E2E Testing of C++ Applications: From Theory to Combat CI/CD

Why "just unit tests" is no longer enough?

You wrote perfect C++ code. Compilation passed, unit tests are green, code review approved. You roll out the release — and an hour later users complain: "the button doesn't work", "the database is not responding", "the GUI freezes". Sound familiar?

The problem is that unit tests check a function, but they don't check a scenario. They are isolated. But in the real world, your C++ code interacts with the network, file system, GPU, external APIs, and other threads. This is where E2E testing of C++ applications (End-to-End testing) comes into play.

In this article, we will break down how to turn the chaos of integration errors into a streamlined system of automated checks. You will learn which tools actually work, how to write tests that won't fall apart on the first run, and how to embed E2E into a CI/CD pipeline. Let's go!

What is E2E testing in the context of C++?

E2E testing is the verification of a complete user scenario "from start to finish". Unlike unit tests, which check a single function, an E2E test runs your application (or its critically important part) in an environment as close to production as possible.

For C++, this could mean:

\n
• Desktop application: launching a window, clicking the "Submit" button, verifying that the data went to the database.
\n
• Server application: sending an HTTP request to your REST API, checking the response and the database state.
\n
• Embedded/Software: applying a signal to the input, checking the output signal after a certain time.
\n

The key difference is the real environment. No mock objects for the database or network. Only live hardware (or Docker).

Toolkit: What actually works for C++ in 2024?

Let's get straight to the point. Here is the stack we use in production:

\n
• Catch2 or Google Test — for writing the test structure (Arrange-Act-Assert).
\n
• Docker — for spinning up a clean environment (DB, queues, dependencies).
\n
• Pytest + subprocess — for orchestration (yes, Python, but it's ideal for gluing C++ binaries together).
\n
• Selenium/Appium — if you have a GUI on Qt or WebView.
\n
• CMake + CTest — for running tests and collecting results.
\n

Don't try to write E2E tests only in C++. It's painful. Use Python as glue — it's faster for integration scenarios.

Practice: Writing an E2E test for a REST API in C++

Let's say we have a C++ server using cpp-httplib (or crow, or drogon). It accepts POST requests at /api/user and saves the user to PostgreSQL.

Let's write an E2E test that:

\n
1. Spins up a Docker container with PostgreSQL.
\n
2. Compiles and runs our server.
\n
3. Sends an HTTP request via curl (or a C++ client).
\n
4. Checks that the user appeared in the database.
\n

Step 1: Docker-compose for the test environment

version: '3.8'\nservices:\n  db:\n    image: postgres:15\n    environment:\n      POSTGRES_DB: testdb\n      POSTGRES_USER: testuser\n      POSTGRES_PASSWORD: testpass\n    ports:\n      - "5432:5432"\n  app:\n    build: .\n    depends_on:\n      - db\n    ports:\n      - "8080:8080"\n    environment:\n      DB_HOST: db\n

Step 2: Test in C++ using Catch2

// test_e2e_api.cpp\n#include <catch2/catch_test_macros.hpp>\n#include <httplib.h>\n#include <nlohmann/json.hpp>\n#include <cstdlib>\n\nTEST_CASE("E2E: Create user via API and verify in DB") {\n    // Arrange\n    httplib::Client cli("localhost", 8080);\n    nlohmann::json user = {\n        {"name", "John Doe"}