Problem
.github/workflows/larql-cli.yml contains this comment at line 107–112:
# Clippy is intentionally skipped: as of 2026-05-10 `larql-cli` carries
# ~82 pre-existing errors under default features and ~112 under
# `--no-default-features` (mostly `large_enum_variant` and `dead_code`
# on metal-only paths). Re-enable once that backlog is cleared; the other
# crates' workflows already enforce `clippy -- -D warnings`.
The clippy step has been commented out since at least 2026-05-10. All other crate CI workflows enforce clippy -- -D warnings. larql-cli is the top-level binary crate that aggregates every subcommand; having it unguarded by clippy means regressions can accumulate silently.
Current state
The larql-cli/src/main.rs has crate-wide #![allow(...)] directives for:
clippy::large_enum_variant
clippy::type_complexity
clippy::too_many_arguments
clippy::doc_overindented_list_items
Some warnings are suppressed via #![allow]; others (particularly dead_code on metal-only paths that are #[cfg(target_os = "macos")]) are only visible under --features gpu on macOS.
Steps to clear the backlog
- Run
cargo clippy -p larql-cli --no-default-features --all-targets -- -D warnings and fix all errors.
- Run
cargo clippy -p larql-cli --features gpu --all-targets -- -D warnings on macOS and fix metal-path dead_code.
- Re-enable the clippy step in
.github/workflows/larql-cli.yml.
- Remove
#![allow(...)] directives from main.rs where the underlying issue is genuinely fixed (vs. silenced).
Notes
- The
--all-targets flag is required to include tests and examples; the convert_moe_to_per_layer example is currently excluded from CI due to an API break, which hides its clippy errors.
- The
large_enum_variant warnings on the Commands enum are by-design (the enum drives the main dispatch table and cannot be split without an intermediate boxing pass); document-and-allow is acceptable for that specific case.
Problem
.github/workflows/larql-cli.ymlcontains this comment at line 107–112:The clippy step has been commented out since at least 2026-05-10. All other crate CI workflows enforce
clippy -- -D warnings.larql-cliis the top-level binary crate that aggregates every subcommand; having it unguarded by clippy means regressions can accumulate silently.Current state
The
larql-cli/src/main.rshas crate-wide#![allow(...)]directives for:clippy::large_enum_variantclippy::type_complexityclippy::too_many_argumentsclippy::doc_overindented_list_itemsSome warnings are suppressed via
#![allow]; others (particularlydead_codeon metal-only paths that are#[cfg(target_os = "macos")]) are only visible under--features gpuon macOS.Steps to clear the backlog
cargo clippy -p larql-cli --no-default-features --all-targets -- -D warningsand fix all errors.cargo clippy -p larql-cli --features gpu --all-targets -- -D warningson macOS and fix metal-path dead_code..github/workflows/larql-cli.yml.#![allow(...)]directives frommain.rswhere the underlying issue is genuinely fixed (vs. silenced).Notes
--all-targetsflag is required to include tests and examples; theconvert_moe_to_per_layerexample is currently excluded from CI due to an API break, which hides its clippy errors.large_enum_variantwarnings on theCommandsenum are by-design (the enum drives the main dispatch table and cannot be split without an intermediate boxing pass); document-and-allow is acceptable for that specific case.