remove unnecessary semicolon which breaks zsh on mac

[craigbox]

zsh, in many default configurations, escapes special characters on paste. The ; in the Gateway API example is replaced with \; which causes the command to fail.

As it's not a load-bearing semi-colon, lets just remove it.

Fixes #16162

[craigbox]

you're kidding me. it is a load bearing semi-colon?

[craigbox]

Now we don't use kustomize should I just nuke the braces?

[kfaseela]

saw the lint failure only after approving :D

[craigbox]

note to self, make Faseela an approver for Ukranian too

[bleggett]

you're kidding me. it is a load bearing semi-colon?

It's required for bash, we are using shellcheck in bash syntax mode, and the snippet is bash syntax.

The actual way to handle stuff like this is have a syntax=sh (POSIX sh subset only) and have shellcheck validate POSIX sh rules versus defaulting to bash syntax (which lots of shells emulate, but not exactly).

POSIX sh is the actual "common" subset to target if you want stuff that runs in any shell.

[craigbox]

I don't think it's "required for bash" in the sense that things work if you remove it, but only because the only remaining command is a }, which is treated as a literal.

I also think our shellcheck scripts aren't smart enough to distinguish between POSIX sh and bash at this point?

Should we go for the trivial fix of changing the command?

kubectl get crd gateways.gateway.networking.k8s.io &> /dev/null || kubectl apply -f https://github.com/kubernetes-sigs/gateway-api/releases/download/v1.2.0/standard-install.yaml

...seems like it will do what we want?

[ilrudie]

yeah, It seems to me that something like this should work

{{< text syntax=bash snip_id=install_crds >}}
$ kubectl get crd gateways.gateway.networking.k8s.io &> /dev/null || \
kubectl apply -f https://github.com/kubernetes-sigs/gateway-api/releases/download/{{< k8s_gateway_api_version >}}/standard-install.yaml
{{< /text >}}

[bleggett]

I don't think it's "required for bash" in the sense that things work if you remove it, but only because the only remaining command is a }, which is treated as a literal.

Right - my point is that we're doing bash-centric linting and CI, so we should expect breakage for non-bash, because bash the subdialect isn't really a standard outside of bash the shell.

I also think our shellcheck scripts aren't smart enough to distinguish between POSIX sh and bash at this point?

Fair enough - shellcheck definitely can distinguish, but our stuff wrapping it may not be able to.