. 127 1. Currently we only provide predicate assertions of arity <= 5. If you need a higher-arity assertion, let us know. 128 129_Availability_: Linux, Windows, Mac 130 131### Using a Function That Returns an AssertionResult ### 132 133While `EXPECT_PRED*()` and friends are handy for a quick job, the 134syntax is not satisfactory: you have to use different macros for 135different arities, and it feels more like Lisp than C++. The 136`::testing::AssertionResult` class solves this problem. 137 138An `AssertionResult` object represents the result of an assertion 139(whether it's a success or a failure, and an associated message). You 140can create an `AssertionResult` using one of these factory 141functions: 142 143``` 144namespace testing { 145 146// Returns an AssertionResult object to indicate that an assertion has 147// succeeded. 148AssertionResult AssertionSuccess(); 149 150// Returns an AssertionResult object to indicate that an assertion has 151// failed. 152AssertionResult AssertionFailure(); 153 154} 155``` 156 157You can then use the `<<` operator to stream messages to the 158`AssertionResult` object. 159 160To provide more readable messages in Boolean assertions 161(e.g. `EXPECT_TRUE()`), write a predicate function that returns 162`AssertionResult` instead of `bool`. For example, if you define 163`IsEven()` as: 164 165``` 166::testing::AssertionResult IsEven(int n) { 167 if ((n % 2) == 0) 168 return ::testing::AssertionSuccess(); 169 else 170 return ::testing::AssertionFailure() << n << " is odd"; 171} 172``` 173 174instead of: 175 176``` 177bool IsEven(int n) { 178 return (n % 2) == 0; 179} 180``` 181 182the failed assertion `EXPECT_TRUE(IsEven(Fib(4)))` will print: 183 184<pre> 185Value of: !IsEven(Fib(4))<br> 186Actual: false (*3 is odd*)<br> 187Expected: true<br> 188</pre> 189 190instead of a more opaque 191 192<pre> 193Value of: !IsEven(Fib(4))<br> 194Actual: false<br> 195Expected: true<br> 196</pre> 197 198If you want informative messages in `EXPECT_FALSE` and `ASSERT_FALSE` 199as well, and are fine with making the predicate slower in the success 200case, you can supply a success message: 201 202``` 203::testing::AssertionResult IsEven(int n) { 204 if ((n % 2) == 0) 205 return ::testing::AssertionSuccess() << n << " is even"; 206 else 207 return ::testing::AssertionFailure() << n << " is odd"; 208} 209``` 210 211Then the statement `EXPECT_FALSE(IsEven(Fib(6)))` will print 212 213<pre> 214Value of: !IsEven(Fib(6))<br> 215Actual: true (8 is even)<br> 216Expected: false<br> 217</pre> 218 219_Availability_: Linux, Windows, Mac; since version 1.4.1. 220 221### Using a Predicate-Formatter ### 222 223If you find the default message generated by `(ASSERT|EXPECT)_PRED*` and 224`(ASSERT|EXPECT)_(TRUE|FALSE)` unsatisfactory, or some arguments to your 225predicate do not support streaming to `ostream`, you can instead use the 226following _predicate-formatter assertions_ to _fully_ customize how the 227message is formatted: 228 229| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 230|:--------------------|:-----------------------|:-------------| 231| `ASSERT_PRED_FORMAT1(`_pred\_format1, val1_`);` | `EXPECT_PRED_FORMAT1(`_pred\_format1, val1_`); | _pred\_format1(val1)_ is successful | 232| `ASSERT_PRED_FORMAT2(`_pred\_format2, val1, val2_`);` | `EXPECT_PRED_FORMAT2(`_pred\_format2, val1, val2_`);` | _pred\_format2(val1, val2)_ is successful | 233| `...` | `...` | `...` | 234 235The difference between this and the previous two groups of macros is that instead of 236a predicate, `(ASSERT|EXPECT)_PRED_FORMAT*` take a _predicate-formatter_ 237(_pred\_formatn_), which is a function or functor with the signature: 238 239`::testing::AssertionResult PredicateFormattern(const char* `_expr1_`, const char* `_expr2_`, ... const char* `_exprn_`, T1 `_val1_`, T2 `_val2_`, ... Tn `_valn_`);` 240 241where _val1_, _val2_, ..., and _valn_ are the values of the predicate 242arguments, and _expr1_, _expr2_, ..., and _exprn_ are the corresponding 243expressions as they appear in the source code. The types `T1`, `T2`, ..., and 244`Tn` can be either value types or reference types. For example, if an 245argument has type `Foo`, you can declare it as either `Foo` or `const Foo&`, 246whichever is appropriate. 247 248A predicate-formatter returns a `::testing::AssertionResult` object to indicate 249whether the assertion has succeeded or not. The only way to create such an 250object is to call one of these factory functions: 251 252As an example, let's improve the failure message in the previous example, which uses `EXPECT_PRED2()`: 253 254``` 255// Returns the smallest prime common divisor of m and n, 256// or 1 when m and n are mutually prime. 257int SmallestPrimeCommonDivisor(int m, int n) { ... } 258 259// A predicate-formatter for asserting that two integers are mutually prime. 260::testing::AssertionResult AssertMutuallyPrime(const char* m_expr, 261 const char* n_expr, 262 int m, 263 int n) { 264 if (MutuallyPrime(m, n)) 265 return ::testing::AssertionSuccess(); 266 267 return ::testing::AssertionFailure() 268 << m_expr << " and " << n_expr << " (" << m << " and " << n 269 << ") are not mutually prime, " << "as they have a common divisor " 270 << SmallestPrimeCommonDivisor(m, n); 271} 272``` 273 274With this predicate-formatter, we can use 275 276``` 277EXPECT_PRED_FORMAT2(AssertMutuallyPrime, b, c); 278``` 279 280to generate the message 281 282<pre> 283b and c (4 and 10) are not mutually prime, as they have a common divisor 2.<br> 284</pre> 285 286As you may have realized, many of the assertions we introduced earlier are 287special cases of `(EXPECT|ASSERT)_PRED_FORMAT*`. In fact, most of them are 288indeed defined using `(EXPECT|ASSERT)_PRED_FORMAT*`. 289 290_Availability_: Linux, Windows, Mac. 291 292 293## Floating-Point Comparison ## 294 295Comparing floating-point numbers is tricky. Due to round-off errors, it is 296very unlikely that two floating-points will match exactly. Therefore, 297`ASSERT_EQ` 's naive comparison usually doesn't work. And since floating-points 298can have a wide value range, no single fixed error bound works. It's better to 299compare by a fixed relative error bound, except for values close to 0 due to 300the loss of precision there. 301 302In general, for floating-point comparison to make sense, the user needs to 303carefully choose the error bound. If they don't want or care to, comparing in 304terms of Units in the Last Place (ULPs) is a good default, and Google Test 305provides assertions to do this. Full details about ULPs are quite long; if you 306want to learn more, see 307[this article on float comparison](http://www.cygnus-software.com/papers/comparingfloats/comparingfloats.htm). 308 309### Floating-Point Macros ### 310 311| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 312|:--------------------|:-----------------------|:-------------| 313| `ASSERT_FLOAT_EQ(`_expected, actual_`);` | `EXPECT_FLOAT_EQ(`_expected, actual_`);` | the two `float` values are almost equal | 314| `ASSERT_DOUBLE_EQ(`_expected, actual_`);` | `EXPECT_DOUBLE_EQ(`_expected, actual_`);` | the two `double` values are almost equal | 315 316By "almost equal", we mean the two values are within 4 ULP's from each 317other. 318 319The following assertions allow you to choose the acceptable error bound: 320 321| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 322|:--------------------|:-----------------------|:-------------| 323| `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 | 324 325_Availability_: Linux, Windows, Mac. 326 327### Floating-Point Predicate-Format Functions ### 328 329Some floating-point operations are useful, but not that often used. In order 330to avoid an explosion of new macros, we provide them as predicate-format 331functions that can be used in predicate assertion macros (e.g. 332`EXPECT_PRED_FORMAT2`, etc). 333 334``` 335EXPECT_PRED_FORMAT2(::testing::FloatLE, val1, val2); 336EXPECT_PRED_FORMAT2(::testing::DoubleLE, val1, val2); 337``` 338 339Verifies that _val1_ is less than, or almost equal to, _val2_. You can 340replace `EXPECT_PRED_FORMAT2` in the above table with `ASSERT_PRED_FORMAT2`. 341 342_Availability_: Linux, Windows, Mac. 343 344## Windows HRESULT assertions ## 345 346These assertions test for `HRESULT` success or failure. 347 348| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 349|:--------------------|:-----------------------|:-------------| 350| `ASSERT_HRESULT_SUCCEEDED(`_expression_`);` | `EXPECT_HRESULT_SUCCEEDED(`_expression_`);` | _expression_ is a success `HRESULT` | 351| `ASSERT_HRESULT_FAILED(`_expression_`);` | `EXPECT_HRESULT_FAILED(`_expression_`);` | _expression_ is a failure `HRESULT` | 352 353The generated output contains the human-readable error message 354associated with the `HRESULT` code returned by _expression_. 355 356You might use them like this: 357 358``` 359CComPtr shell; 360ASSERT_HRESULT_SUCCEEDED(shell.CoCreateInstance(L"Shell.Application")); 361CComVariant empty; 362ASSERT_HRESULT_SUCCEEDED(shell->ShellExecute(CComBSTR(url), empty, empty, empty, empty)); 363``` 364 365_Availability_: Windows. 366 367## Type Assertions ## 368 369You can call the function 370``` 371::testing::StaticAssertTypeEq<T1, T2>(); 372``` 373to assert that types `T1` and `T2` are the same. The function does 374nothing if the assertion is satisfied. If the types are different, 375the function call will fail to compile, and the compiler error message 376will likely (depending on the compiler) show you the actual values of 377`T1` and `T2`. This is mainly useful inside template code. 378 379_Caveat:_ When used inside a member function of a class template or a 380function template, `StaticAssertTypeEq<T1, T2>()` is effective _only if_ 381the function is instantiated. For example, given: 382``` 383template <typename T> class Foo { 384 public: 385 void Bar() { ::testing::StaticAssertTypeEq<int, T>(); } 386}; 387``` 388the code: 389``` 390void Test1() { Foo<bool> foo; } 391``` 392will _not_ generate a compiler error, as `Foo<bool>::Bar()` is never 393actually instantiated. Instead, you need: 394``` 395void Test2() { Foo<bool> foo; foo.Bar(); } 396``` 397to cause a compiler error. 398 399_Availability:_ Linux, Windows, Mac; since version 1.3.0. 400 401## Assertion Placement ## 402 403You can use assertions in any C++ function. In particular, it doesn't 404have to be a method of the test fixture class. The one constraint is 405that assertions that generate a fatal failure (`FAIL*` and `ASSERT_*`) 406can only be used in void-returning functions. This is a consequence of 407Google Test not using exceptions. By placing it in a non-void function 408you'll get a confusing compile error like 409`"error: void value not ignored as it ought to be"`. 410 411If you need to use assertions in a function that returns non-void, one option 412is to make the function return the value in an out parameter instead. For 413example, you can rewrite `T2 Foo(T1 x)` to `void Foo(T1 x, T2* result)`. You 414need to make sure that `*result` contains some sensible value even when the 415function returns prematurely. As the function now returns `void`, you can use 416any assertion inside of it. 417 418If changing the function's type is not an option, you should just use 419assertions that generate non-fatal failures, such as `ADD_FAILURE*` and 420`EXPECT_*`. 421 422_Note_: Constructors and destructors are not considered void-returning 423functions, according to the C++ language specification, and so you may not use 424fatal assertions in them. You'll get a compilation error if you try. A simple 425workaround is to transfer the entire body of the constructor or destructor to a 426private void-returning method. However, you should be aware that a fatal 427assertion failure in a constructor does not terminate the current test, as your 428intuition might suggest; it merely returns from the constructor early, possibly 429leaving your object in a partially-constructed state. Likewise, a fatal 430assertion failure in a destructor may leave your object in a 431partially-destructed state. Use assertions carefully in these situations! 432 433# Death Tests # 434 435In many applications, there are assertions that can cause application failure 436if a condition is not met. These sanity checks, which ensure that the program 437is in a known good state, are there to fail at the earliest possible time after 438some program state is corrupted. If the assertion checks the wrong condition, 439then the program may proceed in an erroneous state, which could lead to memory 440corruption, security holes, or worse. Hence it is vitally important to test 441that such assertion statements work as expected. 442 443Since these precondition checks cause the processes to die, we call such tests 444_death tests_. More generally, any test that checks that a program terminates 445in an expected fashion is also a death test. 446 447If you want to test `EXPECT_*()/ASSERT_*()` failures in your test code, see [Catching Failures](#Catching_Failures.md). 448 449## How to Write a Death Test ## 450 451Google Test has the following macros to support death tests: 452 453| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 454|:--------------------|:-----------------------|:-------------| 455| `ASSERT_DEATH(`_statement, regex_`); | `EXPECT_DEATH(`_statement, regex_`); | _statement_ crashes with the given error | 456| `ASSERT_DEATH_IF_SUPPORTED(`_statement, regex_`); | `EXPECT_DEATH_IF_SUPPORTED(`_statement, regex_`); | if death tests are supported, verifies that _statement_ crashes with the given error; otherwise verifies nothing | 457| `ASSERT_EXIT(`_statement, predicate, regex_`); | `EXPECT_EXIT(`_statement, predicate, regex_`); |_statement_ exits with the given error and its exit code matches _predicate_ | 458 459where _statement_ is a statement that is expected to cause the process to 460die, _predicate_ is a function or function object that evaluates an integer 461exit status, and _regex_ is a regular expression that the stderr output of 462_statement_ is expected to match. Note that _statement_ can be _any valid 463statement_ (including _compound statement_) and doesn't have to be an 464expression. 465 466As usual, the `ASSERT` variants abort the current test function, while the 467`EXPECT` variants do not. 468 469**Note:** We use the word "crash" here to mean that the process 470terminates with a _non-zero_ exit status code. There are two 471possibilities: either the process has called `exit()` or `_exit()` 472with a non-zero value, or it may be killed by a signal. 473 474This means that if _statement_ terminates the process with a 0 exit 475code, it is _not_ considered a crash by `EXPECT_DEATH`. Use 476`EXPECT_EXIT` instead if this is the case, or if you want to restrict 477the exit code more precisely. 478 479A predicate here must accept an `int` and return a `bool`. The death test 480succeeds only if the predicate returns `true`. Google Test defines a few 481predicates that handle the most common cases: 482 483``` 484::testing::ExitedWithCode(exit_code) 485``` 486 487This expression is `true` if the program exited normally with the given exit 488code. 489 490``` 491::testing::KilledBySignal(signal_number) // Not available on Windows. 492``` 493 494This expression is `true` if the program was killed by the given signal. 495 496The `*_DEATH` macros are convenient wrappers for `*_EXIT` that use a predicate 497that verifies the process' exit code is non-zero. 498 499Note that a death test only cares about three things: 500 501 1. does _statement_ abort or exit the process? 502 1. (in the case of `ASSERT_EXIT` and `EXPECT_EXIT`) does the exit status satisfy _predicate_? Or (in the case of `ASSERT_DEATH` and `EXPECT_DEATH`) is the exit status non-zero? And 503 1. does the stderr output match _regex_? 504 505In particular, if _statement_ generates an `ASSERT_*` or `EXPECT_*` failure, it will **not** cause the death test to fail, as Google Test assertions don't abort the process. 506 507To write a death test, simply use one of the above macros inside your test 508function. For example, 509 510``` 511TEST(My*DeathTest*, Foo) { 512 // This death test uses a compound statement. 513 ASSERT_DEATH({ int n = 5; Foo(&n); }, "Error on line .* of Foo()"); 514} 515TEST(MyDeathTest, NormalExit) { 516 EXPECT_EXIT(NormalExit(), ::testing::ExitedWithCode(0), "Success"); 517} 518TEST(MyDeathTest, KillMyself) { 519 EXPECT_EXIT(KillMyself(), ::testing::KilledBySignal(SIGKILL), "Sending myself unblockable signal"); 520} 521``` 522 523verifies that: 524 525 * calling `Foo(5)` causes the process to die with the given error message, 526 * calling `NormalExit()` causes the process to print `"Success"` to stderr and exit with exit code 0, and 527 * calling `KillMyself()` kills the process with signal `SIGKILL`. 528 529The test function body may contain other assertions and statements as well, if 530necessary. 531 532_Important:_ We strongly recommend you to follow the convention of naming your 533test case (not test) `*DeathTest` when it contains a death test, as 534demonstrated in the above example. The `Death Tests And Threads` section below 535explains why. 536 537If a test fixture class is shared by normal tests and death tests, you 538can use typedef to introduce an alias for the fixture class and avoid 539duplicating its code: 540``` 541class FooTest : public ::testing::Test { ... }; 542 543typedef FooTest FooDeathTest; 544 545TEST_F(FooTest, DoesThis) { 546 // normal test 547} 548 549TEST_F(FooDeathTest, DoesThat) { 550 // death test 551} 552``` 553 554_Availability:_ Linux, Windows (requires MSVC 8.0 or above), Cygwin, and Mac (the latter three are supported since v1.3.0). `(ASSERT|EXPECT)_DEATH_IF_SUPPORTED` are new in v1.4.0. 555 556## Regular Expression Syntax ## 557 558On POSIX systems (e.g. Linux, Cygwin, and Mac), Google Test uses the 559[POSIX extended regular expression](http://www.opengroup.org/onlinepubs/009695399/basedefs/xbd_chap09.html#tag_09_04) 560syntax in death tests. To learn about this syntax, you may want to read this [Wikipedia entry](http://en.wikipedia.org/wiki/Regular_expression#POSIX_Extended_Regular_Expressions). 561 562On Windows, Google Test uses its own simple regular expression 563implementation. It lacks many features you can find in POSIX extended 564regular expressions. For example, we don't support union (`"x|y"`), 565grouping (`"(xy)"`), brackets (`"[xy]"`), and repetition count 566(`"x{5,7}"`), among others. Below is what we do support (`A` denotes a 567literal character, period (`.`), or a single `\\` escape sequence; `x` 568and `y` denote regular expressions.): 569 570| `c` | matches any literal character `c` | 571|:----|:----------------------------------| 572| `\\d` | matches any decimal digit | 573| `\\D` | matches any character that's not a decimal digit | 574| `\\f` | matches `\f` | 575| `\\n` | matches `\n` | 576| `\\r` | matches `\r` | 577| `\\s` | matches any ASCII whitespace, including `\n` | 578| `\\S` | matches any character that's not a whitespace | 579| `\\t` | matches `\t` | 580| `\\v` | matches `\v` | 581| `\\w` | matches any letter, `_`, or decimal digit | 582| `\\W` | matches any character that `\\w` doesn't match | 583| `\\c` | matches any literal character `c`, which must be a punctuation | 584| `.` | matches any single character except `\n` | 585| `A?` | matches 0 or 1 occurrences of `A` | 586| `A*` | matches 0 or many occurrences of `A` | 587| `A+` | matches 1 or many occurrences of `A` | 588| `^` | matches the beginning of a string (not that of each line) | 589| `$` | matches the end of a string (not that of each line) | 590| `xy` | matches `x` followed by `y` | 591 592To help you determine which capability is available on your system, 593Google Test defines macro `GTEST_USES_POSIX_RE=1` when it uses POSIX 594extended regular expressions, or `GTEST_USES_SIMPLE_RE=1` when it uses 595the simple version. If you want your death tests to work in both 596cases, you can either `#if` on these macros or use the more limited 597syntax only. 598 599## How It Works ## 600 601Under the hood, `ASSERT_EXIT()` spawns a new process and executes the 602death test statement in that process. The details of of how precisely 603that happens depend on the platform and the variable 604`::testing::GTEST_FLAG(death_test_style)` (which is initialized from the 605command-line flag `--gtest_death_test_style`). 606 607 * On POSIX systems, `fork()` (or `clone()` on Linux) is used to spawn the child, after which: 608 * If the variable's value is `"fast"`, the death test statement is immediately executed. 609 * If the variable's value is `"threadsafe"`, the child process re-executes the unit test binary just as it was originally invoked, but with some extra flags to cause just the single death test under consideration to be run. 610 * On Windows, the child is spawned using the `CreateProcess()` API, and re-executes the binary to cause just the single death test under consideration to be run - much like the `threadsafe` mode on POSIX. 611 612Other values for the variable are illegal and will cause the death test to 613fail. Currently, the flag's default value is `"fast"`. However, we reserve the 614right to change it in the future. Therefore, your tests should not depend on 615this. 616 617In either case, the parent process waits for the child process to complete, and checks that 618 619 1. the child's exit status satisfies the predicate, and 620 1. the child's stderr matches the regular expression. 621 622If the death test statement runs to completion without dying, the child 623process will nonetheless terminate, and the assertion fails. 624 625## Death Tests And Threads ## 626 627The reason for the two death test styles has to do with thread safety. Due to 628well-known problems with forking in the presence of threads, death tests should 629be run in a single-threaded context. Sometimes, however, it isn't feasible to 630arrange that kind of environment. For example, statically-initialized modules 631may start threads before main is ever reached. Once threads have been created, 632it may be difficult or impossible to clean them up. 633 634Google Test has three features intended to raise awareness of threading issues. 635 636 1. A warning is emitted if multiple threads are running when a death test is encountered. 637 1. Test cases with a name ending in "DeathTest" are run before all other tests. 638 1. It uses `clone()` instead of `fork()` to spawn the child process on Linux (`clone()` is not available on Cygwin and Mac), as `fork()` is more likely to cause the child to hang when the parent process has multiple threads. 639 640It's perfectly fine to create threads inside a death test statement; they are 641executed in a separate process and cannot affect the parent. 642 643## Death Test Styles ## 644 645The "threadsafe" death test style was introduced in order to help mitigate the 646risks of testing in a possibly multithreaded environment. It trades increased 647test execution time (potentially dramatically so) for improved thread safety. 648We suggest using the faster, default "fast" style unless your test has specific 649problems with it. 650 651You can choose a particular style of death tests by setting the flag 652programmatically: 653 654``` 655::testing::FLAGS_gtest_death_test_style = "threadsafe"; 656``` 657 658You can do this in `main()` to set the style for all death tests in the 659binary, or in individual tests. Recall that flags are saved before running each 660test and restored afterwards, so you need not do that yourself. For example: 661 662``` 663TEST(MyDeathTest, TestOne) { 664 ::testing::FLAGS_gtest_death_test_style = "threadsafe"; 665 // This test is run in the "threadsafe" style: 666 ASSERT_DEATH(ThisShouldDie(), ""); 667} 668 669TEST(MyDeathTest, TestTwo) { 670 // This test is run in the "fast" style: 671 ASSERT_DEATH(ThisShouldDie(), ""); 672} 673 674int main(int argc, char** argv) { 675 ::testing::InitGoogleTest(&argc, argv); 676 ::testing::FLAGS_gtest_death_test_style = "fast"; 677 return RUN_ALL_TESTS(); 678} 679``` 680 681## Caveats ## 682 683The _statement_ argument of `ASSERT_EXIT()` can be any valid C++ statement 684except that it can not return from the current function. This means 685_statement_ should not contain `return` or a macro that might return (e.g. 686`ASSERT_TRUE()` ). If _statement_ returns before it crashes, Google Test will 687print an error message, and the test will fail. 688 689Since _statement_ runs in the child process, any in-memory side effect (e.g. 690modifying a variable, releasing memory, etc) it causes will _not_ be observable 691in the parent process. In particular, if you release memory in a death test, 692your program will fail the heap check as the parent process will never see the 693memory reclaimed. To solve this problem, you can 694 695 1. try not to free memory in a death test; 696 1. free the memory again in the parent process; or 697 1. do not use the heap checker in your program. 698 699Due to an implementation detail, you cannot place multiple death test 700assertions on the same line; otherwise, compilation will fail with an unobvious 701error message. 702 703Despite the improved thread safety afforded by the "threadsafe" style of death 704test, thread problems such as deadlock are still possible in the presence of 705handlers registered with `pthread_atfork(3)`. 706 707# Using Assertions in Sub-routines # 708 709## Adding Traces to Assertions ## 710 711If a test sub-routine is called from several places, when an assertion 712inside it fails, it can be hard to tell which invocation of the 713sub-routine the failure is from. You can alleviate this problem using 714extra logging or custom failure messages, but that usually clutters up 715your tests. A better solution is to use the `SCOPED_TRACE` macro: 716 717| `SCOPED_TRACE(`_message_`);` | 718|:-----------------------------| 719 720where _message_ can be anything streamable to `std::ostream`. This 721macro will cause the current file name, line number, and the given 722message to be added in every failure message. The effect will be 723undone when the control leaves the current lexical scope. 724 725For example, 726 727``` 72810: void Sub1(int n) { 72911: EXPECT_EQ(1, Bar(n)); 73012: EXPECT_EQ(2, Bar(n + 1)); 73113: } 73214: 73315: TEST(FooTest, Bar) { 73416: { 73517: SCOPED_TRACE("A"); // This trace point will be included in 73618: // every failure in this scope. 73719: Sub1(1); 73820: } 73921: // Now it won't. 74022: Sub1(9); 74123: } 742``` 743 744could result in messages like these: 745 746``` 747path/to/foo_test.cc:11: Failure 748Value of: Bar(n) 749Expected: 1 750 Actual: 2 751 Trace: 752path/to/foo_test.cc:17: A 753 754path/to/foo_test.cc:12: Failure 755Value of: Bar(n + 1) 756Expected: 2 757 Actual: 3 758``` 759 760Without the trace, it would've been difficult to know which invocation 761of `Sub1()` the two failures come from respectively. (You could add an 762extra message to each assertion in `Sub1()` to indicate the value of 763`n`, but that's tedious.) 764 765Some tips on using `SCOPED_TRACE`: 766 767 1. With a suitable message, it's often enough to use `SCOPED_TRACE` at the beginning of a sub-routine, instead of at each call site. 768 1. When calling sub-routines inside a loop, make the loop iterator part of the message in `SCOPED_TRACE` such that you can know which iteration the failure is from. 769 1. Sometimes the line number of the trace point is enough for identifying the particular invocation of a sub-routine. In this case, you don't have to choose a unique message for `SCOPED_TRACE`. You can simply use `""`. 770 1. You can use `SCOPED_TRACE` in an inner scope when there is one in the outer scope. In this case, all active trace points will be included in the failure messages, in reverse order they are encountered. 771 1. The trace dump is clickable in Emacs' compilation buffer - hit return on a line number and you'll be taken to that line in the source file! 772 773_Availability:_ Linux, Windows, Mac. 774 775## Propagating Fatal Failures ## 776 777A common pitfall when using `ASSERT_*` and `FAIL*` is not understanding that 778when they fail they only abort the _current function_, not the entire test. For 779example, the following test will segfault: 780``` 781void Subroutine() { 782 // Generates a fatal failure and aborts the current function. 783 ASSERT_EQ(1, 2); 784 // The following won't be executed. 785 ... 786} 787 788TEST(FooTest, Bar) { 789 Subroutine(); 790 // The intended behavior is for the fatal failure 791 // in Subroutine() to abort the entire test. 792 // The actual behavior: the function goes on after Subroutine() returns. 793 int* p = NULL; 794 *p = 3; // Segfault! 795} 796``` 797 798Since we don't use exceptions, it is technically impossible to 799implement the intended behavior here. To alleviate this, Google Test 800provides two solutions. You could use either the 801`(ASSERT|EXPECT)_NO_FATAL_FAILURE` assertions or the 802`HasFatalFailure()` function. They are described in the following two 803subsections. 804 805 806 807### Asserting on Subroutines ### 808 809As shown above, if your test calls a subroutine that has an `ASSERT_*` 810failure in it, the test will continue after the subroutine 811returns. This may not be what you want. 812 813Often people want fatal failures to propagate like exceptions. For 814that Google Test offers the following macros: 815 816| **Fatal assertion** | **Nonfatal assertion** | **Verifies** | 817|:--------------------|:-----------------------|:-------------| 818| `ASSERT_NO_FATAL_FAILURE(`_statement_`);` | `EXPECT_NO_FATAL_FAILURE(`_statement_`);` | _statement_ doesn't generate any new fatal failures in the current thread. | 819 820Only failures in the thread that executes the assertion are checked to 821determine the result of this type of assertions. If _statement_ 822creates new threads, failures in these threads are ignored. 823 824Examples: 825 826``` 827ASSERT_NO_FATAL_FAILURE(Foo()); 828 829int i; 830EXPECT_NO_FATAL_FAILURE({ 831 i = Bar(); 832}); 833``` 834 835_Availability:_ Linux, Windows, Mac. Assertions from multiple threads 836are currently not supported. 837 838### Checking for Failures in the Current Test ### 839 840`HasFatalFailure()` in the `::testing::Test` class returns `true` if an 841assertion in the current test has suffered a fatal failure. This 842allows functions to catch fatal failures in a sub-routine and return 843early. 844 845``` 846class Test { 847 public: 848 ... 849 static bool HasFatalFailure(); 850}; 851``` 852 853The typical usage, which basically simulates the behavior of a thrown 854exception, is: 855 856``` 857TEST(FooTest, Bar) { 858 Subroutine(); 859 // Aborts if Subroutine() had a fatal failure. 860 if (HasFatalFailure()) 861 return; 862 // The following won't be executed. 863 ... 864} 865``` 866 867If `HasFatalFailure()` is used outside of `TEST()` , `TEST_F()` , or a test 868fixture, you must add the `::testing::Test::` prefix, as in: 869 870``` 871if (::testing::Test::HasFatalFailure()) 872 return; 873``` 874 875Similarly, `HasNonfatalFailure()` returns `true` if the current test 876has at least one non-fatal failure, and `HasFailure()` returns `true` 877if the current test has at least one failure of either kind. 878 879_Availability:_ Linux, Windows, Mac. `HasNonfatalFailure()` and 880`HasFailure()` are available since version 1.4.0. 881 882# Logging Additional Information # 883 884In your test code, you can call `RecordProperty("key", value)` to log 885additional information, where `value` can be either a C string or a 32-bit 886integer. The _last_ value recorded for a key will be emitted to the XML output 887if you specify one. For example, the test 888 889``` 890TEST_F(WidgetUsageTest, MinAndMaxWidgets) { 891 RecordProperty("MaximumWidgets", ComputeMaxUsage()); 892 RecordProperty("MinimumWidgets", ComputeMinUsage()); 893} 894``` 895 896will output XML like this: 897 898``` 899... 900 <testcase name="MinAndMaxWidgets" status="run" time="6" classname="WidgetUsageTest" 901 MaximumWidgets="12" 902 MinimumWidgets="9" /> 903... 904``` 905 906_Note_: 907 * `RecordProperty()` is a static member of the `Test` class. Therefore it needs to be prefixed with `::testing::Test::` if used outside of the `TEST` body and the test fixture class. 908 * `key` must be a valid XML attribute name, and cannot conflict with the ones already used by Google Test (`name`, `status`, `time`, and `classname`). 909 910_Availability_: Linux, Windows, Mac. 911 912# Sharing Resources Between Tests in the Same Test Case # 913 914 915 916Google Test creates a new test fixture object for each test in order to make 917tests independent and easier to debug. However, sometimes tests use resources 918that are expensive to set up, making the one-copy-per-test model prohibitively 919expensive. 920 921If the tests don't change the resource, there's no harm in them sharing a 922single resource copy. So, in addition to per-test set-up/tear-down, Google Test 923also supports per-test-case set-up/tear-down. To use it: 924 925 1. In your test fixture class (say `FooTest` ), define as `static` some member variables to hold the shared resources. 926 1. In the same test fixture class, define a `static void SetUpTestCase()` function (remember not to spell it as **`SetupTestCase`** with a small `u`!) to set up the shared resources and a `static void TearDownTestCase()` function to tear them down. 927 928That's it! Google Test automatically calls `SetUpTestCase()` before running the 929_first test_ in the `FooTest` test case (i.e. before creating the first 930`FooTest` object), and calls `TearDownTestCase()` after running the _last test_ 931in it (i.e. after deleting the last `FooTest` object). In between, the tests 932can use the shared resources. 933 934Remember that the test order is undefined, so your code can't depend on a test 935preceding or following another. Also, the tests must either not modify the 936state of any shared resource, or, if they do modify the state, they must 937restore the state to its original value before passing control to the next 938test. 939 940Here's an example of per-test-case set-up and tear-down: 941``` 942class FooTest : public ::testing::Test { 943 protected: 944 // Per-test-case set-up. 945 // Called before the first test in this test case. 946 // Can be omitted if not needed. 947 static void SetUpTestCase() { 948 shared_resource_ = new ...; 949 } 950 951 // Per-test-case tear-down. 952 // Called after the last test in this test case. 953 // Can be omitted if not needed. 954 static void TearDownTestCase() { 955 delete shared_resource_; 956 shared_resource_ = NULL; 957 } 958 959 // You can define per-test set-up and tear-down logic as usual. 960 virtual void SetUp() { ... } 961 virtual void TearDown() { ... } 962 963 // Some expensive resource shared by all tests. 964 static T* shared_resource_; 965}; 966 967T* FooTest::shared_resource_ = NULL; 968 969TEST_F(FooTest, Test1) { 970 ... you can refer to shared_resource here ... 971} 972TEST_F(FooTest, Test2) { 973 ... you can refer to shared_resource here ... 974} 975``` 976 977_Availability:_ Linux, Windows, Mac. 978 979# Global Set-Up and Tear-Down # 980 981Just as you can do set-up and tear-down at the test level and the test case 982level, you can also do it at the test program level. Here's how. 983 984First, you subclass the `::testing::Environment` class to define a test 985environment, which knows how to set-up and tear-down: 986 987``` 988class Environment { 989 public: 990 virtual ~Environment() {} 991 // Override this to define how to set up the environment. 992 virtual void SetUp() {} 993 // Override this to define how to tear down the environment. 994 virtual void TearDown() {} 995}; 996``` 997 998Then, you register an instance of your environment class with Google Test by 999calling the `::testing::AddGlobalTestEnvironment()` function: 1000 1001``` 1002Environment* AddGlobalTestEnvironment(Environment* env); 1003``` 1004 1005Now, when `RUN_ALL_TESTS()` is called, it first calls the `SetUp()` method of 1006the environment object, then runs the tests if there was no fatal failures, and 1007finally calls `TearDown()` of the environment object. 1008 1009It's OK to register multiple environment objects. In this case, their `SetUp()` 1010will be called in the order they are registered, and their `TearDown()` will be 1011called in the reverse order. 1012 1013Note that Google Test takes ownership of the registered environment objects. 1014Therefore **do not delete them** by yourself. 1015 1016You should call `AddGlobalTestEnvironment()` before `RUN_ALL_TESTS()` is 1017called, probably in `main()`. If you use `gtest_main`, you need to call 1018this before `main()` starts for it to take effect. One way to do this is to 1019define a global variable like this: 1020 1021``` 1022::testing::Environment* const foo_env = ::testing::AddGlobalTestEnvironment(new FooEnvironment); 1023``` 1024 1025However, we strongly recommend you to write your own `main()` and call 1026`AddGlobalTestEnvironment()` there, as relying on initialization of global 1027variables makes the code harder to read and may cause problems when you 1028register multiple environments from different translation units and the 1029environments have dependencies among them (remember that the compiler doesn't 1030guarantee the order in which global variables from different translation units 1031are initialized). 1032 1033_Availability:_ Linux, Windows, Mac. 1034 1035 1036# Value Parameterized Tests # 1037 1038_Value-parameterized tests_ allow you to test your code with different 1039parameters without writing multiple copies of the same test. 1040 1041Suppose you write a test for your code and then realize that your code is affected by a presence of a Boolean command line flag. 1042 1043``` 1044TEST(MyCodeTest, TestFoo) { 1045 // A code to test foo(). 1046} 1047``` 1048 1049Usually people factor their test code into a function with a Boolean parameter in such situations. The function sets the flag, then executes the testing code. 1050 1051``` 1052void TestFooHelper(bool flag_value) { 1053 flag = flag_value; 1054 // A code to test foo(). 1055} 1056 1057TEST(MyCodeTest, TestFooo) { 1058 TestFooHelper(false); 1059 TestFooHelper(true); 1060} 1061``` 1062 1063But this setup has serious drawbacks. First, when a test assertion fails in your tests, it becomes unclear what value of the parameter caused it to fail. You can stream a clarifying message into your `EXPECT`/`ASSERT` statements, but it you'll have to do it with all of them. Second, you have to add one such helper function per test. What if you have ten tests? Twenty? A hundred? 1064 1065Value-parameterized tests will let you write your test only once and then easily instantiate and run it with an arbitrary number of parameter values. 1066 1067Here are some other situations when value-parameterized tests come handy: 1068 1069 * You wan to test different implementations of an OO interface. 1070 * You want to test your code over various inputs (a.k.a. data-driven testing). This feature is easy to abuse, so please exercise your good sense when doing it! 1071 1072## How to Write Value-Parameterized Tests ## 1073 1074To write value-parameterized tests, first you should define a fixture 1075class. It must be derived from `::testing::TestWithParam<T>`, where `T` 1076is the type of your parameter values. `TestWithParam<T>` is itself 1077derived from `::testing::Test`. `T` can be any copyable type. If it's 1078a raw pointer, you are responsible for managing the lifespan of the 1079pointed values. 1080 1081``` 1082class FooTest : public ::testing::TestWithParam<const char*> { 1083 // You can implement all the usual fixture class members here. 1084 // To access the test parameter, call GetParam() from class 1085 // TestWithParam<T>. 1086}; 1087``` 1088 1089Then, use the `TEST_P` macro to define as many test patterns using 1090this fixture as you want. The `_P` suffix is for "parameterized" or 1091"pattern", whichever you prefer to think. 1092 1093``` 1094TEST_P(FooTest, DoesBlah) { 1095 // Inside a test, access the test parameter with the GetParam() method 1096 // of the TestWithParam<T> class: 1097 EXPECT_TRUE(foo.Blah(GetParam())); 1098 ... 1099} 1100 1101TEST_P(FooTest, HasBlahBlah) { 1102 ... 1103} 1104``` 1105 1106Finally, you can use `INSTANTIATE_TEST_CASE_P` to instantiate the test 1107case with any set of parameters you want. Google Test defines a number of 1108functions for generating test parameters. They return what we call 1109(surprise!) _parameter generators_. Here is a summary of them, 1110which are all in the `testing` namespace: 1111 1112| `Range(begin, end[, step])` | Yields values `{begin, begin+step, begin+step+step, ...}`. The values do not include `end`. `step` defaults to 1. | 1113|:----------------------------|:------------------------------------------------------------------------------------------------------------------| 1114| `Values(v1, v2, ..., vN)` | Yields values `{v1, v2, ..., vN}`. | 1115| `ValuesIn(container)` and `ValuesIn(begin, end)` | Yields values from a C-style array, an STL-style container, or an iterator range `[begin, end)`. | 1116| `Bool()` | Yields sequence `{false, true}`. | 1117| `Combine(g1, g2, ..., gN)` | Yields all combinations (the Cartesian product for the math savvy) of the values generated by the `N` generators. This is only available if your system provides the `<tr1/tuple>` header. If you are sure your system does, and Google Test disagrees, you can override it by defining `GTEST_HAS_TR1_TUPLE=1`. See comments in [include/gtest/internal/gtest-port.h](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/internal/gtest-port.h) for more information. | 1118 1119For more details, see the comments at the definitions of these functions in the [source code](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest-param-test.h). 1120 1121The following statement will instantiate tests from the `FooTest` test case 1122each with parameter values `"meeny"`, `"miny"`, and `"moe"`. 1123 1124``` 1125INSTANTIATE_TEST_CASE_P(InstantiationName, 1126 FooTest, 1127 ::testing::Values("meeny", "miny", "moe")); 1128``` 1129 1130To distinguish different instances of the pattern (yes, you can 1131instantiate it more than once), the first argument to 1132`INSTANTIATE_TEST_CASE_P` is a prefix that will be added to the actual 1133test case name. Remember to pick unique prefixes for different 1134instantiations. The tests from the instantiation above will have these 1135names: 1136 1137 * `InstantiationName/FooTest.DoesBlah/0` for `"meeny"` 1138 * `InstantiationName/FooTest.DoesBlah/1` for `"miny"` 1139 * `InstantiationName/FooTest.DoesBlah/2` for `"moe"` 1140 * `InstantiationName/FooTest.HasBlahBlah/0` for `"meeny"` 1141 * `InstantiationName/FooTest.HasBlahBlah/1` for `"miny"` 1142 * `InstantiationName/FooTest.HasBlahBlah/2` for `"moe"` 1143 1144You can use these names in [--gtest\_filter](#Running_a_Subset_of_the_Tests.md). 1145 1146This statement will instantiate all tests from `FooTest` again, each 1147with parameter values `"cat"` and `"dog"`: 1148 1149``` 1150const char* pets[] = {"cat", "dog"}; 1151INSTANTIATE_TEST_CASE_P(AnotherInstantiationName, FooTest, 1152 ::testing::ValuesIn(pets)); 1153``` 1154 1155The tests from the instantiation above will have these names: 1156 1157 * `AnotherInstantiationName/FooTest.DoesBlah/0` for `"cat"` 1158 * `AnotherInstantiationName/FooTest.DoesBlah/1` for `"dog"` 1159 * `AnotherInstantiationName/FooTest.HasBlahBlah/0` for `"cat"` 1160 * `AnotherInstantiationName/FooTest.HasBlahBlah/1` for `"dog"` 1161 1162Please note that `INSTANTIATE_TEST_CASE_P` will instantiate _all_ 1163tests in the given test case, whether their definitions come before or 1164_after_ the `INSTANTIATE_TEST_CASE_P` statement. 1165 1166You can see 1167[these](http://code.google.com/p/googletest/source/browse/trunk/samples/sample7_unittest.cc) 1168[files](http://code.google.com/p/googletest/source/browse/trunk/samples/sample8_unittest.cc) for more examples. 1169 1170_Availability_: Linux, Windows (requires MSVC 8.0 or above), Mac; since version 1.2.0. 1171 1172## Creating Value-Parameterized Abstract Tests ## 1173 1174In the above, we define and instantiate `FooTest` in the same source 1175file. Sometimes you may want to define value-parameterized tests in a 1176library and let other people instantiate them later. This pattern is 1177known as <i>abstract tests</i>. As an example of its application, when you 1178are designing an interface you can write a standard suite of abstract 1179tests (perhaps using a factory function as the test parameter) that 1180all implementations of the interface are expected to pass. When 1181someone implements the interface, he can instantiate your suite to get 1182all the interface-conformance tests for free. 1183 1184To define abstract tests, you should organize your code like this: 1185 1186 1. Put the definition of the parameterized test fixture class (e.g. `FooTest`) in a header file, say `foo_param_test.h`. Think of this as _declaring_ your abstract tests. 1187 1. Put the `TEST_P` definitions in `foo_param_test.cc`, which includes `foo_param_test.h`. Think of this as _implementing_ your abstract tests. 1188 1189Once they are defined, you can instantiate them by including 1190`foo_param_test.h`, invoking `INSTANTIATE_TEST_CASE_P()`, and linking 1191with `foo_param_test.cc`. You can instantiate the same abstract test 1192case multiple times, possibly in different source files. 1193 1194# Typed Tests # 1195 1196Suppose you have multiple implementations of the same interface and 1197want to make sure that all of them satisfy some common requirements. 1198Or, you may have defined several types that are supposed to conform to 1199the same "concept" and you want to verify it. In both cases, you want 1200the same test logic repeated for different types. 1201 1202While you can write one `TEST` or `TEST_F` for each type you want to 1203test (and you may even factor the test logic into a function template 1204that you invoke from the `TEST`), it's tedious and doesn't scale: 1205if you want _m_ tests over _n_ types, you'll end up writing _m\*n_ 1206`TEST`s. 1207 1208_Typed tests_ allow you to repeat the same test logic over a list of 1209types. You only need to write the test logic once, although you must 1210know the type list when writing typed tests. Here's how you do it: 1211 1212First, define a fixture class template. It should be parameterized 1213by a type. Remember to derive it from `::testing::Test`: 1214 1215``` 1216template <typename T> 1217class FooTest : public ::testing::Test { 1218 public: 1219 ... 1220 typedef std::list<T> List; 1221 static T shared_; 1222 T value_; 1223}; 1224``` 1225 1226Next, associate a list of types with the test case, which will be 1227repeated for each type in the list: 1228 1229``` 1230typedef ::testing::Types<char, int, unsigned int> MyTypes; 1231TYPED_TEST_CASE(FooTest, MyTypes); 1232``` 1233 1234The `typedef` is necessary for the `TYPED_TEST_CASE` macro to parse 1235correctly. Otherwise the compiler will think that each comma in the 1236type list introduces a new macro argument. 1237 1238Then, use `TYPED_TEST()` instead of `TEST_F()` to define a typed test 1239for this test case. You can repeat this as many times as you want: 1240 1241``` 1242TYPED_TEST(FooTest, DoesBlah) { 1243 // Inside a test, refer to the special name TypeParam to get the type 1244 // parameter. Since we are inside a derived class template, C++ requires 1245 // us to visit the members of FooTest via 'this'. 1246 TypeParam n = this->value_; 1247 1248 // To visit static members of the fixture, add the 'TestFixture::' 1249 // prefix. 1250 n += TestFixture::shared_; 1251 1252 // To refer to typedefs in the fixture, add the 'typename TestFixture::' 1253 // prefix. The 'typename' is required to satisfy the compiler. 1254 typename TestFixture::List values; 1255 values.push_back(n); 1256 ... 1257} 1258 1259TYPED_TEST(FooTest, HasPropertyA) { ... } 1260``` 1261 1262You can see `samples/sample6_unittest.cc` for a complete example. 1263 1264_Availability:_ Linux, Windows (requires MSVC 8.0 or above), Mac; 1265since version 1.1.0. 1266 1267# Type-Parameterized Tests # 1268 1269_Type-parameterized tests_ are like typed tests, except that they 1270don't require you to know the list of types ahead of time. Instead, 1271you can define the test logic first and instantiate it with different 1272type lists later. You can even instantiate it more than once in the 1273same program. 1274 1275If you are designing an interface or concept, you can define a suite 1276of type-parameterized tests to verify properties that any valid 1277implementation of the interface/concept should have. Then, the author 1278of each implementation can just instantiate the test suite with his 1279type to verify that it conforms to the requirements, without having to 1280write similar tests repeatedly. Here's an example: 1281 1282First, define a fixture class template, as we did with typed tests: 1283 1284``` 1285template <typename T> 1286class FooTest : public ::testing::Test { 1287 ... 1288}; 1289``` 1290 1291Next, declare that you will define a type-parameterized test case: 1292 1293``` 1294TYPED_TEST_CASE_P(FooTest); 1295``` 1296 1297The `_P` suffix is for "parameterized" or "pattern", whichever you 1298prefer to think. 1299 1300Then, use `TYPED_TEST_P()` to define a type-parameterized test. You 1301can repeat this as many times as you want: 1302 1303``` 1304TYPED_TEST_P(FooTest, DoesBlah) { 1305 // Inside a test, refer to TypeParam to get the type parameter. 1306 TypeParam n = 0; 1307 ... 1308} 1309 1310TYPED_TEST_P(FooTest, HasPropertyA) { ... } 1311``` 1312 1313Now the tricky part: you need to register all test patterns using the 1314`REGISTER_TYPED_TEST_CASE_P` macro before you can instantiate them. 1315The first argument of the macro is the test case name; the rest are 1316the names of the tests in this test case: 1317 1318``` 1319REGISTER_TYPED_TEST_CASE_P(FooTest, 1320 DoesBlah, HasPropertyA); 1321``` 1322 1323Finally, you are free to instantiate the pattern with the types you 1324want. If you put the above code in a header file, you can `#include` 1325it in multiple C++ source files and instantiate it multiple times. 1326 1327``` 1328typedef ::testing::Types<char, int, unsigned int> MyTypes; 1329INSTANTIATE_TYPED_TEST_CASE_P(My, FooTest, MyTypes); 1330``` 1331 1332To distinguish different instances of the pattern, the first argument 1333to the `INSTANTIATE_TYPED_TEST_CASE_P` macro is a prefix that will be 1334added to the actual test case name. Remember to pick unique prefixes 1335for different instances. 1336 1337In the special case where the type list contains only one type, you 1338can write that type directly without `::testing::Types<...>`, like this: 1339 1340``` 1341INSTANTIATE_TYPED_TEST_CASE_P(My, FooTest, int); 1342``` 1343 1344You can see `samples/sample6_unittest.cc` for a complete example. 1345 1346_Availability:_ Linux, Windows (requires MSVC 8.0 or above), Mac; 1347since version 1.1.0. 1348 1349# Testing Private Code # 1350 1351If you change your software's internal implementation, your tests should not 1352break as long as the change is not observable by users. Therefore, per the 1353_black-box testing principle_, most of the time you should test your code 1354through its public interfaces. 1355 1356If you still find yourself needing to test internal implementation code, 1357consider if there's a better design that wouldn't require you to do so. If you 1358absolutely have to test non-public interface code though, you can. There are 1359two cases to consider: 1360 1361 * Static functions (_not_ the same as static member functions!) or unnamed namespaces, and 1362 * Private or protected class members 1363 1364## Static Functions ## 1365 1366Both static functions and definitions/declarations in an unnamed namespace are 1367only visible within the same translation unit. To test them, you can `#include` 1368the entire `.cc` file being tested in your `*_test.cc` file. (#including `.cc` 1369files is not a good way to reuse code - you should not do this in production 1370code!) 1371 1372However, a better approach is to move the private code into the 1373`foo::internal` namespace, where `foo` is the namespace your project normally 1374uses, and put the private declarations in a `*-internal.h` file. Your 1375production `.cc` files and your tests are allowed to include this internal 1376header, but your clients are not. This way, you can fully test your internal 1377implementation without leaking it to your clients. 1378 1379## Private Class Members ## 1380 1381Private class members are only accessible from within the class or by friends. 1382To access a class' private members, you can declare your test fixture as a 1383friend to the class and define accessors in your fixture. Tests using the 1384fixture can then access the private members of your production class via the 1385accessors in the fixture. Note that even though your fixture is a friend to 1386your production class, your tests are not automatically friends to it, as they 1387are technically defined in sub-classes of the fixture. 1388 1389Another way to test private members is to refactor them into an implementation 1390class, which is then declared in a `*-internal.h` file. Your clients aren't 1391allowed to include this header but your tests can. Such is called the Pimpl 1392(Private Implementation) idiom. 1393 1394Or, you can declare an individual test as a friend of your class by adding this 1395line in the class body: 1396 1397``` 1398FRIEND_TEST(TestCaseName, TestName); 1399``` 1400 1401For example, 1402``` 1403// foo.h 1404#include <gtest/gtest_prod.h> 1405 1406// Defines FRIEND_TEST. 1407class Foo { 1408 ... 1409 private: 1410 FRIEND_TEST(FooTest, BarReturnsZeroOnNull); 1411 int Bar(void* x); 1412}; 1413 1414// foo_test.cc 1415... 1416TEST(FooTest, BarReturnsZeroOnNull) { 1417 Foo foo; 1418 EXPECT_EQ(0, foo.Bar(NULL)); 1419 // Uses Foo's private member Bar(). 1420} 1421``` 1422 1423Pay special attention when your class is defined in a namespace, as you should 1424define your test fixtures and tests in the same namespace if you want them to 1425be friends of your class. For example, if the code to be tested looks like: 1426 1427``` 1428namespace my_namespace { 1429 1430class Foo { 1431 friend class FooTest; 1432 FRIEND_TEST(FooTest, Bar); 1433 FRIEND_TEST(FooTest, Baz); 1434 ... 1435 definition of the class Foo 1436 ... 1437}; 1438 1439} // namespace my_namespace 1440``` 1441 1442Your test code should be something like: 1443 1444``` 1445namespace my_namespace { 1446class FooTest : public ::testing::Test { 1447 protected: 1448 ... 1449}; 1450 1451TEST_F(FooTest, Bar) { ... } 1452TEST_F(FooTest, Baz) { ... } 1453 1454} // namespace my_namespace 1455``` 1456 1457# Catching Failures # 1458 1459If you are building a testing utility on top of Google Test, you'll 1460want to test your utility. What framework would you use to test it? 1461Google Test, of course. 1462 1463The challenge is to verify that your testing utility reports failures 1464correctly. In frameworks that report a failure by throwing an 1465exception, you could catch the exception and assert on it. But Google 1466Test doesn't use exceptions, so how do we test that a piece of code 1467generates an expected failure? 1468 1469`<gtest/gtest-spi.h>` contains some constructs to do this. After 1470#including this header, you can use 1471 1472| `EXPECT_FATAL_FAILURE(`_statement, substring_`);` | 1473|:--------------------------------------------------| 1474 1475to assert that _statement_ generates a fatal (e.g. `ASSERT_*`) failure 1476whose message contains the given _substring_, or use 1477 1478| `EXPECT_NONFATAL_FAILURE(`_statement, substring_`);` | 1479|:-----------------------------------------------------| 1480 1481if you are expecting a non-fatal (e.g. `EXPECT_*`) failure. 1482 1483For technical reasons, there are some caveats: 1484 1485 1. You cannot stream a failure message to either macro. 1486 1. _statement_ in `EXPECT_FATAL_FAILURE()` cannot reference local non-static variables or non-static members of `this` object. 1487 1. _statement_ in `EXPECT_FATAL_FAILURE()` cannot return a value. 1488 1489_Note:_ Google Test is designed with threads in mind. Once the 1490synchronization primitives in `<gtest/internal/gtest-port.h>` have 1491been implemented, Google Test will become thread-safe, meaning that 1492you can then use assertions in multiple threads concurrently. Before 1493 1494that, however, Google Test only supports single-threaded usage. Once 1495thread-safe, `EXPECT_FATAL_FAILURE()` and `EXPECT_NONFATAL_FAILURE()` 1496will capture failures in the current thread only. If _statement_ 1497creates new threads, failures in these threads will be ignored. If 1498you want to capture failures from all threads instead, you should use 1499the following macros: 1500 1501| `EXPECT_FATAL_FAILURE_ON_ALL_THREADS(`_statement, substring_`);` | 1502|:-----------------------------------------------------------------| 1503| `EXPECT_NONFATAL_FAILURE_ON_ALL_THREADS(`_statement, substring_`);` | 1504 1505# Getting the Current Test's Name # 1506 1507Sometimes a function may need to know the name of the currently running test. 1508For example, you may be using the `SetUp()` method of your test fixture to set 1509the golden file name based on which test is running. The `::testing::TestInfo` 1510class has this information: 1511 1512``` 1513namespace testing { 1514 1515class TestInfo { 1516 public: 1517 // Returns the test case name and the test name, respectively. 1518 // 1519 // Do NOT delete or free the return value - it's managed by the 1520 // TestInfo class. 1521 const char* test_case_name() const; 1522 const char* name() const; 1523}; 1524 1525} // namespace testing 1526``` 1527 1528 1529> To obtain a `TestInfo` object for the currently running test, call 1530`current_test_info()` on the `UnitTest` singleton object: 1531 1532``` 1533// Gets information about the currently running test. 1534// Do NOT delete the returned object - it's managed by the UnitTest class. 1535const ::testing::TestInfo* const test_info = 1536 ::testing::UnitTest::GetInstance()->current_test_info(); 1537printf("We are in test %s of test case %s.\n", 1538 test_info->name(), test_info->test_case_name()); 1539``` 1540 1541`current_test_info()` returns a null pointer if no test is running. In 1542particular, you cannot find the test case name in `TestCaseSetUp()`, 1543`TestCaseTearDown()` (where you know the test case name implicitly), or 1544functions called from them. 1545 1546_Availability:_ Linux, Windows, Mac. 1547 1548# Extending Google Test by Handling Test Events # 1549 1550Google Test provides an <b>event listener API</b> to let you receive 1551notifications about the progress of a test program and test 1552failures. The events you can listen to include the start and end of 1553the test program, a test case, or a test method, among others. You may 1554use this API to augment or replace the standard console output, 1555replace the XML output, or provide a completely different form of 1556output, such as a GUI or a database. You can also use test events as 1557checkpoints to implement a resource leak checker, for example. 1558 1559_Availability:_ Linux, Windows, Mac; since v1.4.0. 1560 1561## Defining Event Listeners ## 1562 1563To define a event listener, you subclass either 1564[testing::TestEventListener](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest.h#855) 1565or [testing::EmptyTestEventListener](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest.h#905). 1566The former is an (abstract) interface, where <i>each pure virtual method<br> 1567can be overridden to handle a test event</i> (For example, when a test 1568starts, the `OnTestStart()` method will be called.). The latter provides 1569an empty implementation of all methods in the interface, such that a 1570subclass only needs to override the methods it cares about. 1571 1572When an event is fired, its context is passed to the handler function 1573as an argument. The following argument types are used: 1574 * [UnitTest](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest.h#1007) reflects the state of the entire test program, 1575 * [TestCase](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest.h#689) has information about a test case, which can contain one or more tests, 1576 * [TestInfo](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest.h#599) contains the state of a test, and 1577 * [TestPartResult](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest-test-part.h#42) represents the result of a test assertion. 1578 1579An event handler function can examine the argument it receives to find 1580out interesting information about the event and the test program's 1581state. Here's an example: 1582 1583``` 1584 class MinimalistPrinter : public ::testing::EmptyTestEventListener { 1585 // Called before a test starts. 1586 virtual void OnTestStart(const ::testing::TestInfo& test_info) { 1587 printf("*** Test %s.%s starting.\n", 1588 test_info.test_case_name(), test_info.name()); 1589 } 1590 1591 // Called after a failed assertion or a SUCCESS(). 1592 virtual void OnTestPartResult( 1593 const ::testing::TestPartResult& test_part_result) { 1594 printf("%s in %s:%d\n%s\n", 1595 test_part_result.failed() ? "*** Failure" : "Success", 1596 test_part_result.file_name(), 1597 test_part_result.line_number(), 1598 test_part_result.summary()); 1599 } 1600 1601 // Called after a test ends. 1602 virtual void OnTestEnd(const ::testing::TestInfo& test_info) { 1603 printf("*** Test %s.%s ending.\n", 1604 test_info.test_case_name(), test_info.name()); 1605 } 1606 }; 1607``` 1608 1609## Using Event Listeners ## 1610 1611To use the event listener you have defined, add an instance of it to 1612the Google Test event listener list (represented by class 1613[TestEventListeners](http://code.google.com/p/googletest/source/browse/trunk/include/gtest/gtest.h#929) 1614- note the "s" at the end of the name) in your 1615`main()` function, before calling `RUN_ALL_TESTS()`: 1616``` 1617int main(int argc, char** argv) { 1618 ::testing::InitGoogleTest(&argc, argv); 1619 // Gets hold of the event listener list. 1620 ::testing::TestEventListeners& listeners = 1621 ::testing::UnitTest::GetInstance()->listeners(); 1622 // Adds a listener to the end. Google Test takes the ownership. 1623 listeners.Append(new MinimalistPrinter); 1624 return RUN_ALL_TESTS(); 1625} 1626``` 1627 1628There's only one problem: the default test result printer is still in 1629effect, so its output will mingle with the output from your minimalist 1630printer. To suppress the default printer, just release it from the 1631event listener list and delete it. You can do so by adding one line: 1632``` 1633 ... 1634 delete listeners.Release(listeners.default_result_printer()); 1635 listeners.Append(new MinimalistPrinter); 1636 return RUN_ALL_TESTS(); 1637``` 1638 1639Now, sit back and enjoy a completely different output from your 1640tests. For more details, you can read this 1641[sample](http://code.google.com/p/googletest/source/browse/trunk/samples/sample9_unittest.cc). 1642 1643You may append more than one listener to the list. When an `On*Start()` 1644or `OnTestPartResult()` event is fired, the listeners will receive it in 1645the order they appear in the list (since new listeners are added to 1646the end of the list, the default text printer and the default XML 1647generator will receive the event first). An `On*End()` event will be 1648received by the listeners in the _reverse_ order. This allows output by 1649listeners added later to be framed by output from listeners added 1650earlier. 1651 1652## Generating Failures in Listeners ## 1653 1654You may use failure-raising macros (`EXPECT_*()`, `ASSERT_*()`, 1655`FAIL()`, etc) when processing an event. There are some restrictions: 1656 1657 1. You cannot generate any failure in `OnTestPartResult()` (otherwise it will cause `OnTestPartResult()` to be called recursively). 1658 1. A listener that handles `OnTestPartResult()` is not allowed to generate any failure. 1659 1660When you add listeners to the listener list, you should put listeners 1661that handle `OnTestPartResult()` _before_ listeners that can generate 1662failures. This ensures that failures generated by the latter are 1663attributed to the right test by the former. 1664 1665We have a sample of failure-raising listener 1666[here](http://code.google.com/p/googletest/source/browse/trunk/samples/sample10_unittest.cc). 1667 1668# Running Test Programs: Advanced Options # 1669 1670Google Test test programs are ordinary executables. Once built, you can run 1671them directly and affect their behavior via the following environment variables 1672and/or command line flags. For the flags to work, your programs must call 1673`::testing::InitGoogleTest()` before calling `RUN_ALL_TESTS()`. 1674 1675To see a list of supported flags and their usage, please run your test 1676program with the `--help` flag. You can also use `-h`, `-?`, or `/?` 1677for short. This feature is added in version 1.3.0. 1678 1679If an option is specified both by an environment variable and by a 1680flag, the latter takes precedence. Most of the options can also be 1681set/read in code: to access the value of command line flag 1682`--gtest_foo`, write `::testing::GTEST_FLAG(foo)`. A common pattern is 1683to set the value of a flag before calling `::testing::InitGoogleTest()` 1684to change the default value of the flag: 1685``` 1686int main(int argc, char** argv) { 1687 // Disables elapsed time by default. 1688 ::testing::GTEST_FLAG(print_time) = false; 1689 1690 // This allows the user to override the flag on the command line. 1691 ::testing::InitGoogleTest(&argc, argv); 1692 1693 return RUN_ALL_TESTS(); 1694} 1695``` 1696 1697## Selecting Tests ## 1698 1699This section shows various options for choosing which tests to run. 1700 1701### Listing Test Names ### 1702 1703Sometimes it is necessary to list the available tests in a program before 1704running them so that a filter may be applied if needed. Including the flag 1705`--gtest_list_tests` overrides all other flags and lists tests in the following 1706format: 1707``` 1708TestCase1. 1709 TestName1 1710 TestName2 1711TestCase2. 1712 TestName 1713``` 1714 1715None of the tests listed are actually run if the flag is provided. There is no 1716corresponding environment variable for this flag. 1717 1718_Availability:_ Linux, Windows, Mac. 1719 1720### Running a Subset of the Tests ### 1721 1722By default, a Google Test program runs all tests the user has defined. 1723Sometimes, you want to run only a subset of the tests (e.g. for debugging or 1724quickly verifying a change). If you set the `GTEST_FILTER` environment variable 1725or the `--gtest_filter` flag to a filter string, Google Test will only run the 1726tests whose full names (in the form of `TestCaseName.TestName`) match the 1727filter. 1728 1729The format of a filter is a '`:`'-separated list of wildcard patterns (called 1730the positive patterns) optionally followed by a '`-`' and another 1731'`:`'-separated pattern list (called the negative patterns). A test matches the 1732filter if and only if it matches any of the positive patterns but does not 1733match any of the negative patterns. 1734 1735A pattern may contain `'*'` (matches any string) or `'?'` (matches any single 1736character). For convenience, the filter `'*-NegativePatterns'` can be also 1737written as `'-NegativePatterns'`. 1738 1739For example: 1740 1741 * `./foo_test` Has no flag, and thus runs all its tests. 1742 * `./foo_test --gtest_filter=*` Also runs everything, due to the single match-everything `*` value. 1743 * `./foo_test --gtest_filter=FooTest.*` Runs everything in test case `FooTest`. 1744 * `./foo_test --gtest_filter=*Null*:*Constructor*` Runs any test whose full name contains either `"Null"` or `"Constructor"`. 1745 * `./foo_test --gtest_filter=-*DeathTest.*` Runs all non-death tests. 1746 * `./foo_test --gtest_filter=FooTest.*-FooTest.Bar` Runs everything in test case `FooTest` except `FooTest.Bar`. 1747 1748_Availability:_ Linux, Windows, Mac. 1749 1750### Temporarily Disabling Tests ### 1751 1752If you have a broken test that you cannot fix right away, you can add the 1753`DISABLED_` prefix to its name. This will exclude it from execution. This is 1754better than commenting out the code or using `#if 0`, as disabled tests are 1755still compiled (and thus won't rot). 1756 1757If you need to disable all tests in a test case, you can either add `DISABLED_` 1758to the front of the name of each test, or alternatively add it to the front of 1759the test case name. 1760 1761For example, the following tests won't be run by Google Test, even though they 1762will still be compiled: 1763 1764``` 1765// Tests that Foo does Abc. 1766TEST(FooTest, DISABLED_DoesAbc) { ... } 1767 1768class DISABLED_BarTest : public ::testing::Test { ... }; 1769 1770// Tests that Bar does Xyz. 1771TEST_F(DISABLED_BarTest, DoesXyz) { ... } 1772``` 1773 1774_Note:_ This feature should only be used for temporary pain-relief. You still 1775have to fix the disabled tests at a later date. As a reminder, Google Test will 1776print a banner warning you if a test program contains any disabled tests. 1777 1778_Tip:_ You can easily count the number of disabled tests you have 1779using `grep`. This number can be used as a metric for improving your 1780test quality. 1781 1782_Availability:_ Linux, Windows, Mac. 1783 1784### Temporarily Enabling Disabled Tests ### 1785 1786To include [disabled tests](#Temporarily_Disabling_Tests.md) in test 1787execution, just invoke the test program with the 1788`--gtest_also_run_disabled_tests` flag or set the 1789`GTEST_ALSO_RUN_DISABLED_TESTS` environment variable to a value other 1790than `0`. You can combine this with the 1791[--gtest\_filter](#Running_a_Subset_of_the_Tests.md) flag to further select 1792which disabled tests to run. 1793 1794_Availability:_ Linux, Windows, Mac; since version 1.3.0. 1795 1796## Repeating the Tests ## 1797 1798Once in a while you'll run into a test whose result is hit-or-miss. Perhaps it 1799will fail only 1% of the time, making it rather hard to reproduce the bug under 1800a debugger. This can be a major source of frustration. 1801 1802The `--gtest_repeat` flag allows you to repeat all (or selected) test methods 1803in a program many times. Hopefully, a flaky test will eventually fail and give 1804you a chance to debug. Here's how to use it: 1805 1806| `$ foo_test --gtest_repeat=1000` | Repeat foo\_test 1000 times and don't stop at failures. | 1807|:---------------------------------|:--------------------------------------------------------| 1808| `$ foo_test --gtest_repeat=-1` | A negative count means repeating forever. | 1809| `$ foo_test --gtest_repeat=1000 --gtest_break_on_failure` | Repeat foo\_test 1000 times, stopping at the first failure. This is especially useful when running under a debugger: when the testfails, it will drop into the debugger and you can then inspect variables and stacks. | 1810| `$ foo_test --gtest_repeat=1000 --gtest_filter=FooBar` | Repeat the tests whose name matches the filter 1000 times. | 1811 1812If your test program contains global set-up/tear-down code registered 1813using `AddGlobalTestEnvironment()`, it will be repeated in each 1814iteration as well, as the flakiness may be in it. You can also specify 1815the repeat count by setting the `GTEST_REPEAT` environment variable. 1816 1817_Availability:_ Linux, Windows, Mac. 1818 1819## Shuffling the Tests ## 1820 1821You can specify the `--gtest_shuffle` flag (or set the `GTEST_SHUFFLE` 1822environment variable to `1`) to run the tests in a program in a random 1823order. This helps to reveal bad dependencies between tests. 1824 1825By default, Google Test uses a random seed calculated from the current 1826time. Therefore you'll get a different order every time. The console 1827output includes the random seed value, such that you can reproduce an 1828order-related test failure later. To specify the random seed 1829explicitly, use the `--gtest_random_seed=SEED` flag (or set the 1830`GTEST_RANDOM_SEED` environment variable), where `SEED` is an integer 1831between 0 and 99999. The seed value 0 is special: it tells Google Test 1832to do the default behavior of calculating the seed from the current 1833time. 1834 1835If you combine this with `--gtest_repeat=N`, Google Test will pick a 1836different random seed and re-shuffle the tests in each iteration. 1837 1838_Availability:_ Linux, Windows, Mac; since v1.4.0. 1839 1840## Controlling Test Output ## 1841 1842This section teaches how to tweak the way test results are reported. 1843 1844### Colored Terminal Output ### 1845 1846Google Test can use colors in its terminal output to make it easier to spot 1847the separation between tests, and whether tests passed. 1848 1849You can set the GTEST\_COLOR environment variable or set the `--gtest_color` 1850command line flag to `yes`, `no`, or `auto` (the default) to enable colors, 1851disable colors, or let Google Test decide. When the value is `auto`, Google 1852Test will use colors if and only if the output goes to a terminal and (on 1853non-Windows platforms) the `TERM` environment variable is set to `xterm` or 1854`xterm-color`. 1855 1856_Availability:_ Linux, Windows, Mac. 1857 1858### Suppressing the Elapsed Time ### 1859 1860By default, Google Test prints the time it takes to run each test. To 1861suppress that, run the test program with the `--gtest_print_time=0` 1862command line flag. Setting the `GTEST_PRINT_TIME` environment 1863variable to `0` has the same effect. 1864 1865_Availability:_ Linux, Windows, Mac. (In Google Test 1.3.0 and lower, 1866the default behavior is that the elapsed time is **not** printed.) 1867 1868### Generating an XML Report ### 1869 1870Google Test can emit a detailed XML report to a file in addition to its normal 1871textual output. The report contains the duration of each test, and thus can 1872help you identify slow tests. 1873 1874To generate the XML report, set the `GTEST_OUTPUT` environment variable or the 1875`--gtest_output` flag to the string `"xml:_path_to_output_file_"`, which will 1876create the file at the given location. You can also just use the string 1877`"xml"`, in which case the output can be found in the `test_detail.xml` file in 1878the current directory. 1879 1880If you specify a directory (for example, `"xml:output/directory/"` on Linux or 1881`"xml:output\directory\"` on Windows), Google Test will create the XML file in 1882that directory, named after the test executable (e.g. `foo_test.xml` for test 1883program `foo_test` or `foo_test.exe`). If the file already exists (perhaps left 1884over from a previous run), Google Test will pick a different name (e.g. 1885`foo_test_1.xml`) to avoid overwriting it. 1886 1887The report uses the format described here. It is based on the 1888`junitreport` Ant task and can be parsed by popular continuous build 1889systems like [Hudson](https://hudson.dev.java.net/). Since that format 1890was originally intended for Java, a little interpretation is required 1891to make it apply to Google Test tests, as shown here: 1892 1893``` 1894<testsuites name="AllTests" ...> 1895 <testsuite name="test_case_name" ...> 1896 <testcase name="test_name" ...> 1897 <failure message="..."/> 1898 <failure message="..."/> 1899 <failure message="..."/> 1900 </testcase> 1901 </testsuite> 1902</testsuites> 1903``` 1904 1905 * The root `<testsuites>` element corresponds to the entire test program. 1906 * `<testsuite>` elements correspond to Google Test test cases. 1907 * `<testcase>` elements correspond to Google Test test functions. 1908 1909For instance, the following program 1910 1911``` 1912TEST(MathTest, Addition) { ... } 1913TEST(MathTest, Subtraction) { ... } 1914TEST(LogicTest, NonContradiction) { ... } 1915``` 1916 1917could generate this report: 1918 1919``` 1920<?xml version="1.0" encoding="UTF-8"?> 1921<testsuites tests="3" failures="1" errors="0" time="35" name="AllTests"> 1922 <testsuite name="MathTest" tests="2" failures="1"* errors="0" time="15"> 1923 <testcase name="Addition" status="run" time="7" classname=""> 1924 <failure message="Value of: add(1, 1)
 Actual: 3
Expected: 2" type=""/> 1925 <failure message="Value of: add(1, -1)
 Actual: 1
Expected: 0" type=""/> 1926 </testcase> 1927 <testcase name="Subtraction" status="run" time="5" classname=""> 1928 </testcase> 1929 </testsuite> 1930 <testsuite name="LogicTest" tests="1" failures="0" errors="0" time="5"> 1931 <testcase name="NonContradiction" status="run" time="5" classname=""> 1932 </testcase> 1933 </testsuite> 1934</testsuites> 1935``` 1936 1937Things to note: 1938 1939 * The `tests` attribute of a `<testsuites>` or `<testsuite>` element tells how many test functions the Google Test program or test case contains, while the `failures` attribute tells how many of them failed. 1940 * The `time` attribute expresses the duration of the test, test case, or entire test program in milliseconds. 1941 * Each `<failure>` element corresponds to a single failed Google Test assertion. 1942 * Some JUnit concepts don't apply to Google Test, yet we have to conform to the DTD. Therefore you'll see some dummy elements and attributes in the report. You can safely ignore these parts. 1943 1944_Availability:_ Linux, Windows, Mac. 1945 1946## Controlling How Failures Are Reported ## 1947 1948### Turning Assertion Failures into Break-Points ### 1949 1950When running test programs under a debugger, it's very convenient if the 1951debugger can catch an assertion failure and automatically drop into interactive 1952mode. Google Test's _break-on-failure_ mode supports this behavior. 1953 1954To enable it, set the `GTEST_BREAK_ON_FAILURE` environment variable to a value 1955other than `0` . Alternatively, you can use the `--gtest_break_on_failure` 1956command line flag. 1957 1958_Availability:_ Linux, Windows, Mac. 1959 1960### Suppressing Pop-ups Caused by Exceptions ### 1961 1962On Windows, Google Test may be used with exceptions enabled. Even when 1963exceptions are disabled, an application can still throw structured exceptions 1964(SEH's). If a test throws an exception, by default Google Test doesn't try to 1965catch it. Instead, you'll see a pop-up dialog, at which point you can attach 1966the process to a debugger and easily find out what went wrong. 1967 1968However, if you don't want to see the pop-ups (for example, if you run the 1969tests in a batch job), set the `GTEST_CATCH_EXCEPTIONS` environment variable to 1970a non- `0` value, or use the `--gtest_catch_exceptions` flag. Google Test now 1971catches all test-thrown exceptions and logs them as failures. 1972 1973_Availability:_ Windows. `GTEST_CATCH_EXCEPTIONS` and 1974`--gtest_catch_exceptions` have no effect on Google Test's behavior on Linux or 1975Mac, even if exceptions are enabled. It is possible to add support for catching 1976exceptions on these platforms, but it is not implemented yet. 1977 1978### Letting Another Testing Framework Drive ### 1979 1980If you work on a project that has already been using another testing 1981framework and is not ready to completely switch to Google Test yet, 1982you can get much of Google Test's benefit by using its assertions in 1983your existing tests. Just change your `main()` function to look 1984like: 1985 1986``` 1987#include <gtest/gtest.h> 1988 1989int main(int argc, char** argv) { 1990 ::testing::GTEST_FLAG(throw_on_failure) = true; 1991 // Important: Google Test must be initialized. 1992 ::testing::InitGoogleTest(&argc, argv); 1993 1994 ... whatever your existing testing framework requires ... 1995} 1996``` 1997 1998With that, you can use Google Test assertions in addition to the 1999native assertions your testing framework provides, for example: 2000 2001``` 2002void TestFooDoesBar() { 2003 Foo foo; 2004 EXPECT_LE(foo.Bar(1), 100); // A Google Test assertion. 2005 CPPUNIT_ASSERT(foo.IsEmpty()); // A native assertion. 2006} 2007``` 2008 2009If a Google Test assertion fails, it will print an error message and 2010throw an exception, which will be treated as a failure by your host 2011testing framework. If you compile your code with exceptions disabled, 2012a failed Google Test assertion will instead exit your program with a 2013non-zero code, which will also signal a test failure to your test 2014runner. 2015 2016If you don't write `::testing::GTEST_FLAG(throw_on_failure) = true;` in 2017your `main()`, you can alternatively enable this feature by specifying 2018the `--gtest_throw_on_failure` flag on the command-line or setting the 2019`GTEST_THROW_ON_FAILURE` environment variable to a non-zero value. 2020 2021_Availability:_ Linux, Windows, Mac; since v1.3.0. 2022 2023## Distributing Test Functions to Multiple Machines ## 2024 2025If you have more than one machine you can use to run a test program, 2026you might want to run the test functions in parallel and get the 2027result faster. We call this technique _sharding_, where each machine 2028is called a _shard_. 2029 2030Google Test is compatible with test sharding. To take advantage of 2031this feature, your test runner (not part of Google Test) needs to do 2032the following: 2033 2034 1. Allocate a number of machines (shards) to run the tests. 2035 1. On each shard, set the `GTEST_TOTAL_SHARDS` environment variable to the total number of shards. It must be the same for all shards. 2036 1. On each shard, set the `GTEST_SHARD_INDEX` environment variable to the index of the shard. Different shards must be assigned different indices, which must be in the range `[0, GTEST_TOTAL_SHARDS - 1]`. 2037 1. Run the same test program on all shards. When Google Test sees the above two environment variables, it will select a subset of the test functions to run. Across all shards, each test function in the program will be run exactly once. 2038 1. Wait for all shards to finish, then collect and report the results. 2039 2040Your project may have tests that were written without Google Test and 2041thus don't understand this protocol. In order for your test runner to 2042figure out which test supports sharding, it can set the environment 2043variable `GTEST_SHARD_STATUS_FILE` to a non-existent file path. If a 2044test program supports sharding, it will create this file to 2045acknowledge the fact (the actual contents of the file are not 2046important at this time; although we may stick some useful information 2047in it in the future.); otherwise it will not create it. 2048 2049Here's an example to make it clear. Suppose you have a test program 2050`foo_test` that contains the following 5 test functions: 2051``` 2052TEST(A, V) 2053TEST(A, W) 2054TEST(B, X) 2055TEST(B, Y) 2056TEST(B, Z) 2057``` 2058and you have 3 machines at your disposal. To run the test functions in 2059parallel, you would set `GTEST_TOTAL_SHARDS` to 3 on all machines, and 2060set `GTEST_SHARD_INDEX` to 0, 1, and 2 on the machines respectively. 2061Then you would run the same `foo_test` on each machine. 2062 2063Google Test reserves the right to change how the work is distributed 2064across the shards, but here's one possible scenario: 2065 2066 * Machine #0 runs `A.V` and `B.X`. 2067 * Machine #1 runs `A.W` and `B.Y`. 2068 * Machine #2 runs `B.Z`. 2069 2070_Availability:_ Linux, Windows, Mac; since version 1.3.0. 2071 2072# Fusing Google Test Source Files # 2073 2074Google Test's implementation consists of ~30 files (excluding its own 2075tests). Sometimes you may want them to be packaged up in two files (a 2076`.h` and a `.cc`) instead, such that you can easily copy them to a new 2077machine and start hacking there. For this we provide an experimental 2078Python script `fuse_gtest_files.py` in the `scripts/` directory (since release 1.3.0). 2079Assuming you have Python 2.4 or above installed on your machine, just 2080go to that directory and run 2081``` 2082python fuse_gtest_files.py OUTPUT_DIR 2083``` 2084 2085and you should see an `OUTPUT_DIR` directory being created with files 2086`gtest/gtest.h` and `gtest/gtest-all.cc` in it. These files contain 2087everything you need to use Google Test. Just copy them to anywhere 2088you want and you are ready to write tests. You can use the 2089[scrpts/test/Makefile](http://code.google.com/p/googletest/source/browse/trunk/scripts/test/Makefile) 2090file as an example on how to compile your tests against them. 2091 2092# Where to Go from Here # 2093 2094Congratulations! You've now learned more advanced Google Test tools and are 2095ready to tackle more complex testing tasks. If you want to dive even deeper, you 2096can read the [FAQ](V1_5_FAQ.md).