virtiofs behavior on Windows is unclear though support is macOS-only

[bobsira]

What Happened?

virtiofs is currently supported only on macOS, but on Windows it can still be exercised and produce errors that look like real failures. This causes confusion because the behavior is from an unsupported platform path, not necessarily a functional regression.

Attach the log file

Running tool: C:\Program Files\Go\bin\go.exe test -test.fullpath=true -timeout 30s -run ^TestVirtiofsValidateMountString$ k8s.io/minikube/pkg/drivers/common/virtiofs

=== RUN   TestVirtiofsValidateMountString
    c:/dev/minikube/pkg/drivers/common/virtiofs/virtiofs_test.go:47: failed to parse mountString "C:\\Users\\bosira\\AppData\\Local\\Temp\\TestVirtiofsValidateMountString3941094040\\001:/mnt/models": host path "C" is not an absolute path
--- FAIL: TestVirtiofsValidateMountString (0.00s)
FAIL
FAIL    k8s.io/minikube/pkg/drivers/common/virtiofs     1.271s

Operating System

Windows

Driver

None

[nirs]

We need to run these tests only on macOS until this code is usable on other OS.
But why the windows unit tests do not fail in the CI?

[bobsira]

We need to run these tests only on macOS until this code is usable on other OS. But why the windows unit tests do not fail in the CI?

Great question. Let me investigate this.

[wasimat404]

I was looking into this and I think windows checks don't fail because this test never runs in CI at all, on any system. The way the test runner picks what to check quietly skips this one (down to how the test file is written), so it's green because it's absent, not because it passes. Agree it should run on mac only until the code works elsewhere.
The failure itself is real (Mac-only code, tripped up by how windows paths look), but CI never caught it because it never ran it. I can either mark it Mac-only and make it actually run on the Mac checks, or also fix the runner so it stops skipping tests written this way.