A couple of bugs I faced today (`dev`, `publish`)

Today on my personal laptop, I used wrangler and faced some issues. If necessary I’ll file separate issues for each of these problems, but just filing these right now while the details are still fresh -

• I made a file index.ts with contents

  export default {
    fetch() {
      return new Response("hello world");
    },
  };
  

  and ran npx wrangler@beta dev index.ts on it. It returned -

  No account id found, quitting...
  

  I think this means that both the oauth_code and refresh_code were invalid. In such a scenario, we should popup the login flow again. I manually ran npx wrangler@beta login and ‘fixed’ this issue.
  • Fix in https://github.com/cloudflare/wrangler2/pull/709
• I then tried to publish it with npx wrangler publish index.js --name devto --latest (and I didn’t have a script named devto already published). It did nothing for a long time (say, 20 seconds), and then this error got logged

  Cannot read properties of undefined (reading 'forEach')
  

  I think this is pointing to this line of code https://github.com/cloudflare/wrangler2/blob/b63efe60646c8c955f4df4f2ce1d87ce9cc84ba3/packages/wrangler/src/cfetch/index.ts#L93
  • We should include stacks in our errors here https://github.com/cloudflare/wrangler2/blob/a5537f147e61b046e141e06d1864ffa62e1f2673/packages/wrangler/src/index.tsx#L2034-L2045 (logging e.stack instead of e.message). If we think these are noisy, then we should handle expected errors properly, and exit with process.exit(1) instead of throwing.
  • It looks like there are failure modes for api calls we haven’t considered. Specifically, it’s possible for the api to not follow the api schema (which, makes sense, networks suck). We should be more defensive in how we handle data structures off the network (In this case, testing whether messages exists on the response)
  • running the command again published the worker successfully.
• When running dev, using the hotkey l to toggle between edge - local - edge logged this error

  17:48:12 Error: listen EADDRINUSE: address already in use :::8787
    at Server.setupListenHandle [as _listen2] (node:net:1334:16)
    at listenInCluster (node:net:1382:12)
    at Server.listen (node:net:1469:7)
    ...
  

  This means that we’re not shutting down the miniflare server cleanly, or maybe we’re trying to startup the remote proxy too soon after switching (before the miniflare server has closed).
  • fix in https://github.com/cloudflare/wrangler2/pull/332

[threepointone]

This should be in a better place now. We still need to handle error messages better and present them in a better manner, tracking that in https://github.com/cloudflare/wrangler2/issues/377. Closing this issue. @tanguyantoine let us know if you face the issue again, thank you!

[tanguyantoine]

Sorry I switched back on something else, I might give it another shot later