Elad Efrat wrote:
> Darren Reed wrote:
>> Elad Efrat wrote:
>>> Darren Reed wrote:
>>>
>>>> Elad, how would you have solved this problem and also
>>>> achieved the side benefit of opening up use of NGROUPS
>>>> to become a tunable rather than a constant?
>>>
>>> since when is the 16bit group limit that horrible?
>>
>> It isn't so much that it is a 16bit limit, but that it is a fixed 
>> limit at
>> compile time.
>
> when was this limit introduced?

It doesn't matter when the limit was introduced,
what matters is architecting a solution that does
not have it.


>>> the two issues are completely orthogonal. the change was done 
>>> originally
>>> to save a malloc. that is a very, very weak argument. the "side 
>>> benefit"
>>> has no effect whatsoever and you know that very well.
>>
>> Well, how would you have done it better?
>
> how about you first revert the changes?

Do you have fixes or changes that solve the problem using the code
base the way it was before these changes were made?  If you do,
lets see it so we can discuss their merits vs what is there today.

Darren