SSH MCP Server

/** * SESSION_BUSY Resource Leak Tests * * Testing the critical resource leak discovered in mcp-ssh-server.ts * where successful MCP command execution NEVER calls completeCommandExecution(), * leaving sessions permanently locked in EXECUTING_COMMAND state. */ import { MCPSSHServer } from '../../src/mcp-ssh-server'; import { SSHConnectionManager } from '../../src/ssh-connection-manager'; import { TerminalSessionStateManager } from '../../src/terminal-session-state-manager'; describe('SESSION_BUSY Resource Leak Prevention', () => { let mcpServer: MCPSSHServer; let sshManager: SSHConnectionManager; let stateManager: TerminalSessionStateManager; const TEST_SESSION_NAME = 'test-resource-leak-session'; beforeEach(async () => { sshManager = new SSHConnectionManager(); stateManager = new TerminalSessionStateManager(); mcpServer = new MCPSSHServer({}, sshManager, stateManager); // Create test SSH session await sshManager.createConnection({ name: TEST_SESSION_NAME, host: 'localhost', username: 'jsbattig', keyFilePath: '/home/jsbattig/.ssh/id_ed25519' }); }); afterEach(async () => { // Let sessions clean up naturally - no explicit cleanup method found }); describe('Resource Leak Reproduction Tests', () => { test('FAILING TEST: Successful MCP command execution leaves session locked', async () => { // This test should FAIL initially due to the resource leak // Execute first MCP command - should succeed const firstResult = await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'echo "first command"' }) as any; // Verify first command succeeded expect(firstResult.success).toBe(true); expect(firstResult.result.stdout).toContain('first command'); // CRITICAL TEST: Second MCP command should work but will fail due to resource leak const secondResult = await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'echo "second command"' }) as any; // This assertion should now PASS with the resource leak fix expect(secondResult.success).toBe(true); expect(secondResult.result.stdout).toContain('second command'); expect(secondResult.error).toBeUndefined(); }); test('FAILING TEST: Session state remains EXECUTING_COMMAND after successful execution', async () => { // Execute MCP command await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'pwd' }); // CRITICAL CHECK: Session should be WAITING_FOR_COMMAND, not EXECUTING_COMMAND const sessionState = stateManager.getSessionState(TEST_SESSION_NAME); expect(sessionState).toBe('WAITING_FOR_COMMAND'); // This will FAIL - should be EXECUTING_COMMAND }); test('FAILING TEST: Multiple sequential MCP commands should all succeed', async () => { const commands = ['pwd', 'whoami', 'date', 'echo "test"']; for (let i = 0; i < commands.length; i++) { const result = await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: commands[i] }) as any; // All commands should succeed, but only first will work due to resource leak expect(result.success).toBe(true); expect(result.error).toBeUndefined(); } }); }); describe('Resource Management Consistency Tests', () => { test('Error path properly cleans up session state', async () => { // Execute command that will fail const result = await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'nonexistent-command-that-will-fail' }) as any; // Error should be handled (command exists but fails) expect(result).toBeDefined(); // Session should be available for next command const nextResult = await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'echo "after error"' }) as any; expect(nextResult.success).toBe(true); expect(nextResult.error).toBeUndefined(); }); test('Session state consistency across execution paths', async () => { // Initial state should be WAITING_FOR_COMMAND let sessionState = stateManager.getSessionState(TEST_SESSION_NAME); expect(sessionState).toBe('WAITING_FOR_COMMAND'); // After successful execution, should return to WAITING_FOR_COMMAND await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'echo "success test"' }); sessionState = stateManager.getSessionState(TEST_SESSION_NAME); expect(sessionState).toBe('WAITING_FOR_COMMAND'); }); }); describe('RAII Pattern Validation', () => { test('Resource acquisition and release symmetry', async () => { const initialState = stateManager.getSessionState(TEST_SESSION_NAME); expect(initialState).toBe('WAITING_FOR_COMMAND'); // Execute command await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'echo "raii test"' }); // State should be restored to initial condition const finalState = stateManager.getSessionState(TEST_SESSION_NAME); expect(finalState).toBe('WAITING_FOR_COMMAND'); }); test('Exception safety - state cleanup on unexpected errors', async () => { // Mock SSH manager to throw unexpected error const originalExecute = sshManager.executeCommand; sshManager.executeCommand = jest.fn().mockRejectedValue(new Error('Unexpected SSH error')); try { await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'echo "error test"' }); } catch (error) { // Expected to throw } // Restore original method sshManager.executeCommand = originalExecute; // Session should still be available for next command const nextResult = await mcpServer.callTool('ssh_exec', { sessionName: TEST_SESSION_NAME, command: 'echo "after exception"' }) as any; expect(nextResult.success).toBe(true); expect(nextResult.error).toBeUndefined(); }); }); });