Fix normalization of hardware key/option

[happz]

A bug was introduced in the patching adding multiple disks to testcloud plugin, and the test was missing.

Pull Request Checklist

• implement the feature

[happz]

@martinhoyer to my horror, I broke hardware normalization, and since I didn't bother to add a unit test, nobody noticed till I used --hardware just this evening. So, I suppose one more to include in 1.34, this time it's a genuine bug fix :/

[martinhoyer]

Nice catch. Thanks for the heads-up.
I'd say we can do the release on Monday anyway. (DevConf)
Unless you really want to do this week :)

[martinhoyer]

It looks good to me, but clearly I don't understand the code enough to be able to click approve with clear conscience.
@happz If you could help me understand, it would be much appreciated 🙏
I get that line 519 is merged[components.name] = [] as there is peer_index, but am not sure why merged[components.name] is otherwise {}. Am I looking correctly at

tmt/tmt/hardware.py

Line 230 in e4991dd

name=groups['name'],

?

[martinhoyer]

@psss @lukaszachy Could you take a look please, so we can do the release? :)

[lukaszachy]

Hm, there is missing fmf on rawhide (as it cannot be built https://kojipkgs.fedoraproject.org//work/tasks/6207/118796207/build.log) which blocks tmt.

Hm, we should monitor bugzilla as well, it was opened in May :/ (https://bugzilla.redhat.com/show_bug.cgi?id=2279994)

[martinhoyer]

Hm, there is missing fmf on rawhide (as it cannot be built https://kojipkgs.fedoraproject.org//work/tasks/6207/118796207/build.log) which blocks tmt.

Hm, we should monitor bugzilla as well, it was opened in May :/ (https://bugzilla.redhat.com/show_bug.cgi?id=2279994)

whoops. I wonder if teemtee/fmf#194 has something to do with it.
Probably shouldn't block this PR though?

[martinhoyer]

whoops. I wonder if teemtee/fmf#194 has something to do with it. Probably shouldn't block this PR though?

actually, I think it's the use of pkg_resources(I'll stop guessing). Looking into it.

[happz]

It looks good to me, but clearly I don't understand the code enough to be able to click approve with clear conscience. @happz If you could help me understand, it would be much appreciated 🙏 I get that line 519 is merged[components.name] = [] as there is peer_index, but am not sure why merged[components.name] is otherwise {}. Am I looking correctly at

tmt/tmt/hardware.py

Line 230 in e4991dd

name=groups['name'],

?

merged is basically this structure:

merged:
  # merged[component.name] = [] for these:
  disk:
    - size: 10GB
    - size: 10GB
  network:
    - type: eth

  # merged[component.name] = {} for these:
  cpu:
    flags:
      - avs
      - avs2
    model: 79
  device:
    vendor: ...

  # and merged[component.name] is whatever the value is for these:
  hostname: foo.bar.com
  memory: 4GB

The issue was in merged[components.name] = [] being used for all three kinds of requirements.

[martinhoyer]

Thanks @happz, much appreciated!