Bug: service goes wrong (including not accessible) when call app.stop and app.start repeatedly

[TimeBather]

最小复现:
向tests/service.spec.js的normal service测试结尾加入下列两行

    app.plugin(Foo)
    expect(app.foo).to.be.instanceOf(Foo)

运行结果:

  31 passing (70ms)
  1 failing

  1) Service
       normal service:
     AssertionError: expected undefined to be an instance of Foo
      at Context.<anonymous> (tests\service.spec.ts:25:27)

版本:

• cordis 1.4.2 (from git clone ,master branch)

[shigma]

It is working as expected. A normal service waits for an asynchronous service.start(), so it is not available until next tick.

app.plugin(Foo)
await new Promise(resolve => setTimeout(resolve, 0))
expect(app.foo).to.be.instanceOf(Foo)

[TimeBather]

那我想请问一下,是不是每一次app.plugin完成之后都需要start
我感觉可能是我刚刚的最小复现给错了

[shigma]

app.start() only needs to be called once.

[TimeBather]

我再去检查一下,这个问题确实是存在的,我使用了多次app.start()和app.stop()之后,service确实没有被启动,constructor也没有被执行,但包裹这个service插件是正常被执行的

[shigma]

I haven't tested whether services continue to work after app.stop(). By current design, you really should not call these two methods repeatedly.

[TimeBather]

OK.I think that I can use the dispose API to dispose all plugins mounted on the instance instead of app.stop().It seems to work.,thank you for your answer.