Skip to content

Latest commit

 

History

History
138 lines (102 loc) · 5.34 KB

README.md

File metadata and controls

138 lines (102 loc) · 5.34 KB

Envoy tests

Envoy uses the Google Test framework for tests, including Google Mock for mocks and matchers. See the documentation on those pages for information on the various classes, macros, and matchers that Envoy uses from those frameworks.

Integration tests

Envoy contains an integration testing framework, for testing downstream-Envoy-upstream communication. See the framework's README for more information.

Custom matchers

Envoy includes some custom Google Mock matchers to make test expectation statements simpler to write and easier to understand.

HeaderValueOf

Tests that a HeaderMap argument contains exactly one header with the given key, whose value satisfies the given expectation. The expectation can be a matcher, or a string that the value should equal.

Examples:

EXPECT_THAT(response->headers(), HeaderValueOf(Headers::get().Server, "envoy"));
using testing::HasSubstr;
EXPECT_THAT(request->headers(),
            HeaderValueOf(Headers::get().AcceptEncoding, HasSubstr("gzip")));

HttpStatusIs

Tests that a HeaderMap argument has the provided HTTP status code. The status code can be passed in as a string or an integer.

Example:

EXPECT_THAT(response->headers(), HttpStatusIs("200"));

HeaderMapEqual and HeaderMapEqualRef

HeaderMapEqualRef tests that two HeaderMap arguments are equal. HeaderMapEqual is the same, but it compares two pointers to HeaderMaps, and the matcher's argument must be a HeaderMapPtr.

Example:

EXPECT_THAT(response->headers(), HeaderMapEqualRef(expected_headers));

ProtoEq, ProtoEqIgnoringField, RepeatedProtoEq

Tests equality of protobufs, with a variant that ignores the value (including presence) of a single named field. Another variant can be used to compare two instances of Protobuf::RepeatedPtrField element-by-element.

Example:

envoy::service::discovery::v3::DeltaDiscoveryRequest expected_request;
// (not shown: set some fields of expected_request...)
EXPECT_CALL(async_stream_, sendMessage(ProtoEqIgnoringField(expected_request, "response_nonce"), false));

response->mutable_resources()->Add();
response->mutable_resources()->Add();
response->mutable_resources()->Add();
// (not shown: do something to populate those empty added items...)
EXPECT_CALL(callbacks_, onConfigUpdate(RepeatedProtoEq(response->resources()), version));

IsSubsetOfHeaders and IsSupersetOfHeaders

Tests that one HeaderMap argument contains every header in another HeaderMap.

Examples:

EXPECT_THAT(response->headers(), IsSubsetOfHeaders(allowed_headers));
EXPECT_THAT(response->headers(), IsSupersetOfHeaders(required_headers));

Controlling time in tests

In Envoy production code, time and timers are managed via Event::TimeSystem, which provides a mechanism for querying the time and setting up time-based callbacks. Bypassing this abstraction in Envoy code is flagged as a format violation in CI.

In tests we use a derivation Event::TestTimeSystem which adds the ability to sleep or do a blocking timed wait on a condition variable. There are two implementations of the Event::TestTimeSystem interface: Event::TestRealTimeSystem, and Event::SimulatedTimeSystem. The latter is recommended for all new tests, as it helps avoid flaky tests on slow machines, and makes tests run faster.

Typically we do not want to have both real-time and simulated-time in the same test; that could lead to hard-to-reproduce results. Thus both implementations have a mechanism to enforce that only one of them can be instantiated at once. A runtime assertion occurs if an Event::TestRealTimeSystem and Event::SimulatedTimeSystem are instantiated at the same time. Once the time-system goes out of scope, usually at the end of a test method, the slate is clean and a new test-method can use a different time system.

There is also Event::GlobalTimeSystem, which can be instantiated in shared test infrastructure that wants to be agnostic to which TimeSystem is used in a test. When no TimeSystem is instantiated in a test, the Event::GlobalTimeSystem will lazy-initialize itself into a concrete TimeSystem. Currently this is TestRealTimeSystem but will be changed in the future to SimulatedTimeSystem.

Benchmark tests

Envoy uses Google Benchmark for microbenchmarks. There are custom bazel rules, envoy_cc_benchmark_binary and envoy_benchmark_test, to execute them locally and in CI environments respectively. envoy_benchmark_test rules call the benchmark binary from a script which runs the benchmark with a minimal number of iterations and skipping expensive benchmarks to quickly verify that the binary is able to run to completion. In order to collect meaningful bechmarks, bazel run -c opt the benchmark binary target on a quiescent machine.

If you would like to detect when your benchmark test is running under the wrapper, call Envoy::benchmark::skipExpensiveBechmarks().