http2: small fixes to compatibility layer #15473

apapirovski · 2017-09-19T15:24:57Z

This is some assorted work on the compatibility layer that doesn't warrant its own PRs:

Expand argument validation through compat API
adjust behaviour of response.end to not throw if stream already closed to match http1
adjust behaviour of response.writeContinue to not throw if stream already closed and other very small tweaks
Add tests for added and fixed behaviour
Add tests for edge case behaviours of setTimeout, createPushResponse, destroy, end and trailers

I'll add some additional comments on the code itself, so see below for those.

As always, happy to implement any feedback. Thanks in advance~

Checklist

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

Affected core subsystem(s)

http2, test

Expand argument validation through compat API, adjust behaviour of response.end to not throw if stream already closed to match http1, adjust behaviour of writeContinue to not throw if stream already closed and other very small tweaks. Add tests for added and fixed behaviour. Add tests for edge case behaviours of setTimeout, createPushResponse, destroy, end and trailers.

apapirovski · 2017-09-19T15:26:33Z

lib/internal/http2/compat.js

      const headers = this[kHeaders];
-      headers[HTTP2_HEADER_STATUS] = state.statusCode;
-      if (stream.finished === true)
-        options.endStream = true;


I could not find any reason that this would be needed. All our tests pass without it and if it is truly needed then there should be a test to cover it. Would appreciate any input!

(I'm referencing lines 520 & 521 above, 519 was just adjusted.)

The stream.finished check here is an optimization to avoid sending an unnecessary empty DATA frame.

@jasnell Right but what's the situation where stream.finished could be true in this particular code? We don't have any test cases that generate a situation like this (as you can see in the coverage since this code path is never hit).

Hmm... is stream.finished even a thing? As far as I can tell, it's just always undefined. There's this.finished or stream._writableState.finished. Can they even respond in either of those cases?

mcollina

LGTM

mcollina · 2017-09-20T08:17:32Z

CI: https://ci.nodejs.org/job/node-test-pull-request/10154/

mcollina · 2017-09-20T08:42:19Z

This PR is failing with the following error when trying to land:

=== release test-http2-compat-serverrequest-settimeout ===
Path: parallel/test-http2-compat-serverrequest-settimeout
(node:64823) ExperimentalWarning: The http2 module is an experimental API.
assert.js:44
  throw new errors.AssertionError({
  ^

AssertionError [ERR_ASSERTION]: function should not have been called
    at ServerHttp2Stream.mustNotCall (/Users/matteo/Repositories/node/test/common/index.js:565:12)
    at Object.onceWrapper (events.js:314:30)
    at emitNone (events.js:105:13)
    at ServerHttp2Stream.emit (events.js:207:7)
    at ServerHttp2Stream.emit (internal/http2/core.js:134:13)
    at _combinedTickCallback (internal/process/next_tick.js:131:7)
    at process._tickCallback (internal/process/next_tick.js:180:9)
Command: out/Release/node --expose-http2 /Users/matteo/Repositories/node/test/parallel/test-http2-compat-serverrequest-settimeout.js

apapirovski · 2017-09-20T11:11:49Z

On it, looks like there's a race condition. Will refactor the test a bit.

apapirovski · 2017-09-20T11:51:25Z

Ok, changed a few places where we were checking for stream === undefined but should've really been checking for state.closed given the changes to when stream becomes undefined. This will need another review & CI run.

Thanks!

mcollina · 2017-09-20T14:45:18Z

CI: https://ci.nodejs.org/job/node-test-pull-request/10163/

apapirovski · 2017-09-20T20:38:50Z

Failures appear unrelated. On Windows, it's inspector/test-bindings and on Linux it's parallel/test-http-writable-true-after-close.

mcollina · 2017-09-21T07:28:34Z

Landed as bc23681

Expand argument validation through compat API, adjust behaviour of response.end to not throw if stream already closed to match http1, adjust behaviour of writeContinue to not throw if stream already closed and other very small tweaks. Add tests for added and fixed behaviour. Add tests for edge case behaviours of setTimeout, createPushResponse, destroy, end and trailers. PR-URL: #15473 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Minwoo Jung <[email protected]>

Expand argument validation through compat API, adjust behaviour of response.end to not throw if stream already closed to match http1, adjust behaviour of writeContinue to not throw if stream already closed and other very small tweaks. Add tests for added and fixed behaviour. Add tests for edge case behaviours of setTimeout, createPushResponse, destroy, end and trailers. PR-URL: nodejs/node#15473 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Minwoo Jung <[email protected]>

Expand argument validation through compat API, adjust behaviour of response.end to not throw if stream already closed to match http1, adjust behaviour of writeContinue to not throw if stream already closed and other very small tweaks. Add tests for added and fixed behaviour. Add tests for edge case behaviours of setTimeout, createPushResponse, destroy, end and trailers. PR-URL: nodejs#15473 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Minwoo Jung <[email protected]>

jasnell · 2017-09-25T18:04:41Z

This is landing clean in v8.x-staging but tests are failing. PTAL @apapirovski

apapirovski · 2017-09-25T18:09:40Z

@jasnell do you happen to have specific reference (/error message) to what's failing by any chance? This should be landing cleanly (incl tests) unless another commit is missing. Thanks!

jasnell · 2017-09-25T19:42:49Z

I got it. There were some semver-major error code commits that this depended on. I've adjusted things in the cherry-picked commit. Should be good to go.

apapirovski · 2017-09-25T19:45:19Z

Ah, awesome, thanks for sorting it! Sorry I couldn't be more help.

Expand argument validation through compat API, adjust behaviour of response.end to not throw if stream already closed to match http1, adjust behaviour of writeContinue to not throw if stream already closed and other very small tweaks. Add tests for added and fixed behaviour. Add tests for edge case behaviours of setTimeout, createPushResponse, destroy, end and trailers. PR-URL: #15473 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Minwoo Jung <[email protected]>

Expand argument validation through compat API, adjust behaviour of response.end to not throw if stream already closed to match http1, adjust behaviour of writeContinue to not throw if stream already closed and other very small tweaks. Add tests for added and fixed behaviour. Add tests for edge case behaviours of setTimeout, createPushResponse, destroy, end and trailers. PR-URL: nodejs/node#15473 Reviewed-By: James M Snell <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Minwoo Jung <[email protected]>

nodejs-github-bot added the http2 Issues or PRs related to the http2 subsystem. label Sep 19, 2017

apapirovski commented Sep 19, 2017

View reviewed changes

jasnell requested a review from mcollina September 19, 2017 15:30

mcollina approved these changes Sep 19, 2017

View reviewed changes

JungMinu approved these changes Sep 20, 2017

View reviewed changes

jasnell approved these changes Sep 20, 2017

View reviewed changes

adjust settimeout behaviour after stream closed

506f9d1

mcollina closed this Sep 21, 2017

apapirovski deleted the patch-http2-compat-fixes-and-tests branch September 21, 2017 18:19

jasnell added the backport-requested-v8.x label Sep 25, 2017

jasnell removed the backport-requested-v8.x label Sep 25, 2017

MylesBorins mentioned this pull request Oct 3, 2017

v8.7.0 proposal #15762

Merged

MylesBorins added the dont-land-on-v6.x label Oct 17, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

http2: small fixes to compatibility layer #15473

http2: small fixes to compatibility layer #15473

apapirovski commented Sep 19, 2017

apapirovski Sep 19, 2017 •

edited

Loading

jasnell Sep 19, 2017

apapirovski Sep 19, 2017

apapirovski Sep 19, 2017 •

edited

Loading

mcollina left a comment

mcollina commented Sep 20, 2017

mcollina commented Sep 20, 2017

apapirovski commented Sep 20, 2017

apapirovski commented Sep 20, 2017 •

edited

Loading

mcollina commented Sep 20, 2017

apapirovski commented Sep 20, 2017

mcollina commented Sep 21, 2017

jasnell commented Sep 25, 2017

apapirovski commented Sep 25, 2017 •

edited

Loading

jasnell commented Sep 25, 2017

apapirovski commented Sep 25, 2017

http2: small fixes to compatibility layer #15473

http2: small fixes to compatibility layer #15473

Conversation

apapirovski commented Sep 19, 2017

Checklist

Affected core subsystem(s)

apapirovski Sep 19, 2017 • edited Loading

Choose a reason for hiding this comment

jasnell Sep 19, 2017

Choose a reason for hiding this comment

apapirovski Sep 19, 2017

Choose a reason for hiding this comment

apapirovski Sep 19, 2017 • edited Loading

Choose a reason for hiding this comment

mcollina left a comment

Choose a reason for hiding this comment

mcollina commented Sep 20, 2017

mcollina commented Sep 20, 2017

apapirovski commented Sep 20, 2017

apapirovski commented Sep 20, 2017 • edited Loading

mcollina commented Sep 20, 2017

apapirovski commented Sep 20, 2017

mcollina commented Sep 21, 2017

jasnell commented Sep 25, 2017

apapirovski commented Sep 25, 2017 • edited Loading

jasnell commented Sep 25, 2017

apapirovski commented Sep 25, 2017

apapirovski Sep 19, 2017 •

edited

Loading

apapirovski Sep 19, 2017 •

edited

Loading

apapirovski commented Sep 20, 2017 •

edited

Loading

apapirovski commented Sep 25, 2017 •

edited

Loading