node 10.0.0 with ts-node throws a TypeError in isInsideNodeModules when an error is thrown by user code #20258

mdouglass · 2018-04-24T20:18:44Z

Version: v10.0.0
Platform: Darwin bigbird.local 17.5.0 Darwin Kernel Version 17.5.0: Mon Mar 5 22:24:32 PST 2018; root:xnu-4570.51.1~1/RELEASE_X86_64 x86_64

Attached sample project (which is a single ts file with just throw new Error()) when run via npm start fails with:

mdouglass$ npm start

> temp@1.0.0 start /Users/mdouglass/kixeye/km/server/temp
> ts-node index.ts

internal/util.js:360
    const filename = frame.getFileName();
                           ^

TypeError: frame.getFileName is not a function
    at isInsideNodeModules (internal/util.js:360:28)
    at showFlaggedDeprecation (buffer.js:149:8)
    at new Buffer (buffer.js:174:3)
    at Array.<anonymous> (/Users/mdouglass/kixeye/km/server/temp/node_modules/source-map-support/source-map-support.js:163:21)
    at /Users/mdouglass/kixeye/km/server/temp/node_modules/source-map-support/source-map-support.js:53:24
    at mapSourcePosition (/Users/mdouglass/kixeye/km/server/temp/node_modules/source-map-support/source-map-support.js:185:21)
    at wrapCallSite (/Users/mdouglass/kixeye/km/server/temp/node_modules/source-map-support/source-map-support.js:357:20)
    at /Users/mdouglass/kixeye/km/server/temp/node_modules/source-map-support/source-map-support.js:392:26
    at Array.map (<anonymous>)
    at Function.prepareStackTrace (/Users/mdouglass/kixeye/km/server/temp/node_modules/source-map-support/source-map-support.js:391:24)

repro-nodejs-10-throw.zip

The text was updated successfully, but these errors were encountered:

jasnell · 2018-04-24T20:23:08Z

Ping @addaleax ... any ideas?

jasnell · 2018-04-24T20:29:53Z

~~This is possibly a bad interaction with make-error but I need to verify~~
I take that back... https://github.com/evanw/node-source-map-support/blob/master/source-map-support.js looks like a better candidate...

(heh... it would help if I actually read the entire stack track ... lol...)

apapirovski · 2018-04-24T21:10:05Z

I'm looking into this but anyone else is also welcome to dig in.

tirthamazumdar · 2018-04-24T21:26:29Z

I do have the same issue. Just now upgraded to 10.0.0 from version 8, and all my tests are failing.

internal/util.js:360
const filename = frame.getFileName();
^
TypeError: frame.getFileName is not a function
at isInsideNodeModules (internal/util.js:360:28)
at showFlaggedDeprecation (buffer.js:149:8)
at new Buffer (buffer.js:174:3)

jasnell · 2018-04-24T21:37:11Z

@tirthamazumdar ... are you also using source-map-support?

apapirovski · 2018-04-24T21:38:40Z

@jasnell It just comes with ts-node. source-map-support is definitely what's causing it but still not sure how to resolve.

tirthamazumdar · 2018-04-24T21:45:49Z

@jasnell yes I am also using source-map-support.. Just now compiled typescript into javascript and the running directly the javascript tests the problem goes away.
Looks like issue is with source-map-support

jasnell · 2018-04-24T22:16:31Z

Ok, @apapirovski @addaleax ... looks like the transpiler here is mucking around with the generation of the stack frames in a way the check in core did not anticipate. I think we should likely do in this case is have isInsideNodeModules() return false if it cannot reliably process the stack. That would cause the Buffer deprecation warning to emit in this case, but that's not really a bad thing.

jasnell · 2018-04-24T22:34:06Z

Attempting to put together a standalone repo test case now but not having much luck yet.

apapirovski · 2018-04-24T22:45:42Z

@jasnell Here's the simplest possible reproduction:

Error.prepareStackTrace = (err, trace) => new Buffer();

new Error().stack;

The error happens because V8 won't call prepareStackTrace if it's already preparing a stack trace (recursive call). I don't think we can fix so, as you said, we'll just have to skip the check and emit the warning in those cases.

PR coming up.

addaleax · 2018-04-24T22:47:55Z

Ouch. Yes, that makes sense – thanks for figuring this out so quick. I am surprised though, it might be nice if V8 supported this – after all, the other Error object does come from a difference VM Context, so one wouldn’t expect that kind of interaction … @nodejs/v8?

jasnell · 2018-04-24T22:53:15Z

I had seen the Buffer use there but hadn't realized it was a recursive check. Good catch.

bnoordhuis · 2018-04-24T22:57:43Z

the other Error object does come from a difference VM Context

Not sure I follow. If it's about @apapirovski's test case, what other Error object and VM context?

apapirovski · 2018-04-24T23:01:47Z

Not sure I follow. If it's about @apapirovski's test case, what other Error object and VM context?

isInsideNodeModules generates a stack-trace yielding function in a VM to get the proper stack trace. See lib/internal/util.js around line ~340. It also uses prepareStackTrace which in V8 is guarded against being called recursively

addaleax · 2018-04-24T23:01:55Z

@bnoordhuis I’m talking about this:

node/lib/internal/util.js

Lines 343 to 352 in ad5307f

    
               getStructuredStack = runInNewContext(`(function() { 
        
                 Error.prepareStackTrace = function(err, trace) { 
        
                   err.stack = trace; 
        
                 }; 
        
                 Error.stackTraceLimit = Infinity; 
        
                 return function structuredStack() { 
        
                   return new Error().stack; 
        
                 }; 
        
               })()`, {}, { filename: 'structured-stack' });

It’s a different context, and while I get the intention on V8’s side, this does seem surprising; it’s not like it’s recusing into the same prepareStackTrace function here as the outer one.

rvagg · 2018-04-24T23:21:28Z

this requires that an error be thrown to show up at all right? so there's no way we could have picked this up by having anything extra in citgm? breaking TS support is becoming a bigger deal as its adoption spreads.

apapirovski · 2018-04-24T23:23:33Z

I think maybe having https://github.com/evanw/node-source-map-support in there would've caught it? But I haven't checked the test suite.

Edit: Yep, would've caught it:

  28 passing (1s)
  1 failing

  1)  should allow for runtime inline source maps:

      AssertionError [ERR_ASSERTION]: 'TypeError: frame.getFileName is not a function' == 'Error: this is the error'

apapirovski · 2018-04-24T23:31:41Z

Thinking about it more, it's probably not the worst module to add. There are a few less common APIs exercised in their module & test suite.

Trott · 2018-04-24T23:32:26Z

I think maybe having https://github.com/evanw/node-source-map-support in there would've caught it? But I haven't checked the test suite.

Edit: Yep, would've caught it:

@nodejs/citgm ☝️

richardlau · 2018-04-25T00:21:39Z

I think maybe having https://github.com/evanw/node-source-map-support in there would've caught it? But I haven't checked the test suite.

Edit: Yep, would've caught it:

@nodejs/citgm ☝️

Pull requests welcome over at https://github.com/nodejs/citgm. Please make sure to document in the PR the fulfilled hard and soft requirements met by the module in question as per: https://github.com/nodejs/citgm/blob/master/CONTRIBUTING.md#submitting-a-module-to-citgm

bnoordhuis · 2018-04-25T11:13:31Z

It’s a different context, and while I get the intention on V8’s side, this does seem surprising; it’s not like it’s recusing into the same prepareStackTrace function here as the outer one.

Okay, I get it now. Wouldn't be terribly hard to fix in V8 at the cost of some additional complexity (tracking the entered contexts.)

That said, the goal of isInsideNodeModules() is to find the first non-core stack frame and there are arguably better ways of doing that than using a custom Error.prepareStackTrace.

Strawman: capture the stack trace with v8::StackTrace::CurrentStackTrace() and find the first frame with a script id that isn't of a built-in module.

We don't currently record our own script ids but that's relatively straightforward to add.

Tests fail due to nodejs/node#20258. We can and should re-enable this once that's been released.

This should resolve the current build failures and can be reverted once nodejs/node#20258 is resolved. See #40.

jy95 · 2018-04-29T20:56:33Z

I found related to source-map-support in my broken build : https://travis-ci.org/jy95/mediaScan/jobs/372789737#L517

yoav-zibin · 2018-05-04T13:35:58Z

What's the solution?

$ node -v
v10.0.0
$ npm -v
6.0.0

TypeError: frame.getFileName is not a function
at isInsideNodeModules (internal/util.js:360:28)
at showFlaggedDeprecation (buffer.js:149:8)
at new Buffer (buffer.js:174:3)
at Array. (/Users/yzibin/GitHub/NewGamePortal/node_modules/source-map-support/source-map-support.js:149:21)
at /Users/yzibin/GitHub/NewGamePortal/node_modules/source-map-support/source-map-support.js:53:24
at mapSourcePosition (/Users/yzibin/GitHub/NewGamePortal/node_modules/source-map-support/source-map-support.js:171:21)
at wrapCallSite (/Users/yzibin/GitHub/NewGamePortal/node_modules/source-map-support/source-map-support.js:343:2

hashseed · 2018-05-04T14:13:18Z

V8 only has a single flag for checking whether prepareStackTrace is recursing. Are you saying having the flag per context or stored on the prepareStackTrace function itself would solve this? @schuay

When isInsideNodeModules gets called while already processing another stack trace, V8 will not call prepareStackTrace again. This used to cause Node.js to just crash — fix it by checking for expected return type of the stack (Array). PR-URL: #20266 Fixes: #20258 Reviewed-By: Anna Henningsen <anna@addaleax.net> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Trivikram Kamat <trivikr.dev@gmail.com>

apapirovski · 2018-05-05T06:00:08Z

@yoav-zibin 10.1.0 should come out soon and have a fix for it.

mariusvw · 2018-05-11T12:24:32Z

Just upgraded, can confirm that after upgrading to 10.1.0 everything works alright on our machines.

addaleax added the buffer Issues and PRs related to the buffer subsystem. label Apr 24, 2018

apapirovski added the confirmed-bug Issues with confirmed bugs. label Apr 24, 2018

apapirovski mentioned this issue Apr 24, 2018

util: fix isInsideNodeModules inside error #20266

Closed

3 tasks

smhxx mentioned this issue Apr 25, 2018

chore(package): update devDependency ts-node to version 6.x smhxx/atom-ts-transpiler#40

Merged

pimterry mentioned this issue Apr 25, 2018

Stick with home-or-tmp version 2, for max engine compatibility balena-io-modules/balena-settings-client#50

Merged

pimterry added a commit to balena-io-modules/balena-settings-client that referenced this issue Apr 25, 2018

Temporarily stop testing in latest Node due to a Node 10 bug

47f1d30

Tests fail due to nodejs/node#20258. We can and should re-enable this once that's been released.

smhxx added a commit to smhxx/atom-ts-transpiler that referenced this issue Apr 26, 2018

chore(travis): temporarily pin Node.js version to 9

3038926

This should resolve the current build failures and can be reverted once nodejs/node#20258 is resolved. See #40.

dcharbonnier mentioned this issue Apr 27, 2018

Update to node 10 in .travis.yml dcharbonnier/hydrated-ws#80

Closed

garethcmurphy mentioned this issue Apr 27, 2018

Version 10 of node.js has been released datacurationproject/CatanieDataLoad#23

Closed

apapirovski closed this as completed in 16aee38 Apr 28, 2018

jakipatryk mentioned this issue Apr 28, 2018

Update dependencies jakipatryk/steemconnect-firebase-functions#58

Merged

williamoliveira mentioned this issue May 1, 2018

Stalls with Yarn and Node 10 kulshekhar/ts-jest#497

Closed

nicolasnoble mentioned this issue May 3, 2018

Unused message properties are being defaulted grpc/grpc-node#189

Closed

jdx mentioned this issue May 6, 2018

"frame.getFileName is not a function" error when passing -h to template hello command oclif/oclif#112

Closed

huan mentioned this issue May 7, 2018

Node v10.0 TypeError: frame.getFileName is not a function huan/hot-import#11

Open

noxan mentioned this issue May 7, 2018

Node version 10 breaks tests and ci picter/gi-cli#35

Open

DanielRuf mentioned this issue May 7, 2018

chore: add Node.js 10 tensorflow/tfjs#273

Merged

victorandree mentioned this issue May 14, 2018

Update npm and yarn images to node 10.1.0 GoogleCloudPlatform/cloud-builders#278

Merged

Sypheos mentioned this issue May 15, 2018

Cannot connect to private network TheThingsArchive/node-app-sdk#84

Closed

mhoeger mentioned this issue Jun 8, 2018

NodeJs v10 support Azure/azure-functions-nodejs-worker#91

Merged

kiprasmel mentioned this issue May 9, 2020

Package is broken with node 14 : fs.watch with recursive is throwing on incompatible os wclr/ts-node-dev#143

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

node 10.0.0 with ts-node throws a TypeError in isInsideNodeModules when an error is thrown by user code #20258

node 10.0.0 with ts-node throws a TypeError in isInsideNodeModules when an error is thrown by user code #20258

mdouglass commented Apr 24, 2018

jasnell commented Apr 24, 2018

jasnell commented Apr 24, 2018 •

edited

Loading

apapirovski commented Apr 24, 2018

tirthamazumdar commented Apr 24, 2018

jasnell commented Apr 24, 2018

apapirovski commented Apr 24, 2018

tirthamazumdar commented Apr 24, 2018

jasnell commented Apr 24, 2018

jasnell commented Apr 24, 2018

apapirovski commented Apr 24, 2018

addaleax commented Apr 24, 2018

jasnell commented Apr 24, 2018

bnoordhuis commented Apr 24, 2018

apapirovski commented Apr 24, 2018

addaleax commented Apr 24, 2018

rvagg commented Apr 24, 2018

apapirovski commented Apr 24, 2018 •

edited

Loading

apapirovski commented Apr 24, 2018

Trott commented Apr 24, 2018

richardlau commented Apr 25, 2018

bnoordhuis commented Apr 25, 2018

jy95 commented Apr 29, 2018

yoav-zibin commented May 4, 2018

hashseed commented May 4, 2018

apapirovski commented May 5, 2018

mariusvw commented May 11, 2018

node 10.0.0 with ts-node throws a TypeError in isInsideNodeModules when an error is thrown by user code #20258

node 10.0.0 with ts-node throws a TypeError in isInsideNodeModules when an error is thrown by user code #20258

Comments

mdouglass commented Apr 24, 2018

jasnell commented Apr 24, 2018

jasnell commented Apr 24, 2018 • edited Loading

apapirovski commented Apr 24, 2018

tirthamazumdar commented Apr 24, 2018

jasnell commented Apr 24, 2018

apapirovski commented Apr 24, 2018

tirthamazumdar commented Apr 24, 2018

jasnell commented Apr 24, 2018

jasnell commented Apr 24, 2018

apapirovski commented Apr 24, 2018

addaleax commented Apr 24, 2018

jasnell commented Apr 24, 2018

bnoordhuis commented Apr 24, 2018

apapirovski commented Apr 24, 2018

addaleax commented Apr 24, 2018

rvagg commented Apr 24, 2018

apapirovski commented Apr 24, 2018 • edited Loading

apapirovski commented Apr 24, 2018

Trott commented Apr 24, 2018

richardlau commented Apr 25, 2018

bnoordhuis commented Apr 25, 2018

jy95 commented Apr 29, 2018

yoav-zibin commented May 4, 2018

hashseed commented May 4, 2018

apapirovski commented May 5, 2018

mariusvw commented May 11, 2018

jasnell commented Apr 24, 2018 •

edited

Loading

apapirovski commented Apr 24, 2018 •

edited

Loading