test(core): stabilize spawned-process IPC descriptor counts

[rwgk]

Description

When test_process_argument_releases_with_the_process_object is the first test to spawn a process, multiprocessing opens a persistent resource-tracker pipe after the file-descriptor baseline. The extra descriptor can produce an order-dependent false leak failure.

Run a no-op child before taking the baseline, keeping the strict descriptor-count assertion unchanged. Use the buffer context manager so an assertion failure also releases its virtual-memory reservation instead of causing subsequent cleanup failures. Follow-up to #3028.

Validation

• A CPU-only spawn reproduction confirms that the first child adds one persistent descriptor and a second child leaves the count unchanged (4 -> 5 -> 5).
• Full pre-commit run --all-files passes locally.
• GPU test execution is delegated to the requested CI run.

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes (test-only fix).

[copy-pr-bot]

Auto-sync is disabled for draft pull requests in this repository. Workflows must be run manually.

Contributors can view more details about this message here.

[rwgk]

/ok to test da0ab5b

[github-actions]

Doc Preview CI
🚀 View preview at https://nvidia.github.io/cuda-python/pr-preview/pr-3146/
https://nvidia.github.io/cuda-python/pr-preview/pr-3146/cuda-core/
https://nvidia.github.io/cuda-python/pr-preview/pr-3146/cuda-bindings/
https://nvidia.github.io/cuda-python/pr-preview/pr-3146/cuda-pathfinder/
Preview will be ready when the GitHub Pages deployment is complete.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

• Configuration used: Repository: NVIDIA/cuda-python/.coderabbit.yaml
• Review profile: CHILL
• Plan: Enterprise
• Run ID: 34655589-3570-4cc2-bf3e-dee6fe9a3fb7

📥 Commits

Reviewing files that changed from the base of the PR and between 576ef68 and da0ab5b.

📒 Files selected for processing: 1

• cuda_core/tests/memory_ipc/test_vmm_ipc.py

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.

📝 Summary

Summary by CodeRabbit

• Tests
  • Updated process-lifecycle checks for file descriptor cleanup, including a more consistent baseline and automatic buffer closure. Assertions still verify that process exit releases the expected resources.

The IPC file descriptor lifetime test starts a no-op child before measuring open file descriptors. It manages the buffer with a context manager and retains the process cleanup, exit-code, and final descriptor-count checks.

Changes: IPC buffer lifetime test

Layer / File(s)	Summary
Descriptor lifetime test cuda_core/tests/memory_ipc/test_vmm_ipc.py	The test starts a no-op child before taking the descriptor baseline. A context manager closes the buffer, while the process cleanup, exit-code, and final descriptor-count checks remain.

Suggested reviewers: andy-jost

Priority: ⬇️ Low

Change: Other

Merge Risk: ⚪ Minimal · up to da0ab

The test stabilizes its descriptor baseline and checks that process-owned IPC resources are released. No material merge risk is evident.

- [ ] Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

[brandon-b-miller]

LGTM. One nit: There's a function that's used for something similar over in test_leaks.py.

https://github.com/NVIDIA/cuda-python/blob/main/cuda_core/tests/memory_ipc/test_leaks.py#L158-L169

Since we have a growing number of tests that hit this safety issue with multiprocessing it might be time to centralize the logic and put a note in the agent skill to reuse it.

[rwgk]

LGTM. One nit: There's a function that's used for something similar over in test_leaks.py.

https://github.com/NVIDIA/cuda-python/blob/main/cuda_core/tests/memory_ipc/test_leaks.py#L158-L169

Since we have a growing number of tests that hit this safety issue with multiprocessing it might be time to centralize the logic and put a note in the agent skill to reuse it.

Thanks @brandon-b-miller, I'll merge this PR and send a follow-on cleanup. (This change is on ctk-next already, out of necessity; therefore merging + follow-on is easier; no merge conflicts.)