Is there value of keeping the remaining playbook?

@LecrisUT yes, I do not want to implement more spagethi fun in the code only the amount is needed. Also the same for FIPS.

Let me rephrase it. What is being done in the playbook now that there is the logic section above?

Edit: or let me put it differently, would this even do what you expect it to do since the install commands are done at build time and the playbook would run immediately without even the plugin available to handle config-manager

Let me rephrase it. What is being done in the playbook now that there is the logic section above?

In the playbook we mostly play around with repository enablement:

https://github.com/teemtee/tmt/blob/c0a3c8994f6be4829632b58e3c13f74dec35e0a8/tmt/steps/prepare/feature/epel-enable.yaml

Edit: or let me put it differently, would this even do what you expect it to do since the install commands are done at build time and the playbook would run immediately without even the plugin available to handle config-manager

The bootc package manager install function should install the missing dependencies to a derived container image, build the new container image and switch to it:

So not sure I follow how it would run immediately. It runs after the missing dependencies are installed the image mode way.

Let me rephrase it. What is being done in the playbook now that there is the logic section above?

In the playbook we mostly play around with repository enablement:

https://github.com/teemtee/tmt/blob/c0a3c8994f6be4829632b58e3c13f74dec35e0a8/tmt/steps/prepare/feature/epel-enable.yaml

That is an answer to the litteral question, not the spirit of the question. The purpose of the question is to prod on whether it makes sense to keep that or not, which is not addressed.

Edit: or let me put it differently, would this even do what you expect it to do since the install commands are done at build time and the playbook would run immediately without even the plugin available to handle config-manager

The bootc package manager install function should install the missing dependencies to a derived container image, build the new container image and switch to it:

tmt/tmt/package_managers/bootc.py

Lines 301 to 316 in c0a3c89

	def install(
	self,
	*installables: Installable,
	options: Optional[Options] = None,
	) -> CommandOutput:
	presence = self.check_presence(*installables)
	missing_installables: set[Installable] = {
	installable for installable, present in presence.items() if not present
	}
	if missing_installables:
	self.engine.install(*missing_installables, options=options)
	return self.build_container() or CommandOutput(stdout=None, stderr=None)
	return CommandOutput(stdout=None, stderr=None)

So not sure I follow how it would run immediately. It runs after the missing dependencies are installed the image mode way.

Please read my comment again and more carefully!

Let me rephrase it. What is being done in the playbook now that there is the logic section above?

In the playbook we mostly play around with repository enablement:
https://github.com/teemtee/tmt/blob/c0a3c8994f6be4829632b58e3c13f74dec35e0a8/tmt/steps/prepare/feature/epel-enable.yaml

That is an answer to the litteral question, not the spirit of the question. The purpose of the question is to prod on whether it makes sense to keep that or not, which is not addressed.

It does for me, I do not want to introduce more python code as necessary. The ansible code alreaedy does decent job doing what is needed except the packages installation.

Edit: or let me put it differently, would this even do what you expect it to do since the install commands are done at build time and the playbook would run immediately without even the plugin available to handle config-manager

The bootc package manager install function should install the missing dependencies to a derived container image, build the new container image and switch to it:

tmt/tmt/package_managers/bootc.py

Lines 301 to 316 in c0a3c89

	def install(
	self,
	*installables: Installable,
	options: Optional[Options] = None,
	) -> CommandOutput:
	presence = self.check_presence(*installables)
	missing_installables: set[Installable] = {
	installable for installable, present in presence.items() if not present
	}
	if missing_installables:
	self.engine.install(*missing_installables, options=options)
	return self.build_container() or CommandOutput(stdout=None, stderr=None)
	return CommandOutput(stdout=None, stderr=None)

So not sure I follow how it would run immediately. It runs after the missing dependencies are installed the image mode way.

Please read my comment again and more carefully!

Which one, this below? Man, I read it like 10 times already :) and I thought I replied well.

Edit: or let me put it differently, would this even do what you expect it to do since the install commands are done at build time

yes, they are done at build time, but they are also applid to the live image right after each install command.

and the playbook would run immediately without even the plugin available to handle config-manager

No, the missing deps are installed and applied on the running system BEFORE the playbook runs.

I just tested it with injecting another dep, because all those ar actually alraedy baked into our images:

❯ tmt -vvv run -a plan --name '/epel/enabled/with-epel-preinstalled' provision --how virtual --image https://artifacts.dev.testing-farm.io/images/CentOS-Stream-9-image-mode-x86_64.qco
w2
...
        how: feature
        order: 50
> /var/home/thrix/git/github.com/teemtee/tmt/worktree/feature-plugins-image-mode/tmt/steps/prepare/feature/epel.py(76)enable()
-> import pdb; pdb.set_trace()
(Pdb)  distro
'centos'
(Pdb) n
> /var/home/thrix/git/github.com/teemtee/tmt/worktree/feature-plugins-image-mode/tmt/steps/prepare/feature/epel.py(77)enable()
-> guest.package_manager.install("neovim")
(Pdb) n
            cmd: rpm -q --whatprovides neovim
            stdout: no package provides neovim
            cmd: mktemp --directory tmp.XXXXXXXXXX
            stdout: tmp.KY02SByTo0
            cmd: /bin/bash -c "(  ( podman pull quay.io/testing-farm/centos-bootc:stream9 || podman pull containers-storage:quay.io/testing-farm/centos-bootc:stream9 )  || bootc image copy-to-storage --target quay.io/testing-farm/centos-bootc:stream9)"
            stderr: Trying to pull quay.io/testing-farm/centos-bootc:stream9...
            stderr: Getting image source signatures
            stderr: Copying blob sha256:9f2741e8a325eabbd2d2fdd05a403342e367cba176db17ad0cee655fd76c2620
            stderr: Copying blob sha256:29b667a8b294984efb1c0aca524ff788af33ee241b4adf543a64e13e23d134ba
            stderr: Copying blob sha256:0f6aaa63cc93223316d8174f136b292152af143f07236da94146f0489a38ea05
....
....
(neovim is installed before the playbook runs)

Edit: or let me put it differently, would this even do what you expect it to do since the install commands are done at build time and the playbook would run immediately without even the plugin available to handle config-manager

...

Please read my comment again and more carefully!

Which one, this below? Man, I read it like 10 times already :) and I thought I replied well.

The conversation that I quoted. My comment there is to indicate that you have misinterpreted what I was saying, i.e. re-read what I wrote and try to find if we are misaligned between what you think I wrote, and what I was trying to write, not about your reply!

To be more precise:

• I did not say that the install commands are done immediately, nor will they ever be able to because dnf install fails. These are done at build time
• I said that the playbooks are the things that are run immediately
• There is discrepancy between these two

yes, they are done at build time, but they are also applid to the live image right after each install command.

What? How can dnf install be applied to a live bootc image? Also be more specific on what you are referencing there, the install step, the ansible step, or something completely different. I have interpreted that you were referring to the install step.

and the playbook would run immediately without even the plugin available to handle config-manager

No, the missing deps are installed and applied on the running system BEFORE the playbook runs.

Huh, but then that means that we are doing the rebuild every time we hit an ansible step? E.g. if we have a prepare/install -> prepare/ansible -> prepare/shell, then it does: add install step to deferred commands -> build container -> run ansible -> add shell step to deferred commands -> build container? That does not fit with the understanding of what was done to support prepare/install + prepare/shell for bootc image. There is a big misalignment in how the bootc support is going on here if this is the case. Please prepare the design document more clearly in #4818 with some references in the PR so that we can all be on the same page of what is going on here.

I just tested it with injecting another dep, because all those ar actually alraedy baked into our images:

❯ tmt -vvv run -a plan --name '/epel/enabled/with-epel-preinstalled' provision --how virtual --image https://artifacts.dev.testing-farm.io/images/CentOS-Stream-9-image-mode-x86_64.qco
w2
...
        how: feature
        order: 50
> /var/home/thrix/git/github.com/teemtee/tmt/worktree/feature-plugins-image-mode/tmt/steps/prepare/feature/epel.py(76)enable()
-> import pdb; pdb.set_trace()
(Pdb)  distro
'centos'
(Pdb) n
> /var/home/thrix/git/github.com/teemtee/tmt/worktree/feature-plugins-image-mode/tmt/steps/prepare/feature/epel.py(77)enable()
-> guest.package_manager.install("neovim")
(Pdb) n
            cmd: rpm -q --whatprovides neovim
            stdout: no package provides neovim
            cmd: mktemp --directory tmp.XXXXXXXXXX
            stdout: tmp.KY02SByTo0
            cmd: /bin/bash -c "(  ( podman pull quay.io/testing-farm/centos-bootc:stream9 || podman pull containers-storage:quay.io/testing-farm/centos-bootc:stream9 )  || bootc image copy-to-storage --target quay.io/testing-farm/centos-bootc:stream9)"
            stderr: Trying to pull quay.io/testing-farm/centos-bootc:stream9...
            stderr: Getting image source signatures
            stderr: Copying blob sha256:9f2741e8a325eabbd2d2fdd05a403342e367cba176db17ad0cee655fd76c2620
            stderr: Copying blob sha256:29b667a8b294984efb1c0aca524ff788af33ee241b4adf543a64e13e23d134ba
            stderr: Copying blob sha256:0f6aaa63cc93223316d8174f136b292152af143f07236da94146f0489a38ea05
....
....
(neovim is installed before the playbook runs)

Please provide a simple tmt plan to properly follow. I do not follow what you showed there

This is wrong. epel-next is only for centos

This should be handled similarly with copr_plugin. We also have a handling of _base_debuginfo_command that can be generalized or implemented in a similar way