Modernize child_process import in reconnecting socket test

[dynst]

One-third of #1724. Closes #1724

Looks like this test code was inherited from #246

https://github.com/webpack/webpack/releases/v5.92.0

webpack added a feature that adds node: prefixes "in runtime code," but still compile-time rejects it if it's already present in the code, and the target is a browser. Even if it's a dynamic import, which can fail.

[webmaster128]

Nice. Could you update to latest main and remove the dependency changes from the PR? Then we can get this in and close #1724 – great stuff

[webmaster128]

I think it is a bit less boilerplate and more natural to use (done) => with an inner unnamed async function. See e.g.

it("can subscribe to block header events", (done) => {
    pendingWithoutTendermint();

    const testStart = ReadonlyDate.now();

    (async () => {
      const events: responses.NewBlockHeaderEvent[] = [];
      const client = Tendermint37Client.create(rpcFactory());
      const stream = client.subscribeNewBlockHeader();
      expect(stream).toBeTruthy();
      // ...

in this codebase

[dynst]

Wait, how would that even work?

And ideally all this code would be refactored to be Promise-y instead of callback-y to avoid this boilerplate kludge that combines the 2 styles, do you not think this is a step in the right direction and a reminder to more fully rewrite the test later?

[webmaster128]

No super strong opinion. I was just remembering that we had those cases before and here it looks like a lot of boilerplate. But at the end of the day I am happy either way

[webmaster128]

See https://github.com/cosmos/cosmjs/blob/v0.34.0/packages/tendermint-rpc/src/comet38/comet38client.spec.ts#L633-L688 for the full test