tests: use start_network instead of ethos directly

[miri64]

Contribution description

On some platforms (e.g. Arch Linux or newer Ubuntu versions) the TAP interface does not come up once connected and must be set up manually. This is already done in the start_network.sh script so let's just use that instead of pure ethos.

Testing procedure

The tests in question should still work for boards supporting ethos:

cd tests/${TEST}
export BOARD=samr21-xpro
make flash
sudo make test

Issues/PRs references

See #13734 (comment)

[benpicco]

Hard to tell if the test was already broken on master 😕

% sudo make BOARD=saml10-xpro PORT=/dev/ttyACM1 test 
[sudo] password for benpicco: 
make[1]: Nothing to be done for 'all'.
Timeout in expect script at "child.expect(r"HWaddr: (?P<hwaddr>[A-Fa-f:0-9]+)\s")" (tests/gnrc_ipv6_ext/tests/01-run.py:598)

make: *** [/home/benpicco/dev/RIOT/tests/gnrc_ipv6_ext/../../Makefile.include:733: test] Error 1

[benpicco]

Ah now I see my mistake - the PORT=/dev/ttyACM1 does not get exported into the sudo environment.

I unplugged the other board occupying /dev/ttyACM0, now the test is progressing further

sudo make BOARD=samr21-xpro test
make[1]: Nothing to be done for 'all'.
......FAILED
Timeout in expect script at "child.expect(r"~~ SNIP  0 - size:\s+(\d+) byte, type: NETTYPE_\w+ \(\d+\)")" (tests/gnrc_ipv6_ext/tests/01-run.py:135)

make: *** [/home/benpicco/dev/RIOT/tests/gnrc_ipv6_ext/../../Makefile.include:733: test] Error 1

But that failure should be unrelated to this PR.

The tap0 interface keeps lingering around after the failed test though.

[miri64]

The tap0 interface keeps lingering around after the failed test though.

Again: this should be removed by the cleanup function in the start_network.sh script.

[miri64]

The tap0 interface keeps lingering around after the failed test though.

Again: this should be removed by the cleanup function in the start_network.sh script.

Mkay.. if testrunner just calls SIGKILL without warning the best script cleanup does not help.....

Anyway... #12994 (comment) the issue that triggered this PR does not seem to be about ethos after all... The cleanup issue persists, but for now, I am closing this PR.