Hey Lucas! seems a bit wrong isn't it ? ... maybe some test-cases would help :

2.3.0 :002 > 112.bit_length
 => 7 
2.3.0 :003 > -112.bit_length
 => 7 

@kares

>> 112.bit_length
=> 7
>> -112.bit_length
=> 0

It seems to be working for me - I'll send a PR with tests to rubyspecs.

-112 should not have a bit length of 0 ... if it does at your MRI its platform dependent or a bug.
... if you think of it it would be a pretty weird outcome to expect all < 0 to have a bit length of 0

You're right, it seems pretty weird - I was assuming MRI was right - my bad. Done @kares

Great, thanks Lucas.

Thanks for improvement, but for some values still return different value from MRI.

In MRI implementation uses bit inversion, (-(2**x)).bit_lengthbecomes x in MRI (and x + 1 for this implementation)

@jkr2255 It will be fixed by #3833