CI Switch to strict xfail policy for common estimator checks

[betatim]

This runs the scikit-learn common estimator checks in strict mode. This means a test marked as xfail that passes leads to a failure. This helps us keep the xfail list current and notice flaky checks.

fixes #7346

[csadorf]

Is this ready for review?

[betatim]

Ready for review (modulo figuring out the CI failures which seem to be unrelated to the PR content)

[coderabbitai]

📝 Walkthrough

Summary by CodeRabbit

• Chores
  • Raised scikit-learn minimum version to 1.8 for test compatibility.
• Tests
  • Enforced stricter handling of expected test failures, reducing the set of checks marked as expected failures.
  • Retained targeted workarounds and a few disabled checks for known issues to keep test stability.

Walkthrough

Enabled stricter xfail enforcement in scikit-learn compatibility tests by adding xfail_strict=True to parametrize_with_checks, raised scikit-learn minversion to 1.8, and removed many previously-expected xfail entries from the per-estimator xfail list.

Changes

Cohort / File(s)	Summary
Test decorator & xfail list python/cuml/tests/test_sklearn_compatibility.py	Added xfail_strict=True to estimator_checks.parametrize_with_checks; updated importorskip minversion from 1.7 to 1.8; pruned many entries from the PER_ESTIMATOR_XFAIL_CHECKS (removed numerous check_* items across many estimators), while keeping a few targeted workarounds (e.g., AgglomerativeClustering, UMAP.nan_inf).

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~25 minutes

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Title check	✅ Passed	The title accurately describes the main change: switching to strict xfail policy for sklearn estimator checks.
Description check	✅ Passed	The description is directly related to the changeset, explaining the purpose of switching to strict mode and referencing the linked issue.
Linked Issues check	✅ Passed	The PR fully addresses issue #7346 by updating parametrize_with_checks to include xfail_strict=True and removing no-longer-failing checks.
Out of Scope Changes check	✅ Passed	All changes are directly related to enabling strict xfail handling in test_sklearn_compatibility.py as required by issue #7346.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment

Important

Action Needed: IP Allowlist Update

If your organization protects your Git platform with IP whitelisting, please add the new CodeRabbit IP address to your allowlist:

• ✨ 136.113.208.247/32 (new)
• 34.170.211.100/32
• 35.222.179.152/32

Reviews will stop working after February 8, 2026 if the new IP is not added to your allowlist.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[coderabbitai]

Actionable comments posted: 1

🤖 Fix all issues with AI agents

In `@python/cuml/tests/test_sklearn_compatibility.py`:
- Around line 42-44: Remove the blanket skip by deleting
pytest.importorskip("sklearn", minversion="1.8") and instead detect sklearn
version at runtime (using sklearn.__version__ or packaging.version.parse) to set
a module-level flag (e.g., SKLEARN_SUPPORTS_STRICT_XFAIL). Use that flag when
calling or decorating with parameterize_with_checks (or when passing
xfail_strict) so xfail_strict=True is only used when
SKLEARN_SUPPORTS_STRICT_XFAIL is True, while allowing the rest of the sklearn
compatibility tests (parameterize_with_checks) to run on older versions; update
any uses of xfail_strict to reference this flag.

🧹 Nitpick comments (1)

python/cuml/tests/test_sklearn_compatibility.py (1)

596-622: Remove commented-out xfail entries or add a rationale.

Leaving disabled keys as comments makes the strict-xfail list harder to audit. Consider removing them entirely (or add an explicit rationale if you want to keep the history in-file).

♻️ Suggested cleanup

-        # "check_no_attributes_set_in_init": "GaussianRandomProjection sets attributes during init",
...
-        # "check_parameters_default_constructible": "GaussianRandomProjection parameters are mutated on init",

[csadorf]

/merge