tests: Avoid timeout in slirp4netns-no-unmount.sh

[mbargull]

Cherry-picked 189e1ed .
Discussion at #234 (review) :

---

@mbargull:

Without this the tests (also on Ubuntu) are prolonged by 40s due to the 2 x 40 x 0.5s waiting from wait_for_network_namespace and wait_for_network_device.

@AkihiroSuda

This seems to need investigation. I don't think we should need -r here.

cc @giuseppe

@giuseppe:

the user should already have enough privileges to run these tests, why does the new user namespace make a difference?

@mbargull

I have no idea, unfortunately (not my area of expertise). I only noticed this while running the tests.
I reverted the associated commit for now.
@AkihiroSuda, @giuseppe can you open an issue to track this?
(Asking you because you can give more context than unknowledgeable me.)

[AkihiroSuda]

CI seems passed without unshare -r: #234

Do we still need unshare -r?

[mbargull]

nsenter --preserve-credentials -U -n --target=$1 <cmd> in the loops from

slirp4netns/tests/common.sh

Lines 3 to 27 in 4bce8d8

	function wait_for_network_namespace {
	# Wait that the namespace is ready.
	COUNTER=0
	while [ $COUNTER -lt 40 ]; do
	if nsenter --preserve-credentials -U -n --target=$1 true; then
	break
	else
	sleep 0.5
	fi
	let COUNTER=COUNTER+1
	done
	}
	function wait_for_network_device {
	# Wait that the device appears.
	COUNTER=0
	while [ $COUNTER -lt 40 ]; do
	if nsenter --preserve-credentials -U -n --target=$1 ip addr show $2; then
	break
	else
	sleep 0.5
	fi
	let COUNTER=COUNTER+1
	done
	}

does not finish successfully when triggered at

slirp4netns/tests/slirp4netns-no-unmount.sh

Line 17 in 4bce8d8

wait_for_network_namespace $child

and

slirp4netns/tests/slirp4netns-no-unmount.sh

Line 27 in 4bce8d8

wait_for_network_device $child tun11

.

The tests complete fine because this waiting does not affected the outcome of the test, it just delays its completion.

---

Your preliminary conclusion was

This seems to need investigation. I don't think we should need -r here.

which makes sense given @giuseppe's comment

the user should already have enough privileges to run these tests, why does the new user namespace make a difference?

---

(I'm not able to help with this investigation though, unfortunately. But do feel free to discuss and amend this PR how you see fit!)

[AkihiroSuda]

nsenter was called with -U even when unshare was called without -r, and yet the failure of nsenter was not caught in the test script.

Being fixed in #239

[mbargull]

nsenter was called with -U even when unshare was called without -r,

Out of curiosity: Is it really the -U or the --preserve-credentials (combination?) that might be the cause.
(Sorry, if this is a silly question -- I'm just trying to understand the cause without knowing much context, really.)

and yet the failure of nsenter was not caught in the test script.

Ah, it was supposed to fail -- now all of that makes much more sense to me :).

Being fixed in #239

Thanks!