run-e2e-tests

name	run-e2e-tests
description	Run the end-to-end integration test suite for the CodeScene MCP Server. Use when you need to verify that all MCP tools work correctly against the real CLI and server infrastructure.
metadata	{"audience":"contributors","language":"rust"}

Purpose

Use this skill when you need to run the full e2e test suite to verify MCP server behavior end-to-end. These tests launch the compiled MCP server binary, communicate over JSON-RPC stdio, and exercise real tool calls against the CodeScene API.

Prerequisites

• Rust toolchain (cargo)
• CS_ACCESS_TOKEN environment variable set with a valid CodeScene access token
• git available in PATH
• Docker (only for Docker backend)

How to Run

Full suite (default — static/Cargo backend)

cargo test --test e2e

This builds the release binary (or uses CS_MCP_EXECUTABLE if set) and runs all registered test modules.

Run a specific test

cargo test --test e2e test_business_case_basic_response

Run all tests matching a pattern

cargo test --test e2e test_analytics

Include ignored tests (e.g., stress tests)

cargo test --test e2e -- --ignored

Run with a specific backend

# Static/Cargo backend (default)
CS_MCP_BACKEND=static cargo test --test e2e

# Docker backend
CS_MCP_BACKEND=docker cargo test --test e2e

# npm backend
CS_MCP_BACKEND=npm cargo test --test e2e

Use a pre-built executable (skip build)

CS_MCP_EXECUTABLE=target/release/cs-mcp cargo test --test e2e

Test Modules

Module	What it tests
analyze_change_set	analyze_change_set tool
analytics_environment_override	CS_ENVIRONMENT override
analytics_tracking	Analytics event sending and enrichment
bundled_docs	explain_code_health and explain_code_health_productivity tools
business_case	code_health_refactoring_business_case tool
cloudfront_headers	CloudFront API client headers
configure	get_config and set_config tools
enabled_tools	Tool filtering via CS_ENABLED_TOOLS
error_logging	Error telemetry redaction and file logging
git_subtree	Git subtree repository support
git_worktree	Git worktree repository support
platform_specific	Path handling (absolute, relative, symlinks, spaces, unicode)
relative_paths	Relative path resolution
require_access_token	Access token validation
shutdown_during_handshake	Graceful shutdown (stdin close + SIGTERM)
skill_resources	Skill resources, listing, reading, and syncing
ssl_api_ca_bundle	SSL CA bundle configuration for API calls
ssl_cli_truststore	SSL truststore args for CLI
standalone_license	Standalone license tool filtering
stress_code_health_review	Stress test (250 iterations, #[ignore])
version_check	Version check tool and background fetch

All modules are declared in tests/e2e/tests/mod.rs and wrappers are in tests/e2e/main.rs.

Infrastructure

File	Role
tests/e2e/main.rs	Entry point, #[test] wrappers, setup(), find_or_build_executable(), make_client()
tests/e2e/tests/mod.rs	Module declarations and infrastructure re-exports
tests/e2e/mcp_client.rs	MCPClient — JSON-RPC over stdio
tests/e2e/server_backends.rs	ServerBackend trait + 3 backend implementations
tests/e2e/file_utils.rs	create_git_repo(), create_temp_dir()
tests/e2e/response_parsers.rs	extract_result_text(), extract_code_health_score()
tests/e2e/fixtures.rs	Sample code with known Code Health characteristics
tests/e2e/tests/fake_http_server.rs	FakeHttpServer for intercepting API calls
tests/e2e/tests/fake_https_server.rs	FakeHttpsServer for SSL tests

Interpreting Results

Standard cargo test output with pass/fail per test:

test test_business_case_basic_response ... ok
test test_analytics_events_are_sent ... ok
test test_platform_symlinks ... ignored  (Docker)

Exit code is 0 if all pass, non-zero otherwise.

Troubleshooting

• CS_ACCESS_TOKEN not set: Export it before running. Some tests intentionally use invalid tokens but the env var must exist.
• Build failures: Run cargo build --release separately to isolate build errors.
• Timeout on tool calls: Default timeout is 60s per tool call. CodeScene API latency may cause timeouts — not a code bug.
• Docker tests fail: Ensure Docker is running and host.docker.internal resolves.
• Tests marked ignored: Stress tests are #[ignore] by default. Run with --ignored to include them.
• Platform-specific tests skip in Docker: Tests using skip_if_docker() print a message and return early — this is expected.