fix: we have one failing test

[NoNameProvided]

One of our test is failing in the last few commit. We should investigate and fix that.

[MichalLytek]

I use node v8.1.4 and don't have any failing test. I think that the newest version of node have broken some internals API so we should try to update our dependencies to the newest versions.

[NoNameProvided]

Here is the latest build: https://travis-ci.org/pleerock/routing-controllers/jobs/261758848

Our @Session(param) should throw required error when param is empty tests fail with a Socket hang up error.

[MichalLytek]

Ok, I've updated local node version and on v8.2.1 I have the test failing. But I have no idea why the socket hang, tried deps update but no success 😞

[pleerock]

If this issue with @Session then @19majkel94 you have to figure out solution since its your creation 😆

[MichalLytek]

😆
I've only done case "session": return req.session. I don't find myself guilty as the test mystically fail with closing server, socket and connection 😨

[patrickhousley]

Just quickly looking at the issue, I found that commenting out line https://github.com/pleerock/routing-controllers/blob/master/src/driver/express/ExpressDriver.ts#L336 results in the test passing.

[NoNameProvided]

I don't find myself guilty as the test mystically fail with closing server, socket and connection 😨

No body is guilty of anything in OSS. 🙂 @pleerock simply said you should check it out because you wrote it, so probably you are the one who can find out where the issue commings from.

Just quickly looking at the issue, I found that commenting out line

Thanks for looking into this @patrickhousley do you want to follow up on your findings and create a PR to fix it?

[patrickhousley]

Honestly, the fix would be to not call next if routing-controllers handles the error. This is standard for all express error handlers. Simply put, you should never try to modify the response, send headers, or send more data after the first call to send. The fact that this works in Node 8.1 and not in 8.2 is a fluke. They updated 8.2 here to prevent this.

[patrickhousley]

This is the actual PR that was merged into Node.

[patrickhousley]

@19majkel94 How would you like to proceed on this one. It seems it is going to need a, possibly drastic and breaking, change to the code base.

[NoNameProvided]

Honestly, the fix would be to not call next if routing-controllers handles the error. This is standard for all express error handlers.

I cannot find it now, but this was discussed. It is a needed bad coming from some implementation detail of the library.

It seems it is going to need a, possibly drastic and breaking, change to the code base.

Just +1 reason to rewrite the middleware execution.

I think we should move forward with this after making our decision on #107, but that will need some more discussion. For now only @19majkel94 and me made our decision.