[expr.rel] It is unclear whether positive zeros compare equal to negative zeros

[eisenwave]

Reference (section label): [expr.rel]

Issue description

The C++ standard acknowledges the existence of positive and negative zeros such as in [atomics.types.float], but does not specify whether these distinct values compare equal. Note that std::signbit lets the user observe a difference between these values.

C23 6.5.8 [Relational operators] paragraph 4 clarifies this:

Positive zeros compare equal to negative zeros.

Suggested resolution

Change in [expr.spaceship] paragraph 4 bullet 3 as follows:

 Otherwise, the operands have floating-point type,
 and the result is of type std​::​partial_ordering.
 The expression a <=> b yields
 std​::​partial_ordering​::​less if a is less than b,
 std​::​partial_ordering​::​greater if a is greater than b,
 std​::​partial_ordering​::​equivalent if a is equivalent to b, and
 std​::​partial_ordering​::​unordered otherwise.
+Positive zeros are equivalent to negative zeros.

Change in [expr.rel] paragraph 6 as follows:

 If both operands (after conversions) are of arithmetic or enumeration type,
 each of the operators shall yield true if the specified relationship is true and
 false if it is false.
+Positive zeros compare equal to negative zeros.

Change in [expr.eq] paragraph 8 as follows:

 If both operands are of arithmetic or enumeration type,
 the usual arithmetic conversions ([expr.arith.conv]) are performed on both operands;
 each of the operators shall yield true if the specified relationship is true and
 false if it is false.
+Positive zeros compare equal to negative zeros.

[t3nsor]

I think we should also acknowledge the possible existence of signed zeroes in [basic.fundamental]

[eisenwave]

I generally agree; it seems like something that could easily be done editorially though. A note ought to do the job.

[jwakely]

I think we should also acknowledge the possible existence of signed zeroes in [basic.fundamental]

Yes, and say whether they exist at all is implementation-defined. Otherwise this could be interpreted to require them (the library wording doesn't require anything, it's only a recommendation saying how to handle them if they exist, and if the library wording needs improvement, that should be done there).

[eisenwave]

After a quick detour through the /draft repo, there is now #819 for clarifying that negative zero exists in the first place.