test(cron): skip POSIX mode assertions on Windows

[cryptoyasenka]

tests/cron/test_file_permissions.py asserts exact POSIX modes (0o700 / 0o600). Windows chmod() only toggles the read-only bit, so os.stat() never reports those values and all six tests in the two permission classes fail on a Windows checkout.

Both classes assert nothing but modes, so they are gated at class level with the same idiom the suite already uses elsewhere: os.name == "nt" with reason POSIX mode bits not enforced on Windows. TestSecureHelpers is untouched and still runs everywhere.

Verified on Windows 11, Python 3.13, cherry-picked onto current main:

• before: 6 failed, 1 passed
• after: 1 passed, 6 skipped

Green ubuntu run for this head: https://github.com/cryptoyasenka/hermes-agent/actions/runs/31056703155

Test only, no source changes. ruff check clean.

[monerostar]

Native Win11 verification (monerostar)

Host: Windows 11 build 10.0.26200, Hermes venv Python 3.11.15, pytest 9.1.1 (-o addopts=).

Compared current main vs this PR head for tests/cron/test_file_permissions.py only.

Results

Tree	Result
main (d71033a40)	6 failed, 2 passed
this PR (1d3ba1837)	2 passed, 6 skipped

Failure shape on main (expected)

Mode assertions compare stat bits to POSIX targets; on NTFS chmod does not produce 0o700/0o600:

• AssertionError: 511 != 448 (0o777 vs 0o700)
• AssertionError: 438 != 384 (0o666-ish vs 0o600)

Failing classes: TestCronFilePermissions (3) + TestConfigFilePermissions (3). The two TestSecureHelpers cases still pass on both trees.

Notes

• Skip gate uses the suite’s existing os.name == "nt" idiom; no production code change.
• Matches issue [Windows] backup import test fails: os.chmod(..., 0o600) is a no-op on Windows #56923 class of Windows permission test noise.
• Non-blocking nit: if maintainers prefer sys.platform.startswith("win") for consistency with some newer tests, either form is fine on this host.

Verdict: correct, minimal, green on live Win11. Evidence only — not a maintainer approve.

[monerostar]

monerostar native Windows live-verify

Environment

• Windows 11 native (Build 26200) · Python 3.11 hermes venv
• Control: install tests/cron/test_file_permissions.py (no nt skip)
• PR HEAD: 1d3ba1837 — worktree pr-71428
• Note: no CI checks reported on this branch head (stale/behind?); local proof below

Why main fails here

Windows chmod only toggles the read-only bit. stat.S_IMODE never reports POSIX 0o700 / 0o600. Live on main:

pytest tests/cron/test_file_permissions.py -q -o addopts=
→ 6 failed, 2 passed

Failures are pure mode equality (511 != 448 = 0o777 vs 0o700, 438 != 384 = 0o666 vs 0o600). TestSecureHelpers (2 tests) still pass.

PR

Class-level @unittest.skipIf(os.name == "nt", "POSIX mode bits not enforced on Windows") on the two mode-asserting classes — same idiom used elsewhere in the suite.

pytest tests/cron/test_file_permissions.py -q -o addopts=   # PR tree
→ 2 passed, 6 skipped

Assessment

• Correct test-portability fix; does not weaken production permission helpers on POSIX
• +2/−0, focused
• Optional follow-up (not blocking): Win-specific assertions on readonly bit if you want coverage rather than skip-only

Verdict: LGTM for landing so Windows contributors get a green local suite on this file.

Formal Approve: blocked for external collaborator — comment review only.

[teknium1]

Thanks for the focused Windows test-portability fix. Current main still makes exact POSIX-mode assertions in tests/cron/test_file_permissions.py:37-40, :56-74, and :94-118; the two proposed class decorators guard exactly those mode-only classes. TestSecureHelpers remains outside both decorators at tests/cron/test_file_permissions.py:121.

This matches an existing suite convention at tests/test_honcho_client_config.py:74 and is a minimal salvage onto current main.

Automated hermes-sweeper review.

[cryptoyasenka]

@monerostar thanks for the native Win11 run here too, and for listing the actual assertion values rather than just a pass/fail count. Independent hardware evidence on these Windows PRs is exactly what they were missing.

On the nit: I kept os.name == "nt" because that is the idiom the surrounding suite already uses, for example tests/test_honcho_client_config.py, and switching only this file to sys.platform.startswith("win") would make it the odd one out. Happy to flip it if maintainers prefer the newer form.

No changes needed on this one after the review, and it still reports no conflict against current main.

[alt-glitch]

This was generated by AI during triage.

Related: #60888 covers the same Windows cron permission-test family more broadly. This is the narrower class-level assertion gate, so retain both for maintainer comparison.