Deprecate accepting out of range values for unsigned integers in PyArg_Parse

Feature or enhancement

For unsigned integer formats (B, H, I, k, K) in PyArg_Parse, there is no overflow check. If the Python integer value is out of range of the corresponding C type, only the lowest bits are stored, and the higher bits are silently dropped. There are two reasons for this:

• The signess of the C type accepted by the C API (like uid_t, etc) is not always known. We need to accept values that fit in corresponding both signed and unsigned C types.
• Even if the type is unsigned, some small negative integers (casted to the unsigned tyep) can have special meaning. It is convenient if we can pass -1 instead of 0xffff_ffff or 0xffff_ffff_ffff_ffff, depending on platform.

But values that cannot be represented neither in unsigned not in signed C type, are obviously invalid. Silently accepting them can provoke bugs.

At first, I propose to emit a deprecation warning for such values. Later they will became errors.

Linked PRs

• gh-132630

先閱讀 PyArg_Parse 對無號整數格式 B、H、I、k 和 K 的處理，然後檢視相關的 PR gh-132630，了解已在進行的工作。當既不適合有號 C 型別也不適合無號 C 型別的超出範圍值發出所提議的棄用警告，同時受支援的負值保留其預期行為時，即表示完成。