Skip to content

Conversation

@giortzisg
Copy link
Contributor

Description

Since the addition of Telemetry Buffers moves the serialization to a background worker, user provided attributes that can be mutated should be deep copied, to avoid panics during serialization.

@giortzisg giortzisg requested a review from lcian November 27, 2025 14:50
@giortzisg giortzisg self-assigned this Nov 27, 2025
@codecov
Copy link

codecov bot commented Nov 27, 2025

Codecov Report

❌ Patch coverage is 28.84615% with 74 lines in your changes missing coverage. Please review.
✅ Project coverage is 85.83%. Comparing base (34261f3) to head (647e8aa).

Files with missing lines Patch % Lines
deepcopy.go 24.32% 50 Missing and 6 partials ⚠️
client.go 0.00% 18 Missing ⚠️
Additional details and impacted files
@@            Coverage Diff             @@
##           master    #1148      +/-   ##
==========================================
- Coverage   86.85%   85.83%   -1.02%     
==========================================
  Files          62       63       +1     
  Lines        6092     6184      +92     
==========================================
+ Hits         5291     5308      +17     
- Misses        587      656      +69     
- Partials      214      220       +6     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Comment on lines 330 to 341
clone.breadcrumbs = make([]*Breadcrumb, len(scope.breadcrumbs))
copy(clone.breadcrumbs, scope.breadcrumbs)
clone.breadcrumbs = make([]*Breadcrumb, 0, len(scope.breadcrumbs))
for _, b := range scope.breadcrumbs {
clone.breadcrumbs = append(clone.breadcrumbs, deepCopyBreadcrumb(b))
}
clone.attachments = make([]*Attachment, len(scope.attachments))
copy(clone.attachments, scope.attachments)
for key, value := range scope.tags {
clone.tags[key] = value
}
for key, value := range scope.contexts {
clone.contexts[key] = cloneContext(value)
}
for key, value := range scope.extra {
clone.extra[key] = value
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Correct me if I'm wrong, but:
this was here already and was not causing problems, and it's a separate code path from the buffer/transport, so I don't think this should be changed at all.
I assume this was deliberately done this way because scope forking happens frequently and so you wouldn't want to do an expensive deep copy.

// a proper deep copy: if some context values are pointer types (e.g. maps),
// they won't be properly copied.
func cloneContext(c Context) Context {
res := make(Context, len(c))
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

same as above

client.go Outdated
if event.User.Data != nil {
eventToBuffer.User.Data = deepCopyMapStringString(event.User.Data)
}

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Nested in Event there are other things we might care about, for example a Span can have Data (and apparently tags and extra), same about Log attributes, etc. Do we care about those?

@aldy505
Copy link
Contributor

aldy505 commented Dec 30, 2025

...or we can serialize it up front, and then send them to telemetry buffer after serialization? I would assume doing deep copy would inflict in more RAM usage than just serialize it up front.

@giortzisg
Copy link
Contributor Author

giortzisg commented Jan 7, 2026

...or we can serialize it up front, and then send them to telemetry buffer after serialization?

I was weighting both approaches, but i think that pre-serialization has more drawbacks:

  • the scheduler needs the whole event, to get the trace context to build the envelope (we would need to pass metadata around, along with the serialized event)
  • batching logic for logs/spans would also require metadata and more complicated
  • performance wise, it's better to serialize on the background for:
    • request latency for the user is low
    • when rate limited we drop immediately, skipping serialization completely

I would assume doing deep copy would inflict in more RAM

That is a valid concern but we only keep the copy around till we flush the event (short lifespan) and other SDKs already copy some event data.

Comment on lines 443 to 449
}

for key, value := range scope.extra {
event.Extra[key] = value
event.Extra[key] = deepCopyValue(value)
}
}

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Bug: Breadcrumbs are not deep-copied when added to the scope. If a user mutates a breadcrumb's Data map after adding it, a race condition can occur during serialization.
Severity: CRITICAL

🔍 Detailed Analysis

The AddBreadcrumb function appends breadcrumbs by reference, but breadcrumbs can contain a mutable Data map. If a user adds a breadcrumb and later modifies its Data map, this can create a race condition between the user's code and the background serialization worker that reads the map. This can lead to a runtime panic from concurrent map access. This behavior is inconsistent with the pull request's goal of deep-copying other mutable user-provided attributes like extra and context to prevent this exact issue.

💡 Suggested Fix

Deep-copy the Breadcrumb.Data map when a breadcrumb is added to the scope via AddBreadcrumb. This would align the handling of breadcrumbs with that of extra and context data, preventing potential race conditions during background serialization.

🤖 Prompt for AI Agent
Review the code at the location below. A potential bug has been identified by an AI
agent.
Verify if this is a real issue. If it is, propose a fix; if not, explain why it's not
valid.

Location: scope.go#L443-L449

Potential issue: The `AddBreadcrumb` function appends breadcrumbs by reference, but
breadcrumbs can contain a mutable `Data` map. If a user adds a breadcrumb and later
modifies its `Data` map, this can create a race condition between the user's code and
the background serialization worker that reads the map. This can lead to a runtime panic
from concurrent map access. This behavior is inconsistent with the pull request's goal
of deep-copying other mutable user-provided attributes like `extra` and `context` to
prevent this exact issue.

Did we get this right? 👍 / 👎 to inform future reviews.
Reference ID: 8391684

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants