[conv.rank] Issues in defining integer conversion rank

[Halalaluyafail3]

Full name of submitter (unless configured in github; will be published with the issue): Jay Ghiron

Reference (section label): [conv.rank]/1

Issue description:

The first rule defining that no two signed integer types have the same conversion rank says that char is a signed integer type:

No two signed integer types other than char and signed char (if char is signed) have the same rank, even if they have the same representation.

However, char is never a signed integer type even if it is signed. Even if it was a signed integer type, wchar_t can be signed too so it should be mentioned too.

The rule defining the conversion rank of bool allows it to have a greater rank than extended integer types, even those with a width greater than one.

The rank of bool is less than the rank of all standard integer types.

It seems unintended to allow this as a possibility. C has adjusted its wording recently to require bool have the lowest conversion rank, see N3747.

Suggested resolution:

Modify [conv.rank]/1.1:

No two signed integer types other than char and signed char (if char is signed) have the same rank, even if they have the same representation.

Modify [conv.rank]/1.7:

The rank of bool is less than the rank of all standardother integer types.

Add a new footnote which is linked to at the end of [conv.rank]/1.1:

char and wchar_t are not signed integer types, even if they are signed.

[frederick-vs-ja]

Perhaps it would be more consistent to also expand [conv.rank] p1.8 as:

(1.8) The ranks of char, char8_t, char16_t, char32_t, and wchar_t equal the ranks of their underlying types ([basic.fundamental]).

I wonder whether we should add a note. If we add a note, perhaps it would be more consistent to say "bool, char8_t, char16_t, and char32_t are not unsigned integer types even though their underlying types are" (note that there're some differences between C and C++).

[Halalaluyafail3]

Perhaps it would be more consistent to also expand [conv.rank] p1.8 as:

(1.8) The ranks of char, char8_t, char16_t, char32_t, and wchar_t equal the ranks of their underlying types ([basic.fundamental]).

Considering [conv.rank]/1.6 says "The rank of char equals the rank of signed char and unsigned char" it seems redundant to restate that. A simpler approach to connect them would be to swap [conv.rank]/1.6 and [conv.rank]/1.7. Alternatively since all character types would be included with what you suggested, it could be modified like so:

The ranks of char8_t, char16_t, char32_t, and wchar_tthe character types equal the ranks of their underlying types ([basic.fundamental]).)

This would be redundantly specifying the rank of char, but not in such an obvious way.

I wonder whether we should add a note. If we add a note, perhaps it would be more consistent to say "bool, char8_t, char16_t, and char32_t are not unsigned integer types even though their underlying types are" (note that there're some differences between C and C++).

Considering the text ended up being incorrect, I think it is worth specifying so there is no confusion.