Repository navigation
cargo check fires when rustc lints use - instead of _ #17341
Description
Problem
All the names of lints and lint groups use - instead of _ in the output of rustc -Whelp:
[zack@laptop ~]$ rustc --version
rustc 1.99.0-nightly (1a98b1e13 2026-08-07)
[zack@laptop ~]$ rustc -Whelp
Available lint options:
-W <foo> Warn about <foo>
-A <foo> Allow <foo>
-D <foo> Deny <foo>
-F <foo> Forbid <foo> (deny <foo> and all attempts to override)
Lint checks provided by rustc:
name default meaning
---- ------- -------
absolute-paths-not-starting-with-crate allow fully qualified paths that start with a module name instead of `crate`, `self`, or an extern crate name
⋮Despite the "official names" using - instead of _ at least per rustc in -Whelp and diagnostic messages (e.g., note: requested on the command line with `-D unused-results`), cargo check complains that the use of such names is deprecated and won't work in a future edition:
[zack@laptop example]$ cargo check
warning: Cargo.toml: `lints.rust.ambiguous-negative-literals` is deprecated in favor of `lints.rust.ambiguous_negative_literals` and will not work in a future edition
warning: Cargo.toml: `lints.rust.closure-returning-async-block` is deprecated in favor of `lints.rust.closure_returning_async_block` and will not work in a future edition
warning: Cargo.toml: `lints.rust.dead-code-pub-in-binary` is deprecated in favor of `lints.rust.dead_code_pub_in_binary` and will not work in a future edition
warning: Cargo.toml: `lints.rust.deprecated-in-future` is deprecated in favor of `lints.rust.deprecated_in_future` and will not work in a future edition
warning: Cargo.toml: `lints.rust.deprecated-safe` is deprecated in favor of `lints.rust.deprecated_safe` and will not work in a future edition
warning: Cargo.toml: `lints.rust.deref-into-dyn-supertrait` is deprecated in favor of `lints.rust.deref_into_dyn_supertrait` and will not work in a future edition
warning: Cargo.toml: `lints.rust.ffi-unwind-calls` is deprecated in favor of `lints.rust.ffi_unwind_calls` and will not work in a future edition
warning: Cargo.toml: `lints.rust.future-incompatible` is deprecated in favor of `lints.rust.future_incompatible` and will not work in a future edition
warning: Cargo.toml: `lints.rust.impl-trait-redundant-captures` is deprecated in favor of `lints.rust.impl_trait_redundant_captures` and will not work in a future edition
warning: Cargo.toml: `lints.rust.keyword-idents` is deprecated in favor of `lints.rust.keyword_idents` and will not work in a future edition
warning: Cargo.toml: `lints.rust.let-underscore` is deprecated in favor of `lints.rust.let_underscore` and will not work in a future edition
warning: Cargo.toml: `lints.rust.linker-info` is deprecated in favor of `lints.rust.linker_info` and will not work in a future edition
warning: Cargo.toml: `lints.rust.macro-use-extern-crate` is deprecated in favor of `lints.rust.macro_use_extern_crate` and will not work in a future edition
warning: Cargo.toml: `lints.rust.meta-variable-misuse` is deprecated in favor of `lints.rust.meta_variable_misuse` and will not work in a future edition
warning: Cargo.toml: `lints.rust.missing-copy-implementations` is deprecated in favor of `lints.rust.missing_copy_implementations` and will not work in a future edition
warning: Cargo.toml: `lints.rust.missing-debug-implementations` is deprecated in favor of `lints.rust.missing_debug_implementations` and will not work in a future edition
warning: Cargo.toml: `lints.rust.missing-docs` is deprecated in favor of `lints.rust.missing_docs` and will not work in a future edition
warning: Cargo.toml: `lints.rust.non-ascii-idents` is deprecated in favor of `lints.rust.non_ascii_idents` and will not work in a future edition
warning: Cargo.toml: `lints.rust.nonstandard-style` is deprecated in favor of `lints.rust.nonstandard_style` and will not work in a future edition
warning: Cargo.toml: `lints.rust.redundant-imports` is deprecated in favor of `lints.rust.redundant_imports` and will not work in a future edition
warning: Cargo.toml: `lints.rust.redundant-lifetimes` is deprecated in favor of `lints.rust.redundant_lifetimes` and will not work in a future edition
warning: Cargo.toml: `lints.rust.refining-impl-trait` is deprecated in favor of `lints.rust.refining_impl_trait` and will not work in a future edition
warning: Cargo.toml: `lints.rust.rust-2018-compatibility` is deprecated in favor of `lints.rust.rust_2018_compatibility` and will not work in a future edition
warning: Cargo.toml: `lints.rust.rust-2018-idioms` is deprecated in favor of `lints.rust.rust_2018_idioms` and will not work in a future edition
warning: Cargo.toml: `lints.rust.rust-2021-compatibility` is deprecated in favor of `lints.rust.rust_2021_compatibility` and will not work in a future edition
warning: Cargo.toml: `lints.rust.rust-2024-compatibility` is deprecated in favor of `lints.rust.rust_2024_compatibility` and will not work in a future edition
warning: Cargo.toml: `lints.rust.single-use-lifetimes` is deprecated in favor of `lints.rust.single_use_lifetimes` and will not work in a future edition
warning: Cargo.toml: `lints.rust.trivial-casts` is deprecated in favor of `lints.rust.trivial_casts` and will not work in a future edition
warning: Cargo.toml: `lints.rust.trivial-numeric-casts` is deprecated in favor of `lints.rust.trivial_numeric_casts` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unit-bindings` is deprecated in favor of `lints.rust.unit_bindings` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unknown-or-malformed-diagnostic-attributes` is deprecated in favor of `lints.rust.unknown_or_malformed_diagnostic_attributes` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unnameable-types` is deprecated in favor of `lints.rust.unnameable_types` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unreachable-pub` is deprecated in favor of `lints.rust.unreachable_pub` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unsafe-code` is deprecated in favor of `lints.rust.unsafe_code` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unstable-features` is deprecated in favor of `lints.rust.unstable_features` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unused-crate-dependencies` is deprecated in favor of `lints.rust.unused_crate_dependencies` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unused-import-braces` is deprecated in favor of `lints.rust.unused_import_braces` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unused-lifetimes` is deprecated in favor of `lints.rust.unused_lifetimes` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unused-qualifications` is deprecated in favor of `lints.rust.unused_qualifications` and will not work in a future edition
warning: Cargo.toml: `lints.rust.unused-results` is deprecated in favor of `lints.rust.unused_results` and will not work in a future edition
warning: `example` (manifest) generated 40 warningsProposed Solution
Lints should be allowed to have - instead of _ in [lints.rust] and [lints.clippy].
Notes
I believe #17298 is the sudden cause of these I have been informed this is due to #17051.warnings.
Activity
[-]`renamed-and-removed-lints` fires when lints use `-` instead of `_` on `nightly`[/-][+]`cargo check` fires when lints use `-` instead of `_` on `nightly`[/+][-]`cargo check` fires when lints use `-` instead of `_` on `nightly`[/-][+]`cargo check` fires when `rustc` lints use `-` instead of `_` on `nightly`[/+]weihanglo commentedon Aug 8, 2026
In #17298 (comment):
However, down a few lines:
https://rustc-dev-guide.rust-lang.org/diagnostics.html#lint-naming
See also the rationale: #13943 (comment)
zacknewman commentedon Aug 8, 2026
Hm, I don't find that very convincing though. I understand Cargo wants to require
_for its own lints, but I fail to see how that applies to a different table likelints.rust. Additionally, it seems inconsistent that on a package level no hard requirement is made for packages to only contain one or the other; and even more bizarre is the fact that if such a decision were made it'd likely be in favor of-over_based on my experience with package names at least.While I realize one can technically have two
rustclints that only differ by-or_, that's also true for dependencies. Just like dependencies, Cargo transforms the-one to the_one when there are no "duplicates". When there are "duplicates", it doesn't complain (unless they are truly identical) and doesn't transform the-one to the_one forcing one to handle the-one manually by giving it an alias. Because the most popular registry, crates.io, treats packages as the same if they only differ by-and_, it's rare when one has two dependencies inCargo.tomlthat differ by-and_alone. It can still happen though especially with local crates.One difference I imagine is that the use of
-in package names is far more common than people using-forrustclints, so this breaking change is not as impactful. Still sucks that I have to update every crate I've ever written since all of them contain at least onerustclint or lint group with a name that contains-. I chose to reference such lints with-because I thought it would never matter, and I assumed the one with-was "official" seeing how it's the version you always see in diagnostic messages and-Whelp.tgross35 commentedon Aug 26, 2026
I understand that the duplication in #13943 is a problem but I don't think forcing one behavior is a great solution. Why not only warn on duplicates after normalization rather than forcing the underscore style?
Lints are documented using
-, including CLI examples https://doc.rust-lang.org/rustc/lints/groups.html. It's pretty rough that Cargo is deviating here, especially when all other builtin manifest keys use kebab case (dev-dependenciesand others).zacknewman commentedon Aug 26, 2026
Yeah, I really don't understand or agree with this change. It stinks of overreach and adds to the unfortunate reputation that Rustaceans are a little too evangelical. Can you imagine if
rustcwas changed to no longer compile code that contained types whose names didn't conform to the standard style even if such change happened on an edition boundary? That's obviously an extreme analogy, but I also thought this kind of change wouldn't happen. Again, the biggest problem isn't even the warning but the actual deprecation that will occur.I don't even consider duplication a real problem and certainly not a bigger problem than the problem this change creates. As stated multiple times above, there are a bunch of sources that use the
-ed form for lints; so it's quite understandable if one does as well. Does this change how lints can be supplied on the command line? If not, that's inconsistent. Cargo is burying its head pretending to not know what a lint means when it's inCargo.tomlbut magically knows its meaning when passed on the command line.Last, the solution to the "problem" of duplication is far bigger than it needs to be. Why not complain when there actually exists a duplicate? Thus entities that choose to conform with how
rustc, online documentation for lints, etc. spell lints (i.e., using-instead of_) can still do so; and for those that have duplicates, they get the deprecation warning saying all lints that only differ by-and_will be ignored in a future release.[-]`cargo check` fires when `rustc` lints use `-` instead of `_` on `nightly`[/-][+]`cargo check` fires when `rustc` lints use `-` instead of `_` on `beta`[/+]tgross35 commentedon Oct 14, 2026
I'm pretty sure this is referring to snake case for the symbol name, as opposed to
MyLintName. They are still referred to using hyphens in most documentation that isn't a code example.[-]`cargo check` fires when `rustc` lints use `-` instead of `_` on `beta`[/-][+]`cargo check` fires when `rustc` lints use `-` instead of `_`[/+]