feat(crashtracking): collect all native stacks for unhandled exception and also crashing thread - #2155
Merged
gyuheon0h merged 2 commits intoJun 24, 2026
Conversation
Contributor
Author
This stack of pull requests is managed by Graphite. Learn more about stacking. |
Contributor
🎉 All green!🧪 All tests passed 🎯 Code Coverage (details) 🔗 Commit SHA: 8ba90b9 | Docs | Datadog PR Page | Give us feedback! |
Contributor
Clippy Allow Annotation ReportComparing clippy allow annotations between branches:
Summary by Rule
Annotation Counts by File
Annotation Stats by Crate
About This ReportThis report tracks Clippy allow annotations for specific rules, showing how they've changed in this PR. Decreasing the number of these annotations generally improves code quality. |
gyuheon0h
changed the title
gyuheon0h
changed the title
Contributor
📚 Documentation Check Results📦
|
Contributor
🔒 Cargo Deny Results📦
|
gyuheon0h
marked this pull request as ready for review
June 23, 2026 15:24
gyuheon0h
force-pushed
the
gyuheon0h/PROF-15198-collect-crashing-thread-receiver
branch
from
June 23, 2026 15:25
504d740 to
be53e9d
Compare
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 504d740bd5
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
Contributor
Artifact Size Benchmark Reportaarch64-alpine-linux-musl
aarch64-unknown-linux-gnu
libdatadog-x64-windows
libdatadog-x86-windows
x86_64-alpine-linux-musl
x86_64-unknown-linux-gnu
|
gleocadie
approved these changes
Jun 24, 2026
gleocadie
left a comment
gleocadie
left a comment
Contributor
There was a problem hiding this comment.
one question, other than that LGTM
Comment on lines
-569
to
-573
| // Use the remaining receiver budget | ||
| // When this expires, stream_thread_contexts stops and returns the threads | ||
| // collected so far, which are still emitted in the report. | ||
| let context_timeout = budget; | ||
|
|
Contributor
There was a problem hiding this comment.
Not in this PR but general question: do we have test(s) to validate that crashtracker adjust the crash report construction based the budget ? like a truncated report is sent with a comment/message "..truncated...incomplete.."?
Contributor
Author
There was a problem hiding this comment.
No we do not. We need to have this though. I'll add a task to do this.
gyuheon0h
deleted the
gyuheon0h/PROF-15198-collect-crashing-thread-receiver
branch
June 24, 2026 13:53
taegyunkim
pushed a commit
that referenced
this pull request
Jul 8, 2026
…n and also crashing thread (#2155) # What does this PR do? Previously, when a runtime reported an unhandled exception, only the runtime-provided stack trace was captured; no native stacks were collected for any thread because `PR_SET_PTRACER` was not set for unhandled exceptions. Additionally, when all-thread collection was enabled for signal-based crashes, the crashing thread was excluded from the threads array. This change: 1. Enables native stack collection for all threads (including the crashing thread) when reporting unhandled exceptions with `collect_all_threads` enabled. The receiver now has the necessary ptrace permissions to unwind every thread in the process. 2. Includes the crashing thread in the `error.threads` array (marked with `crashed: true`) for both signal-based crashes and unhandled exceptions. Its native stack is collected by the receiver using ptrace, same as every other thread. The runtime-provided stack remains in `error.stack` as the canonical crash context, while `error.threads` provides a uniform native view of all threads at the time of the crash. This gives us two immediate benefits: 1. Unhandled exception crash reports now include native stacks for all threads, providing visibility into what the process was doing at the OS level; not just the managed runtime view. 2. The crashing thread's native stack is now collected in the receiver alongside all other threads, establishing a path toward consolidating all stack collection in the receiver process in the future. # Motivation What inspired you to submit this pull request? # Additional Notes Anything else we should know when reviewing? # How to test the change? Describe here in detail how the change can be validated. Signed-off-by: Taegyun Kim <taegyun.kim@datadoghq.com>
hoolioh
added a commit
that referenced
this pull request
Aug 19, 2026
# Release proposal for libdd-crashtracker, libdd-gotter and their dependencies This PR contains version bumps based on public API changes and commits since last release. ## libdd-gotter **Next version:** `1.0.0` **Semver bump:** `major` **Tag:** `libdd-gotter-v1.0.0` **Warning:** this is an initial release. Please verify that the version and commits included are correct. ## libdd-crashtracker **Next version:** `2.0.0` **Semver bump:** `major` **Tag:** `libdd-crashtracker-v2.0.0` ###⚠️ major bump forced due to: - `libdd-common`: ^1.0.0 → ^5.2.0 - `libdd-telemetry`: ^1.0.0 → ^7.0.0 ### Commits - feat(crashtracking)!: retrieve c assert message for linux when `__assert_fail` is dynamically loaded (#2268) - test: skip/shorten slow miri jobs (#2331) - feat(crashtracking): send debug log when no data is received at all (#2321) - fix(crashtracking): check fields and exclude uuid for `has_data` (#2322) - refactor: migrate to workspace dependencies, phase 4 (#2296) - refactor: migrate to workspace dependencies, phase 3 (#2283) - chore: moving to workspace-level dependencies, phase 2 (#2270) - refactor: consolidate core dependencies at workspace level (phase 1) (#2253) - test: stabilize flaky tests (#2256) - feat(telemetry)!: make telemetry worker wasm-compatible for the TraceExporter (#2172) - chore(crashtracking): harden multi thread ptrace collection (#2216) - fix(sidecar): don't double-encode file:// telemetry endpoints (#2230) - fix: new clippy lints (#2219) - refactor(sidecar)!: Avoid a dedicated socket for crashtracker (#2179) - refactor(clippy): prefer core and alloc imports (#2196) - refactor(libdd-telemetry)!: avoid leaking libdd-common types in the public API (#2152) - feat(crashtracking): collect all native stacks for unhandled exception and also crashing thread (#2155) - fix(crashtracking): sanitize type and message for unhandled exceptions (#2148) - fix(crashtracking): multi thread collection centos flakes harden (#2113) - chore(crashtracking)!: remove frame count experimental field (#2114) - fix(crashtracking): authenticate peer granted socket ptrace access (#2098) - feat(crashtracking)!: add experimental frame count field (#2099) - fix(crashtracker): increase test_waitall_nohang timeout to 500ms (#2097) - fix(crashtracker): support socket based receiver for all thread collection (#2080) - fix(crashtracker): set failed thread stack collection as incomplete empty stack (#2079) - fix(crashtracking)!: flatten all threads object into a list of `ThreadData` (#2054) - refactor(crashtracker): update imports to linux only (#2036) - fix(crashtracker): move preload logger marking after recursive guard (#2023) - fix(libdd-telemetry): restore previous Cargo.toml version (#1993) - feat!: integrate obfuscation to the stats exporter [APMSP-2764] (#1819) - chore(crashtracking)!: create errorsintake crash ping directly from telemetry (#1963) - feat(telemetry)!: include dependencies and integrations in app-extended-heartbeat (#1962) - feat(crashtracking)!: collect all threads (#1878) - chore(crashtracking): bump libdatadog-libunwind to v1.0.2 (#1942) - chore(crashtracking): default errors intake crash report upload to be on (#1902) - chore(libdd-crashtracker): remove path reference for libdd-libunwind-sys (#1877) - fix(crashtracking): handle new lines in client submitted exception message (#1836) - fix(crashtracker): Fix bin_tests in gitlab (#1832) - feat(crashtracking)!: improve parity between errors intake payload and telemetry intake payload (#1823) - feat(crashtracking)!: emit ucontext registers as structured data (#1787) - fix(ci): use single threaded for all tests that mutate signal state and use non-fatal signal (#1812) - fix(crashtracking): fix SIGCHLD signal guarding while in CT signal handler (#1807) - fix(ci): use single threaded to avoid race conditions for sa guard tests (#1800) - fix(crashtracking): guard sigchld and sigpipe during crashtracker signal handler execution (#1771) - chore(crashtracking): preserve errno for crashtracker (#1767) - chore(crashtracker): use weaker mem ordering for OP_COUNTERS (#1744) - chore(crashtracking): rename target triple to runtime platform (#1747) - chore(ci): give libdd-libunwind-sys its own version (#1743) - chore(crashtracking): add tag for target triple (#1741) - refactor(libdd-crashtracker)!: avoid leaking Endpoint through the public API (#1705) - chore(cargo): bump to 29.0.0 (#1702) - fix(crashtracking): use libunwind to unwind frames (#1663) - chore(deps): bump blazesym to 0.2.3 and blazesym-c to 0.1.7 (#1654) - chore(ci): fix crashtracker receiver binary rpath setting (#1652) - chore(crashtracking): emit a best effort stacktrace for Mac (#1645) - chore(crashtracker): use default-features=false for aws-lc-sys (#1625) - feat(crashtracking): report unhandled exceptions (#1596) - refactor(ddcommon)!: remove direct dependency on hyper client everywhere in common (#1604) - feat(crashtracking): include `Kind` in crash ping and clarify requirements (#1595) - fix(crashtracking): add process_tags to application field (#1576) - ci: update nightly in CI to 2026-02-08 (#1539) - fix(telemetry)!: fix logs payload format [APMSP-2590] (#1498) - chore(crashtracking): bump os_info crate to 3.14 (#1507) - Don't bail (#1494) - feat(crashtracking)!: emit crashing thread name in crash report for linux crashes (#1485) - test(crashtracking): add minimal LD preload test for crashtracker collector (#1428) - chore: release libddcommon-v1.1.0 (#1456) - chore: prepare libdd-telemetry-v2.0.0 (#1457) - [crashtracker] Retrieve panic message when crashing (#1361) - fix(sidecar): AWS lambda also can return EACCESS for shm_open (#1446) - chore(crashtracking): add `is_crash_debug` tag to crashtracker receiver debug logs (#1445) - [crashtracker] Log errors in crashtracker receiver (#1395) - chore: add changelog for every published crate (#1396) - Fix CI (#1389) - [crashtracker] Set OS info in the crash info builder when receiving report (#1388) - crashtracker: support cxx bindings for crashinfo (#1379) [APMSP-2764]: https://datadoghq.atlassian.net/browse/APMSP-2764?atlOrigin=eyJpIjoiNWRkNTljNzYxNjVmNDY3MDlhMDU5Y2ZhYzA5YTRkZjUiLCJwIjoiZ2l0aHViLWNvbS1KU1cifQ --------- Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com> Co-authored-by: hoolioh <107922352+hoolioh@users.noreply.github.com>
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
2 participants
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 does this PR do?
Previously, when a runtime reported an unhandled exception, only the runtime-provided stack trace was captured; no native stacks were collected for any thread because
PR_SET_PTRACERwas not set for unhandled exceptions. Additionally, when all-thread collection was enabled for signal-based crashes, the crashing thread was excluded from the threads array.This change:
collect_all_threadsenabled. The receiver now has the necessary ptrace permissions to unwind every thread in the process.error.threadsarray (marked withcrashed: true) for both signal-based crashes and unhandled exceptions. Its native stack is collected by the receiver using ptrace, same as every other thread.The runtime-provided stack remains in
error.stackas the canonical crash context, whileerror.threadsprovides a uniform native view of all threads at the time of the crash.This gives us two immediate benefits:
Motivation
What inspired you to submit this pull request?
Additional Notes
Anything else we should know when reviewing?
How to test the change?
Describe here in detail how the change can be validated.