Re: [Boost-users] boost::ptr_vector null check bloat

From: Miles Bader (miles.bader_at_[hidden])
Date: 2007-09-20 09:43:18

Thorsten Ottosen <thorsten.ottosen_at_[hidden]> writes:
> Just about any boost library has the implicit assumption that assertions 
> are compiled away *if the optimal efficiency/smallest code is to be gained*.
...
> Anyway, to justify why assertions should be disabled in release mode,
> I recommed:
>
> http://www.artima.com/cppsource/deepspace.html
>
> In particular, read the section on "The Principle of Removability".

Well I certainly agree with "The Principle of Removability" (who
wouldn't?).  However, there seems to be an implicit assumption that
software is always in one of two states:  "debugging/development" (where
speed is not that important, and extra careful checking is needed) and
"release" (where speed is very important, and extra careful checking
would be wasted or even harmful).

I find that for much of the software I write, that isn't an accurate
model -- I need speed, but I don't want to forgo debuggability/checking
in cases where the speed hit is not excessive.  Because of this, I think
it's useful to have more than just "all on" or "all off", because the
cost of various debugging code varies a lot...

Thanks,

-Miles

-- 
We have met the enemy, and he is us.  -- Pogo