Repository navigation
Wrong explanation about the Exception Interface in docs. #8871
Description
Is there an existing issue for this?
- I have checked for existing issues https://github.com/getsentry/sentry-javascript/issues
- I have reviewed the documentation https://docs.sentry.io/
- I am using the latest SDK release https://github.com/getsentry/sentry-javascript/releases
How do you use Sentry?
Self-hosted/on-premise
Which SDK are you using?
@sentry/browser
SDK Version
7.57.0
Framework Version
7.57.0
Link to Sentry event
No response
SDK Setup
Sentry.init({
dsn: <redacted>,
integrations: integrations,
release: `${$BUILD_FLAVOR}-${$REVISION}`,
environment: $BUILD_FLAVOR,
transport: (transportOptions: BrowserTransportOptions) =>
makeFetchTransport(transportOptions, fetchImpl),
maxBreadcrumbs: 100,
sendClientReports: false,
beforeSend: (event: Event, hint?: EventHint) => {
<redacted>
/**
* Essentially, here if the event.exception.values[0] is of the type that we want to skip, we return null,
* otherwise we just return the event.
*/
},
});Steps to Reproduce
Throw an error from a function (let's call it MessageRenderer) that returns React components based on certain conditions, and in one of the conditions it throws an error.
The return of this function is wrapped in sentry react's ErrorBoundary component.
Expected Result
When looking at the exception.values list in the event object in the beforeSend function, based on the docs (link: https://develop.sentry.dev/sdk/event-payloads/exception/), I was expecting the original error thrown from the function (MessageRenderer from my example above) to be present in the first index, followed by the ErrorBoundary error.
Actual Result
But, the first index (index 0) is taken up by the React ErrorBoundary Error object and the second index (index 1) contains the original thrown error. This is opposite of what is mentioned in the docs.
Am I missing something or is Error Boundary a unique case where this is not true?
Activity
Lms24 commentedon Aug 28, 2023
Hi @bhargavcp just to confirm, you're using
@sentry/react, correct?Also, did you set up our ErrorBoundary as described here?
Since you're on self-hosted Sentry, would you mind sharing a (minimal) JSON of the event you're describing?
Afaict, we set the original error (sticking with your example, from
MessageRenderer) ascausefor the React error boundary error. Another integration of ours,LinkedErrorswill take care of correctly adding the cause (i.e. the original error) to the array of errors inexception.values. Maybe there's indeed a problem with ordering.Gonna tag @AbhiPrasad as you have the most context around the error boundary - do you think we need to adjust something here?
AbhiPrasad commentedon Aug 28, 2023
The ordering here is on purpose. Please see #7557 (comment) for more details.
bhargavcp commentedon Aug 28, 2023
@AbhiPrasad @Lms24 thank you for the reply. I'm just wondering, does this ordering apply for all types of chained exceptions or is React Error Boundary a special case?
So in the Python snippet from the docs
should the order of error values be [ValueError from e, Exception]? i.e. reverse of what is mentioned in the docs?
@Lms24 yes we're using @sentry/react and it is set up as mentioned in the doc you linked. Here's what the exception.values array looks like.
{ "values": [ { "type": "React ErrorBoundary Error", "value": "nonParticipantBotResponse", "stacktrace": { "frames": [....] }, "mechanism": { "type": "chained", "handled": true, "source": "cause", "exception_id": 1, "parent_id": 0 } }, { "type": "UnsupportedMessageError", "value": "nonParticipantBotResponse", "stacktrace": { "frames": [....] }, "mechanism": { "type": "generic", "handled": true, "is_exception_group": true, "exception_id": 0 } } ] }AbhiPrasad commentedon Aug 28, 2023
The react error boundary is a special case and is inconsistent on purpose. If the quality of the error produced react gets higher in a future version of react we'll re-adjust this to make it consistent with everything else.
bhargavcp commentedon Aug 28, 2023
Thanks @AbhiPrasad , which means that the Python example in the doc is correct and we can expect non-React boundary errors to behave as mentioned in the docs?
AbhiPrasad commentedon Aug 28, 2023
yes you should expect the other errors to behave as documented. If they don't please let us know and we'll fix asap!