Proper way to stop everything?

[cstoneham]

Thank you guys for the library, been looking for a good alternative to hardhat given that I use a viem / foundry stack, so happy to come across this.

--

On my global test teardown I have this:

  // close all of the connections
  console.log("closing anvil connections and stopping the server");
  server.closeAllConnections();
  await server.stop();

Is this the proper way to shut everything down? What's the difference between closing connections and stopping the server?

[cstoneham]

An related question I have is around best practices for managing orphaned connections.

Basically my setup is that I use prool to start/stop a global instance of a network fork:

vitest.config.mts

export default defineProject({
  test: {
    globals: true,
    setupFiles: ["./vitest.setup.ts"],
    globalSetup: ["./vitest.globalSetup.ts"],
...

vitest.globalSetup.ts

export async function teardown(): Promise<void> {
  const prisma = new PrismaClient({
    datasources: {
      db: { url: TEST_DATABASE_URL },
    },
  });

  // close all of the connections
  console.log("closing anvil connections and stopping the server");
  server.closeAllConnections();
  await server.stop();

  // close database connection
  console.log("disconnecting from the database");
  await prisma.$disconnect();
}

the issue is that if there's some exception in my tests, the teardown doesn't get called and there's a hanging network fork on 8545 which wasn't torn down, so when i start the tests again there's an exception on startup (port already in use).

at first I tried lsof -t -i:8545 | xargs kill to kill the process manually between tests, but then i noticed there was some parent respawning the connection on that port, so I'd get the PID from lsof -t -i:8545 => find the parent: ps -o ppid= -p <PID> => kill the parent and the child.

it's doable, but seems like I'm taking a very roundabout approach to dealing with connections, any suggestions?