Skip unit test when prjquota not supported

[oggstr]

Closes #51065.

- What I did

Added an additional prerequisite check before running daemon/internal/quota units tests. This fixes an issue where the tests could run in environments where they could never succeed.

In particular, I extended the CanTestQuota to also attempt a mount using the prjquota option. This ensures that XFS actually supports the quota option before running tests.

- How I did it

Added a function called supportsQuota that attempts a mount. It reuses some existing logic to prepare the XFS image (PrepareQuotaTestImage).

- How to verify it

Run the quota unit tests

TESTDIRS='./daemon/internal/quota/' ./hack/test/unit

The outcome depends on what the backing filesystem supports

• No XFS support: tests skipped (old check)
• XFS support, but missing quota support: tests skipped (new check)
• XFS support and quota support: tests run

Unfortunately I don't have easy access to a system where XFS support quotas, so I've been unable to verify the last scenario.

- Human readable description for the release notes

- A picture of a cute animal (not mandatory but encouraged)

Baby moose

[AkihiroSuda]

Doesn't seem necessary to pass t *testing.T just for t.TempDir()

Suggested change

	func CanTestQuota(t *testing.T) (string, bool) {
	func CanTestQuota() (string, bool) {

Alternatively you can just omit return values and call t.Skip() from this function

Suggested change

	func CanTestQuota(t *testing.T) (string, bool) {
	func RequireSupported(t *testing.T) {

[thaJeztah]

Just a quick blurb / thinking out loud; I haven't checked exactly how we set things up for tests, but I think the location used for daemons to run tests is configurable; in that case, we may need to pass the location to test, as it could be on a different filesystem.

(I recall we had some checks in graph-drivers that created a tempDir to test if they were supported, but didn't account for (e.g. /var/lib/docker/XXX to be a different storage / mount attached to the machine)

[oggstr]

Sorry for the delay – I've been under the weather for a few days.

@thaJeztah

I'm not sure I understand your comment fully.

I found this issue (that you submitted to GO) linked in a comment (daemon/daemon_test.go:345). Seems like there's some trickiness with temp directories in MacOS. Is this related to what you're talking about?

The code at this particular comment resolves symlinks using filepath.EvalSymlinks(t.TempDir()). However, this seems to be the only case where this is done. As far as I can tell, there seems to be no special handling of t.TempDir() anywhere else.

Would you mind expanding on your point?

@AkihiroSuda

My thinking was that it simplifies clean-up. Calling t.Skip() from inside the helper would be hiding control flow in my eyes, but I think with a re-name to RequireSupported (or something similar) it's fine enough.

I'll go ahead and change it once I hear back from @thaJeztah

[oggstr]

Went ahead and updated the PR. Function now skips the test directly.

[robmry]

Just a quick blurb / thinking out loud; I haven't checked exactly how we set things up for tests, but I think the location used for daemons to run tests is configurable; in that case, we may need to pass the location to test, as it could be on a different filesystem.

@thaJeztah - maybe you're thinking of $DOCKER_INTEGRATION_DAEMON_DEST? If so, it's probably ok .. the temp dir here is only used as a mount point, to check the kernel can mount a filesystem with the prjquota option. The filesystem it's mounted in to isn't important?

[oggstr]

Hope you don't mind be pinging you @thaJeztah, but would you mind having a look at Rob's comment above?

I'd be happy to look in to your question, but I would need some more info to go on.

[robmry]

LGTM - but will wait to see if @thaJeztah's question is answered before merging.