Normalize virtual grain for LXC

[piterpunk]

What does this PR do?

Fixes the virtual grains in LXC containers. Now it always set grains["virtual"] = "container" and grains["virtual_subtype"] = "LXC", independent of the command used for detection and even running inside another virtual machine.

What issues does this PR fix or reference?

Fixes #53868
Fixes #59573

Previous Behavior

The grains["virtual"] is inconsistent depending with the available commands for probing:

root@example:~# which systemd-detect-virt
root@example:~# which virt-what
root@example:~# salt-call grains.get virtual
local:
    kvm
root@example:~# which systemd-detect-virt
/usr/bin/systemd-detect-virt
root@example:~# which virt-what
root@example:~# salt-call grains.get virtual
local:
    LXC
root@example:~# which systemd-detect-virt
/usr/bin/systemd-detect-virt
root@example:~# which virt-what
/usr/sbin/virt-what
root@example:~# salt-call grains.get virtual
local:
    kvm

New Behavior

The grains["virtual"] and grains["virtual_subtype"] shows always the same output regardless the probe command.

root@example:~# which systemd-detect-virt
root@example:~# which virt-what
root@example:~# salt-call grains.get virtual
local:
    container
root@example:~# salt-call grains.get virtual_subtype
local:
    LXC
root@example:~# which systemd-detect-virt
/usr/bin/systemd-detect-virt
root@example:~# which virt-what
root@example:~# salt-call grains.get virtual
local:
    container
root@example:~# salt-call grains.get virtual_subtype
local:
    LXC
root@example:~# which systemd-detect-virt
/usr/bin/systemd-detect-virt
root@example:~# which virt-what
/usr/sbin/virt-what
root@example:~# salt-call grains.get virtual
local:
    container
root@example:~# salt-call grains.get virtual_subtype
local:
    LXC

Merge requirements satisfied?

[NOTICE] Bug fixes or features added to Salt require tests.

• Changelog - https://docs.saltstack.com/en/master/topics/development/changelog.html
• Tests written/updated

Commits signed with GPG?

No

[sagetherage]

@piterpunk
Re-running those mac tests, but they have been failing all over and so unlikely that it is due to your PR - I will check on the fix for those tests tomorrow and request reviews, today and will check back tomorrow, here. Thank you for your contribution!

[piterpunk]

re-run pr-debian-9-amd64-py3-pytest

[waynew]

I don't think it /needs/ changing, but a couple of comments around the tests.

Looks good 👍

[waynew]

suggestion (non-blocking) patch actually creates a MagicMock already. Also it's typically a good idea to create_autospec. Also in this case, patching object is probably less-desirable. All-in-all, this could be written like so:

with patch("os.path.isdir", autospec=True, return_value=False):
    with patch("os.path.isfile", autospec=True, side_effect=lambda x: x in (...)): 

(note, no need to True if <something> else False - just <something> or bool(something) [but, thing in stuff produces a bool value anyway)

[piterpunk]

Hi @waynew ,

I just used the test_lxc_virtual as basis to create the new tests, that's why the code looks this way. Take a look in lines 1218 to 1223.