install: Compile (but error at runtime) on non-x86_64/aarch64

[jmarrero]

We already are checking for target_arch when using the constants. Just instantiating them when we are in a target arch makes the build fail when we are building on a arch that that hits the constraint we created for initiating the constant.

closes #112

[openshift-ci]

Skipping CI for Draft Pull Request.
If you want CI signal for your change, please convert it to an actual PR.
You can still manually trigger a test run with /test all

[jmarrero]

draft while I test this locally and a debug pod. This is just a theory atm... I can't see anything else that could be wrong.

[jmarrero]

It builds ok locally and I started up a s390x debug pod, ran:

cosa shell
sudo dnf install openssl-devel glib2-devel ostree-devel
cargo build

and got a good build:

   Compiling bootc v0.1.0 (/home/jenkins/agent/workspace/debug-pod/bootc/cli)
    Finished dev [optimized + debuginfo] target(s) in 2m 26s
[coreos-assembler]$ uname -a
Linux 1830b359aaec 6.3.12-200.fc38.s390x #1 SMP Thu Jul  6 03:39:39 UTC 2023 s390x GNU/Linux

[cgwalters]

A procedural note, we don't usually include the file suffix ".rs" in the topic prefix. This is about the install code (module). And it's usually best in the subject message to explain a bit more what the goal is. Something like:

install: Compile (but error at runtime) on non-x86_64/aarch64

This all said, while I'm not opposed to this fix...thinking about this a bit more, what seems it'd be better (and about equally easy) is to change the build configuration in Cargo.toml to only enable the install feature on x86_64 and ppc64le.

Then we can drop many of the architecture conditionals in that code. Conceptually we'd be going from bootc install on those arches saying "This isn't implemented" to "command not found" which I think is probably more right.

Hmm...I'm not totally sure Cargo has support for conditionally enabling a crate feature for a given arch (without resorting to a build script). It does support this for dependencies https://doc.rust-lang.org/cargo/reference/specifying-dependencies.html#platform-specific-dependencies

Well...in the end, let's just do some cleanup for this commit message and merge!

(Of course the real fix is to actually implement the install flow for those arches, but that's more involved)

[cgwalters]

BTW just for reference this install logic is a Rust rewrite of https://github.com/coreos/coreos-assembler/blob/main/src/create_disk.sh

[jmarrero]

Thanks, yeah my main intent here is just to not break rpm-ostree builds by vendoring bootc for now. I found https://github.com/coreos/coreos-assembler/blob/main/src/create_disk.sh when looking for EFIPN which I am still not sure what it is 🥲