Repository navigation
fix(sentry): give stackless console events a stack and stop issue explosion - #2810
Merged
Merged
Conversation
…losion
Three related defects surfaced by PostHog's $cymbal_errors filter, which
flags $exception events arriving with an empty exception list. Those events
were reaching Sentry fine — but as message events with no frames, and with
grouping keys that forked a new issue per URL variant and per device.
captureConsoleIntegration only calls captureException when one of the console
args is an Error instance; otherwise it calls captureMessage, which carries no
stack. Set attachStacktrace so every message event gets one, including the
deliberate captureMessage calls in fetchWithSentry and native-auth-capture.
That covers third-party console noise too, where we cannot touch the call site.
A stack cannot recover a payload the call site already destroyed, so fix the
first-party sites that discarded theirs: SetupPasskey passed err.name and
err.message instead of err (the largest family, ~126/wk); qr-scanner rejects
with bare strings; the Sumsub loader had no Error at all; and the crypto
withdraw and useSendMoney handlers were logging [object Object] in
money-moving flows. Adds utils/to-error, replacing what were already three
ad-hoc copies of the same normalization, and serializing plain objects so the
shape survives into the title.
sanitizeUrl normalized numeric query values but not the rest, so one 429 on
/bridge/exchange-rate was three issues (accountType=iban/clabe/gb). Normalize
every value, with a lookahead so already-correct fingerprints keep their
existing issue instead of forking on deploy. Anchoring the numeric rule fixes
a bug it caught: unanchored, it ate the leading 0 of an 0x-prefixed address
and left {id}xaf88d065, so every wallet still got its own fingerprint.
OneSignal inlines a per-device onesignalId into its op-failure message, minting
a Sentry issue per device. Collapse those onto one fingerprint per op name
rather than dropping them — push-subscription writes failing is the signal that
was missing while iOS push was dead on a mis-provisioned APNs key. Its worker
messenger's no-service-worker warning is pure noise and is ignored.
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
Contributor
|
Warning Review limit reached
On-demand reviews are free for the next 27 days. After that, they cost $0.25 per reviewed file. Or wait 31 minutes for your next included review. View limit detailsLimit details: You’ve used all 2 included reviews currently available. Your 55 included PR review attempts over the past 7 days set your current allowance at 2 reviews per hour. Review configuration: ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (12)
Comment |
Collaborator
Author
|
@coderabbitai review |
Contributor
|
Contributor
Code-analysis diffPainscore total: 7425.47 → 7431.14 (+5.67) 🆕 New findings (41)
…and 21 more. ✅ Resolved (39)
…and 19 more. 📈 Painscore deltas (top movers)
|
Contributor
🧪 UI test report — ✅ all greenSuites
📊 Coverage (unit)
⏱ 10 slowest test cases
|
innolope-dev
added a commit
that referenced
this pull request
Aug 24, 2026
#2810 added collapseNoisyFingerprint at the same spot this branch added isTransientCapgoNoise. Independent additions, both kept: Capgo's transient updater logs are suppressed in shouldIgnoreError, OneSignal's per-device op failures are re-fingerprinted in beforeSendHandler.
This branch was successfully deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
What
Started from this PostHog view —
$exceptionevents where$cymbal_errorsis set. That property is PostHog's flag for "empty exception list": the event arrived with a message but noexception.values, so no type and no stack.Those events were reaching Sentry fine. The problem is the shape they arrived in, and how they grouped. 433 events / 255 people over 7d (~3.4% of exception volume).
Why they were stackless
captureConsoleIntegrationonly callscaptureExceptionwhen one of the console args is anErrorinstance; otherwise it callscaptureMessage, which carries no stack at all (captureconsole.js). Every event on that page was either aconsole.errorwhose args held noError, or one of our deliberatecaptureMessagecalls.Changes
1.
attachStacktrace: true— synthesizes a stack for every message event, includingfetchWithSentry/native-auth-capture, and third-party console noise where we can't touch the call site. One line, covers all 433.2. Call sites that destroyed their payload — a stack can't recover an error the call site already threw away:
SetupPasskey.tsxerr.name, err.messageinstead oferr— largest family, ~126/wkuseQRScanner.ts×2'Camera not found.')SumsubKycWrapper.tsxErrorobject at allwithdraw/crypto/page.tsx[object Object]— money-moving flowuseSendMoney.ts[object Object]— money-moving flowAdds
utils/to-error.ts, replacing what were already three ad-hoc copies of the same normalization. It JSON-serializes plain objects so the shape survives into the title instead of collapsing.Passkey filtering is unchanged — the
AbortError/not allowed by the uservariants still match, via exception type rather than message.3.
sanitizeUrlfingerprint gap — it normalized numeric query values but not the rest, so one upstream 429 was three separate issues:exchange-rate?accountType=ibanexchange-rate?accountType=clabeexchange-rate?accountType=gbNow normalizes every value. A lookahead skips values the numeric pass already replaced, so fingerprints that were already correct keep their existing issue rather than forking a new one on deploy.
While adding tests for this I found a pre-existing bug: the numeric rule was unanchored, so
?address=0xaf88d065had its leading0replaced and became{id}xaf88d065— every wallet still got its own fingerprint and the value was unreadable. Anchored with(?=&|$).4. OneSignal issue explosion — its op-failure message inlines a per-device
onesignalId, so every device minted its own Sentry issue. These are fingerprinted, not dropped: push-subscription writes failing is exactly the signal that was missing while iOS push was dead on a mis-provisioned APNs key. Collapsed onto one issue per op name. The worker messenger's no-service-worker warning is pure noise and is ignored.Note
attachStacktracelands onthreads, notexception.values, so PostHog's mirror will still flag the remaining third-party ones as an empty exception list. That flag only clears where we now pass a realError. Sentry is fully fixed; the PostHog view shrinks but won't empty out.Testing
to-error.test.ts(8 cases),sentry-fingerprint-url.test.ts(7 cases, incl. the no-refingerprint regression guard)sentry.utils.test.tswith the OneSignal fingerprint + ignore casestoErrortests and 272 tests across the touched components pass; prettier and eslint cleanOverlap
Touches
sentry.utils.ts, as does #2735 (chained-error matching + Capgo noise). No textual conflict — different regions, andcollapseNoisyFingerprintcomposes with itsshouldIgnoreErrorchanges — but whichever lands second deserves a re-read.