fix data race in span lifecycle

[mrproliu]

No description provided.

[Copilot]

Pull request overview

This PR hardens the core tracing implementation against span/segment lifecycle concurrency misuse by introducing per-span synchronization and an explicit “freeze” point for safe lock-free reporting, plus a synchronized correlation context to prevent fatal concurrent map access during propagation. It also adds targeted recovery in long-lived goroutines (metrics/profile collection and reporter transforms) and introduces comprehensive race/e2e regression tests and a benchmark for the new hot path.

Changes:

• Add per-span locking + freeze semantics to make concurrent mutation/end safe and to publish immutable span data to the collector/reporter.
• Replace segment correlation storage with a synchronized CorrelationContext and propagate via snapshots to avoid concurrent map iteration/writes.
• Add recover guards and new stress/race/e2e tests + a benchmark to prevent regressions.

Reviewed changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
plugins/core/tracing.go	Snapshot correlation on exit-span injection; clone runtime on ContinueContext to avoid shared-map races.
plugins/core/test_base.go	Make test reporter storage thread-safe; initialize correlation defaults in test tracer.
plugins/core/span_tracing.go	Update segment collector lifecycle and span end/async-finish paths for freeze + safe collection.
plugins/core/span_default.go	Introduce opLock + freeze (ended) to serialize span mutations and drop late writes safely.
plugins/core/correlation.go	Add synchronized correlation store with snapshot/clone support.
plugins/core/context.go	Optimize runtime context cloning and make runtime map lazy-allocated.
plugins/core/metrics.go	Add per-iteration recover to prevent user meter panics from killing the process.
plugins/core/profile.go	Add per-chunk recover and tighten locking around profile task state updates.
plugins/core/propagating.go	Clarify that wire-format SpanContext.CorrelationContext remains a plain map for encode/decode.
plugins/core/reporter/grpc/grpc.go	Register recover before transform + send to avoid collector-goroutine crashes.
plugins/core/reporter/kafka/kafka.go	Register recover before transform + send to avoid collector-goroutine crashes.
plugins/core/reporter/grpc/send_tracing_recover_test.go	New test ensuring SendTracing recover wraps transform panics.
plugins/core/reporter/kafka/send_tracing_recover_test.go	New test ensuring SendTracing recover wraps transform panics.
plugins/core/span_freeze_test.go	New unit tests for freeze semantics, correlation snapshots, async behavior, and runtime cloning.
plugins/core/segment_datarace_test.go	New -race regression tests covering concurrent mutators, freeze vs reporting, double-End, async finish, etc.
plugins/core/span_hostile_workload_test.go	New hostile e2e workload exercising real pipeline under concurrency abuse patterns.
plugins/core/span_crash_e2e_test.go	Subprocess e2e test asserting no panic/runtime fatal under hostile workload with runtime hardening.
plugins/core/span_sync_bench_test.go	New benchmark to track performance impact of span lifecycle locking.
CHANGES.md	Release note entries for the race fixes and added recover protections.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 20 out of 20 changed files in this pull request and generated no new comments.