http: DRY ClientRequest.prototype._deferToConnect

[ahoym]

Came across some repetitive lines when reviewing code.

Logic for calling the passed in socket method and/or cb was duplicated. Extracted these calls into a separate function, callSocketMethod, and invoked this new function in place of the replaced code.

[cjihrig]

Can you make this function callSocketMethod()

[ahoym]

Comments have been addressed. Of course, any more are welcome.

[kahnvex]

🍇

[brendanashworth]

While you're here, self doesn't seem to be necessary. Mind removing it? It can be replaced with, as far as I can tell, a var socket = this.socket. The bottom reference to self can just be used as this.

[ahoym]

I converted the code as you suggested (good catch btw), and that caused the error

cannotreadproperty'writable'ofnullif(socket.writable){// socket used to be self.socket
...

This seems to stem from the initial _deferToConnect call, where this.socket is undefined. When the ClientRequest catches the 'socket' event, the socket variable in onSocket() refers to the initial undefined value which was scoped in. As a result, everything blows up when trying to read the writable property of that.

A work around that I found was to have onSocket() accept a sock argument (given by the socket event), and update socket with that parameter. So this piece of code works:

ClientRequest.prototype._deferToConnect=function(method,arguments_,cb){// ...varsocket=this.socket;functioncallSocketMethod(){if(method)socket[method].apply(socket,arguments_);if(typeofcb==='function')cb();}varonSocket=function(sock){// Now accepts the socket from the 'socket' event.socket=sock||socket;// Update socket reference.if(socket.writable){callSocketMethod();}else{socket.once('connect',callSocketMethod);}};if(!socket){this.once('socket',onSocket);}else{onSocket();}};

However I'm not completely sure on the repercussions of this. Maybe leaving the existing implementation would be for the better? I'm interested in what you think

[brendanashworth]

Hmm, you're right, that is something I missed originally. I'd say let's keep the current implementation, as that new one wouldn't really improve the quality of the code. Sorry about the run around!

[ahoym]

Sounds good, then I'll leave this PR as it is. And no need to apologize, we got some documentation and a deeper understanding of the code here 😄

[brendanashworth]

CI: https://ci.nodejs.org/view/iojs/job/node-test-pull-request/283/

[ahoym]

Seems like the build failed?

Looks like only the node-test-commit-arm(#603) job failed.
Digging deeper, the pi1-raspbian-wheezy-1_of_2 and pi2-raspbian-wheezy-2_of_2 builds are both erroring out with this:

returned status code 128:
stdout: stderr: error: object directory /Users/rvagg/git/iojs/io.js/.git/objects does not exist; check .git/objects/info/alternates.
error: refs/remotes/origin/archived-io.js-v0.10 does not point to a valid object!
error: refs/remotes/origin/archived-io.js-v0.12 does not point to a valid object!
error: refs/remotes/origin/mergev010/base does not point to a valid object!
...

[rvagg]

https://ci.nodejs.org/job/node-test-pull-request/302/

those arm errors were fixed yesterday

[brendanashworth]

CI is happy, this LGTM. @cjihrig ?

[cjihrig]

Yep, LGTM

[ahoym]

Bump

[cjihrig]

Thanks! Landed in 79d2c4e after squashing.