Re: [PATCH] Add BUILD_BUG_ON_NOT_POWER_OF_2()

[Stefan Richter <stefanr@s5r6.in-berlin.de>]

Robert P. J. Day wrote:
>   my normal pedantry coming to the surface, but can we at least assume
> that people will use this test to *legitimately* test whether or not
> something is a power of two, and not that there is a single bit set
> (in the case of mask bits where all settings must be mutually
> exclusive)?
> 
>   there's a lot of this sort of thing throughout the kernel:
> 
>     (n) != 0 && (((n) & ((n) - 1))
> 
> but it's sometimes unclear whether someone is testing for a) power of
> two, or b) single bit set.  if you're going to introduce that kind of
> BUILD BUG (which is a good idea), let's try to not immediately abuse
> it semantically. :-)

It's merely about math, not about semantics.  Plus, its application is
restricted to build-time checks (of defined constants) anyway.

Hence I would argue that "check at build time whether a defined
bitmask's Hamming weight is 1; abort build if it isn't" can be
legitimately and sufficiently readably implemented by means of the new
BUILD_BUG_ON_NOT_POWER_OF_2().
-- 
Stefan Richter
-=====-==-=- ---= --===
http://arcgraph.de/sr/