CI: https://ci.nodejs.org/job/node-test-pull-request/35099/

LGTM if all the added tests pass reliably.

Thinking out loud about the tests, I am wondering if this behavior could cause problems. In the first two cases res emits close after req, while in the last one is the other way around, as already pointed out. In the previous implementation, it seems that the order was always the same.

I added a few more comments.

I still think this behavior is ok. I would like to know a few arguments on why this is bad (minus it was a breaking change as a minor).

I still think this behavior is ok. I would like to know a few arguments on why this is bad (minus it was a breaking change as a minor).

My concern is with framework/libraries that might expect a strict ordering, always. I don't know if that is a good practice or not. This behavior broke hapi, but I tested fastify out of curiosity, and it didn't suffer the same issue. TLDR, I hope this doesn't complicate things too much in codebases that broke because of this change (i.e., when a fix would have to be implemented on their side). That's all I can say. I don't have a deep knowledge of hapi internals to expand more on that atm. Maybe I just got a little worried after yesterday.

@cjihrig wdyt? I don't really understand how Hapi relies on this specific ordering of things.

I did the analysis for hapi and can elaborate.

Hapi adds a 'close' listener to the http.IncomingRequest and will terminate the response processing when this is emitted. This broke since req would suddenly close as soon as all incoming data was processed.

The specific ordering of the req & res 'close' emits are irrelevant, since hapi only listens to the req event. Other frameworks are not affected because they don't take the same holistic approach that hapi does.

Note that for hapi I tried to change this to listen on the 'close' event of the http.ServerResponse instead. This fixed the issue, except for an upload socket close test that is most likely wrong (it uses synthesized req / res and events).

Thank you @kanongil for the analysis. It was very helpful.

@kanongil could you confirm that there is no significant issue with that change shipping in Node.js v16?

CI: https://ci.nodejs.org/job/node-test-pull-request/35228/

I can't see on https://ci.nodejs.org/ why those 3 jobs are failing, unfortunately.

CI: https://ci.nodejs.org/job/node-test-pull-request/35251/

- Loading data for nodejs/node/pull/36645
✔  Done loading data for nodejs/node/pull/36645
----------------------------------- PR info ------------------------------------
Title      http: add test case for req-res close ordering (#36645)
Author     Daniele Belardi  (@dnlup)
Branch     dnlup:fix/regression_test_http_incoming -> nodejs:master
Labels     author ready, test
Commits    1
 - http: add test case for req-res close ordering
Committers 1
 - Daniele Belardi 
PR-URL: https://github.com/nodejs/node/pull/36645
Reviewed-By: Rich Trott 
Reviewed-By: Robert Nagy 
Reviewed-By: Matteo Collina 
------------------------------ Generated metadata ------------------------------
PR-URL: https://github.com/nodejs/node/pull/36645
Reviewed-By: Rich Trott 
Reviewed-By: Robert Nagy 
Reviewed-By: Matteo Collina 
--------------------------------------------------------------------------------
   ⚠  Commits were pushed since the last review:
   ⚠  - http: add test case for req-res close ordering
   ✔  Last GitHub Actions successful
   ℹ  Last Full PR CI on 2021-01-04T21:19:02Z: https://ci.nodejs.org/job/node-test-pull-request/35251/
- Querying data for job/node-test-pull-request/35251/
✔  Build data downloaded
   ✔  Last Jenkins CI successful
   ℹ  This PR was created on Sun, 27 Dec 2020 18:06:44 GMT
   ✔  Approvals: 3
   ✔  - Rich Trott (@Trott) (TSC): https://github.com/nodejs/node/pull/36645#pullrequestreview-558953069
   ✔  - Robert Nagy (@ronag): https://github.com/nodejs/node/pull/36645#pullrequestreview-558953108
   ✔  - Matteo Collina (@mcollina) (TSC): https://github.com/nodejs/node/pull/36645#pullrequestreview-558962536
--------------------------------------------------------------------------------
   ✔  Aborted `git node land` session in /home/runner/work/node/node/.ncu

Landed in ddac3bd