Add support for adding Linuxrc parameters and AY variables

[imobachgs]

Additional parameters for Linuxrc

To check VNC installations, we need to pass additional arguments to Linuxrc when booting the media. So I've decided to implement support for adding those arguments. Given a test sles12.rb, a new file should be created containing the list of parameters to add. For example:

vnc=1 VNCPassword=12345678 y2debug

Variables for the AutoYaST profile

At this time, only the %IP% placeholder is supported in AutoYaST profiles. So we're adding support for reading those variables from a file (form example, sles12.vars):

REPO1_URL=http://{{IP}}:{{PORT}}/static/repos/sles12
NAME=some name

You can use {{IP}} and {{PORT}} placeholders in that file. In the AutoYaST profile, just use placeholders like this:

<rule>
  <name>{{NAME}}</name>
</rule>

[imobachgs]

aytests 1.0.1 will use this new mechanism.

[ancorgs]

Is there any situation in which you want to call build with a different build_environment?

I mean, why not just leave build without parameters and call build_environment inside?

Same applies to autoinst_path.sub_ext(".linuxrc"). Do you ever use a different file? If not, I would move that part to the build_environment method.

[imobachgs]

It's part of my 'semi-functional' mindset :) For example, it makes the method easier to test. But I can change it anyway.

[ancorgs]

Well, actually you need to pass autoinst, which I see fine because is quite simetric to the call to setup_autoinst.

[imobachgs]

Ok, I could pass autoinst and do the transformation inside.

[ancorgs]

To be honest, this doesn't look like the most explicitly declarative spec ever. Would it be possible to have things like

it "adds the variables read from the linuxrc file" do
  #some expect
end

instead of a "reimplementation" of the steps were such checks are simply implicit?

[imobachgs]

Yes, it would be better. There's a lot of room for improvement in these tests (too many private methods, too much mocking, etc.).

[imobachgs]

If you don't mind, I'll leave this as it is right now to finish the PBI. I'll come back to work on this right after the sprint.

[ancorgs]

Ok. But I take it as a promise (it's written) 😉

[imobachgs]

Yes, it's a promise :)

[ancorgs]

I like it better now. But I actually wanted to go one step further in the same direction having

build(autoinst)

[ancorgs]

Because the same thing happens again, build is always called with build_environment as only parameter (and even the name of the method seems to indicate that it really only makes sense in the context of a build).

[ancorgs]

LGTM

[imobachgs]

Thanks.