gh-149202: Implement PEP 831 – Frame Pointers Everywhere: Enabling System-Level Observability for Python

[pablogsal]

• Issue: Implement PEP 831 – Frame Pointers Everywhere: Enabling System-Level Observability for Python #149202

Closes: #96174#149202

[pablogsal]

CC: @savannahostrowski@Fidget-Spinner@diegorusso

[read-the-docs-community]

Documentation build overview

📚 cpython-previews | 🛠️ Build #32502418 | 📁 Comparing af6c75e against main (bdbb55c)

🔍 Preview build

21 files changed · ± 21 modified

±Modified

• contents.html
• download.html
• genindex-B.html
• genindex-C.html
• genindex-E.html
• genindex-H.html
• genindex-I.html
• genindex-L.html
• genindex-O.html
• genindex-P.html
• and 11 more...

[savannahostrowski]

I think this also closes out on #96174 👀

[pablogsal]

@savannahostrowski We need a what's new entry as well i can add one tomorrow if you don't beat me to it. Also we need to tell people to add the option in their custom build systems as well

[Fidget-Spinner]

I read the configure.ac code in detail, and looked at the docs as well.

[corona10]

lgtm!

[emmatyping]

One minor configure wording question, but otherwise looks good!

[savannahostrowski]

Thanks for the reviews, folks!

@pablogsal I've added a What's New and guidance for downstream folks too!

[hugovk]

Updating main to include #149248 to fix CI.

[hugovk]

test_manual_unwind_respects_frame_pointers is failing on the S390x buildbots:

https://buildbot.python.org/#/release_status

test_manual_unwind_respects_frame_pointers (test.test_frame_pointer_unwind.FramePointerUnwindTests.test_manual_unwind_respects_frame_pointers) ... test_manual_unwind_respects_frame_pointers (test.test_frame_pointer_unwind.FramePointerUnwindTests.test_manual_unwind_respects_frame_pointers) (env={'PYTHON_JIT': '0'}) ... FAIL
Stdout:
#00 0x3ff7fa759d0 -> other
{"length": 1, "python_frames": 0, "jit_frames": 0, "other_frames": 1, "jit_backend": null}
======================================================================
FAIL: test_manual_unwind_respects_frame_pointers (test.test_frame_pointer_unwind.FramePointerUnwindTests.test_manual_unwind_respects_frame_pointers) (env={'PYTHON_JIT': '0'})
----------------------------------------------------------------------
Traceback (most recent call last):
File "/var/lib/buildbot/worker/cstratak-fedora-stable-s390x/3.x.cstratak-fedora-stable-s390x/build/Lib/test/test_frame_pointer_unwind.py", line 205, in test_manual_unwind_respects_frame_pointersself.assertGreater(
~~~~~~~~~~~~~~~~~~^
python_frames,
^^^^^^^^^^^^^^0,
^^f"expected to find Python frames on {self.machine} with env {env}",
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^AssertionError: 0 not greater than 0 : expected to find Python frames on s390x with env {'PYTHON_JIT': '0'}Stdout:
#00 0x3ff7fa759d0 -> other
{"length": 1, "python_frames": 0, "jit_frames": 0, "other_frames": 1, "jit_backend": null}

[encukou]

It's also failing on ARM Raspbian (on aarch64 with env {'PYTHON_JIT': '0'}).

[diegorusso]

I've seen the failures. Today I cannot see what's going on. It's bank holiday in the UK. I think I can replicate a Raspbian OS. In case I need is there a way to login on the S390x?
What I notice though is that the classifications of the frames is broken on these platforms.

Stdout:
#00 0x3ff7fa759d0 -> other
{"length": 1, "python_frames": 0, "jit_frames": 0, "other_frames": 1, "jit_backend": null}

[encukou]

Are these the platforms without -mno-omit-leaf-frame-pointer?

checking whether C compiler accepts -mno-omit-leaf-frame-pointer... no

In case I need is there a way to login on the S390x?

That would be a question for @vstinner

I was able to reproduce in a container with QEMU (podman run --rm -it --arch s390x -v $PWD:/workspace:O fedora bash; install deps), but, it's late so I'll stop here today.

[pablogsal]

Pls check #149362 for an explanation of the problem and a fix

[pablogsal]

But long story short: a mix of the test is wrong and we are missing flags.