Files
viem/test/setup.global.ts
jxomandGitHub 940db54cc9 feat: viem/tempo (#4147)
* feat: tempo

* chore: fmt

* wip

* chore: up

* chore: u

* chore: u

* chore: u

* chore: up

* refactor: defineChain; rm assignKeyAuthorization

* fmt

* refactor: test config; nonce key store

* chore: fmt

* chore: fmt

* chore: up

* chore: fmt

* wip: tests

* chore: knip; up tests

* chore: knip; up tests

* wip: require account on walletNamespaceCompat

* ci: fix

* ci: up

* ci: up

* ci: up

* tests: up

* tests: up

* chore: up vocs

* tests: up

* fix: inference

* chore: up nonce

* chore: up

* chore: up

* chore: up

* chore: submodule + agents

* chore: up biome

* chore: up

* chore: up

* ci: up

* chore: up

* chore: up

* chore: up

* feat: enshrine signature envelope verification

* docs

* chore: up

* chore: up

* chore: changeset

* docs: chain

* chore: bump size
2025-12-16 16:27:28 +11:00

31 lines
1.6 KiB
TypeScript

import * as executionInstances from './src/anvil.js'
import * as bundlerInstances from './src/bundler.js'
export default async function () {
if (process.env.SKIP_GLOBAL_SETUP) return
// Using this proxy, we can parallelize our test suite by spawning multiple "on demand" anvil
// instances and proxying requests to them. Especially for local development, this is much faster
// than running the tests serially.
//
// In vitest, each thread is assigned a unique, numerical id (`process.env.VITEST_POOL_ID`). We
// append this id to the local rpc url (e.g. `http://127.0.0.1:8545/<ID>`).
//
// Whenever a request hits the proxy server at this url, it spawns (or reuses) an anvil instance
// at a randomly assigned port and proxies the request to it. The anvil instance is added to a
// [id:port] mapping for future request and is kept alive until the test suite finishes.
//
// Since each thread processes one test file after the other, we don't have to worry about
// non-deterministic behavior caused by multiple tests hitting the same anvil instance concurrently
// as long as we avoid `test.concurrent()`.
//
// We still need to remember to reset the anvil instance between test files. This is generally
// handled in `setup.ts` but may require additional resetting (e.g. via `afterAll`), in case of
// any custom per-test adjustments that persist beyond `anvil_reset`.
const shutdown = await Promise.all([
...Object.values(executionInstances).map((instance) => instance.start()),
...Object.values(bundlerInstances).map((instance) => instance.start()),
])
return () => Promise.all(shutdown.map((fn) => fn()))
}