What causes "❌ line({node.lineno}): {message}" in vllm-project/vllm?

Open the file at the reported line number and fix the flagged construct (move logic out of the @config class body, keep only field/schema definitions).

vllm-project/vllm · warning · ValueError

❌ line({node.lineno}): {message}

Error message

❌ line({node.lineno}): {message}

What it means

tools/pre_commit/validate_config.py walks the AST of files under vllm/ and tests/ that use the @config decorator and enforces rules on config classes (e.g. disallowed statements in decorated classes). fail() raises ValueError with the offending line number; main() prints it and exits 1, failing the pre-commit run.

Source

Thrown at tools/pre_commit/validate_config.py:150


def validate_file(file_path: str):
    try:
        print(f"Validating {file_path} config dataclasses ", end="")
        with open(file_path, encoding="utf-8") as f:
            source = f.read()

        tree = ast.parse(source, filename=file_path)
        validate_ast(tree)
    except ValueError as e:
        print(e)
        raise SystemExit(1) from e
    else:
        print("✅")


def fail(message: str, node: ast.stmt):
    raise ValueError(f"❌ line({node.lineno}): {message}")


def main():
    for filename in sys.argv[1:]:
        # Only run for Python files in vllm/ or tests/
        if not re.match(r"^(vllm|tests)/.*\.py$", filename):
            continue
        # Only run if the file contains @config
        with open(filename, encoding="utf-8") as f:
            if "@config" in f.read():
                validate_file(filename)


if __name__ == "__main__":
    main()

View on GitHub (pinned to c794754062)

Solutions

Open the file at the reported line number and fix the flagged construct (move logic out of the @config class body, keep only field/schema definitions).
Run `pre-commit run validate-config --files <yourfile>` locally to iterate quickly before pushing.
Mirror the style of existing @config classes in vllm/config.py (declarative fields only).

Example fix

# before (vllm/my_config.py)
@config
class MyConfig:
    total = a + b  # ❌ line(8): dynamic statement in @config class

# after
@config
class MyConfig:
    a: int = 1
    b: int = 2

Defensive patterns

Strategy: try-catch

Validate before calling

# Fast local gate before committing @config changes
import subprocess, sys
rc = subprocess.run(["python", "tools/pre_commit/validate_config.py", "vllm/config.py"]).returncode
if rc != 0:
    raise SystemExit("Fix @config violations before committing")

Try / catch

try:
    validate_file(path)  # tools/pre_commit/validate_config.py
except ValueError as e:
    print(e)  # message carries the offending line number
    raise SystemExit(1)

Prevention

Run the validate_config pre-commit hook locally before every push touching @config classes.
Keep @config class bodies declarative; put logic elsewhere.
Copy the structure of existing config classes in vllm/config.py.

When it happens

Trigger: Adding or editing a class decorated with @config in vllm/ or tests/ in a way the validator forbids — e.g. statements inside the config class body that the AST validator rejects — then running pre-commit (validate_config hook) on that file.

Common situations: Contributing a new config dataclass to vllm/config.py and writing logic (assignments, imports, function defs) the validator prohibits; refactoring an existing @config class so it trips a rule that previously passed.

Related errors

AI-assisted analysis of vllm-project/vllm@c794754062 (2026-08-14). Data as JSON: /api/errors/e4bc54068b9bbfba. Report an issue: GitHub.