Michal and I are working on a bug with network configuration using
AutoYaST. The fix is rather intrusive so we want to add tests and
check for regressions.
The basic idea for such tests is to start with a specific AY profile
and a matching virtual hardware (network cards), run the
installation, and check the resulting filesystem for expected
One test case:
- autoyast profile
- feature being tested (eg. NIC naming)
- helpers for the automated test (multiple pkg repos,
disable sig check, final_halt)
- a standard installation repo (SP2)
- new yast package(s) to test
- to be put to a DUD and an add-on repo (on a CD)
- virtual hardware configuration (platform: libvirt)
- feature-wise (number of NICs and their MACs)
- helpers (second CDROM)
- assertions to check after the run
- simple way: check that some files have excactly some expected
Automating multiple runs of multiple test cases: TBD
If this sounds too abstract, here's our existing manual test plan
for the bug:
http://idontknow.suse.cz/install/ (internal, hrefs work)
https://gist.github.com/4194824 (external, hrefs broken online)
AFAIK we don't have a framework for such tests, so we want to build
one. Well, I know of some related work:
- Jenkins CI
- AutoYaST tests that I don't know of?
So that's why I'm sharing the design and want to check with you:
Are we reinventing the wheel?
Can we reuse the existing projects?
Martin Vidner, Cloud & Systems Management Team
Kuracke oddeleni v restauraci je jako fekalni oddeleni v bazenu