core/testing

testing

Types

20

Options

Options :: struct { // Equivalent to the TEST_NAMES compile-time definition, but used dynamically at runtime. test_names: string, }Source

Stop_Reason

Stop_Reason :: enum int { Unknown = 0, Successful_Stop = 1, Illegal_Instruction = 2, Arithmetic_Error = 3, Segmentation_Fault = 4, Unhandled_Trap = 5, }Source

T

T :: struct { error_count: int, // If your test needs to perform random operations, it's advised to use // this value to seed a local random number generator rather than relying // on the non-thread-safe global one. // // This way, your results will be deterministic. // // This value is chosen at startup of the test runner, logged, and may be // specified by the user. It is the same for all tests of a single run. seed: u64, channel: Update_Channel_Sender, cleanups: [dynamic]Internal_Cleanup, // This allocator is shared between the test runner and its threads for // cloning log strings, so they can outlive the lifetime of individual // tests during channel transmission. _log_allocator: runtime.Allocator, _fail_now_called: bool, }Source

Constants

27

FANCY_OUTPUT

FANCY_OUTPUT :: bool = #config(ODIN_TEST_FANCY, true)Source

Show the fancy animated progress report. This requires terminal color support, as well as STDOUT to not be redirected to a file.

GO_TO_ERROR

GO_TO_ERROR :: bool = #config(ODIN_TEST_GO_TO_ERROR, false)Source

Print the full file path for failed test cases on a new line in a way that's friendly to regex capture for an editor's "go to error".

SHARED_RANDOM_SEED

SHARED_RANDOM_SEED :: u64 = #config(ODIN_TEST_RANDOM_SEED, 0)Source

This is the random seed that will be sent to each test. If it is unspecified, it will be set to the system cycle counter at startup.

SIGTRAP

SIGTRAP :: 5Source

Windows does not appear to have a SIGTRAP, so this is defined here, instead of in the libc package, just so there's no confusion about it being available there.

TEST_NAMES

TEST_NAMES :: string = #config(ODIN_TEST_NAMES, "")Source

Select a specific set of tests to run by name. Each test is separated by a comma and may optionally include the package name. This may be useful when running tests on multiple packages with -all-packages. The format is: package.test_name,test_name_only,...

Procedures

39

cleanup

cleanup :: proc(t: ^T, procedure: proc(), user_data: rawptr)Source

cleanup registers a procedure and user_data, which will be called when the test, and all its subtests, complete. Cleanup procedures will be called in LIFO (last added, first called) order.

Each procedure will use a copy of the context at the time of registering, and if the test failed due to a timeout, failed assertion, panic, bounds-checking error, memory access violation, or any other signal-based fault, this procedure will run with greater privilege in the test runner's main thread.

That means that any cleanup procedure absolutely must not fail in the same way, or it will take down the entire test runner with it. This is for when you need something to run no matter what, if a test failed.

For almost every usual case, defer should be preferable and sufficient.

expect_assert_from

expect_assert_from :: proc(t: ^T, expected_place: runtime.Source_Code_Location, caller_loc = #caller_location)Source

Let the test runner know that it should expect an assertion failure from a specific location in the source code for this test.

In the event that an assertion fails, a debug message will be logged with its exact message and location in a copyable format to make it convenient to write tests which use this API.

This procedure may be called up to 5 times with different locations.

This is a limitation for the sake of simplicity in the implementation, and you should consider breaking up your tests into smaller procedures if you need to check for asserts in more than 2 places.

expect_assert_message

expect_assert_message :: proc(t: ^T, expected_message: string, caller_loc = #caller_location)Source

Let the test runner know that it should expect an assertion failure with a specific message for this test.

In the event that an assertion fails, a debug message will be logged with its exact message and location in a copyable format to make it convenient to write tests which use this API.

This procedure may be called up to 5 times with different messages.

This is a limitation for the sake of simplicity in the implementation, and you should consider breaking up your tests into smaller procedures if you need to check for more than a couple different assertion messages.

expect_signal

expect_signal :: proc(t: ^T, sig: i32)Source

Let the test runner know that it should expect a signal to be raised within this test.

This API is for advanced users, as arbitrary signals will not be caught; only the ones already handled by the test runner, such as

  • SIGINT, (interrupt)
  • SIGTERM, (polite termination)
  • SIGILL, (illegal instruction)
  • SIGFPE, (arithmetic error)
  • SIGSEGV, and (segmentation fault)
  • SIGTRAP (only on POSIX systems). (trap / debug trap)

Note that only one signal can be expected per test.

fail_now

fail_now :: proc(t: ^T, msg: untyped string = "", loc = #caller_location) -> ()Source

fail_now will cause a test to immediately fail and abort, much in the same way a failed assertion or panic call will stop a thread.

It is for when you absolutely need a test to fail without calling any of its deferred statements. It will be cleaner than a regular assert or panic, as the test runner will know to expect the signal this procedure will raise.

Procedure Groups

1

Reference search

Find anything

Documentation preferences

Settings

System theme variants

Used only while Theme is set to System.