test: fix flaky interop tests - address already in use

packages/integration-tests/.aegir.js

@@ -1,4 +1,7 @@
 import { execa } from 'execa'
+import fs from 'node:fs'
+import net from 'node:net'
+import { randomUUID } from 'node:crypto'
 import pDefer from 'p-defer'
 
 /** @type {import('aegir').PartialOptions} */
@@ -169,11 +172,11 @@ async function createGoLibp2pRelay () {
   const { defaultLogger } = await import('@libp2p/logger')
 
   const log = defaultLogger().forComponent('go-libp2p')
-  const controlPort = Math.floor(Math.random() * (50000 - 10000 + 1)) + 10000
-  const apiAddr = multiaddr(`/ip4/127.0.0.1/tcp/${controlPort}`)
+  const { controlSocketPath, apiAddr, daemonListenAddr } = await getControlEndpoint(multiaddr)
   const deferred = pDefer()
+
   const proc = execa(p2pd(), [
-    `-listen=${apiAddr.toString()}`,
+    `-listen=${daemonListenAddr}`,
     // listen on TCP, WebSockets and WebTransport
     '-hostAddrs=/ip4/127.0.0.1/tcp/0,/ip4/127.0.0.1/tcp/0/ws,/ip4/127.0.0.1/udp/0/quic-v1/webtransport',
     '-noise=true',
@@ -185,8 +188,17 @@ async function createGoLibp2pRelay () {
       GOLOG_LOG_LEVEL: 'debug'
     }
   })
+
   proc.catch(() => {
     // go-libp2p daemon throws when killed
+  }).finally(() => {
+    if (controlSocketPath != null) {
+      fs.rmSync(controlSocketPath, { force: true })
+    }
+  })
+
+  proc.once('exit', code => {
+    deferred.reject(new Error(`go-libp2p daemon exited before startup (code: ${code ?? 'unknown'})`))
   })
 
   proc.stdout?.on('data', (buf) => {
@@ -202,6 +214,7 @@ async function createGoLibp2pRelay () {
 
     log(str)
   })
+
   await deferred.promise
 
   const daemonClient = createClient(apiAddr)
@@ -214,3 +227,52 @@ async function createGoLibp2pRelay () {
     proc
   }
 }
+
+async function getControlEndpoint (multiaddr) {
+  if (process.platform === 'win32') {
+    const controlPort = await getAvailablePort()
+    const apiAddr = multiaddr(`/ip4/127.0.0.1/tcp/${controlPort}`)
+
+    return {
+      controlPort,
+      apiAddr,
+      daemonListenAddr: apiAddr.toString()
+    }
+  }
+
+  const controlSocketPath = `/tmp/p2pd-${randomUUID()}.sock`
+  const apiAddr = multiaddr(`/unix/${encodeURIComponent(controlSocketPath)}`)
+
+  return {
+    controlSocketPath,
+    apiAddr,
+    daemonListenAddr: `/unix${controlSocketPath}`
+  }
+}
+
+async function getAvailablePort () {
+  return new Promise((resolve, reject) => {
+    const server = net.createServer()
+
+    server.once('error', reject)
+    server.listen(0, '127.0.0.1', () => {
+      const addr = server.address()
+
+      if (addr == null || typeof addr === 'string') {
+        server.close(() => {
+          reject(new Error('could not allocate control port'))
+        })
+        return
+      }
+
+      server.close(err => {
+        if (err != null) {
+          reject(err)
+          return
+        }
+
+        resolve(addr.port)
+      })
+    })
+  })
+}

packages/integration-tests/test/interop.ts

@@ -1,4 +1,6 @@
 import fs from 'fs'
+import { randomUUID } from 'node:crypto'
+import net from 'node:net'
 import { noise } from '@chainsafe/libp2p-noise'
 import { yamux } from '@chainsafe/libp2p-yamux'
 import { circuitRelayServer, circuitRelayTransport } from '@libp2p/circuit-relay-v2'
@@ -38,14 +40,20 @@ import type { Libp2pOptions, ServiceFactoryMap } from 'libp2p'
  * ```
  */
 
+interface ControlEndpoint {
+  controlSocketPath?: string
+  controlPort?: number
+  apiAddr: ReturnType<typeof multiaddr>
+  daemonListenAddr: string
+}
+
 async function createGoPeer (options: SpawnOptions): Promise<Daemon> {
-  const controlPort = Math.floor(Math.random() * (50000 - 10000 + 1)) + 10000
-  const apiAddr = multiaddr(`/ip4/127.0.0.1/tcp/${controlPort}`)
+  const { controlSocketPath, controlPort, apiAddr, daemonListenAddr } = await getControlEndpoint()
 
-  const log = logger(`go-libp2p:${controlPort}`)
+  const log = logger(`go-libp2p:${controlSocketPath ?? controlPort}`)
 
   const opts = [
-    `-listen=${apiAddr.toString()}`
+    `-listen=${daemonListenAddr}`
   ]
 
   if (options.noListen === true) {
@@ -92,13 +100,21 @@ async function createGoPeer (options: SpawnOptions): Promise<Daemon> {
     opts.push('-muxer=yamux')
   }
 
-  const deferred = pDefer()
+  const deferred = pDefer<void>()
   const proc = execa(p2pd(), opts, {
     env: {
       GOLOG_LOG_LEVEL: 'debug'
     }
   })
 
+  proc.catch(() => {
+    // go-libp2p daemon throws when killed
+  })
+
+  proc.once('exit', code => {
+    deferred.reject(new Error(`go-libp2p daemon exited before startup (code: ${code ?? 'unknown'})`))
+  })
+
   proc.stdout?.on('data', (buf: Buffer) => {
     const str = buf.toString()
     log(str)
@@ -119,8 +135,63 @@ async function createGoPeer (options: SpawnOptions): Promise<Daemon> {
     client: createClient(apiAddr),
     stop: async () => {
       proc.kill()
+
+      if (controlSocketPath != null) {
+        fs.rmSync(controlSocketPath, { force: true })
+      }
+    }
+  }
+}
+
+async function getControlEndpoint (): Promise<ControlEndpoint> {
+  // Use tcp control ports on windows
+  if (process.platform === 'win32') {
+    const controlPort = await getAvailablePort()
+    const apiAddr = multiaddr(`/ip4/127.0.0.1/tcp/${controlPort}`)
+
+    return {
+      controlPort,
+      apiAddr,
+      daemonListenAddr: apiAddr.toString()
     }
   }
+
+  // Use unix domain sockets on non-windows - avoids port clashes
+  const controlSocketPath = `/tmp/p2pd-${randomUUID()}.sock`
+  const apiAddr = multiaddr(`/unix/${encodeURIComponent(controlSocketPath)}`)
+
+  return {
+    controlSocketPath,
+    apiAddr,
+    daemonListenAddr: `/unix${controlSocketPath}`
+  }
+}
+
+async function getAvailablePort (): Promise<number> {
+  return new Promise((resolve, reject) => {
+    const server = net.createServer()
+
+    server.once('error', reject)
+    server.listen(0, '127.0.0.1', () => {
+      const addr = server.address()
+
+      if (addr == null || typeof addr === 'string') {
+        server.close(() => {
+          reject(new Error('could not allocate control port'))
+        })
+        return
+      }
+
+      server.close(err => {
+        if (err != null) {
+          reject(err)
+          return
+        }
+
+        resolve(addr.port)
+      })
+    })
+  })
 }
 
 async function createJsPeer (options: SpawnOptions): Promise<Daemon> {