There should be a comment somewhere (Integer class comment?) explicitly stating, that you cannot rely on two complement semantics for LargeNegativeIntegers regarding bit logic (bit logic restricted to positive Integers is correct).

I fear, changing the fundamental design decision to use magnitude representation for LargeNegativeIntegers - e.g. by introducing some kind of two complement representation - would imply a lot of work...


Regards,
Stephan


On 26.01.2008 04:12, nicolas cellier wrote:
> 
> SmallInteger < 0 are stored directly in two complement native format 
> (well, except the tag bits)
> 
> But LargeNegativeInteger are stored same as their absolute value.
> 
> This does simplify some algorithms. But this leads to quite tricky ones 
> when bit logic is to be performed based on two complement.
> And whenever tricks are used, bugs frequency generally tends to 
> increase. I believe long life bugs are lying in such swamps.
> 
> That's why i'm very proud of finding accidentaly 
> http://bugs.squeak.org/view.php?id=6874 via 
> http://bugs.squeak.org/view.php?id=6873.
> 
> As usual, most of you won't ever care of it.
> Yeah, one more useless bugfix i am specialized in since the probability 
> you bump into it must not exceed 1.0e-8 per hour of Smalltalking.
> 
> But who knows, it's a kind of insurance in case a silly engineer had the 
> idea to run nearest nuclear plant in Squeak!
> 
> And well, it's also for the beauty of mathematics!
> 
> Nicolas
> 
> 
> 

-- 
Stephan Rudlof (sr at evolgo.de)
   "Genius doesn't work on an assembly line basis.
    You can't simply say, 'Today I will be brilliant.'"
    -- Kirk, "The Ultimate Computer", stardate 4731.3