Tests fail on OpenBSD for unix feature

[lcheylus]

After merge of PR #9158 and #9185, build OK on OpenBSD amd64/current with Rust 1.90.0.

But there are some errors during tests with unix feature => 4 tests FAIL

$ rustc -vV
rustc 1.90.0 (1159e78c4 2025-09-14) (built from a source tarball)
binary: rustc
commit-hash: 1159e78c4747b02ef996e55082b704c09b970588
commit-date: 2025-09-14
host: x86_64-unknown-openbsd
release: 1.90.0
LLVM version: 19.1.7

$ cargo test -v --no-default-features --features unix
(...)
failures:
    test_cp::test_cp_current_directory_preserve_attributes
    test_df::test_type_option_with_file
    test_expr::test_long_input
    test_hostname::test_hostname_ip

test result: FAILED. 3946 passed; 4 failed; 46 ignored; 0 measured; 0 filtered out; finished in 158.05s

[lcheylus]

After some debug:

• tests/by-util/test_cp.rs: test test_cp_current_directory_preserve_attributes must be disabled via cfg for OpenBSD (same as FreeBSD and Windows)
• tests/by-util/test_df.rs: test test_type_option_with_file must be disabled via cfg for OpenBSD (same as FreeBSD and Windows)

There is also an error with test test_hostname_ip in tests/by-util/test_hostname.rs using dns_lookup crate => debugging and tests in progress...

[naoNao89]

The CI doesn't test on OpenBSD, which is why these platform-specific test failures weren't caught. The unix feature flag differs from the platform-specific features used in CI, creating test environment mismatches that only surface when developers test locally on unsupported platforms like OpenBSD.

[lcheylus]

After some tests/debug, error in tests/by-util/test_hostname.rs for test_hostname_ip is due to an unknown hostname => solved with modification of /etc/hosts file (add entry for hostname/local IP).

[lcheylus]

After some tests, error for test_expr::test_long_input is due to a limitation on OpenBSD.

With an expression with 4000 entries,

coreutils/tests/by-util/test_expr.rs

Lines 583 to 588 in 690405d

	fn test_long_input() {
	// Giving expr an arbitrary long expression should succeed rather than end with a segfault due to a stack overflow.
	#[cfg(not(windows))]
	const MAX_NUMBER: usize = 40000;
	#[cfg(not(windows))]
	const RESULT: &str = "800020000\n";

test fails with segfaut:

$ ./target/debug/coreutils expr $(cat ~/tmp/expr1.out)
Segmentation fault         (core dumped) ./target/debug/coreutils expr $(cat ~/tmp/expr1.out)

If I modified code for test_expr::test_long_input to use 10000 values (instead of 40000), no error on OpenBSD and the test is successful.