Work around pixi lockfile oscillation in lockfile refresh as well.

[jpascucci-nv]

Description

refs pr #3059

Lockfile oscillation is blocking refresh.

Checklist

• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[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: 3accbb27-889c-48ec-8e1d-2d950c7859b5

---

📥 Commits

Reviewing files that changed from the base of the PR and between df752de and 1c14fd3.

---

📒 Files selected for processing (2)

• .github/workflows/ci-pixi-lockfile-refresh.yml
• ci/README.md

---

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

• Chores
  • The lockfile refresh process can now skip a specifically detected known issue in the cuda_core workspace and continue updating other workspaces. Unrecognized errors still cause the refresh to fail, and existing lockfile stability checks remain in place.
• Documentation
  • Updated CI guidance to describe the refresh behavior and the conditions required for the known issue to be considered resolved.

Walkthrough

The refresh workflow skips a failed cuda_core lock check only when the known samples-source pruning detector confirms the defect. It restores the committed lockfile, continues with other workspaces, verifies lockfile stability, and reports the skip.

Changes

Pixi lockfile refresh

Layer / File(s)	Summary
Detect and restore the known defect .github/workflows/ci-pixi-lockfile-refresh.yml	The workflow records the updated lockfile blob and runs the detector after a failed cuda_core check. On a confirmed defect, it restores the committed lockfile and continues. Detector errors, unconfirmed defects, and failures in other workspaces remain failures.
Verify stability and report results .github/workflows/ci-pixi-lockfile-refresh.yml, ci/README.md	The workflow checks the second lock check’s status and compares lockfile blobs to report instability. PR descriptions and job summaries report when the committed cuda_core lockfile is retained. The README documents the detector and the condition for removing it.

---

Priority: ➖ Normal

Change: Bug fix

Merge Risk: ⚪ Minimal · up to 1c14f

The refresh workaround appears ready to merge after normal checks. Only the recognized cuda_core lockfile defect is skipped; other failures still stop the refresh.

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

---

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

[github-actions]

Doc Preview CI
Preview removed because the pull request was closed or merged.

[rwgk]

codex GPT-6.1-Sol ultra:

No blocking findings in PR #3062; I’d approve it.

The important distinction: this makes refresh usable by postponing all cuda_core updates, not by repairing its lockfile. The restore-and-skip branch recognizes the exact defect, restores the pre-update lock, and leaves every other failure fatal. That is a reasonable temporary policy.

Validation:

• Jason’s run exercised the exception and created PR #3063, changing exactly the other five lockfiles.
• All 30 detector tests and 11 executable shell-path checks passed.
• The generated refresh remains compatible with #3059’s freshness exception.

Only non-blocking gap: the new workflow branches lack committed regression tests; our temporary diagnostic harness covers them.