Bug #14014: NaN.finite?

• Status: Closed
• Assignee: -
• Target version: 2.5
• ruby -v: 2.4.2
• Backport: 2.3: DONTNEED, 2.4: DONE

[marcandre (Marc-Andre Lafortune)]

Ruby gives contradictory answers for NaN:

> (0/0.0).finite?
 => false 
> Complex(0/0.0, 0).finite?
 => true 

Note that (0/0.0).infinite? returns nil, so the float answer of false looks even more wrong.

The two solutions I see are either changing Float#finite? to return true for NaN, or to raise in both cases.

I'd lean towards raising in both cases, as NaN can not be said to be finite or infinite

[nobu (Nobuyoshi Nakada)]

• Status changed from Open to Rejected

NaN is not a (either infinite or finite) number, definitely.
Also it is not positive nor negative, so its signedness is nonsense too.
Complex(0/0.0, 0).finite? returns false in the trunk already.

[Eregon (Benoit Daloze)]

Has this been backported to the 2.4 branch? I think it should.
What commit fixed it?

[nobu (Nobuyoshi Nakada)]

Not yet, it's r60040.

[marcandre (Marc-Andre Lafortune)]

Thanks, sorry I should have tested trunk.

[Eregon (Benoit Daloze)]

Should this be marked as closed rather than rejected and a test be added for NaN for the backport to be performed then?

[nobu (Nobuyoshi Nakada)]

• Status changed from Rejected to Closed
• Backport changed from 2.3: UNKNOWN, 2.4: UNKNOWN to 2.3: DONTNEED, 2.4: REQUIRED

[nagachika (Tomoyuki Chikanaga)]

• Backport changed from 2.3: DONTNEED, 2.4: REQUIRED to 2.3: DONTNEED, 2.4: DONE

ruby_2_4 r62173 merged revision(s) 60040,60188.