Get port of e2e test | NestJS | Page 1

latent swallow Jul 15, 2025, 5:25 AM

#

During my e2e test I am using supertest together with the nestjs application. This is not starting a Webserver, because I want to collect coverage.

However I need to publish metadata (openid connect information) which includes also the URL of the service. Right now these values are defined in the constructor and passed via env.

So I see two approaches:

instead of a random port, bind it to a static one that is known (like 3000)
pass the value after the start to set it.

Creating the response based on the request by using the host paramter would be one approach, but I am not sure if this will work when using a proxy.

spiral moss Jul 15, 2025, 5:55 AM

#

You can extract the port from the Express instance. I don't remember exactly, but app.getHttpServer() gives you the express instance, on which you could call server.address().port

#

But address() will only be available if you call app.listen(xxx) with a specified port. If you call app.init(), it will be null, because it doesn't really start an actual server

#

I think it's best to set a known port as an environment variable for e2e testing

latent swallow Jul 15, 2025, 6:23 AM

#

I thought making a request to the server in the beforeall and the get the port via response.request.host would work. But it seems that it is using a new random port for every new request:

stdout | test/issuance.e2e-spec.ts > Issuance
[127.0.0.1:57646] /

stdout | test/issuance.e2e-spec.ts > Issuance > create oid4vci offer
[127.0.0.1:57648] /vci/offer

stdout | test/issuance.e2e-spec.ts > Issuance > create oid4vci offer
[127.0.0.1:57650] /session/be009687-bd47-443c-af67-dcadf040961b

stdout | test/issuance.e2e-spec.ts > Issuance > ask for an invalid oid4vci offer
[127.0.0.1:57652] /vci/offer

stdout | test/issuance.e2e-spec.ts > Issuance > get credential from oid4vci offer
[127.0.0.1:57654] /vci/offer

#

Since I want to get the coverage, start the app with .listen(3000 would give me the fixed port, but it will not be able to collect coverage

#

Nevermind, I read something wrong. I thought when running app.listen vitest would not be able to get to know which lines are executed.
But supertest that is used for this will start a webserver on an ephermal port when there is none specific. So running app.listen(3000) before will make sure the port is always the same. And coverage will be generated

#

Thank you @spiral moss for responding so fast!!!!

spiral moss Jul 15, 2025, 6:35 AM

#

You would not be ale to collect coverage if you started the server in a different process than the one the test runs in.

#

Since you do it jn the same process, it doesn't matter if the requests go through the network layer, or if you call methods directly

#

But I'll give you another unprompted advice - generating coverage for e2e tests is a waste of resources (your effort, time and CI credits).

#Get port of e2e test