Hi Mark,
Mark Kettenis wrote on Fri, Sep 25, 2015 at 11:37:17PM +0200:
> Ingo Schwarze wrote:
>> Mark Kettenis wrote on Fri, Sep 25, 2015 at 02:27:52PM -0600:
>>> CVSROOT: /cvs
>>> Module name: src
>>> Changes by: kette...@cvs.openbsd.org 2015/09/25 14:27:52
>>>
>>> Modified files:
>>> sys/dev/pci/drm: drm_linux.h
>>>
>>> Log message:
>>> Apparently 0 is not a power of 2
>>> (despite <sys/param.h> claiming that it is).
>>> Fixes inteldrm(4) on the GM45 chipset.
>> Heh. It also claims that INT_MIN is a power of 2,
>> which isn't true either...
>>
>> Then again, INT_MIN is weird anyway, given that
>> INT_MIN == -INT_MIN && INT_MIN != 0.
>>
>> Or do you think
>>
>> (x > 0 && (((x) - 1) & (x)) == 0)
>>
>> would be even safer?
> Well, the is_power_of_2() macro in drm_linux.h is now bug-compatible
> with Linux, which is what matters ;).
>
> And perhaps it is intentional that powerof2(0) is true. Even Linux
> has it that way. I'd say this is better left like it is now.
I am not surprised that you chose with deliberation,
which is why i didn't send a patch in the first place...
It just seemed so weird that
(1) is_power_of_2() and powerof2() are intentionally different and
(2) intentionally, neither of them is mathematically correct
for the complete range of whole numbers of all common types
that i thought it might make sense to ask, just to make sure that's
indeed how it's all intended and nothing was accidentally overlooked.
Case closed, thanks :)
Ingo
