e2e: rethink setenforce-0

[edsantiago]

Found by accident while reviewing #18544:

podman/test/e2e/login_logout_test.go

Lines 45 to 50 in 493aac6

	se := SystemExec("setenforce", []string{"0"})
	se.WaitWithDefaultTimeout()
	if se.ExitCode() != 0 {
	Skip("Cannot disable selinux, this may cause problem for reading cert files inside container.")
	}
	defer SystemExec("setenforce", []string{"1"})

podman/test/e2e/push_test.go

Lines 248 to 252 in 493aac6

	se := SystemExec("setenforce", []string{"0"})
	Expect(se).Should(Exit(0))
	defer func() {
	se2 := SystemExec("setenforce", []string{"1"})
	Expect(se2).Should(Exit(0))

Let's understand if/why these are needed.

• If not needed: remove
• If needed: explain in comments, and Serialize.

[vrothberg]

Commit 3a3c558 added these lines but there's no indication of why SELinux would play a role. Maybe because the registry setup mounts HOME directories into the registry container?

I wonder if we couldn't just nuke all this code and use the hack/podman-registry-go bindings instead to create/start/stop a local registry?

[Luap99]

Yes please nuke this, I was also very surprised when I saw it. This by definition also means it skips all these test as rootless as setenfore is only allowed as root.

[vrothberg]

I didn't volunteer (yet) to do the work as time is scarce. But I had a quick look and it seems like a bit of work to get the entire file migrated.

[Luap99]

Given that this doesn't work in a parallel context anyway (TOCTOU) I am just going to remove the setenfore calls and see if CI passes (#18607). We can look into using hack/podman-registry-go another time.