tests: add core revert test #3348
Conversation
Codecov Report
@@ 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
Continue to review full report at Codecov.
|
* do not fail on broken connection (auto reboot after refresh/revert) * wait for revert change to be done
|
The test case finally reproduces the seccomp issue, with the current status and executed setting @niemeyer if you could take a look that would be great |
tests/lib/nested.sh
Outdated
| # create ubuntu-core image | ||
| mkdir -p /tmp/work-dir | ||
| snap install --devmode --beta ubuntu-image | ||
| trap 'snap remove ubuntu-image' EXIT |
There was a problem hiding this comment.
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.
There was a problem hiding this comment.
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.
| QEMU="$(which qemu-system-x86_64)" | ||
| ;; | ||
| i386) | ||
| QEMU="$(which qemu-system-i386)" |
There was a problem hiding this comment.
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?
There was a problem hiding this comment.
Sure, will do in a followup
| esac | ||
|
|
||
| # create ubuntu-core image | ||
| mkdir -p /tmp/work-dir |
There was a problem hiding this comment.
I'd prefer if this was a temporary directory.
There was a problem hiding this comment.
Ok, noted for the followup
| @@ -0,0 +1,28 @@ | |||
| summary: core revert test | |||
|
|
|||
| systems: [ubuntu-16.04-64] | |||
There was a problem hiding this comment.
Can this test happen on other distributions?
There was a problem hiding this comment.
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.
| @@ -0,0 +1,28 @@ | |||
| summary: core revert test | |||
There was a problem hiding this comment.
Could you please document the intent and mechanics of this test.
There was a problem hiding this comment.
Sure, I'll add a details section
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