tests: add core revert test

[fgimenez]

This test exercises a core revert on a nested ubuntu-core system, currently only a preinstalled snap is set (network-manager) and the check after the revert only takes into account network configuration. It can be executed with: SPREAD_CORE_CHANNEL=stable spread -v qemu:ubuntu-16.04-64:tests/nested/core-revert.

This PR also includes a refactor of reusable functions from a previous nested task, the actual test is https://github.com/snapcore/snapd/pull/3348/files#diff-4825fd024a235e7895b109ad37072e68

[codecov-io]

Codecov Report

Merging #3348 into master will increase coverage by <.01%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master    #3348      +/-   ##
==========================================
+ Coverage   77.55%   77.55%   +<.01%     
==========================================
  Files         371      371              
  Lines       25519    25519              
==========================================
+ Hits        19790    19792       +2     
+ Misses       3978     3976       -2     
  Partials     1751     1751

Impacted Files	Coverage Δ
cmd/snap/cmd_aliases.go	96% <0%> (+2%)	⬆️
interfaces/sorting.go	93.33% <0%> (+3.33%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update a020a0f...78802c4. Read the comment docs.

[fgimenez]

The test case finally reproduces the seccomp issue, with the current status and executed setting SPREAD_CORE_CHANEL=stable it gets stuck here https://github.com/snapcore/snapd/pull/3348/files#diff-4825fd024a235e7895b109ad37072e68R41 (after reverting to stable network-manager doesn't come up and the testbed is not accessible through ssh).

@niemeyer if you could take a look that would be great

[mvo5]

Looks good, one (most unrelated for the core of this PR) comment about trap

[mvo5]

Trap will replace other existing traps on the same signal. So if we alread use a trap ... EXIT anwhere this is overriden now. We may consider switching to a helper (add_trap) that stacks the cleanups if we already use traps. Probably material for a new PR though.

[fgimenez]

Thanks @mvo5 ! Indeed, for the time being, and given that all the tasks in the nested suite are going to use ubuntu-image, I've moved its install/unistall at the suite level, and btw also removed a trap used for apt, please take another look.

[morphis]

LGTM

[zyga]

A review that I apparently never submitted

[zyga]

QEMU can expand to "". This is not a big deal but I'd rather get a qemu-system-x86_64: command not found than a more confusing message below when the first argument of qemu becomes the command. Could we please drop the which?

[fgimenez]

Sure, will do in a followup

[zyga]

I'd prefer if this was a temporary directory.

[fgimenez]

Ok, noted for the followup

[zyga]

Could you please document the intent and mechanics of this test.

[fgimenez]

Sure, I'll add a details section

[zyga]

Can this test happen on other distributions?

[fgimenez]

I think so, in fact the interesting part happens on the nested machine, if we can install qemu all should be good. I'll try locally with debian/fedora anyway.