gpl: change size_t to int

[openroad-ci]

Fix bug introduced in PR #9534

[gemini-code-assist]

Code Review

This pull request attempts to fix a bug related to destroyCbkGCell's return value. However, changing the return type to int contradicts the established pattern for size_t indices and static_cast<size_t>(-1) as a failure sentinel, as outlined in our rules. This change also introduces potential narrowing conversions. I've provided a comment suggesting adherence to the existing pattern or a more robust refactoring using std::optional.

[gemini-code-assist]

The rule for NesterovBase::destroyCbkGCell states that its return value, which is an index of type size_t, should be checked against static_cast<size_t>(-1) for failure. Therefore, changing the return type's second element to int to handle a -1 sentinel is not aligned with this established pattern and introduces a potential narrowing conversion (lines 3630 and 3632) where a size_t index is converted to an int. For very large designs where the number of g-cells could exceed INT_MAX, this might lead to data loss or undefined behavior.

If the intention is to use -1 as a sentinel, it should be static_cast<size_t>(-1) when returning a size_t.

A more robust long-term solution would be to refactor the return type to avoid using a sentinel value like static_cast<size_t>(-1). For example, using std::optional<std::pair<odb::dbInst*, size_t>> could provide a safer and clearer interface.

References

1. The NesterovBase::destroyCbkGCell method returns a size_t index, and its failure value is static_cast<size_t>(-1).

[github-actions]

clang-tidy review says "All clean, LGTM! 👍"