promise: warning on unhandled rejection #8217

jasnell · 2016-08-21T22:40:46Z

Checklist

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

Affected core subsystem(s)

promises

Description of change

Emit warnings on unhandled rejection. This is a completion of @benjamingr's PR here that's been sitting dormant for a bit.

/cc @Fishrock123 @trevnorris

jasnell · 2016-08-21T22:41:25Z

CI: https://ci.nodejs.org/job/node-test-pull-request/3780/

chrisdickinson · 2016-08-22T00:04:18Z

lib/internal/process/promises.js

-          // TODO(petkaantonov) Take some default action, see #830
+          process.emitWarning('Possibly unhandled promise rejection ' +
+                              `(rejection id: ${uid}): ${reason}`,
+                              'UnhandledPromiseRejectionWarning');


My quibble from the last discussion holds here — at the time of logging it is definitely unhandled.

I still agree.

chrisdickinson · 2016-08-22T00:25:51Z

It's been a bit since I've been in the promises-in-core headspace, but, IIRC, I believe where I settled after the discussion last spring was thus:

We should definitely warn on unhandled rejection in the absence of other listeners by default for at least one major version.
We should definitely fix the interaction between Promises and AsyncWrap/Domains.
We should consider crashing on unhandled rejection in the absence of listeners in a subsequent major.

This PR addresses point (1) and, my nits around the wording of the error aside, LGTM. Point (2) is tangential to this conversation, except for release timing relating to point (3). Point (3) relates to this conversation in that depending on the nature of the warning we log, we could introduce the breaking behavior of unhandled rejection sooner — that is, if we log a deprecation warning in v6, we could introduce crash-on-unhandled-rejection in v7.

My question is: do we want to — or need to — make a decision on that in advance of landing this PR? I note that with this change, we're taking a subset of programs that would have previously been "quiet" and making them noisy. This implicitly disincentivizes users from designing applications and packages that rely on handling previously unhandled rejections. The work necessary to correct the new logging behavior is the same as the work necessary to correct the proposed crashing behavior. My sense is that folks who object to crashing on unhandled rejection may also object to warning on unhandled rejection, but I would like to check that assumption.

jasnell · 2016-08-22T00:41:29Z

Adding this as a deprecation warning is a semver-major, which means the
warning would not go into a release until v7. The earliest we could throw
is v8.

Adding this as a regular warning now, then changing it to a deprecation
warning later may cause some confusion for those handling the warnings, so
we should decide now which we want to do. Is this a regular warning or a
deprecation warning?

On Sunday, August 21, 2016, Chris Dickinson notifications@github.com
wrote:

It's been a bit since I've been in the promises-in-core headspace, but,
IIRC, I believe where I settled after the discussion last spring was thus:

We should definitely warn on unhandled rejection in the absence
of other listeners by default for at least one major version.

We should definitely fix the interaction between Promises and
AsyncWrap/Domains.

We should consider crashing on unhandled rejection in the absence
of listeners in a subsequent major.

This PR addresses point (1) and, my nits around the wording of the error
aside, LGTM. Point (2) is tangential to this conversation, except for
release timing relating to point (3). Point (3) relates to this
conversation in that depending on the nature of the warning we log, we
could introduce the breaking behavior of unhandled rejection sooner — that
is, if we log a deprecation warning in v6, we could introduce
crash-on-unhandled-rejection in v7.

My question is: do we want to — or need to — make a decision on that in
advance of landing this PR? I note that with this change, we're taking a
subset of programs that would have previously been "quiet" and making them
noisy. This implicitly disincentivizes users from designing applications
and packages that rely on handling previously unhandled rejections. The
work necessary to correct the new logging behavior is the same as the work
necessary to correct the crashing behavior. My sense is that folks who
object to crashing on unhandled rejection may also object to warning on
unhandled rejection, but I would like to check that assumption.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#8217 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/AAa2eedtrz47vt_qdhtM_wgm_Bkuj59Oks5qiOyWgaJpZM4JpaK0
.

Fishrock123 · 2016-08-22T00:58:59Z

We should definitely fix the interaction between Promises and AsyncWrap/Domains.

Point (2) is tangential to this conversation, except for release timing relating to point (3)

@fhinkel has been helping me get @trevnorris in touch with the V8 team about the MicroTaskQueue here: https://bugs.chromium.org/p/v8/issues/detail?id=4643

I definitely think we should float patches to expose what we need for the time being.

My question is: do we want to — or need to — make a decision on that in advance of landing this PR?

Maybe in regards to documentation?

If we are going to "crash", ideally this message would say that we will in the future. Documentation would also be helpful in warning of that in advance.

My sense is that folks who object to crashing on unhandled rejection may also object to warning on unhandled rejection, but I would like to check that assumption.

Probably. You win some, you loose some -- I think we are working within our boundaries here.

jasnell · 2016-08-22T04:08:18Z

Thinking about this a bit further... we could actually add a second, separate DeprecationWarning here that would only be printed on the first unhandled rejection.

jasnell · 2016-08-22T04:19:39Z

Added a third deprecation commit. The first two can easily be ported into v6 as a semver-minor, adding the regular warning that will be emitted on every unhandled rejection. The third is a semver-major that would go only into v7. It prints the one time deprecation warning on the first unhandled rejection.

jasnell · 2016-08-22T04:20:30Z

I'll work on a documentation update next.

benjamingr · 2016-08-22T05:51:25Z

Thanks for picking up my slack and sorry!

benjamingr · 2016-08-22T05:53:23Z

Changes themselves LGTM

yorkie · 2016-08-22T08:20:52Z

lib/internal/process/promises.js

    addPendingUnhandledRejection(promise, reason);
  }

  function rejectionHandled(promise) {
    var hasBeenNotified = hasBeenNotifiedProperty.get(promise);
    if (hasBeenNotified !== undefined) {
      hasBeenNotifiedProperty.delete(promise);
+      const uid = promiseToGuidProperty.get(promise);


Could we move this variable to define at line 36?

Fishrock123 · 2016-08-22T13:15:34Z

Added a third deprecation commit. The first two can easily be ported into v6 as a semver-minor, adding the regular warning that will be emitted on every unhandled rejection. The third is a semver-major that would go only into v7. It prints the one time deprecation warning on the first unhandled rejection.

@jasnell could we do that in a separate PR? Doing majors alongside non-majors is a recurring pain for releasers of the Current branch.

jasnell · 2016-08-22T13:46:06Z

Yep, absolutely. Alternatively I can simply put a backport pr together for
v6. Either way should work.

On Monday, August 22, 2016, Jeremiah Senkpiel notifications@github.com
wrote:

Added a third deprecation commit. The first two can easily be ported into
v6 as a semver-minor, adding the regular warning that will be emitted on
every unhandled rejection. The third is a semver-major that would go only
into v7. It prints the one time deprecation warning on the first unhandled
rejection.

@jasnell https://github.com/jasnell could we do that in a separate PR?
Doing majors alongside non-majors is a recurring pain for releasers of the
Current branch.

—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
#8217 (comment), or mute
the thread
https://github.com/notifications/unsubscribe-auth/AAa2eQdkkX6y7u1OWR-eQ4Exdoh4RWopks5qiaD9gaJpZM4JpaK0
.

Fishrock123 · 2016-08-22T13:49:38Z

test/parallel/test-promises-warning-on-unhandled-rejection.js

+var b = 0;
+
+process.on('warning', common.mustCall((warning) => {
+  switch (b) {


more robust to do b++ in the switch

benjamingr · 2016-08-23T16:54:32Z

Still LGTM.

jasnell · 2016-08-23T17:36:22Z

@chrisdickinson and @Fishrock123 ... PTAL. The commits have been updated with a more precise warning message.

Fishrock123 · 2016-08-23T23:22:28Z

lib/internal/process/promises.js

+            deprecationWarned = true;
+            process.emitWarning(
+              'Unhandled promise rejections are deprecated. In the future, ' +
+              'an error will be thrown when a promise rejection is not ' +


No error will be thrown.

There is no way to catch it so that you cannot listen to GC.
It will exit with the rejection formatted how we format regular uncaught exceptions.

'In the future, the Node process will exit if promise rejections are not handled before garbage collection.'

That may be slightly confusing since this warning message is not on GC... not sure how to clarify.

process.emit('error') could be used to propagate the error but understood. Perhaps,

In the future, promise rejections that are not handled before garbage collection will be handled as exceptions that may terminate the Node.js process.

That gives us some wiggle room in how we handle it later on.

@jasnell We won't do it if it must expose GC via a publicly accessible API. Legit everyone will yell at us.

If we do it, it will terminate the process.

Tangentially, it will also produce error output and a non-zero exit code if the process exits normally with an unhanded rejection that has not been GC'd (for completeness)

'In the future, the Node process will exit with a non-zero exit code if a promise rejection is not handled and the promise is no longer referenced by anything.'

We can say what we mean by garbage collection without saying garbage collection which should be fine :)

@Fishrock123 ... mentioning garbage collection in the message does not imply that we have to expose garbage collection via a public API..

In the future, promise rejections that are not handled before garbage collection will be handled as exceptions that will terminate the Node.js process with a non-zero exit code.

I'm confused by garbage collection entering the conversation here: my understanding is that we would crash on unhandled rejection as soon as it happened if an unhandledRejection listener were not installed.

Any comments on my terminology? We can say not referenced without saying GC

Your suggested terminology is fine, I just think there's still some disagreement on exactly when the error should eventually happen.

OK, I'm back up to date. The conversation about "when" the error crashes — whether on GC, or immediately — should probably be left for later, to see how the community at large reacts to the warning messages. If, by this time next year, there's still a lot of folks leaning on handling previously unhandled rejections, we can lean towards crashing on GC, otherwise we can crash immediately. To that end, a slight spin on @jasnell's wording:

In the future, promise rejections that are not handled will terminate the Node.js process with a non-zero exit code.

(Without a specific "when" — whether at GC, or at unhandled-rejection-time)

Works for me.

jasnell · 2016-08-24T23:41:12Z

On other thought on this... in the emitted warning, it might be a good idea to attach the uid as a standalone property on the warning object in addition to having it in the error message.

Notable Changes: * Buffer * Passing invalid input to Buffer.byteLength will now throw an error [nodejs#8946](nodejs#8946). * Calling Buffer without new is now deprecated and will emit a process warning [nodejs#8169](nodejs#8169). * Passing a negative number to allocUnsafe will now throw an error [nodejs#7079](nodejs#7079). * Child Process * The fork and execFile methods now have stronger argument validation [nodejs#7399](nodejs#7399). * Cluster * The worker.suicide method is deprecated and will emit a process warning [nodejs#3747](nodejs#3747). * Deps * V8 has been updated to 5.4.500.36 [nodejs#8317](nodejs#8317), [nodejs#8852](nodejs#8852), [nodejs#9253](nodejs#9253). * NODE_MODULE_VERSION has been updated to 51 [nodejs#8808](nodejs#8808). * File System * A process warning is emitted if a callback is not passed to async file system methods [nodejs#7897](nodejs#7897). * Intl * Intl.v8BreakIterator constructor has been deprecated and will emit a process warning [nodejs#8908](nodejs#8908). * Promises * Unhandled Promise rejections have been deprecated and will emit a process warning [nodejs#8217](nodejs#8217). * Punycode * The `punycode` module has been deprecated [nodejs#7941](nodejs#7941). * URL * An Experimental WHATWG URL Parser has been introduced [nodejs#7448](nodejs#7448).

Notable Changes: * Buffer * Passing invalid input to Buffer.byteLength will now throw an error [#8946](#8946). * Calling Buffer without new is now deprecated and will emit a process warning [#8169](#8169). * Passing a negative number to allocUnsafe will now throw an error [#7079](#7079). * Child Process * The fork and execFile methods now have stronger argument validation [#7399](#7399). * Cluster * The worker.suicide method is deprecated and will emit a process warning [#3747](#3747). * Deps * V8 has been updated to 5.4.500.36 [#8317](#8317), [#8852](#8852), [#9253](#9253). * NODE_MODULE_VERSION has been updated to 51 [#8808](#8808). * File System * A process warning is emitted if a callback is not passed to async file system methods [#7897](#7897). * Intl * Intl.v8BreakIterator constructor has been deprecated and will emit a process warning [#8908](#8908). * Promises * Unhandled Promise rejections have been deprecated and will emit a process warning [#8217](#8217). * Punycode * The `punycode` module has been deprecated [#7941](#7941). * URL * An Experimental WHATWG URL Parser has been introduced [#7448](#7448). PR-URL: #9099

Notable Changes: * Buffer * Passing invalid input to Buffer.byteLength will now throw an error [#8946](nodejs/node#8946). * Calling Buffer without new is now deprecated and will emit a process warning [#8169](nodejs/node#8169). * Passing a negative number to allocUnsafe will now throw an error [#7079](nodejs/node#7079). * Child Process * The fork and execFile methods now have stronger argument validation [#7399](nodejs/node#7399). * Cluster * The worker.suicide method is deprecated and will emit a process warning [#3747](nodejs/node#3747). * Deps * V8 has been updated to 5.4.500.36 [#8317](nodejs/node#8317), [#8852](nodejs/node#8852), [#9253](nodejs/node#9253). * NODE_MODULE_VERSION has been updated to 51 [#8808](nodejs/node#8808). * File System * A process warning is emitted if a callback is not passed to async file system methods [#7897](nodejs/node#7897). * Intl * Intl.v8BreakIterator constructor has been deprecated and will emit a process warning [#8908](nodejs/node#8908). * Promises * Unhandled Promise rejections have been deprecated and will emit a process warning [#8217](nodejs/node#8217). * Punycode * The `punycode` module has been deprecated [#7941](nodejs/node#7941). * URL * An Experimental WHATWG URL Parser has been introduced [#7448](nodejs/node#7448). Signed-off-by: Ilkka Myller <ilkka.myller@nodefield.com>

vitaly-t · 2016-11-01T02:13:06Z

Will the generic error handler be usable for unhandled rejections?

Fishrock123 · 2016-11-01T02:22:37Z

@vitaly-t see https://nodejs.org/dist/latest-v7.x/docs/api/process.html#process_event_warning

benjamingr · 2016-11-02T12:40:15Z

@vitaly-t at the moment - a warning is emitted and you need to add an unhandledRejection event handler separately. In the future GC based event handling will be performed.

I intend to open a PR clarifying the wording of unhandled rejections soon (I'm between laptops :( )

leodutra · 2016-11-11T13:16:37Z

Thank you so much for this. The other behavior was a kick in the balls.

danShumway · 2016-11-24T01:34:47Z

Highly approve of this change.

If people aren't explicitly handling errors, it's easy for promises to eat them which makes debugging much harder, since there's no easy way to trace which promise has forgotten its error callback or what the original error was in the first place.

What is the timeline for actually removing the behavior though? I have libraries where I'd like errors within callbacks or methods to actually break things like they'd normally do - specifically in situations where I don't want to necessarily force the user to think about the fact that they're working with Promises behind the scenes.

I'm trying to decide if I can just leave things as they are in anticipation of the full removal, or if I should try to polyfill a similar behavior in the meantime.

constructPromise().catch(function (err) {
    console.log(err);
    process.exit(1);
});

benjamingr · 2016-12-01T15:04:21Z

What is the timeline for actually removing the behavior though? I have libraries where I'd like errors within callbacks or methods to actually break things like they'd normally do - specifically in situations where I don't want to necessarily force the user to think about the fact that they're working with Promises behind the scenes.

Either use a library like bluebird and add a onPossiblyUnhandledRejection for only your promises (which would not cause issues) or if it's your app add a "unhandledRejection" event handler on process.

As for the actual change - not really sure where it stands with @Fishrock123 but IIRC there was an issue with V8 being addressed.

As discovered in 18F#51, the `UnhandledPromiseRejectionWarning` and `PromiseRejectionHandledWarning` warnings were apparently added in v6.6.0 (https://nodejs.org/en/blog/release/v6.6.0/); specifically it was added by nodejs/node#8223. See also: nodejs/node#6439 nodejs/node#8217 https://nodejs.org/dist/latest-v6.x/docs/api/process.html#process_event_unhandledrejection https://nodejs.org/dist/latest-v6.x/docs/api/process.html#process_event_rejectionhandled https://nodejs.org/dist/latest-v6.x/docs/api/process.html#process_event_warning Test failures from `test/integration-test.js` after upgrading to Node v6.9.1 showed extra output such as: ``` (node:39696) UnhandledPromiseRejectionWarning: Unhandled promise rejection (rejection id: 2): null (node:39696) PromiseRejectionHandledWarning: Promise rejection was handled asynchronously (rejection id: 2) ``` This was happening because `scripts/slack-github-issues.js` created a Hubot listener that returned a `Promise` so that the integration test could use `.should.be.rejectedWith` assertions. Rather than jump through hoops to quash the warnings, this change now causes the listener's `catch` handler to return the result of the rejected `Promise`, converting it to a fulfilled `Promise` in the process. Since Hubot never used the result anyway, the only effect it has in production is to eliminate the warning messages. The tests now just check that the `Promise` returned by the listener callback is fulfilled with the expected error result, with practically no loss in clarity.

Backported from mbland/hubot-slack-github-issues#7. As discovered in 18F/hubot-slack-github-issues#51, the `UnhandledPromiseRejectionWarning` and `PromiseRejectionHandledWarning` warnings were apparently added in v6.6.0 (https://nodejs.org/en/blog/release/v6.6.0/); specifically it was added by nodejs/node#8223. See also: nodejs/node#6439 nodejs/node#8217 https://nodejs.org/dist/latest-v6.x/docs/api/process.html#process_event_unhandledrejection https://nodejs.org/dist/latest-v6.x/docs/api/process.html#process_event_rejectionhandled https://nodejs.org/dist/latest-v6.x/docs/api/process.html#process_event_warning A test failure from `solutions/06-integration/test/integration-test.js` after upgrading to Node v6.9.1 showed output such as: ``` - "(node:87412) UnhandledPromiseRejectionWarning: Unhandled promise rejection (rejection id: 14): Error: failed to create a GitHub issue in 18F/handbook: received 500 response from GitHub API: {\"message\":\"test failure\"}\n" ``` This was happening because `scripts/slack-github-issues.js` ignored the return value from `Middleware.execute()`, whether it was undefined or a `Promise`. For consistency's sake (and to provide a clearer upgrade path to the current state of mbland/slack-github-issues), `Middleware.execute()` always returns a `Promise`, and `scripts/slack-github-issues.js` handles and ignores any rejected Promises. This fixed the `integration-test.js` error, but also required minor updates to `solutions/{05-middleware,complete}/test/middleware-test.js`. The next commit will update the tutorial narrative to account for this change.

jasnell added semver-minor PRs that contain new features and should be released in the next minor version. process Issues and PRs related to the process subsystem. promises Issues and PRs related to ECMAScript promises. labels Aug 21, 2016

jasnell mentioned this pull request Aug 21, 2016

promise: warn on unhandled rejections #6439

Closed

chrisdickinson reviewed Aug 22, 2016
View reviewed changes

jasnell force-pushed the warning-on-unhandled-rejection branch from b0daba2 to ffe66d7 Compare August 22, 2016 04:18

jasnell added semver-major PRs that contain breaking changes and should be released in the next major version. and removed semver-minor PRs that contain new features and should be released in the next minor version. labels Aug 22, 2016

yorkie reviewed Aug 22, 2016
View reviewed changes

Fishrock123 reviewed Aug 22, 2016
View reviewed changes

jasnell force-pushed the warning-on-unhandled-rejection branch 2 times, most recently from 6ed481f to bbb0dce Compare August 22, 2016 16:25

jasnell mentioned this pull request Aug 22, 2016

promise: warning on unhandled rejection (v6.x) #8223

Closed

3 tasks

Fishrock123 reviewed Aug 23, 2016
View reviewed changes

jasnell force-pushed the warning-on-unhandled-rejection branch from bbb0dce to 2e3eb6a Compare August 24, 2016 23:48

mattgrill mentioned this pull request Oct 29, 2016

Unhandled promise rejection acquia/waterwheel.js#47

Closed

domenic mentioned this pull request Oct 30, 2016

Unhandled rejection warnings when expecting a rejection domenic/chai-as-promised#173

Closed

bminer mentioned this pull request Dec 1, 2016

Default Unhandled Rejection Detection Behavior #830

Closed

mbland mentioned this pull request Dec 1, 2016

Integration test fixes in light of new Promise warnings mbland/hubot-slack-github-issues#7

Merged

bnoordhuis mentioned this pull request Feb 13, 2017

Don't kill node.js / REPL #11342

Closed

shiftkey mentioned this pull request Mar 29, 2017

update tooling to node v7 and npm v4 desktop/desktop#1085

Closed

misterdjules mentioned this pull request Apr 19, 2017

Support awaiting values through util.awaitable nodejs/CTC#12

Closed

gibfahn mentioned this pull request Jun 15, 2017

Auditing for 6.11.1 nodejs/Release#230

Closed

3 tasks

benjamingr mentioned this pull request Apr 25, 2018

src: exit on gc unhandled rejection #20097

Closed

4 tasks

tniessen added the deprecations Issues and PRs related to deprecations. label Sep 7, 2018

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

promise: warning on unhandled rejection #8217

promise: warning on unhandled rejection #8217

jasnell commented Aug 21, 2016

jasnell commented Aug 21, 2016

chrisdickinson Aug 22, 2016

Fishrock123 Aug 22, 2016

chrisdickinson commented Aug 22, 2016 •

edited

jasnell commented Aug 22, 2016

Fishrock123 commented Aug 22, 2016 •

edited

jasnell commented Aug 22, 2016

jasnell commented Aug 22, 2016

jasnell commented Aug 22, 2016

benjamingr commented Aug 22, 2016

benjamingr commented Aug 22, 2016

yorkie Aug 22, 2016

Fishrock123 commented Aug 22, 2016

jasnell commented Aug 22, 2016

Fishrock123 Aug 22, 2016

benjamingr commented Aug 23, 2016

jasnell commented Aug 23, 2016

Fishrock123 Aug 23, 2016

jasnell Aug 24, 2016

Fishrock123 Aug 24, 2016

benjamingr Aug 24, 2016

jasnell Aug 24, 2016

chrisdickinson Aug 24, 2016

benjamingr Aug 24, 2016

jasnell Aug 24, 2016

chrisdickinson Aug 24, 2016

jasnell Aug 24, 2016

jasnell commented Aug 24, 2016

vitaly-t commented Nov 1, 2016

Fishrock123 commented Nov 1, 2016

benjamingr commented Nov 2, 2016

leodutra commented Nov 11, 2016

danShumway commented Nov 24, 2016 •

edited

benjamingr commented Dec 1, 2016

promise: warning on unhandled rejection #8217

promise: warning on unhandled rejection #8217

Conversation

jasnell commented Aug 21, 2016

Checklist

Affected core subsystem(s)

Description of change

jasnell commented Aug 21, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chrisdickinson commented Aug 22, 2016 • edited

jasnell commented Aug 22, 2016

Fishrock123 commented Aug 22, 2016 • edited

jasnell commented Aug 22, 2016

jasnell commented Aug 22, 2016

jasnell commented Aug 22, 2016

benjamingr commented Aug 22, 2016

benjamingr commented Aug 22, 2016

Choose a reason for hiding this comment

Fishrock123 commented Aug 22, 2016

jasnell commented Aug 22, 2016

Choose a reason for hiding this comment

benjamingr commented Aug 23, 2016

jasnell commented Aug 23, 2016

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

jasnell commented Aug 24, 2016

vitaly-t commented Nov 1, 2016

Fishrock123 commented Nov 1, 2016

benjamingr commented Nov 2, 2016

leodutra commented Nov 11, 2016

danShumway commented Nov 24, 2016 • edited

benjamingr commented Dec 1, 2016

chrisdickinson commented Aug 22, 2016 •

edited

Fishrock123 commented Aug 22, 2016 •

edited

danShumway commented Nov 24, 2016 •

edited