From ff7745bb8794c4cc57fcda9f5dcb04c5f87f0bb4 Mon Sep 17 00:00:00 2001 From: Cole Lightfighter Date: Sun, 1 Oct 2017 12:11:59 -0600 Subject: [PATCH] Edited test readme for accuracy and depth Removed some erroneous instructions, and added information about different test suites. Signed-off-by: Cole Lightfighter --- tests/README.md | 183 ++++++++++++++++-------------------------------- 1 file changed, 61 insertions(+), 122 deletions(-) diff --git a/tests/README.md b/tests/README.md index b76d3507d..48a6c41a7 100644 --- a/tests/README.md +++ b/tests/README.md @@ -1,178 +1,117 @@ -# Crypto tests +# Running all tests -## Running crypto Perl tests - -Crypto tests require the Math::GMP Perl library, make sure it is installed on you system before running the tests. - -Installing dependencies (using cpan): +To run all tests, run: ``` -cpan -cpan> install Math::BigInt::GMP -cpan> install Digest::Keccak +cd /path/to/monero +make [-jn] debug-test # where n is number of compiler processes ``` -Running tests: +To test a release build, replace `debug-test` with `release-test` in the previous command. -``` -TESTPATH=/path/to/monero/tests -cd $TESTPATH -perl -I $TESTPATH cryptotest.pl -``` +# Core tests -Important: must include test path for perl to import cryptolib.pl +Core tests take longer than any other Monero tests, due to the high amount of computational work involved in validating core components. -## Writing new crypto tests +Tests are located in `tests/core_tests/`, and follow a straightforward naming convention. Most cases cover core functionality (`block_reward.cpp`, `chaingen.cpp`, `rct.cpp`, etc.), while some cover basic security tests (`double_spend.cpp` & `integer_overflow.cpp`). -[TODO] +To run only Monero's core tests (after building): -# Core tests +``` +cd build/debug/tests/core +ctest +``` -## Running core tests +To run the same tests on a release build, replace `debug` with `release`. -Monero uses the Google C++ Testing Framework (`gtest`) to write unit, integration and functional tests for core and other features of the project. -`gtest` runs on top of cmake, and you can run all tests by: -``` -cd /path/to/monero -make [-jn] debug-test # where n is number of compiler processes -``` +# Crypto Tests -To test a release build, replace `debug-test` with `release-test` in the previous command. +Crypto tests are located under the `tests/crypto` directory. + +- `crypto-tests.h` contains test harness headers +- `main.cpp` implements the driver for the crypto tests -One can also run individual test suites by building monero, then running `ctest` in test suite folders. +Tests correspond to components under `src/crypto/`. A quick comparison reveals the pattern, and new tests should continue the naming convention. -Run only the hash tests: +To run only Monero's crypto tests (after building): ``` -cd /path/to/monero -make [-j#] debug -cd build/debug/tests/hash +cd build/debug/tests/crypto ctest ``` -To run the same tests on a release build, replace `debug` with `release` in previous commands. - -## Writing new tests +To run the same tests on a release build, replace `debug` with `release`. -Based on local tests and Google's guide on creating [simple tests with gtest](https://github.com/google/googletest/blob/master/googletest/docs/Primer.md#simple-tests) +# Daemon tests -Tests consist of a test harness (defined with the TEST() macro), and the test body consisting of gtest assertions. - -Example of a test harness: +[TODO] -``` -TEST(test_case_name, test_name) { - ... test body ... +# Functional tests -} -``` +[TODO] -As an example in Monero's [crypto unit test](./unit_tests/crypto.cpp): +# Fuzz tests -``` -TEST(Crypto, Ostream) -{ - EXPECT_TRUE(is_formatted()); - EXPECT_TRUE(is_formatted()); - EXPECT_TRUE(is_formatted()); - EXPECT_TRUE(is_formatted()); - EXPECT_TRUE(is_formatted()); - EXPECT_TRUE(is_formatted()); - EXPECT_TRUE(is_formatted()); -} +Fuzz tests are written using American Fuzzy Lop (AFL), and located under the `tests/fuzz` directory. -``` +An additional helper utility is provided `contrib/fuzz_testing/fuzz.sh`. AFL must be installed, and some additional setup may be necessary for the script to run properly. -The assertions inside the test harness are a bit complex, but fairly straightforward. +# Hash tests -- `is_formatted()` is a polymorphic function that accepts the various types of structs defined in [crypto/hash.h](../src/crypto/hash.h). +Hash tests exist under `tests/hash`, and include a set of target hashes in text files. -Just above the test harness, we have the definition for `is_formatted`: +To run only Monero's hash tests (after building): ``` - template - bool is_formatted() - { - T value{}; - - static_assert(alignof(T) == 1, "T must have 1 byte alignment"); - static_assert(sizeof(T) <= sizeof(source), "T is too large for source"); - static_assert(sizeof(T) * 2 <= sizeof(expected), "T is too large for destination"); - std::memcpy(std::addressof(value), source, sizeof(T)); - - std::stringstream out; - out << "BEGIN" << value << "END"; - return out.str() == "BEGIN<" + std::string{expected, sizeof(T) * 2} + ">END"; - } +cd build/debug/tests/hash +ctest ``` -`T value {}` produces the data member of the struct (`hash8` has `char data[8]`), which runs a number of tests to ensure well structured hash data. +To run the same tests on a release build, replace `debug` with `release`. -Let's write a new test for the keccak function: +# Libwallet API tests -``` - bool keccak_harness() - { - size_t inlen = sizeof(source); - int mdlen = (int)sizeof(md); - int ret = keccak(source, inlen, md, mdlen); - if (md[0] != 0x00) - { - return true; - } - else if (!ret) - { - return true; - } - else - { - return false; - } - } -``` +[TODO] -This is a basic test that ensures `keccak()` returns successfully when given proper input. It reuses the `source` array for input, and a new byte array `md` for storing the hash digest. Full source is in the [crypto unit test](./unit_tests/crypto.cpp). +# Net Load tests -Now let's create a new test harness: +[TODO] -``` -TEST(Crypto, Keccak) -{ - # ... - EXPECT_TRUE(keccak_harness()); -} +# Performance tests -``` +Performance tests are located in `tests/performance_tests`, and test features for performance metrics on the host machine. -This creates a new test under the `Crypto` test case named `Keccak`. The harness includes one assertion `EXPECT_TRUE(keccak_harness())`, which invokes `keccak_harness()`. More complex logic can be added to test various functionality of the `Keccak` library. - -To run the new test: +To run only Monero's performance tests (after building): ``` -cd /path/to/monero -make -jn debug # if no debug build exists -cd build/debug/tests/unit_test -make -jn -make -jn test +cd build/debug/tests/performance_tests +./performance_tests ``` -# Fuzz tests +If the `performance_tests` binary does not exist, try running `make` in the `build/debug/tests/performance_tests` directory. -## Running fuzz tests +To run the same tests on a release build, replace `debug` with `release`. -``` -cd /path/to/monero -make [-jn] fuzz # where n is number of compiler processes -``` +# Unit tests -or +Unit tests are defined under the `tests/unit_tests` directory. Independent components are tested individually to ensure they work properly on their own. + +To run only Monero's unit tests (after building): ``` -cd path/to/monero -./contrib/fuzz_testing/fuzz.sh +cd build/debug/tests/unit_tests +ctest ``` +To run the same tests on a release build, replace `debug` with `release`. + +# Writing new tests + +## Test hygiene + +When writing new tests, please implement all functions in `.cpp` or `.c` files, and only put function headers in `.h` files. This will help keep the fairly complex test suites somewhat sane going forward. + ## Writing fuzz tests [TODO]