[apollo-server-cloudflare] Worker deployment fails when using listen()

[maggo]

apollo-server-cloudflare overrides the handy listen() method to set up addEventListener('fetch', …) but wrangler publish doesn't like this and fails with

Error: Something went wrong! Status: 400 Bad Request, Details {
  "result": null,
  "success": false,
  "errors": [
    {
      "code": 10021,
      "message": "No event handlers were registered. This script does nothing.\n"
    }
  ],
  "messages": []
}

I guess wrangler always requires a top-level addEventListener that is not registered in an async function.

I was able to fix this by implementing my own top-level event handler and copying the code from listen()

Example

https://github.com/maggo/apollo-cloudflare-issue

Clone, yarn, add account_id in wrangler.toml and wrangler publish. You'll receive the same error

[glasser]

As of Apollo Server 4, we maintain a stable API for building web framework integrations instead of a specific integration with CloudFlare Workers (which, let's be honest, was an experimental integration that one person wrote over the weekend 4.5 years ago and we never touched again — it's impressive that it worked as well as it did for anyone!). We hope to see CloudFlare Workers users build and maintain an integration against our new API!