Repository navigation
Conversation
Scan-Build Report
Bug Summary
Filter Results by FileReports
|
|||||||||||||||||||||||||||||||||||||||||||||||||
CI Test ResultsRun: #45069475120 | Commit:
Status Overview
Legend: ✅ passed | ❌ failed | ⚪ skipped | 🚫 cancelled Summary: Total: 32 | Passed: 32 | Failed: 0 Updated: 2026-11-22 21:24:56 UTC |
❌ ErrorsYour PR has failed checks. Please review the issues below and take necessary action before merging. 🚦 4 Pipeline jobs failed
Useful? React with 👍 / 👎 This comment will be updated automatically if new data arrives.🔗 Commit SHA: 95ac270 | Docs | View more details | Give us feedback! |
|
@DataDog review |
There was a problem hiding this comment.
The new sanitizer guards fail preprocessing under GCC 13, blocking native-test compilation. Both affected files require the same portability fix.
🤖 Bits Code Review · Commit 44ca954
44ca954 to
cca182f
Compare
d45495f to
59b62d4
Compare
|
🐶 babysit-pr is watching this PR — CI breakages and mechanical findings are fixed and pushed automatically; design decisions are always brought to you. This run
👀 watching · since 20:28 UTC · CI green on Commits pushed this run: Findings answered (one reply each, never resolved): stale Bits P2 on mallocAdmission_ut.cpp (4603337880), Bits P2 SysV chain cap (4640551607), Bits P1 GCC 4.8 stand-in (4641105194). Left for the author/humans: Bits re-triggers posted for heads |
59b62d4 to
c321077
Compare
|
@DataDog review |
There was a problem hiding this comment.
A valid SysV hash chain containing exactly 65,536 entries is treated as exhausted, so native-memory profiling can reject startup for an allocator image without ddprof hooks.
🤖 Bits Code Review · Commit c321077
|
@DataDog review |
There was a problem hiding this comment.
EL7 functional runs compile the new JNI helper with GCC 4.8.5, which lacks its overflow builtins. Affected native test invocations will terminate with unresolved-symbol errors.
🤖 Bits Code Review · Commit 531fcb3
|
@DataDog review |
There was a problem hiding this comment.
More details
Runtime compatibility with released ddprof binaries and Linux architecture/libc variants remains unverified; confidence is limited to static inspection.
🤖 Bits Code Review · Commit 9636bde
|
❌ 38 passed, 0 failed, 2 without a result out of 40 expected configurations Test Matrix
No result producedThese configurations wrote no validation log, so their outcome is unknown and
Links
|
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 9636bde76e
ℹ️ About Codex in GitHub
Your team has set up Codex to 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 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
More details
The previously reported hash-chain boundary and libc-banner issues are already addressed at this head. Compatibility with released ddprof binaries remains unverified because this review was static.
🤖 Bits Code Review · Commit 9636bde
|
@DataDog review |
There was a problem hiding this comment.
Installing the test stand-in after a profiling session exposes allocator hooks before their forwarding pointers are initialized. Persistent JVM callbacks can enter that window and crash the child JVM.
🤖 Bits Code Review · Commit ef100a0
The rewritten LoadedImage::readable() only walks PT_LOAD segments with PF_R set, as a loader reports them; the misaligned dynamic-section fixtures left p_flags zeroed, so imageFingerprint() rejected their import layout and returned 0. Model the real mapping with PF_R|PF_X.
lookupSysv() read the terminator word inside the loop but only tested index at the top: a valid chain of exactly MAX_HASH_CHAIN_STEPS (65536) entries left the loop holding the terminator 0 and was reported CHAIN_LIMIT, which refuses native-memory profiling startup for the image. Test the final read before reporting exhaustion; a chain that still has an entry at the cap is unchanged, and the nchain cycle bound still bounds loops. Boundary test mirrors GnuChainLimitBoundary: 65536 entries classify DDPROF, 65537 report CHAIN_LIMIT.
6857f7d to
316e037
Compare
There was a problem hiding this comment.
More details
Static review identified no actionable defect in the specified range. Compatibility with released ddprof binaries and loader lifetime behavior remain unverified by runtime validation on this head.
🤖 Bits Code Review · Commit 316e037
|
@DataDog review |
There was a problem hiding this comment.
More details
The SysV chain-boundary defect described in the PR body is already fixed in the reviewed head. Compatibility with released ddprof binaries and shared-library lifetime behavior remains unverified by execution.
🤖 Bits Code Review · Commit 93c951e
|
@DataDog review |
There was a problem hiding this comment.
The PR rejects ddprof-owned allocator targets at native-memory startup, and this update initializes stand-in forwarders before publishing hooks. One issue remains: the new workload lacks a readiness handshake, allowing a scheduling-dependent lifecycle-test failure; released-ddprof interoperability and loader lifetime behavior still lack runtime validation.
🤖 Bits Code Review · Commit 327e118
| private static void startSpinThread() { | ||
| if (spinThread != null) { | ||
| throw new IllegalStateException("spin thread already running"); | ||
| } | ||
| spinRunning = true; | ||
| spinThread = new Thread(ExternalLauncher::runSpinWorkload, "admission-spin"); | ||
| spinThread.start(); | ||
| } |
There was a problem hiding this comment.
Wait for the spin workload before continuing the scenario
If admission-spin is delayed until after INSTALL/rejection/RESTORE, SPINSTOP clears spinRunning before the workload enters its loop. spinRounds remains zero, so the new lifecycle assertion fails despite correct hook installation. Thread.start() does not guarantee readiness; wait for workload progress before continuing the scenario.
| private static void startSpinThread() { | |
| if (spinThread != null) { | |
| throw new IllegalStateException("spin thread already running"); | |
| } | |
| spinRunning = true; | |
| spinThread = new Thread(ExternalLauncher::runSpinWorkload, "admission-spin"); | |
| spinThread.start(); | |
| } | |
| private static void startSpinThread() throws InterruptedException { | |
| if (spinThread != null) { | |
| throw new IllegalStateException("spin thread already running"); | |
| } | |
| spinRounds = 0; | |
| spinRunning = true; | |
| spinThread = new Thread(ExternalLauncher::runSpinWorkload, "admission-spin"); | |
| spinThread.start(); | |
| long deadline = System.nanoTime() + java.util.concurrent.TimeUnit.SECONDS.toNanos(5); | |
| while (spinRounds == 0) { | |
| if (!spinThread.isAlive() || System.nanoTime() - deadline >= 0) { | |
| spinRunning = false; | |
| throw new IllegalStateException("spin workload did not start within 5s"); | |
| } | |
| Thread.sleep(1); | |
| } | |
| } |
React: 👍 useful · 👎 wrong · 😕 unclear · 👀 right, but not acting on it
What each reaction means
- 👍 Useful: correct, and you're fixing it or will.
- 👎 Wrong: the finding is incorrect, for example the bug can't happen or the code was misread.
- 😕 Unclear: you can't tell what it means or why it matters.
- 👀 Right, not acting: correct, but intended, out of scope for this PR, a known risk, or too minor.
🤖 Bits Code Review · Open Bits AI session
There was a problem hiding this comment.
Fixed by 18add5e + 95ac270 (rebased on the main merge): startSpinThread() now waits for the first completed round (5s deadline, 1ms poll) before INSTALL runs, so spinRounds can no longer stay 0 when SPINSTOP races the workload start; every failed start path (timeout, dead thread, interrupted wait) now clears spinRunning and joins the thread, and a scenario abort after SPINSTART stops the workload in a finally instead of leaving a non-daemon thread hanging the child JVM.
There was a problem hiding this comment.
More details
The admission checks account for both current allocator imports and saved forwarding targets across restarts. The completed static review found no reportable defect; runtime compatibility with released ddprof binaries was not established.
🤖 Bits Code Review · Commit 327e118
…th-the-native-ixsr
There was a problem hiding this comment.
The PR rejects ddprof-owned allocator targets at startup and fixes stand-in hook publication ordering. Two test defects remain: the previously reported spin-workload readiness race and a non-atomic loader-thread stop flag; compatibility with released ddprof binaries and loader lifetime behavior still lack runtime validation.
🤖 Bits Code Review · Commit 242daa7
| if (!e) _exit(CHILD_START_UNEXPECTED); | ||
| if (strcmp(e.message(), DDPROF_LITERAL) != 0) _exit(CHILD_FAIL); | ||
| } | ||
| g_loader_loop_stop = 1; |
There was a problem hiding this comment.
Make the loader-thread stop flag atomic
When the concurrent loader scenario runs, loaderLoop reads g_loader_loop_stop while the scenario thread writes it before joining. volatile sig_atomic_t provides no synchronization between pthreads, so the test itself has a C++ data race and can trigger TSan reports in configurations that execute it. Include and use std::atomic, or use atomic builtins for both accesses.
React: 👍 useful · 👎 wrong · 😕 unclear · 👀 right, but not acting on it
What each reaction means
- 👍 Useful: correct, and you're fixing it or will.
- 👎 Wrong: the finding is incorrect, for example the bug can't happen or the code was misread.
- 😕 Unclear: you can't tell what it means or why it matters.
- 👀 Right, not acting: correct, but intended, out of scope for this PR, a known risk, or too minor.
🤖 Bits Code Review · Open Bits AI session
Thread.start() does not guarantee the admission-spin workload entered its loop before SPINSTOP runs; wait for the first completed round (5s deadline) so the rounds>0 lifecycle assertion cannot race the thread scheduler.
Clean up the spin-thread state on every startSpinThread failure path (timeout, dead thread, interrupted wait) and stop a running workload in a scenario-abort finally, so an aborted child JVM cannot hang waiting for the non-daemon admission-spin thread.
|
@DataDog review |
There was a problem hiding this comment.
More details
The completed static review found no reportable defect. Compatibility with released ddprof binaries and Linux loader lifetime behavior still requires runtime acceptance.
🤖 Bits Code Review · Commit 95ac270
What does this PR do?:
Rejects Java native-memory profiling startup when the allocator's current or retained forwarding targets resolve to ddprof hooks. It adds ELF-image admission checks and startup tests for conflicting profilers.
Motivation:
Prevent simultaneous allocation profilers from routing samples through ddprof-owned hooks, which can corrupt attribution or count the same allocation through multiple profilers.
Additional Notes:
This PR remains a draft because the Linux acceptance tests have not run on the pushed head. Human review is mandatory for the shared-library identity and lifetime behavior.
The committed-range review found no Critical or High findings. It found one Medium defect: a valid SysV hash chain with exactly
MAX_HASH_CHAIN_STEPSentries is rejected. It also identified two validation gaps: integration tests use the stand-in rather than released ddprof binaries, and a GNU libc banner without the literalglibccan omit the IFUNC fixture and make its test skip. These remain open on the pushed head.How to test the change?:
qualityChecks: passed before the rebase at44ca9543(build/logs/20261117-164019-qualityChecks.log); not rerun on the pushed head.buildReleaseand:ddprof-test:testClasses: passed at pre-rebase candidate9d054557; not verified on the pushed head.:ddprof-lib:gtestDebug: passed locally before the rebase (build/logs/20261118-155739-_ddprof-lib_gtestDebug.log) on a worktree that also contained uncommitted review fixes; not evidence for the pushed head.For Datadog employees:
Unsure? Have a question? Request a review!