test: pin fail-soft contract for rootdir symlink probe

[piyushrajyadav]

Problem / Motivation

The symlink capability probe exists in two places with the same fail-soft contract:

• _can_create_symlink in test/conftest.py
• _root_can_create_real_symlink in repo-root conftest.py

While _can_create_symlink was pinned in #5943 via test/test_symlink_capability_probe.py, its rootdir sibling _root_can_create_real_symlink remained unpinned against regression.

Why it matters

Both probes run at conftest import time (_HAS_SYMLINKS and _ROOT_HAS_REAL_SYMLINKS). If an unanticipated filesystem error (OSError, missing os.symlink attribute, or NotImplementedError) is allowed to propagate, it results in a whole-suite collection error taking down the entire test run, rather than failing soft and skipping only the symlink-dependent tests.

What changed (motivation → approach → change)

• Extended test/test_symlink_capability_probe.py using a fixture (symlink_probe) that parametrizes all test cases across both test/conftest.py's _can_create_symlink and repo-root conftest.py's _root_can_create_real_symlink (accessed via pytest's plugin manager).
• Pinned fail-soft behavior across:
  • Standard decline errnos (EPERM, EACCES, EOPNOTSUPP, ENOSYS)
  • Windows privilege absence (winerror == 1314 with unrelated errno)
  • Unanticipated filesystem OSError (e.g. EIO)
  • Missing/unsupported symlink runtime (NotImplementedError and missing os.symlink attribute)

Tests

• Added parametrized test coverage in test/test_symlink_capability_probe.py covering 16 test permutations across both probes.
• Ran pytest test/test_symlink_capability_probe.py — all 16 tests passed.

Manual verification

N/A — unit test coverage directly tests the probe functions and failure modes.

Related Issues

Fixes #5946

Checklist

• At most two commits (one is the norm), with a Conventional Commits title (feat|fix|docs|refactor|perf|test|chore|ci|build|revert: ...)
• Existing tests pass and new tests added for new functionality
• Self-review completed; code follows project style guidelines
• Documentation updated (if applicable)
• No secrets, credentials, or internal references in the diff

[dwu96]

👋 Hi! This PR's description is missing some required sections from our PR template. Workflow runs won't be auto-approved until the description is updated.

Missing sections:

• ## Problem / Motivation
• ## Why it matters
• ## What changed
• ## Tests

Please update your PR description to include these sections, then push or re-save the description. The workflows will be approved on the next cycle.

[piyushrajyadav]

@dwu96 Thanks for the heads up! I've updated the PR description with all the required sections from the PR template (Problem / Motivation, Why it matters, What changed, Tests, and checklist).

[github-actions]

GPT 5.6 Review (fork) — ✅ no blocking findings

Reviewed a4ce33a474ccf124bf32b0a6a6d0f8468597d1df via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] a4ce33a

[github-actions]

Design Review (Fable 5, fork) — ✅ PASS

Design-level review of a4ce33a474ccf124bf32b0a6a6d0f8468597d1df via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Both probes in the base already catch (OSError, NotImplementedError, AttributeError), so the new tests pin existing behavior rather than demand new code — and the pre-existing test file was actually exercising test/conftest.py's probe despite claiming to cover the rootdir one, which this PR corrects. The fixture reaches the live rootdir conftest via the plugin manager (correctly avoiding a re-import that would re-run the conftest's import-time side effects), and the two probes genuinely differ (target_is_directory=True on the rootdir one), so parametrizing tests over both rather than deduplicating the probes is the right shape. Test-only, no contract or reversibility concerns.

Design-Verdict: PASS

Pins a real, previously-unpinned fail-soft contract on the rootdir probe with a proportionate, test-only change that also fixes the old test's mislabeled target.

[DESIGN-REVIEWED] a4ce33a

[github-actions]

Opus 4.8 Review (fork) — ✅ no blocking findings

Reviewed a4ce33a474ccf124bf32b0a6a6d0f8468597d1df via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] a4ce33a

[iamwhatever]

Tier 1 auto-approve: test (1 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: parameterises the existing symlink-capability-probe regression tests over both the test/ and rootdir conftest probes and adds a missing-os.symlink case; single test file, no runtime code. CodeQL is not applicable on this fork PR (default-setup emits no check-run); SAST coverage is Semgrep only, latest run success with 0 annotations.

[bolichen97]

Tier 1 auto-approve: test (1 file). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, SAST annotations clean, security checklist all-NO, AI reviewers green. Category: test: pin fail-soft contract for rootdir symlink probe — test-file-only change extending existing probe regression tests to both conftests. CodeQL is not applicable on this fork PR (default-setup emits no check-run); SAST coverage is Semgrep only, latest run success with 0 annotations. Note: auto-merge could not be armed from this run (agent command posture blocks gh pr merge); merge remains gated by branch protection and a maintainer.