Function process_file

scripts/code-verify.py:1677–1800

(path: Path, fix: bool)

def process_file(path: Path, fix: bool) -> tuple[list[Violation], str | None]:
    # Read as bytes first so CRLF detection isn't masked by Python's universal
    # newline translation in text mode — read_text() silently rewrites \r\n
    # and bare \r to \n, hiding the very thing we want to flag.
    raw_bytes = path.read_bytes()
    raw_text = raw_bytes.decode("utf-8")
    raw_lines = raw_text.splitlines()

    first_party = _is_first_party(path)
    is_qml = first_party and path.suffix == ".qml"
    violations: list[Violation] = []
    new_raws: list[str] = list(raw_lines)
    qml_changed = False
    id_blanks: list[int] = []

    fence_mask = _compute_fence_mask(raw_lines)

    # CRLF/CR line endings get normalized to LF on rewrite. `splitlines()`
    # already discards \r\n / \r / \n uniformly, so the round-trip via
    # "\n".join() at the bottom of this function strips them — we just need
    # to register the change so the file actually gets written out.
    crlf_changed = b"\r" in raw_bytes
    if crlf_changed:
        violations.append(
            Violation(path, 1, "line-endings", "CRLF/CR line endings normalized to LF")
        )

    # Comment-style and AI-narration checks apply to first-party C/C++/QML
    # only — vendored sources keep their upstream comment style.
    if first_party and path.suffix in _BRACE_FREE_SUFFIXES:
        violations.extend(find_comment_style_violations(raw_lines, path, fence_mask))
        violations.extend(find_ai_narration_violations(raw_lines, path, fence_mask))
        violations.extend(find_non_ascii_violations(raw_lines, path, fence_mask))
        violations.extend(
            find_qml_inline_comment_violations(raw_lines, path, fence_mask)
        )
        violations.extend(
            find_qml_underscore_property_violations(raw_lines, path, fence_mask)
        )
        violations.extend(find_interrupt_guard_violations(raw_lines, path, fence_mask))
        violations.extend(find_todouble_violations(raw_lines, path, fence_mask))

        # Static-analysis rules (Qt/C++ semantic checks + QML conventions).
        # The rules module degrades gracefully when tree-sitter is missing.
        if _SEMANTIC_RULES is not None:
            for f in _SEMANTIC_RULES.analyze(path, raw_text, fence_mask):
                violations.append(Violation(path, f.line, f.kind, f.message))

    # Christmas-tree property sort + id-placement check are QML-specific.
    if is_qml:
        lines = tokenize(raw_lines)
        id_violations, id_blanks = check_id_placement(lines, path)
        violations.extend(id_violations)

        runs = find_property_runs(lines)
        sortable_runs: list[tuple[int, int]] = []
        for start, end in runs:
            run_lines = lines[start:end]