Passing options object to http.createServer stops server responding

[brokencube]

• Version: node 8.12.0
• Platform: macOS Sierra 10.12.6 (16G1510)
• uname -a: Darwin [USERNAME] 16.7.0 Darwin Kernel Version 16.7.0: [DATETIME]; root:xnu-3789.73.14~1/RELEASE_X86_64 x86_64
• Subsystem: http

Example:

"use strict";
const http = require('http');
const options = {};

const server = http.createServer(options, (req, res) => {
    console.log('got here', req, res);
});
server.listen(4000);

# telnet localhost 4000
GET /

Expected: Console.log of "got here"
Actual: No output.

It also doesn't work is options is set to:

const options = {
  IncomingMessage: http.IncomingMessage,
  ServerResponse: http.ServerResponse
}

Is working as expected in v10.13.0

Using node inspect the difference in execution between v8 and v10 seems to be this line:
_http_common.js:117
return parser.onIncoming(parser.incoming, shouldKeepAlive);

Which should be returning a IncomingMessage object, but instead returns nothing. But that's as far as I could get with this.

Calling without the options object - i.e.:

const server = http.createServer((req, res) => {})

Works as expected

[Trott]

@nodejs/http

[oyyd]

node/doc/api/http.md

Lines 1779 to 1786 in bd8c107

	## http.createServer([options][, requestListener])
	<!-- YAML
	added: v0.1.13
	changes:
	- version: v9.6.0
	pr-url: https://github.com/nodejs/node/pull/15752
	description: The `options` argument is supported now.
	-->

In fact, The options argument is added in v9.6.0 so that it don't pass requestListener to http.Server in v8.12.0.

[brokencube]

Looks like it was backported for v8.12.0:
#21593

and from the Release notes from v8.12.0:

add options to http.createServer() (Peter Marton) #15752

[oyyd]

Yeah, but the code of v8.x-staging seems not:

node/lib/http.js

Lines 33 to 35 in 26d145a

	function createServer(requestListener) {
	return new Server(requestListener);
	}

I'm not sure if I have missed something.

[brokencube]

Ah, apparently the http change got backed out of the merge for v8.12.0: #20456 (comment)

[apapirovski]

As expected given this is not a valid signature for that function.

[brokencube]

Well there is still an issue here... either:

1. That change that was backed out needs to be merged in so that the current 8.x documentation is correct, OR
2. The documentation for 8.x needs to be updated to have the correct (unchanged) function signature.

[apapirovski]

@brokencube Thanks for an update. That's helpful in moving this forward. I'm going to reopen and assign the appropriate labels.