Broken gtest support

[yfinkelstein]

There are 2 issues in version 0.5.0:

1. The header files tokenizer.hpp and utils.hpp included from assert-gtest.hpp are missing - they are not exported in conan package
2. Worse - there is compilation error:

ASSERT(1 + 1 == 3, "my assert");

error: void value not ignored as it ought to be
  195 |     ASSERT(1 + 1 == 3, "my assert");
      |     ^

A more detailed info provided by clangd:

Cannot initialize return object of type 'int' with an rvalue of type 'void'. 
clang(init_conversion_failed)

macro ASSERT
provided by <libassert/assert-gtest.hpp>

#define ASSERT(...)                        \
    do {                                   \
        try {                              \
            LIBASSERT_ASSERT(__VA_ARGS__); \
            SUCCEED();                     \
        } catch (std::exception & e) {     \
            FAIL() << e.what();            \
        }                                  \
    } while (false)

// Expands to
do {
    try {
        do {
            auto libassert_decomposer = libassert ::detail ::expression_decomposer(
                libassert ::detail ::expression_decomposer{} << 1 + 1 == 3);
            if (__builtin_expect_with_probability((!static_cast<bool>(libassert_decomposer.get_value())),
                    (0),
                    1)) {
                libassert ::ERROR_ASSERTION_FAILURE_IN_CONSTEXPR_CONTEXT();
                using libassert_params_t = libassert ::detail ::assert_static_parameters;
                const libassert_params_t *libassert_params = []() -> const libassert_params_t * {
                    static constexpr std ::string_view libassert_arg_strings[] = {"\"my assert\"", "", ""};
                    static constexpr libassert_params_t _libassert_params = {
                        "ASSERT",
                        libassert ::assert_type ::assertion,
                        "1 + 1 == 3",
                        {},
                        {libassert_arg_strings, sizeof(libassert_arg_strings) / sizeof(std ::string_view)},
                    };
                    return &_libassert_params;
                }();
                if constexpr (sizeof libassert_decomposer > 32) {
                    libassert ::detail ::process_assert_fail(libassert_decomposer,
                        libassert_params,
                        "my assert",
                        libassert ::detail ::pretty_function_name_wrapper{__extension__ __PRETTY_FUNCTION__});
                } else {
                    libassert ::detail ::process_assert_fail_n(std ::move(libassert_decomposer),
                        libassert_params,
                        "my assert",
                        libassert ::detail ::pretty_function_name_wrapper{__extension__ __PRETTY_FUNCTION__});
                }
            }
        } while (false);
        ::testing ::internal ::AssertHelper(::testing ::TestPartResult ::kSuccess,
            "/home/yfinkelstein/baydb/storage_server/src/tests/test-fiber-uring-grpc.cpp",
            195,
            "Succeeded") = ::testing ::Message();
    } catch (std ::exception& e) {
        return ::testing ::internal ::AssertHelper(::testing ::TestPartResult ::kFatalFailure,
                   "/home/yfinkelstein/baydb/storage_server/src/tests/test-fiber-uring-grpc.cpp",
                   195,
                   "Failed") = ::testing ::Message() << e.what();
    }
} while (false)

[yfinkelstein]

The problem is with return under catch and the bottom of the above. Just remove this return and it compiles

[yfinkelstein]

Looks like the return comes from gtest FAIL(). This compiles and works as expected: EXPECT(1 + 1 == 3, "my assert");

[yfinkelstein]

gtest macro chain: FAIL() -> GTEST_FAIL()->GTEST_FATALFAILURE("Failed") -> return GTEST_MESSAGE_(message, ::testing::TestPartResult::kFatalFailure)

[yfinkelstein]

Ouh, this is because I was having this ASSERT outside of the gtest unit test code! It does work inside of unit test body. So, it's just not intended to be used like that. Just add a disclosure. Not a bug as such.

By missing headers in conan pckage is a real problem

[jeremy-rifkin]

Hello, thanks for opening this issue! The missing includes in the gtest header are a copy-paste mistake on my part, I’ll fix this asap.

[jeremy-rifkin]

Fixed on main, will do a 2.0.1 patch at some point soon