babyworm/rtl-agent-team/lint-tool-profiles

Lint Tool Profiles

Common Contract

• Prefer wrapper: lint/scripts/run_lint.sh
• All profiles must normalize output into:
• errors, warnings, tool, log_path, replay_path
• Gate decision is based on normalized errors == 0

Open-Source Baseline

• verilator:
• lint/scripts/run_lint.sh --tool verilator -f rtl/filelist_top.f --outdir lint/lint
• verible:
• lint/scripts/run_lint.sh --tool verible -f rtl/filelist_top.f --outdir lint/lint
• slang:
• lint/scripts/run_lint.sh --tool slang -f rtl/filelist_top.f --outdir lint/lint
• Auto-detects RTL vs TB based on source file paths:
• RTL (rtl/): runs with -Weverything for maximum strictness — catches always_ff multi-driver violations (VCS ICPD), uninitialized variables, width mismatches, etc.
• TB (sim/): runs with --allow-dup-initial-drivers — permits initial + always_ff on same signal (common testbench pattern)
• Only slang catches IEEE 1800 §9.2.2.4 multi-driver violations; Verilator and Verible do not

Commercial Profile

• spyglass (binary: sg_shell, config key: sg_shell, script flag: --tool spyglass):
• lint/scripts/run_lint.sh --tool spyglass --top <top> -f rtl/filelist_top.f --outdir lint/lint
• Note: rat_config.json uses sg_shell as tool key; runner scripts accept --tool spyglass

Severity Mapping

• Tool error/fatal -> normalized error
• Tool warning -> normalized warning
• Unknown parse line -> keep raw log reference and mark parse caution