How to stop scanning the entire workspace for Cargo.toml files?

[guw]

I am setting up rules_rust in a bzlmod workspace.

In my MODULE.bazel I list the Cargo.toml files explicitely.

# Rust toolchain
bazel_dep(name = "rules_rust", version = "0.61.0")

rust = use_extension("@rules_rust//rust:extensions.bzl", "rust")
rust.toolchain(
    edition = "2024",
    versions = ["1.85.0"],
)

# Crate Universe dependencies
crate = use_extension("@rules_rust//crate_universe:extensions.bzl", "crate")
crate.from_cargo(
    name = "crates",
    cargo_lockfile = "//:Cargo.lock",
    manifests = [
        "//:Cargo.toml",
        "//foo/cli:Cargo.toml",
        "//bar/lib:Cargo.toml",
    ],
    supported_platform_triples = [
        "aarch64-apple-darwin",
        "aarch64-unknown-linux-gnu",
        "x86_64-unknown-linux-gnu",
    ],
)
use_repo(crate, "crates")

When I run bazel mod deps (or some other Bazel command) it fails with:

Error: Failed to splice workspace

Caused by:
    0: Failed to walk filesystem finding workspace Cargo.toml files
    1: IO error for operation on ../data/postgres/16: Permission denied (os error 13)
    2: Permission denied (os error 13)

This is expected. The data directory is listed in .bazelignore. It's also not listed in Cargo.toml.

Why is rules_rust trying to scan for Cargo.toml files? I listed all I want explicitly in the manifests list. Is there a way to exclude the data folder from scanning?

[illicitonion]

Hi! Sorry, this sounds frustrating!

The purpose behind the scanning is to make sure your MODULE.bazel is actually up to date - if you add a Cargo.toml file to your workspace, and don't remember to update the workspace Cargo.toml or the MODULE.bazel, we'd otherwise start giving incorrect results.

One way of fixing this would be to start respecting .bazelignore in the scanning (#3425).

Another could be adding a "no_really_trust_me_my_crates_manifest_list_is_up_to_date" option to the repository rule to just disable the scanning entirely.

Yet another could be to make sure the repository rule can properly detect and track the Cargo.toml files based just on the workspace Cargo.toml, and avoid needing to list the transitive ones at all. It's possible we're already doing this (or something approaching it) with this code - if someone can run some experiments where they do a resolve, create a new Cargo.toml file in the workspace, rebuild, and make sure we re-do the resolve with no intervention, we could potentially remove this entirely.

[guw]

The purpose behind the scanning is to make sure your MODULE.bazel is actually up to date - if you add a Cargo.toml file to your workspace, and don't remember to update the workspace Cargo.toml or the MODULE.bazel, we'd otherwise start giving incorrect results.

This is an understandable requirement but the implementation seems to violate a few assumptions/best practices how Bazel works. bazel mod tidy or a specific Gazelle invocation is the recommend best practice here, i.e. that's where I would expect Bazel (and rules) to du such work. But not during regular invocation for a build or test or query. I filed an issue for this request (#3515).

Thanks for providing more context and the links to other issues to track.