promises: more robust stringification #13784

TimothyGu · 2017-06-19T08:25:02Z

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)

promises

bnoordhuis · 2017-06-19T08:47:16Z

Swallowing exceptions seems like an - wait for it - exceptionally bad idea to me. The test case in #13771 works the way I would expect it to.

jasnell · 2017-06-19T16:01:02Z

I'm not too keen on this either. I'd almost prefer to go in the other direction and just make it:

const warning = new Error('Unhandled promise rejection ' +
                          `(rejection id: ${uid}): ${reason}`);

TimothyGu · 2017-06-20T01:56:09Z

@bnoordhuis While I would generally agree with that sentiment, this case is a special one, since the thrown error is uncatchable.

@jasnell That would mean reverting #11640.

bnoordhuis · 2017-06-20T09:39:24Z

All the more reason to let it bubble up. If .toString() throws an exception, something is probably deeply wrong with the application. Moving on as if nothing happened is just bad.

The more I think about it, the more I'm -1.

benjamingr · 2017-06-20T12:51:28Z

test/parallel/test-promises-unhandled-unstringifyable-rejections.js

+'use strict';
+const common = require('../common');
+
+const expectedDeprecationWarning = 'Unhandled promise rejections are ' +


I don't think we should test this part, the below is good.

benjamingr · 2017-06-20T12:53:24Z

All the more reason to let it bubble up. If .toString() throws an exception, something is probably deeply wrong with the application. Moving on as if nothing happened is just bad.

The issue isn't that toString threw an exception - is that Node using your code to give you a warning invoked toString which threw an error.

If no node side promise hooks were involved - the code wouldn't throw and the server wouldn't crash. Here, adding instrumentation can crash the server which is not a great idea.

benjamingr · 2017-06-20T12:55:43Z

I would prefer to have something like <toString() threw exception ${ex}> with an error instance that preserves both the original and new error.

That said, what we should do is align with browser behavior (I don't think this is good enough reason to diverge) - did anyone check what chrome's hooks do?

TimothyGu · 2017-06-20T13:04:12Z

@benjamingr

I would prefer to have something like <toString() threw exception ${ex}>

Here, you are coercing ex to a string again… and so you see the challenge.

did anyone check what chrome's hooks do?

Chrome's Dev Tools offers to inspect the object and doesn't invoke any methods to serialize it.

addaleax

LGTM, I’m good with this change for the reasons @benjamingr mentioned

benjamingr · 2017-06-21T20:41:32Z

Here, you are coercing ex to a string again… and so you see the challenge.

@TimothyGu I'm sorry if it wasn't clear from my comment, but I'm +1 on these changes and the pull request.

It is unreasonable to swallow exceptions when the user causes them but it is unacceptable to crash the process for something the user didn't do for our internal tooling

bnoordhuis

Making my -1 explicit so it's not overlooked.

Swallowing arbitrary exceptions is tantamount to process.on('uncaughtException', () => {}), it leaves the process in an undefined state.

It's bad enough when users do it. It has no place in core.

addaleax · 2017-06-22T11:21:40Z

@nodejs/ctc

mcollina

I am -0 on this matter, as swallowing the exception does not give me the reason for that exception and that stacktrace. So it would total to 2 exceptions that should have crashed my process and didn't.

However this does make the situation somewhat better, but I'd prefer to have node 9 enforcing that deprecation, which would lead to this code being removed anyway.

In case anyone is interested, I wrote this: https://github.com/mcollina/make-promises-safe

mcollina · 2017-06-22T13:41:59Z

lib/internal/process/promises.js

+  try {
+    return savedString(value);
+  } catch (err) {
+    return '<toString() threw exception>';


can you please include the message of that error?

@mcollina Keep in mind that err can be any JS value, and practically any operation we do with it can lead to yet another exception.

I'm ok for crashing badly in that case. It's always better to crash in those situations.

If even the error thrown can not be stringified safely then IMO it's reasonable to crash with what we've got. Not including the error message makes it more confusing than what it is now if you have no idea whose toString() is throwing...

What can a promise be rejected with that can't either be stringified in some way or util/inspect-ed?

I suppose if toString() throws we could either inspect the object or just <Object>.prototype.toString()?

@Fishrock123 A promise can be rejected with any JS value, and inspecting that is not foolproof.

Also, since we have [Symbol.toStringTag], Object.prototype.toString() may throw as well.

This message, as it is, will not be clear to the users. Maybe something like, "stringifying promise rejection reason failed" would be better?

@thefourtheye The final error message already contains the information that it’s about promise rejection reasons, so that would be duplicating a lot. Also, I think V8 reports the same literal string in similar scenarios.

benjamingr · 2017-06-22T14:31:01Z

I want to point out that this is not swallowing an error, this is Node's instrumentation causing an error and then swallowing that internal error.

If we didn't have the promise hooks in place the error would not occur in the first place. It just happens to be an error. If no exception was thrown would that help with your objection @bnoordhuis ? (We can check explicitly for what error is)

Check if it's an object.
If it's not - String it and return.
Check if it is an Error object.
If it is, assume it has a stack and print it and its stack.
If it's not, go over the keys and print the keys (does util.inspect do that?). This is what chrome does.

However, I think that's way more complicated than checking if it threw an exception and catching that - it's an exception we're throwing and we're catching and I'm having a really hard time coming with a realistic case where I'd prefer to crash the user's server for something they didn't do.

mcollina · 2017-06-22T14:46:09Z

However, I think that's way more complicated than checking if it threw an exception and catching that - it's an exception we're throwing and we're catching and I'm having a really hard time coming with a realistic case where I'd prefer to crash the user's server for something they didn't do.

We are not throwing that exception, a user custom toString() method is. IMHO we should not call toString() at all. util.inspect might be a good idea.

I would note that we are already swallowing an error and not crashing. It's the full point of 'unhandledRejection' and this deprecation. It's always better to crash vs swallowing/not handling an exception, as swallowing can make file descriptors and other native resources leak.

TimothyGu · 2017-06-22T15:45:27Z

util.inspect might be a good idea.

Unfortunately util.inspect may crash also because it calls property getters that may throw. Heck, even Object.prototype.toString might throw if the Symbol.toStringTag property is implemented as a getter. Unlike Dev Tools, we are simply not in a position to print anything about the object without worrying about side effects. In this light, toString is the best we can do other than not trying to do anything about the object more than printing rejection value: object, which would be a great shame as most of promise rejection values are just plain Error objects.

mcollina · 2017-06-22T15:50:06Z

I like the proposal @benjamingr did in #13784 (comment). It's complex, but I can see that topic coming up more and more often.

bnoordhuis · 2017-06-24T09:34:22Z

If no exception was thrown would that help with your objection

If you mean inspecting the value in a way that no user code is invoked, making it impossible for exceptions to occur - yes, that's a good approach.

bnoordhuis · 2017-07-31T10:22:01Z

I'm removing the ctc-review label, please add it back if you think that is a mistake.

addaleax · 2017-08-01T19:22:43Z

@bnoordhuis @nodejs/ctc I do think we should come to a conclusion on this, and I do think this is the right thing to do, so I’m adding it to the CTC agenda.

hashseed · 2017-12-19T18:06:08Z

git blame says it was last touched in 2015, and if memory serves me correctly we had that long before that.

MylesBorins · 2017-12-19T18:13:43Z

@TimothyGu this will need to be manually backported if we are going to land on v6.x. Please lmk if you want to do this, as an alternative we can decide not to land

bnoordhuis · 2017-12-19T20:42:40Z

does anyone know which version of v8 v8::Value::ToDetailString()was introduced in?

I think it's been in V8 since its first public release. I know for a fact it was in node.js v0.0.1!

TimothyGu · 2017-12-19T21:47:35Z

This should definitely go into v6.x. But, I'll have limited time and availability over the break so it would be great if someone else could take care of this.

PR-URL: nodejs#13784 Fixes: nodejs#13771 Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: Rod Vagg <[email protected]> Reviewed-By: Michaël Zasso <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Ruben Bridgewater <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Benjamin Gruenbaum <[email protected]> Reviewed-By: James M Snell <[email protected]>

TimothyGu · 2017-12-23T05:56:28Z

Backported to v6.x in #17833.

Backport-PR-URL: #17833 PR-URL: #13784 Fixes: #13771 Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: Rod Vagg <[email protected]> Reviewed-By: Michaël Zasso <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Ruben Bridgewater <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Benjamin Gruenbaum <[email protected]> Reviewed-By: James M Snell <[email protected]>

This LTS release comes with 109 commits, 17 of which are considered Semver-Minor. This includes 32 which are doc related, 29 which are test related, 8 which are build / tool related and 1 commit which updates a dependency. Notable Changes: * console: - added console.count() and console.clear() (James M Snell) #12678 * crypto: - expose ECDH class (Bryan English) #8188 - added cypto.randomFill() and crypto.randomFillSync() (Evan Lucas) #10209 - warn on invalid authentication tag length (Tobias Nießen) #17566 * deps: - upgrade libuv to 1.16.1 (cjihrig) #16835 * dgram: - added socket.setMulticastInterface() (Will Young) #7855 * http: - add agent.keepSocketAlive and agent.reuseSocket as to allow overridable keep-alive behavior of `Agent` (Fedor Indutny) #13005 * lib: - return this from net.Socket.end() (Sam Roberts) #13481 * module: - add builtinModules api that provides list of all builtin modules in Node (Jon Moss) #16386 * net: - return this from getConnections() (Sam Roberts) #13553 * promises: - more robust stringification for unhandled rejections (Timothy Gu) #13784 * repl: - improve require() autocompletion (Alexey Orlenko) #14409 * src: - add openssl-system-ca-path configure option (Daniel Bevenius) #16790 - add --use-bundled-ca --use-openssl-ca check (Daniel Bevenius) #12087 - add process.ppid (cjihrig) #16839 * tls: - accept `lookup` option for `tls.connect()` (Fedor Indutny) #12839 * tools, build: - a new macOS installer! (JP Wesselink) #15179 * url: - WHATWG URL api support (James M Snell) #7448 * util: - add %i and %f formatting specifiers (Roman Reiss) #10308 PR-URL: #18342

Backport-PR-URL: #17833 PR-URL: #13784 Fixes: #13771 Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: Rod Vagg <[email protected]> Reviewed-By: Michaël Zasso <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Ruben Bridgewater <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Benjamin Gruenbaum <[email protected]> Reviewed-By: James M Snell <[email protected]>

This LTS release comes with 109 commits, 17 of which are considered Semver-Minor. This includes 32 which are doc related, 29 which are test related, 8 which are build / tool related and 1 commit which updates a dependency. Notable Changes: * console: - added console.count() and console.clear() (James M Snell) #12678 * crypto: - expose ECDH class (Bryan English) #8188 - added cypto.randomFill() and crypto.randomFillSync() (Evan Lucas) #10209 - warn on invalid authentication tag length (Tobias Nießen) #17566 * deps: - upgrade libuv to 1.16.1 (cjihrig) #16835 * dgram: - added socket.setMulticastInterface() (Will Young) #7855 * http: - add agent.keepSocketAlive and agent.reuseSocket as to allow overridable keep-alive behavior of `Agent` (Fedor Indutny) #13005 * lib: - return this from net.Socket.end() (Sam Roberts) #13481 * module: - add builtinModules api that provides list of all builtin modules in Node (Jon Moss) #16386 * net: - return this from getConnections() (Sam Roberts) #13553 * promises: - more robust stringification for unhandled rejections (Timothy Gu) #13784 * repl: - improve require() autocompletion (Alexey Orlenko) #14409 * src: - add openssl-system-ca-path configure option (Daniel Bevenius) #16790 - add --use-bundled-ca --use-openssl-ca check (Daniel Bevenius) #12087 - add process.ppid (cjihrig) #16839 * tls: - accept `lookup` option for `tls.connect()` (Fedor Indutny) #12839 * tools, build: - a new macOS installer! (JP Wesselink) #15179 * url: - WHATWG URL api support (James M Snell) #7448 * util: - add %i and %f formatting specifiers (Roman Reiss) #10308 PR-URL: #18342

Backport-PR-URL: #17833 PR-URL: #13784 Fixes: #13771 Reviewed-By: Anna Henningsen <[email protected]> Reviewed-By: Rod Vagg <[email protected]> Reviewed-By: Michaël Zasso <[email protected]> Reviewed-By: Matteo Collina <[email protected]> Reviewed-By: Ruben Bridgewater <[email protected]> Reviewed-By: Colin Ihrig <[email protected]> Reviewed-By: Benjamin Gruenbaum <[email protected]> Reviewed-By: James M Snell <[email protected]>

This LTS release comes with 109 commits, 17 of which are considered Semver-Minor. This includes 32 which are doc related, 29 which are test related, 8 which are build / tool related and 1 commit which updates a dependency. Notable Changes: * console: - added console.count() and console.clear() (James M Snell) #12678 * crypto: - expose ECDH class (Bryan English) #8188 - added cypto.randomFill() and crypto.randomFillSync() (Evan Lucas) #10209 - warn on invalid authentication tag length (Tobias Nießen) #17566 * deps: - upgrade libuv to 1.16.1 (cjihrig) #16835 * dgram: - added socket.setMulticastInterface() (Will Young) #7855 * http: - add agent.keepSocketAlive and agent.reuseSocket as to allow overridable keep-alive behavior of `Agent` (Fedor Indutny) #13005 * lib: - return this from net.Socket.end() (Sam Roberts) #13481 * module: - add builtinModules api that provides list of all builtin modules in Node (Jon Moss) #16386 * net: - return this from getConnections() (Sam Roberts) #13553 * promises: - more robust stringification for unhandled rejections (Timothy Gu) #13784 * repl: - improve require() autocompletion (Alexey Orlenko) #14409 * src: - add openssl-system-ca-path configure option (Daniel Bevenius) #16790 - add --use-bundled-ca --use-openssl-ca check (Daniel Bevenius) #12087 - add process.ppid (cjihrig) #16839 * tls: - accept `lookup` option for `tls.connect()` (Fedor Indutny) #12839 * tools, build: - a new macOS installer! (JP Wesselink) #15179 * url: - WHATWG URL api support (James M Snell) #7448 * util: - add %i and %f formatting specifiers (Roman Reiss) #10308 PR-URL: #18342

This LTS release comes with 112 commits, 17 of which are considered Semver-Minor. This includes 32 which are doc related, 30 which are test related, 8 which are build / tool related and 1 commit which updates a dependency. Notable Changes: * console: - added console.count() and console.clear() (James M Snell) #12678 * crypto: - expose ECDH class (Bryan English) #8188 - added cypto.randomFill() and crypto.randomFillSync() (Evan Lucas) #10209 - warn on invalid authentication tag length (Tobias Nießen) #17566 * deps: - upgrade libuv to 1.16.1 (cjihrig) #16835 * dgram: - added socket.setMulticastInterface() (Will Young) #7855 * http: - add agent.keepSocketAlive and agent.reuseSocket as to allow overridable keep-alive behavior of `Agent` (Fedor Indutny) #13005 * lib: - return this from net.Socket.end() (Sam Roberts) #13481 * module: - add builtinModules api that provides list of all builtin modules in Node (Jon Moss) #16386 * net: - return this from getConnections() (Sam Roberts) #13553 * promises: - more robust stringification for unhandled rejections (Timothy Gu) #13784 * repl: - improve require() autocompletion (Alexey Orlenko) #14409 * src: - add openssl-system-ca-path configure option (Daniel Bevenius) #16790 - add --use-bundled-ca --use-openssl-ca check (Daniel Bevenius) #12087 - add process.ppid (cjihrig) #16839 * tls: - accept `lookup` option for `tls.connect()` (Fedor Indutny) #12839 * tools, build: - a new macOS installer! (JP Wesselink) #15179 * url: - WHATWG URL api support (James M Snell) #7448 * util: - add %i and %f formatting specifiers (Roman Reiss) #10308 PR-URL: #18342

This LTS release comes with 112 commits, 17 of which are considered Semver-Minor. This includes 32 which are doc related, 30 which are test related, 8 which are build / tool related and 1 commit which updates a dependency. Notable Changes: * console: - added console.count() and console.clear() (James M Snell) nodejs#12678 * crypto: - expose ECDH class (Bryan English) nodejs#8188 - added cypto.randomFill() and crypto.randomFillSync() (Evan Lucas) nodejs#10209 - warn on invalid authentication tag length (Tobias Nießen) nodejs#17566 * deps: - upgrade libuv to 1.16.1 (cjihrig) nodejs#16835 * dgram: - added socket.setMulticastInterface() (Will Young) nodejs#7855 * http: - add agent.keepSocketAlive and agent.reuseSocket as to allow overridable keep-alive behavior of `Agent` (Fedor Indutny) nodejs#13005 * lib: - return this from net.Socket.end() (Sam Roberts) nodejs#13481 * module: - add builtinModules api that provides list of all builtin modules in Node (Jon Moss) nodejs#16386 * net: - return this from getConnections() (Sam Roberts) nodejs#13553 * promises: - more robust stringification for unhandled rejections (Timothy Gu) nodejs#13784 * repl: - improve require() autocompletion (Alexey Orlenko) nodejs#14409 * src: - add openssl-system-ca-path configure option (Daniel Bevenius) nodejs#16790 - add --use-bundled-ca --use-openssl-ca check (Daniel Bevenius) nodejs#12087 - add process.ppid (cjihrig) nodejs#16839 * tls: - accept `lookup` option for `tls.connect()` (Fedor Indutny) nodejs#12839 * tools, build: - a new macOS installer! (JP Wesselink) nodejs#15179 * url: - WHATWG URL api support (James M Snell) nodejs#7448 * util: - add %i and %f formatting specifiers (Roman Reiss) nodejs#10308 PR-URL: nodejs#18342

mscdex added promises Issues and PRs related to ECMAScript promises. process Issues and PRs related to the process subsystem. labels Jun 19, 2017

benjamingr reviewed Jun 20, 2017

View reviewed changes

addaleax approved these changes Jun 21, 2017

View reviewed changes

bnoordhuis previously requested changes Jun 22, 2017

View reviewed changes

addaleax added the ctc-review label Jun 22, 2017

mcollina reviewed Jun 22, 2017

View reviewed changes

Trott mentioned this pull request Jun 25, 2017

Node.js Foundation Core Technical Committee (CTC) Meeting 2017-06-28 nodejs/CTC#148

Closed

Trott mentioned this pull request Jul 2, 2017

Node.js Foundation Core Technical Committee (CTC) Meeting 2017-07-05 nodejs/CTC#153

Closed

Trott mentioned this pull request Jul 16, 2017

Node.js Foundation Core Technical Committee (CTC) Meeting 2017-07-19 nodejs/CTC#154

Closed

TimothyGu mentioned this pull request Jul 17, 2017

repl: handle null thrown #14306

Closed

3 tasks

Trott mentioned this pull request Jul 31, 2017

Node.js Foundation Core Technical Committee (CTC) Meeting 2017-08-02 nodejs/CTC#160

Closed

bnoordhuis removed the ctc-review label Jul 31, 2017

addaleax added the ctc-agenda label Aug 1, 2017

MylesBorins added the backport-requested-v6.x label Dec 19, 2017

TimothyGu mentioned this pull request Dec 23, 2017

[backport v6.x] promises: more robust stringification #17833

Closed

3 tasks

MylesBorins added backported-to-v6.x and removed backport-requested-v6.x labels Jan 18, 2018

MylesBorins mentioned this pull request Jan 24, 2018

v6.13.0 proposal #18342

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

promises: more robust stringification #13784

promises: more robust stringification #13784

TimothyGu commented Jun 19, 2017

bnoordhuis commented Jun 19, 2017

jasnell commented Jun 19, 2017

TimothyGu commented Jun 20, 2017

bnoordhuis commented Jun 20, 2017

benjamingr Jun 20, 2017

benjamingr commented Jun 20, 2017

benjamingr commented Jun 20, 2017

TimothyGu commented Jun 20, 2017

addaleax left a comment

benjamingr commented Jun 21, 2017

bnoordhuis left a comment

addaleax commented Jun 22, 2017

mcollina left a comment

mcollina Jun 22, 2017

addaleax Jun 22, 2017

mcollina Jun 22, 2017

joyeecheung Jun 24, 2017

Fishrock123 Aug 14, 2017

addaleax Aug 14, 2017

thefourtheye Aug 14, 2017

addaleax Aug 14, 2017

benjamingr commented Jun 22, 2017 •

edited

Loading

mcollina commented Jun 22, 2017

TimothyGu commented Jun 22, 2017

mcollina commented Jun 22, 2017

bnoordhuis commented Jun 24, 2017

bnoordhuis commented Jul 31, 2017

addaleax commented Aug 1, 2017

hashseed commented Dec 19, 2017 •

edited

Loading

MylesBorins commented Dec 19, 2017

bnoordhuis commented Dec 19, 2017

TimothyGu commented Dec 19, 2017 •

edited

Loading

TimothyGu commented Dec 23, 2017

promises: more robust stringification #13784

promises: more robust stringification #13784

Conversation

TimothyGu commented Jun 19, 2017

Checklist

Affected core subsystem(s)

bnoordhuis commented Jun 19, 2017

jasnell commented Jun 19, 2017

TimothyGu commented Jun 20, 2017

bnoordhuis commented Jun 20, 2017

Choose a reason for hiding this comment

benjamingr commented Jun 20, 2017

benjamingr commented Jun 20, 2017

TimothyGu commented Jun 20, 2017

addaleax left a comment

Choose a reason for hiding this comment

benjamingr commented Jun 21, 2017

bnoordhuis left a comment

Choose a reason for hiding this comment

addaleax commented Jun 22, 2017

mcollina left a 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

benjamingr commented Jun 22, 2017 • edited Loading

mcollina commented Jun 22, 2017

TimothyGu commented Jun 22, 2017

mcollina commented Jun 22, 2017

bnoordhuis commented Jun 24, 2017

bnoordhuis commented Jul 31, 2017

addaleax commented Aug 1, 2017

hashseed commented Dec 19, 2017 • edited Loading

MylesBorins commented Dec 19, 2017

bnoordhuis commented Dec 19, 2017

TimothyGu commented Dec 19, 2017 • edited Loading

TimothyGu commented Dec 23, 2017

benjamingr commented Jun 22, 2017 •

edited

Loading

hashseed commented Dec 19, 2017 •

edited

Loading

TimothyGu commented Dec 19, 2017 •

edited

Loading