stream: pipeline should use req.abort() to destroy response #31054

ronag · 2019-12-21T11:12:19Z

destroy(err) on http response will propagate the error to the
request causing 'error' to be unexpectedly emitted. Furthermore,
response.destroy() unlike request.abort() does not _dump buffered
data.

Fixes a breaking change applied in 6480882.

Prefer res.req.abort() over res.destroy() until this situation is
clarified.

This is a bugfix and I believe it should be semver minor.

Fixes #31029.

Checklist

make -j4 test (UNIX), or vcbuild test (Windows) passes
tests and/or benchmarks are included
documentation is changed or added
commit message follows commit guidelines

ronag · 2019-12-21T11:16:10Z

Credit to @szmarczak #31029 (comment)

szmarczak · 2019-12-21T11:22:33Z

I have been just running git clone ... right now :P The speed is about 500KB/s (speedtest results show 2.5MB/s) - that's quite slow. I don't know if this is because of my location or something else... Thanks for taking care of this :)

Edit 1: I guess it's just my location, on my VPS I get 10MB/s.
Edit 2: I set up a forward proxy, it's running at full speed now :D

test/parallel/test-stream-pipeline.js

destroy(err) on http response will propagate the error to the request causing 'error' to be unexpectedly emitted. Furthermore, response.destroy() unlike request.abort() does not _dump buffered data. Fixes a breaking change introduced in nodejs@6480882. Prefer res.req.abort() over res.destroy() until this situation is clarified. Fixes: nodejs#31029 Refs: nodejs@6480882

ronag · 2019-12-21T11:41:46Z

@mcollina: You ok with this instead of reverting 6480882?

test/parallel/test-stream-pipeline.js

nodejs-github-bot · 2019-12-21T11:51:12Z

CI: https://ci.nodejs.org/job/node-test-pull-request/27855/

mcollina

lgtm

nodejs-github-bot · 2019-12-23T01:10:39Z

CI: https://ci.nodejs.org/job/node-test-pull-request/27882/

BridgeAR · 2019-12-24T15:38:49Z

test/parallel/test-stream-pipeline.js

+        res,
+        stream,
+        common.mustCall((err) => {
+          server.close();


Does it make sense to also validate the error here?

I agree. At least that err is truthy.

Fixed. Unfortunately we get an unexpected/different error (ERR_STREAM_DESTROYED) due to the behavior discussed in #31060. So I made it just check for truthyness.

I can't understand how this is related to #31060. AFAIK none of the _destroy() implementations is async.

It seems a regression on master to me

const stream = require('stream'); const data = Buffer.alloc(1024); { const readable = new stream.Readable({ read() { this.push(data); } }); const writable = new stream.Writable({ write(chunk, encoding, callback) { callback(); } }); writable.on('error', console.error); readable.pipe(writable); writable.destroy(new Error('Oops')); } { const readable = new stream.Readable({ read() { this.push(data); } }); const writable = new stream.Writable({ write(chunk, encoding, callback) { callback(); } }); stream.pipeline(readable, writable, console.error); writable.destroy(new Error('Oops')); }

This prints

Error: Oops at Object.<anonymous> (/Users/luigi/Desktop/pipe.js:23:20) at Module._compile (internal/modules/cjs/loader.js:1139:30) at Object.Module._extensions..js (internal/modules/cjs/loader.js:1159:10) at Module.load (internal/modules/cjs/loader.js:988:32) at Function.Module._load (internal/modules/cjs/loader.js:896:14) at Function.executeUserEntryPoint [as runMain] (internal/modules/run_main.js:71:12) at internal/main/run_main_module.js:17:47 Error: Oops at Object.<anonymous> (/Users/luigi/Desktop/pipe.js:41:20) at Module._compile (internal/modules/cjs/loader.js:1139:30) at Object.Module._extensions..js (internal/modules/cjs/loader.js:1159:10) at Module.load (internal/modules/cjs/loader.js:988:32) at Function.Module._load (internal/modules/cjs/loader.js:896:14) at Function.executeUserEntryPoint [as runMain] (internal/modules/run_main.js:71:12) at internal/main/run_main_module.js:17:47

on Node.js 13.5.0 and

Error [ERR_STREAM_DESTROYED]: Cannot call write after a stream was destroyed at Writable.write (_stream_writable.js:321:17) at Readable.ondata (_stream_readable.js:779:22) at Readable.emit (events.js:320:20) at Readable.read (_stream_readable.js:579:10) at flow (_stream_readable.js:1052:34) at resume_ (_stream_readable.js:1033:3) at processTicksAndRejections (internal/process/task_queues.js:84:21) { code: 'ERR_STREAM_DESTROYED' } Error [ERR_STREAM_DESTROYED]: Cannot call write after a stream was destroyed at Writable.write (_stream_writable.js:321:17) at Readable.ondata (_stream_readable.js:779:22) at Readable.emit (events.js:320:20) at Readable.read (_stream_readable.js:579:10) at flow (_stream_readable.js:1052:34) at resume_ (_stream_readable.js:1033:3) at processTicksAndRejections (internal/process/task_queues.js:84:21) { code: 'ERR_STREAM_DESTROYED' }

on Node.js master

Bisecting points to 67ed526.

If that _destroy() completed with an error yes, totally.

Cool. I’ll add it to the proposals list.

How can this thread be resolved? We need a fix with the bug this PR aims to solve for node v13, or better revert the few commits that caused the problem in the first place.

I don't think this specific thread needs to be urgently resolved (or at the least it's a different issue). This PR in its current form resolves the original issue.

Agreed, this discussion should not block the PR. The issue discussed here is caused by a semver-major change that will not be included in v13.x

mcollina

lgtm

test/parallel/test-stream-pipeline.js

Co-Authored-By: Ruben Bridgewater <ruben@bridgewater.de>

nodejs-github-bot · 2019-12-25T12:17:10Z

CI: https://ci.nodejs.org/job/node-test-pull-request/27937/

nodejs-github-bot · 2019-12-25T18:29:33Z

CI: https://ci.nodejs.org/job/node-test-pull-request/27940/

nodejs-github-bot · 2019-12-25T20:02:11Z

CI: https://ci.nodejs.org/job/node-test-pull-request/27943/

Trott · 2019-12-25T20:50:52Z

Landed in c852f7e

destroy(err) on http response will propagate the error to the request causing 'error' to be unexpectedly emitted. Furthermore, response.destroy() unlike request.abort() does not _dump buffered data. Fixes a breaking change introduced in 6480882. Prefer res.req.abort() over res.destroy() until this situation is clarified. Fixes: #31029 Refs: 6480882 PR-URL: #31054 Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: Stephen Belanger <admin@stephenbelanger.com> Reviewed-By: Rich Trott <rtrott@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>

targos · 2020-01-14T13:32:49Z

This lands without issues on v12.x-staging after #30869 but the new test fails because the 'error' event is emitted twice.

destroy(err) on http response will propagate the error to the request causing 'error' to be unexpectedly emitted. Furthermore, response.destroy() unlike request.abort() does not _dump buffered data. Fixes a breaking change introduced in nodejs@6480882. Prefer res.req.abort() over res.destroy() until this situation is clarified. Fixes: nodejs#31029 Refs: nodejs@6480882 PR-URL: nodejs#31054 Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: Stephen Belanger <admin@stephenbelanger.com> Reviewed-By: Rich Trott <rtrott@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>

destroy(err) on http response will propagate the error to the request causing 'error' to be unexpectedly emitted. Furthermore, response.destroy() unlike request.abort() does not _dump buffered data. Fixes a breaking change introduced in 6480882. Prefer res.req.abort() over res.destroy() until this situation is clarified. Fixes: #31029 Refs: 6480882 PR-URL: #31054 Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: Matteo Collina <matteo.collina@gmail.com> Reviewed-By: Stephen Belanger <admin@stephenbelanger.com> Reviewed-By: Rich Trott <rtrott@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>

ronag force-pushed the pipeline-response-destroy branch from b4a4544 to 16b4ed4 Compare December 21, 2019 11:14

szmarczak reviewed Dec 21, 2019

View reviewed changes

test/parallel/test-stream-pipeline.js Outdated Show resolved Hide resolved

lpinca reviewed Dec 21, 2019

View reviewed changes

test/parallel/test-stream-pipeline.js Outdated Show resolved Hide resolved

test/parallel/test-stream-pipeline.js Outdated Show resolved Hide resolved

ronag force-pushed the pipeline-response-destroy branch from 16b4ed4 to 6dd8500 Compare December 21, 2019 11:39

lpinca approved these changes Dec 21, 2019

View reviewed changes

test/parallel/test-stream-pipeline.js Outdated Show resolved Hide resolved

fixup: nit

b7a0f75

szmarczak approved these changes Dec 21, 2019

View reviewed changes

mcollina approved these changes Dec 21, 2019

View reviewed changes

lpinca added http Issues or PRs related to the http subsystem. stream Issues and PRs related to the stream subsystem. labels Dec 22, 2019

Qard approved these changes Dec 22, 2019

View reviewed changes

Trott approved these changes Dec 23, 2019

View reviewed changes

Trott added the author ready PRs that have at least one approval, no pending requests for changes, and a CI started. label Dec 23, 2019

BridgeAR approved these changes Dec 24, 2019

View reviewed changes

fixup

039b944

ronag mentioned this pull request Dec 25, 2019

stream: destroy(err) predictability #31060

Closed

mcollina approved these changes Dec 25, 2019

View reviewed changes

BridgeAR reviewed Dec 25, 2019

View reviewed changes

test/parallel/test-stream-pipeline.js Outdated Show resolved Hide resolved

Update test/parallel/test-stream-pipeline.js

14c816d

Co-Authored-By: Ruben Bridgewater <ruben@bridgewater.de>

Trott closed this Dec 25, 2019

BridgeAR mentioned this pull request Jan 7, 2020

v13.6.0 proposal #31238

Merged

targos added the backport-requested-v12.x label Jan 14, 2020

targos mentioned this pull request Jan 14, 2020

stream: make all streams error in a pipeline #30869

Closed

4 tasks

szmarczak mentioned this pull request Mar 5, 2020

Pipeline aborts all HTTP requests #32105

Closed

targos added backport-open-v12.x and removed backport-requested-v12.x labels Apr 25, 2020

codebytere mentioned this pull request Sep 28, 2020

v12.19.0 proposal #35401

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

stream: pipeline should use req.abort() to destroy response #31054

stream: pipeline should use req.abort() to destroy response #31054

ronag commented Dec 21, 2019 •

edited

Loading

ronag commented Dec 21, 2019

szmarczak commented Dec 21, 2019 •

edited

Loading

ronag commented Dec 21, 2019

nodejs-github-bot commented Dec 21, 2019

mcollina left a comment

nodejs-github-bot commented Dec 23, 2019

BridgeAR Dec 24, 2019

szmarczak Dec 24, 2019

ronag Dec 25, 2019 •

edited

Loading

lpinca Dec 25, 2019 •

edited

Loading

lpinca Dec 25, 2019

lpinca Dec 25, 2019 •

edited

Loading

ronag Dec 25, 2019

mcollina Dec 25, 2019

ronag Dec 25, 2019 •

edited

Loading

lpinca Dec 25, 2019

mcollina left a comment

nodejs-github-bot commented Dec 25, 2019

nodejs-github-bot commented Dec 25, 2019

nodejs-github-bot commented Dec 25, 2019

Trott commented Dec 25, 2019

targos commented Jan 14, 2020

stream: pipeline should use req.abort() to destroy response #31054

stream: pipeline should use req.abort() to destroy response #31054

Conversation

ronag commented Dec 21, 2019 • edited Loading

Checklist

ronag commented Dec 21, 2019

szmarczak commented Dec 21, 2019 • edited Loading

ronag commented Dec 21, 2019

nodejs-github-bot commented Dec 21, 2019

mcollina left a comment

Choose a reason for hiding this comment

nodejs-github-bot commented Dec 23, 2019

BridgeAR Dec 24, 2019

Choose a reason for hiding this comment

szmarczak Dec 24, 2019

Choose a reason for hiding this comment

ronag Dec 25, 2019 • edited Loading

Choose a reason for hiding this comment

lpinca Dec 25, 2019 • edited Loading

Choose a reason for hiding this comment

lpinca Dec 25, 2019

Choose a reason for hiding this comment

lpinca Dec 25, 2019 • edited Loading

Choose a reason for hiding this comment

ronag Dec 25, 2019

Choose a reason for hiding this comment

mcollina Dec 25, 2019

Choose a reason for hiding this comment

ronag Dec 25, 2019 • edited Loading

Choose a reason for hiding this comment

lpinca Dec 25, 2019

Choose a reason for hiding this comment

mcollina left a comment

Choose a reason for hiding this comment

nodejs-github-bot commented Dec 25, 2019

nodejs-github-bot commented Dec 25, 2019

nodejs-github-bot commented Dec 25, 2019

Trott commented Dec 25, 2019

targos commented Jan 14, 2020

ronag commented Dec 21, 2019 •

edited

Loading

szmarczak commented Dec 21, 2019 •

edited

Loading

ronag Dec 25, 2019 •

edited

Loading

lpinca Dec 25, 2019 •

edited

Loading

lpinca Dec 25, 2019 •

edited

Loading

ronag Dec 25, 2019 •

edited

Loading