template://podman-rootful: provision scripts should not depend on LIMA_CIDATA_*

[AkihiroSuda]

lima/examples/podman-rootful.yaml

Line 44 in b453ded

echo "d /run/podman 0700 ${LIMA_CIDATA_USER} -" > /etc/tmpfiles.d/podman.conf

The provision scripts should probably use template like {{.User}}
#2052 (comment)

[afbjorklund]

Also used in docker-rootful, I think

[jandubois]

@AkihiroSuda Can you explain why template variables are preferable to the env variables?

It makes no big difference, but env variables are slightly more flexible if you want to work on the scripts in a running instance because the variables are not yet filled in (you can make changes to the script and copy it back into lima.yaml).

Another thing is that lima.env can be used to look up what information is available. But I guess we can document all the template variables in default.yaml as well.

Anyways, no strong feelings either way.

[AkihiroSuda]

Can you explain why template variables are preferable to the env variables?

Because these variables are implementation details of the init scripts, and can be removed even in a patch release.

[jandubois]

Because these variables are implementation details of the init scripts, and can be removed even in a patch release.

Oh, so the template would only make a subset of the information available? I thought you wanted to apply the cidata templating function to the scripts, which is functionally equivalent to exposing the shell variables; lima.env is essentially the mapping function.

So what are the template variables that should be exposed via the template?

[AkihiroSuda]

So what are the template variables that should be exposed via the template?

Just same as the template for the sockets?

lima/examples/default.yaml

Line 364 in b453ded

# # "hostSocket" can include {{.Home}}, {{.Dir}}, {{.Name}}, {{.UID}}, and {{.User}}.

[jandubois]

Just same as the template for the sockets?

Ok, that works for me.

Would the LIMA_* variables still remain available for the system provisioning scripts (for backwards compatibility), or would you want to block them?

I guess we could do a string match for LIMA_CIDATA on the expanded script template and generate a warning?

[AkihiroSuda]

I guess we could do a string match for LIMA_CIDATA on the expanded script template and generate a warning?

👍