Allow a no-cargo setup for bzlmod

[dzbarsky]

Usage looks like so:

crate = use_extension("@rules_rust//crate_universe:extension.bzl", "crate")
crate.spec(package = "anyhow", version = "1.0.77")
....
crate.from_specs(name = "crates")

It might make sense to merge the annotation attributes into the spec so we don't have to duplicate things, but we can probably iterate on this in the future, this API is still experimental, yeah?

[dzbarsky]

@matts1 You're probably best-equipped to review this if you have the cycles

[matts1]

A few nitpicks, but generally looks good to me

[matts1]

Nit: I've recently been made aware of a much better way to do this.

Can you add bazel_dep(name = "aspect_bazel_lib", version = "1.36.0") to MODULE.bazel, then add the following to the build file:

load("@aspect_bazel_lib//lib:run_binary.bzl", "run_binary)
load("@bazel_skylib//rules:diff_test.bzl", "diff_test")

run_binary(
  name = "run_hello_world",
  tool = ":hello_world",
)

diff_test(
    name = "hello_world_diff_test",
    file1 = ":run_hello_world",
    file2 = "testdata/hello_world.out",
)

[dzbarsky]

This would need more changes to the binary to direct the output to a file, or another wrapper. But really I don't even need to run it, I'll just switch this to a build_test. As long as it compiles we're ok :)

[matts1]

That's a good point, sounds good to me. Don't know if we even need a build_test. bazel test :all actually builds all non-test targets as well IIRC.

[dzbarsky]

that's true, I'll just leave the rust_binary

[matts1]

Nit: Can we make most of the files in the directory a symlink to examples/bzlmod/hello_world/*? That should help us keep them in sync - the only file that should actually be different is MODULE.bazel, but the rest should be the same.

[dzbarsky]

I added --noenable_workspace to the .bazelrc, so now we only have BUILD.bazel, MODULE.bazel, .bazelrc and src/main.rs. The last one is the only one that is the same. I tried symlinking it but Bazel rust_binary seems to not pick up the symlinked source file. Given the source is pretty trivial, it didn't seem worth poking at it more, but let me know if you feel strongly!

[matts1]

That's very curious, but yeah, don't think it's worth digging into.

[matts1]

Nit: Looks like a lot of duplicated code between this and the above section. Could we pull the common stuff into a function?

[dzbarsky]

good call, I rearranged this a bit to share the repo validation. So the duplication now is the 3 lines for annotations. I'm hesitating a bit because sharing this either requires using a lambda or a function that takes module_annotations, repo_specific_annotations, and cfg.name. Alternately, it can look like this_repo_annotations = update_annotations(module_annotations, repo_specific_annotations.get(cfg.name, {})) or something.

Do you still think it's worth pulling out or keep it as-is?

[matts1]

I quite like the repo_specific_annotations idea. I think that'd probably be the cleanest.

[dzbarsky]

I factored that out, PTAL

[matts1]

Nit: This looks like it needs to be updated

[dzbarsky]

yep good catch, ty!

[dzbarsky]

@UebelAndre @illicitonion can we merge this?

[UebelAndre]

Thanks! A couple of nits and can you add a build in CI that runs on this workspace? Or some kind of regression test somewhere else?

I'd also love to know if @matts1 approves too. (They do!)

[dzbarsky]

Thanks! A couple of nits and can you add a build in CI that runs on this workspace? Or some kind of regression test somewhere else?

Oh I had assumed we were already testing the examples in CI. Should we fix that instead?

[UebelAndre]

Oh I had assumed we were already testing the examples in CI. Should we fix that instead?

A new workspace requires a new job. Had this been an expansion of an existing case the it would have been picked up

[dzbarsky]

Oh I had assumed we were already testing the examples in CI. Should we fix that instead?

A new workspace requires a new job. Had this been an expansion of an existing case the it would have been picked up

Ah I see, the CI framework makes it hard to iterate over all the example workspaces, fair enough. Updated with all your comments, PTAL!

[UebelAndre]

Nice! I had just one more request to fix the .bazelignore file and then I think it's good to go!

[UebelAndre]

Thank you!