Skip to content

COMPARISON.md

Latest commit

 

History

History
162 lines (119 loc) · 5.43 KB

COMPARISON.md

File metadata and controls

162 lines (119 loc) · 5.43 KB
Raw

Clippy Lint Configuration Comparison Report

Executive Summary

This report compares the two most comprehensive Rust Clippy lint configurations found:

  1. OpenAgentsInc/openagents - The "viral tweet" config from @notnotstorm (enhanced version)
  2. Mockapapella/tenex - Self-described "every single restriction" config

Finding: Neither config is a strict superset of the other. Each has valuable lints the other is missing. The combined "best of both" config is what rust-magic-linter provides.

Side-by-Side Comparison

Lint Groups Enabled

Lint Group OpenAgentsInc Tenex Winner
all not set deny Tenex
pedantic warn deny Tenex
nursery not set deny Tenex
cargo not set deny Tenex

Tenex wins on breadth - it enables more lint groups and at deny level.

Lints in OpenAgentsInc MISSING from Tenex

These are critical lints for AI agents that Tenex lacks:

# === ASYNC SAFETY (Critical for modern Rust) ===
await_holding_lock = "deny"      # Prevent deadlocks in async code
# This is CRITICAL - async code with locks is a common footgun

# === PROCESS/MEMORY SAFETY ===
exit = "deny"                    # Library code should never call exit()
mem_forget = "deny"              # std::mem::forget can leak resources

# === STACK/PERFORMANCE ===
large_stack_arrays = "warn"      # Catch huge stack allocations
panic_in_result_fn = "warn"      # Functions returning Result shouldn't panic

# === OUTPUT CONTROL ===
print_stdout = "deny"            # No println! in library code
print_stderr = "deny"            # No eprintln! in library code
# These force structured logging - important for production

Why these matter for AI agents:

  • await_holding_lock - AI agents write async code frequently and this is a subtle bug
  • print_stdout/stderr - Forces agents to use proper logging instead of debug prints
  • panic_in_result_fn - Catches when agent writes fn foo() -> Result<...> but panics inside
  • exit - Prevents agents from adding std::process::exit() which breaks composability

Lints in Tenex MISSING from OpenAgentsInc

Tenex has these that the tweet config lacks:

# === ADDITIONAL LINT GROUPS ===
all = { level = "deny", priority = -1 }      # Base clippy lints
nursery = { level = "deny", priority = -1 }  # Experimental lints
cargo = { level = "deny", priority = -1 }    # Dependency/manifest hygiene

# === ALLOW ATTRIBUTE CONTROL ===
allow_attributes_without_reason = "deny"     # If you DO allow, document why

Rust Lints (Beyond Clippy)

Tenex has comprehensive [lints.rust] that OpenAgentsInc lacks:

[lints.rust]
unsafe_code = "forbid"                    # No unsafe anywhere
warnings = "deny"                         # All warnings are errors
rust_2018_idioms = { level = "deny", priority = -1 }
future_incompatible = { level = "deny", priority = -1 }
nonstandard_style = { level = "deny", priority = -1 }
missing_docs = "deny"                     # Public items must be documented
missing_debug_implementations = "deny"    # Types must impl Debug
missing_copy_implementations = "deny"     # Types that can be Copy should be
non_ascii_idents = "forbid"               # No unicode in identifiers

[lints.rustdoc]
all = { level = "deny", priority = -1 }   # Documentation must be valid

OpenAgentsInc only has:

[workspace.lints.rust]
missing_docs = "allow"        # Explicitly allows missing docs
unexpected_cfgs = "allow"

What's Unique to Each Config

OpenAgentsInc's Unique Value

  1. Async-specific guards (await_holding_lock, exit, mem_forget)
  2. Output control (print_stdout, print_stderr)
  3. Stack safety (large_stack_arrays, panic_in_result_fn)
  4. Curated pedantic overrides with detailed explanations

Tenex's Unique Value

  1. Full lint group coverage (all, nursery, cargo at deny level)
  2. Comprehensive Rust lints (beyond clippy)
  3. Documentation enforcement (missing_docs, missing_debug_implementations)
  4. Stricter overall (deny vs warn)

Recommendation for AI Agent Workflows

For maximum AI guardrails, prioritize these categories:

Must-Have (Prevent Bad Code)

unwrap_used = "deny"           # No hidden panics
expect_used = "deny"           # No hidden panics
panic = "deny"                 # No explicit panics
allow_attributes = "deny"      # Can't silence warnings
todo = "deny"                  # No incomplete code
unimplemented = "deny"         # No incomplete code
dbg_macro = "deny"             # No debug leftovers

Should-Have (Prevent Subtle Bugs)

await_holding_lock = "deny"    # Async deadlock prevention
panic_in_result_fn = "deny"    # Logic error detection
exit = "deny"                  # Composability
mem_forget = "deny"            # Resource leak prevention

Nice-to-Have (Code Quality)

print_stdout = "deny"          # Proper logging
print_stderr = "deny"          # Proper logging
nursery = "warn"               # Catch more issues
missing_docs = "warn"          # Force documentation

The Combined Configuration

rust-magic-linter's standard preset combines the best of both:

  • OpenAgentsInc's async safety and output control
  • Tenex's comprehensive lint groups
  • Carefully curated relaxations for practical use

See skills/rust-magic-linter/assets/ for the complete configurations.