What's Changed

• ci: run benchmarks only for the top of a PR stack by @jbachorik in #822
• feat(nativemem): measure allocator overhead instead of assuming a blanket factor by @rkennke in #759
• walkVM attribution addresses (PROF-15955) + one shared DWARF step (PROF-16059) by @rkennke in #813
• Update README.md for optional -PenableFaultInjection and -PenableSamplerPerf builds by @zhengyu123 in #823
• Emit one addressing convention for every frame by @rkennke in #830
• Hand back raw return addresses from the unwind helpers by @rkennke in #831
• fix: plug remote symbolication enum to JFR by @kaahos in #828
• Improve AArch64 runtime stub unwinding with precomputed per-stub unwind info by @jbachorik in #816
• Compute the attribution pc with integer arithmetic by @rkennke in #835
• Build the gtest helper libraries again by @rkennke in #837
• Reference-chain tracker and leak-signal engine by @jbachorik in #797
• Write saved GOT slots only while their library is loaded by @rkennke in #838
• Fix recurring nightly sanitized failures (UBSan null-pc, fuzz harness compile, jmethodID churn test flake) by @jbachorik in #833
• Fix aarch64 stub unwinding in I2C/C2I adapters; fix StubUnwindCpuTest slowness on debug builds by @rkennke in #842
• Never reclaim dictionary storage after a timed-out drain by @rkennke in #839
• Fix duplicate and biased thread picks in the wall-clock reservoir sampler by @kaahos in #844
• Use a small table in PutWithExistingIdNoInfiniteLoopWhenFull by @rkennke in #847
• Close guard-scan and reused-buffer gaps in dictionary reclamation by @rkennke in #845
• fix: avoid UB left shift of negative ADRP immediate in decodeAdrTarget by @kaahos in #848
• Never free a call-trace table a put() may still hold after a drain timeout by @rkennke in #840
• Fix flaky ASan itable-stub tests and musl apk setup by @jbachorik in #849
• Liveness epoch/urgency fixes, deferred chain invalidation, and debug JNI test seams by @jbachorik in #798
• Keep one id per key across StringDictionary::rotate() by @rkennke in #857
• Fix live trace preservation deduplication by trace ID (PROF-16138) by @rkennke in #860
• Test preserved traces across several dumps (PROF-16144) by @rkennke in #861
• Fix native socket hook timestamp sequencing (PROF-16151) by @rkennke in #862
• Sort PerfEvents thread ids with a valid qsort comparator by @kaahos in #863
• Stop build-and-summarize from leaking a tail process per run by @rkennke in #865
• Let container tests clone the repo under rootful Docker by @rkennke in #869
• Reference chains: recycled leak-tag identity and Java integration tests by @jbachorik in #800
• Reference-chain test-stack updates, seam fixes, and repro/sweep tooling by @jbachorik in #801
• Reference-chains design and architecture docs by @jbachorik in #802
• Add fjmethodid=lazy: profile HotSpot without jmethodIDs by @rkennke in #868
• Fix broken async-profiler 2.8.3 release link in README by @jbachorik in #881
• Remove unreachable dead code in t07_body musl variant (issue #592) by @jbachorik in #882
• Require trace IDs to increase in creation order during calibration by @rkennke in #885
• Make TEST_LOG async-signal-safe by @rkennke in #883
• ci: upload test reports when a watchdog-dumped job times out by @rkennke in #880
• ci: list a run's jobs 50 per page in the test summary by @rkennke in #886
• Fix continuation metadata lookup and frame alignment by @kaahos in #870
• Route GitLab CI Java dependency downloads through Depot by @zihaoyu in #866
• Quarantine flaky reference-chain slow e2e tests (PROF-16305) by @jbachorik in #888
• Fix Kotlin DSL plugin resolution with restored Gradle caches by @zihaoyu in #891
• Restructure AGENTS.md into a thin index and correct the doc set against the code by @jbachorik in #892
• Revert "Fix Kotlin DSL plugin resolution with restored Gradle caches" by @zihaoyu in #894
• Revert "Route GitLab CI Java dependency downloads through Depot" by @zihaoyu in #896

New Contributors

• @zihaoyu made their first contribution in #866

Full Changelog: v_1.51.0...v_1.52.0

What's Changed

• fix: omit source_sha in retry mode by @jbachorik in #764
• Fix release.sh script by @zhengyu123 in #776
• fix(nativemem): refresh NM_NATIVE_SYMBOLS where the tables change by @rkennke in #753
• feat(nativemem): emit post-serialization counters so flush cost is visible by @rkennke in #754
• refactor(profiling): Signal handler boilerplate by @yaronguro-datadog in #756
• Possible out-of-bounds Context read by @zhengyu123 in #774
• refactor(sampling): replace std::random_device with the existing xorshift64 by @rkennke in #794
• ci: move the shipped binaries to AlmaLinux 8, get the test images off EOL Debian by @rkennke in #790
• Avoid UBSan wrap in ElfSection vaddr arithmetic by @jbachorik in #795
• Model crashNow's intentional crash as a trap under scan-build by @jbachorik in #804
• ci: count integration configurations with no result as failures by @rkennke in #792
• ci: quarantine flaky tests against a ticket instead of retrying past them by @rkennke in #777
• JFR event types and emission for reference chains by @jbachorik in #796
• Fix: PROF-15934 - native stack walker return-address attribution by @rkennke in #786
• Restore ucontext on faulted paths by @zhengyu123 in #781
• Add performance counters/metrics for samplers by @zhengyu123 in #811
• Fix SanityCheckTest OOM on JDK 8 and a false crash in flake_report by @rkennke in #815
• ci: restore EL7 stresstest coverage, add nightly EL7 functional tests by @rkennke in #805
• ci: wire the EL7 functional job into #777's quarantine machinery by @rkennke in #810
• Purge unsupported platforms by @zhengyu123 in #818
• fix(test): stop -O3 tail-calling away the return-address trampolines by @rkennke in #819
• ci: run the C++ unit tests in the release configuration by @rkennke in #821
• fix: address argument parsing silent drops by @kaahos in #817
• Possible UAF of otel_thread_ctx_v1 by @zhengyu123 in #820

New Contributors

• @yaronguro-datadog made their first contribution in #756

Full Changelog: v_1.50.0...v_1.51.0

What's Changed

• Remove dead Lookup::_class_cache by @rkennke in #723
• Remove duplicated SpinLock Exclusive/Shared Locker by @zhengyu123 in #731
• Add sanity checks for profiler by @jbachorik in #48
• Return JMETHODID_NOT_WALKABLE instead of nullptr for unpopulated jmethodID slots by @jbachorik in #733
• feat(nativemem): instrument remaining allocation sites by @rkennke in #724
• Make ProfiledThread's UnwindFailures field debug only by @zhengyu123 in #734
• Fix entry resolution for main(String[]) method by @zhengyu123 in #740
• Improve detecting subclass of Thread by @zhengyu123 in #741
• Restore priming and fix races on acquire/release ProfiledThread by @zhengyu123 in #713
• Serialize stale jmethodID frames as instead of jvmtiError by @jbachorik in #744
• Fix CollectedHeapWrapper stale layout on JDK 25+ by @jbachorik in #746
• Protect HotspotSupport::resolve() with longjmp by @zhengyu123 in #743
• Adopt JmpCtxScope for setting stack walk siglongjmp protection by @zhengyu123 in #758
• Protect Lookup::resolveMethod() with siglongjmp by @zhengyu123 in #752
• fix(nativemem): count NM_CALLTRACE residency, not reserved chunk capacity by @rkennke in #757

Full Changelog: v_1.49.0...v_1.50.0

What's Changed

• [Automated] Bump dev version to 1.49.0 by @dd-octo-sts[bot] in #690
• Missing libgcc_s.so.1 on the host should not result in crash by @zhengyu123 in #689
• fix(reliability): stop the reliability CI jobs from failing on every run by @jbachorik in #697
• Avoid retaining invalid VM method pointers by @jbachorik in #699
• Remove legacy DirectByteBuffer-based OTEP context storage (Phase 3) by @rkennke in #696
• chore(release): fix release bump auto-merge by @kaahos in #698
• fix: restore trivial approval action pins by @kaahos in #710
• [fault-injection] Profiler::recordSample() stack walk is unprotected by @zhengyu123 in #687
• Carve slow/e2e tests into separate testSlow tasks by @jbachorik in #700
• fix(ci): bump IBM Java 8 JRE URL to 8.0.8.70 by @jbachorik in #717
• fix: skip jobs needing build/deploy-artifact on release/* branches by @jbachorik in #716
• Add disabled regression test for unprimed native threads by @jbachorik in #712
• fix(ci): stop JAVA_TOOL_OPTIONS from silently killing integration tests by @jbachorik in #718
• Fix memory leak in JfrMetadata::reset() by @jbachorik in #694
• Guard Recording::writeElement against SIGSEGV on a corrupted metadata tree by @jbachorik in #692
• Replace JMC's JFR loader with a lightweight, streaming jafar-backed API by @jbachorik in #719
• fix: don't disable profiler on Alpine/musl over missing libgcc_s.so.1 by @jbachorik in #720
• Never patch the profiler's own import table by @hisener in #721
• Add prepare-patch.sh to backport pending PRs before a patch release by @jbachorik in #722

New Contributors

• @hisener made their first contribution in #721

Full Changelog: v_1.48.0...v_1.49.0

What's Changed

• [Automated] Bump dev version to 1.48.2 by @kaahos in #715

Full Changelog: v_1.48.1...v_1.48.2

What's Changed

• chore(release): backport release automation to 1.48 by @kaahos in #709
• fix: backport libgcc and invalid method pointer fixes to 1.48 by @kaahos in #711

Full Changelog: v_1.48.0...v_1.48.1

What's Changed

• [Automated] Bump dev version to 1.48.0 by @dd-octo-sts[bot] in #684
• fix(codecache): make memoryUsage() accurate and live by @rkennke in #677
• chore: update token for version bump auto-merge by @kaahos in #685
• feat(profiler): native malloc/socket stack capture with profiler-hook frame skipping by @jbachorik in #648
• fix(codecache): report the real runtime-stubs cache size by @rkennke in #681
• feat(nativemem): categorized native-memory accounting — first cut by @rkennke in #669
• [fault-injection] Startup should setup signal handlers before initializing VMStructs and protect VMFlag::name() call by @zhengyu123 in #674
• Use sigsetjmp/siglongjmp instead of setjmp/longjmp by @zhengyu123 in #686

Full Changelog: v_1.47.0...v_1.48.0

What's Changed

• Use ThreadLocal to setup protection longjmp buffer and fix jmp_buf chaining by @zhengyu123 in #615
• ci: fix codeql-action version mismatch between init and analyze by @rkennke in #638
• fix(build): link vmStructs inline definitions in assertion-enabled builds by @rkennke in #641
• ci: raise ASan test JVM heap to fix flaky nightly OOMs by @jbachorik in #640
• build-logic: link sanitizer runtime for the fuzzer shared lib by @rkennke in #646
• feat(context): all-native OTEL context write API — Phase 1 (PROF-15271) by @rkennke in #631
• test: coarsen ASan sampling rate/workload to stop nightly heap OOMs by @jbachorik in #653
• test: add fuzzer for thread local by @zhengyu123 in #645
• feat: add matrix tests to containers by @kaahos in #637
• Bound class-map growth to fix reliability-chaos OOMs by @jbachorik in #654
• (fix): gate FrameType::isRawPointer on HotSpot by @jbachorik in #658
• Profiler's fuzz tests store corpus files under wrong directories by @zhengyu123 in #656
• fix(profiler): guard line-number-table copy against stale jmethodID by @jbachorik in #647
• Optimize SafeAccess:safeCopy() implementation in assembly by @zhengyu123 in #660
• fix: prevent duplicate trace_id after CallTraceHashTable expansion by @jbachorik in #659
• Gate FrameType::isRawPointer on ENCODED_MASK and HotSpot by @jbachorik in #667
• chaos: vthread-context-cascade antagonist + glibc heap-corruption defaults by @jbachorik in #642
• Share object across fuzz targets & increase build timeout by @edznux-dd in #662
• Adopt ThreadLocal to store ProfiledThread by @zhengyu123 in #639
• Add compile-time fault-injection infrastructure for stack-walker recovery paths by @zhengyu123 in #661
• Fix native OOM from unbounded CallTraceHashTable expansion by @jbachorik in #670
• chore(ci): pin benchmarking base image to :newest tag by @faydef in #679
• chore(ci): fix ci images tags by @kaahos in #682
• chore: address release script by @kaahos in #683

New Contributors

• @faydef made their first contribution in #679

Full Changelog: v_1.46.0...v_1.47.0

What's Changed

Fixes

• An alternative TLS implementation to fix a crash on musl/aarch64 by @zhengyu123 in #600
• Don't preload jmethodIDs when cstack=vm with hotspot JVM by @zhengyu123 in #549
• fix(context): scope OTEL ThreadContext storage to the carrier thread (PROF-15271) by @rkennke in #625
• [PROF-15201] fix(profiler): close two TOCTOU races between SIGPROF handler and JFR lifecycle by @r1viollet in #614

Features

• feat(profiler): wall-clock signal suppression for idle threads by @kaahos in #560

Tests

• test: thread lifecycle × profiler-state coverage by @jbachorik in #624
• fix(test): reduce ASAN wall-clock rate in BoundMethodHandleProfilerTest to avoid OOM by @rkennke in #630
• fix(test): forkEvery for asan Test task to bound cumulative heap growth by @rkennke in #632

Full Changelog: v_1.45.0...v_1.46.0

What's Changed

• Fix duplicated cookies by @zhengyu123 in #564
• fix(wallclock): add signal-origin guard to WallClockJvmti by @jbachorik in #569
• fix: prevent overflow and misaligned read in build-id parsing by @edznux-dd in #567
• fix: bound DWARF .eh_frame_hdr parsing to the section by @edznux-dd in #571
• fix: bounds-check ELF section and symbol-table parsing by @edznux-dd in #568
• fix(spinlock): bound OptionalSharedLockGuard spin in signal handler paths by @jbachorik in #572
• Native malloc fixes and cleanup by @jbachorik in #550
• feat(chaos): production-pattern antagonists + Java-25 thread-churn crash simulation by @jbachorik in #558
• Refactor argument matching with case insensitive string comparison by @zhengyu123 in #580
• fix(calltrace): fix three SIGSEGV-inducing defects in clearTableOnly/collect (PROF-14889) by @jbachorik in #585
• feat(fuzz): add GitLab fuzzing infra setup by @edznux-dd in #587
• feat: native socket I/O tracking via PLT hooks (PROF-10637) by @jbachorik in #488
• fix: switch container runtime to podman by @kaahos in #588
• fix(tsan): suppress false-positive race from VA reuse in LinearAllocator by @jbachorik in #593
• feat(otep) Publish thread context attribute keys at process start by @ivoanjo in #576
• Fix crash at thread exit by @zhengyu123 in #591
• feat(sframe): complete SFrame V2 parser — replace stub with full implementation by @jbachorik in #471
• fix(profiler): name native threads periodically (PROF-15139) by @jbachorik in #601
• fix(jfr): prevent duplicate method-pool ids (PROF-15130) by @jbachorik in #602
• fix(chaos): rate-limit tight-loop antagonists to prevent long-run OOM by @jbachorik in #604
• feat(otep): add batch reapply-by-id-and-bytes context API by @jbachorik in #598
• Potentially infinite loop and store trace in unowned slot in CallTraceHashTable::putWithExistingId() by @zhengyu123 in #578
• fix(jfr): skip emitting empty constant pools by @jbachorik in #603

New Contributors

• @edznux-dd made their first contribution in #567
• @kaahos made their first contribution in #588

Full Changelog: v_1.44.0...v_1.45.0