[ARROW-12768] [C++] Add sign bit checks to floating-point arithmetic kernels tests

• Type: Test
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Fix Version/s: 7.0.0
• Component/s: C++
• Labels: pull-request-available
• External issue URL: https://github.com/apache/arrow/issues/28509
• Assignee: Antoine Pitrou
• Reporter: Eduardo Ponce
• Created: 12/May/21 22:56
• Updated: 11/Jan/23 08:28
• Resolved: 13/Dec/21 18:53

Description

There are special cases in arithmetic kernels with floating-point arguments that can result in signed values which are equivalent to their unsigned counterpart when compared with the equality operator. To make the comparisons correct, one needs to also compare the sign bit. For example, -0.0 == 0.0 is true while std::signbit(-0.0) == std::signbit(0.0) is false. This issue also applies to Inf and NaN values.

Issue Links

• links to: GitHub Pull Request #11940