test(profiling): unwind_greenlets RSS test on main without the fix [DO NOT MERGE] by taegyunkim · Pull Request #18422 · DataDog/dd-trace-py

taegyunkim · 2026-06-02T20:33:29Z

Purpose — DO NOT MERGE

This is an experiment to validate the regression test added in the buffer-reuse work (branch taegyunkim/prof-14423-unwind-greenlets-buffer-reuse).

It applies only the new test test_gevent_unwind_greenlets_rss_stable on top of main, without the C++ unwind_greenlets buffer-reuse fix.

Question being answered

A code review flagged that the test asserts on RSS growth (< 20 MB over 10s), but the regression it guards is per-sample allocation churn / heap-live-size. A recycling allocator can absorb that churn without net RSS growth, so the test might pass even on unfixed code — a false-negative.

✅ Test FAILS here → it's a genuine regression guard.
❌ Test PASSES here → it's a false-negative and needs strengthening (e.g. assert on allocation count / heap-live-size, or a baseline-vs-fixed comparison).

Scope

Single file changed: tests/profiling/collector/test_stack.py (+87 lines, the new test only).
No native/profiling source changes (verified identical to main).
The test is gated behind DD_PROFILE_TEST_GEVENT, so it runs only in the gevent profiling CI job.

🤖 Generated with Claude Code

… fix) Experimental branch: contains ONLY the new gevent unwind_greenlets RSS stability test from PR for taegyunkim/prof-14423-unwind-greenlets-buffer-reuse, applied on top of main WITHOUT the C++ buffer-reuse fix. Purpose: verify whether the test actually fails on unfixed code, i.e. whether it is a genuine regression guard or a false-negative (a code review flagged that an RSS-delta assertion may pass even when the per-sample allocation churn regression is present). Not for merge. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

cit-pr-commenter-54b7da · 2026-06-02T20:34:47Z

Codeowners resolved as

tests/profiling/collector/test_greenlet_buffer_reuse.py                 @DataDog/profiling-python

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: caf0b6218f

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".

chatgpt-codex-connector · 2026-06-02T20:35:48Z

+        p.stop()
+
+    peak_rss = max(samples)
+    growth_bytes = peak_rss - rss_after_warmup


Measure allocation churn instead of post-warmup RSS

Because the baseline is taken only after the 3s warmup, this test can pass on the unfixed path whenever the allocator has already grown/cached enough memory during warmup to satisfy the repeated StackInfo/deque allocations during the 10s measurement window. In that scenario the per-sample churn still exists, but peak_rss - rss_after_warmup stays under 20 MB, so the regression guard produces the false negative the test is meant to rule out; assert on allocation/heap-live-size churn or compare fixed-vs-baseline behavior instead of post-warmup RSS growth.

Useful? React with 👍 / 👎.

datadog-datadog-prod-us1 · 2026-06-02T20:36:32Z

✨ Fix all issues with BitsAI

⚠️ Warnings

🚦 14 Pipeline jobs failed

DataDog/apm-reliability/dd-trace-py | profiling/profile 1/24

See error
AttributeError: module 'ddtrace.internal.datadog.profiling.stack._stack' has no attribute '_greenlet_buffer_alloc_count' in test_greenlet_unwind_buffer_reuse.
🧪 1 Test failed
test_greenlet_unwind_buffer_reuse from test_greenlet_buffer_reuse.py (Fix with Cursor)
Expected status 0, got 1.
=== Captured STDOUT ===
=== End of captured STDOUT ===
=== Captured STDERR ===
Traceback (most recent call last):
  File &#34;tests/profiling/collector/test_greenlet_buffer_reuse.py&#34;, line 76, in &lt;module&gt;
    c1 = stack._stack._greenlet_buffer_alloc_count()
AttributeError: module &#39;ddtrace.internal.datadog.profiling.stack._stack&#39; has no attribute &#39;_greenlet_buffer_alloc_count&#39;
=== End of captured STDERR ===
DataDog/apm-reliability/dd-trace-py | profiling/profile 23/24

See error
AttributeError: module 'ddtrace.internal.datadog.profiling.stack._stack' has no attribute '_greenlet_buffer_alloc_count' in tests/profiling/collector/test_greenlet_buffer_reuse.py.
🧪 1 Test failed
test_greenlet_unwind_buffer_reuse from test_greenlet_buffer_reuse.py (Fix with Cursor)
Expected status 0, got 1.
=== Captured STDOUT ===
=== End of captured STDOUT ===
=== Captured STDERR ===
Failed to register thread: 7220ccfa4fc0 (2784) MainThread
Traceback (most recent call last):
  File &#34;tests/profiling/collector/test_greenlet_buffer_reuse.py&#34;, line 76, in &lt;module&gt;
    c1 = stack._stack._greenlet_buffer_alloc_count()
AttributeError: module &#39;ddtrace.internal.datadog.profiling.stack._stack&#39; has no attribute &#39;_greenlet_buffer_alloc_count&#39;
=== End of captured STDERR ===
DataDog/apm-reliability/dd-trace-py | profiling/profile 3/24

See error
1 failed test: AttributeError: module 'ddtrace.internal.datadog.profiling.stack._stack' has no attribute '_greenlet_buffer_alloc_count' in tests/profiling/collector/test_greenlet_buffer_reuse.py.
🧪 1 Test failed
test_greenlet_unwind_buffer_reuse from test_greenlet_buffer_reuse.py (Fix with Cursor)
Expected status 0, got 1.
=== Captured STDOUT ===
=== End of captured STDOUT ===
=== Captured STDERR ===
Traceback (most recent call last):
  File &#34;tests/profiling/collector/test_greenlet_buffer_reuse.py&#34;, line 76, in &lt;module&gt;
    c1 = stack._stack._greenlet_buffer_alloc_count()
         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
AttributeError: module &#39;ddtrace.internal.datadog.profiling.stack._stack&#39; has no attribute &#39;_greenlet_buffer_alloc_count&#39;
=== End of captured STDERR ===
View all 14 failed jobs.

🧪 1 Test failed in 1 job

DataDog/apm-reliability/dd-trace-py | profiling/profile 15/24

test_greenlet_unwind_buffer_reuse from test_greenlet_buffer_reuse.py

(Fix with Cursor)

Expected status 0, got 1.
=== Captured STDOUT ===
=== End of captured STDOUT ===
=== Captured STDERR ===
Traceback (most recent call last):
  File &#34;tests/profiling/collector/test_greenlet_buffer_reuse.py&#34;, line 76, in &lt;module&gt;
    c1 = stack._stack._greenlet_buffer_alloc_count()
AttributeError: module &#39;ddtrace.internal.datadog.profiling.stack._stack&#39; has no attribute &#39;_greenlet_buffer_alloc_count&#39;
=== End of captured STDERR ===

ℹ️ Info

No other issues found (see more)

❄️ No new flaky tests detected

🔄 Datadog retried 1 test - 0 passed on retry

Useful? React with 👍 / 👎

_{This comment will be updated automatically if new data arrives.

🔗 Commit SHA: a73dcff | Docs | Datadog PR Page | Give us feedback!}

LD_PRELOAD malloc-counting interposer; prints MALLOC_DELTA over a fixed greenlet-sampling window. Experiment to find a signal that separates the buffer-reuse fix from unfixed code (RSS does not). Always passes for now. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Same LD_PRELOAD malloc-counting experiment as #18422, but on top of the buffer-reuse C++ fix. Compare MALLOC_DELTA against the unfixed run. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Asserts stack._greenlet_buffer_alloc_count() plateaus after warmup. On main (no buffer-reuse fix) the counter symbol does not exist, so this fails -- it is the regression guard that accompanies the fix. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…odule

taegyunkim requested a review from a team as a code owner June 2, 2026 20:33

taegyunkim added the changelog/no-changelog A changelog entry is not required for this PR. label Jun 2, 2026

taegyunkim requested a review from gyuheon0h June 2, 2026 20:33

taegyunkim added the changelog/no-changelog A changelog entry is not required for this PR. label Jun 2, 2026

taegyunkim marked this pull request as draft June 2, 2026 20:33

taegyunkim removed the request for review from gyuheon0h June 2, 2026 20:33

chatgpt-codex-connector Bot reviewed Jun 2, 2026

View reviewed changes

taegyunkim mentioned this pull request Jun 2, 2026

test(profiling): malloc-churn experiment ON fixed build [DO NOT MERGE] #18426

Draft

taegyunkim force-pushed the taegyunkim/prof-14423-test-regression-check branch from 53910d7 to 41d2a4d Compare June 2, 2026 22:35

taegyunkim and others added 2 commits June 3, 2026 17:40

test(profiling): access greenlet buffer alloc counter via _stack subm…

a73dcff

…odule

taegyunkim force-pushed the taegyunkim/prof-14423-test-regression-check branch from 78b4e52 to a73dcff Compare June 3, 2026 18:30

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

test(profiling): unwind_greenlets RSS test on main without the fix [DO NOT MERGE]#18422

test(profiling): unwind_greenlets RSS test on main without the fix [DO NOT MERGE]#18422
taegyunkim wants to merge 4 commits into
mainfrom
taegyunkim/prof-14423-test-regression-check

taegyunkim commented Jun 2, 2026

Uh oh!

cit-pr-commenter-54b7da Bot commented Jun 2, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Jun 2, 2026

Uh oh!

datadog-datadog-prod-us1 Bot commented Jun 2, 2026 •

edited by datadog-prod-us1-6 Bot

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

taegyunkim commented Jun 2, 2026

Purpose — DO NOT MERGE

Question being answered

Scope

Uh oh!

cit-pr-commenter-54b7da Bot commented Jun 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codeowners resolved as

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Jun 2, 2026

Choose a reason for hiding this comment

Uh oh!

datadog-datadog-prod-us1 Bot commented Jun 2, 2026 • edited by datadog-prod-us1-6 Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

⚠️ Warnings

ℹ️ Info

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

cit-pr-commenter-54b7da Bot commented Jun 2, 2026 •

edited

Loading

datadog-datadog-prod-us1 Bot commented Jun 2, 2026 •

edited by datadog-prod-us1-6 Bot

Loading