RE: [boost] alignment?

From: scleary_at_[hidden]
Date: 2001-05-04 08:23:58

> -----Original Message-----
> From: Jens Maurer [mailto:Jens.Maurer_at_[hidden]]
> Sent: Friday, May 04, 2001 2:54 AM
> To: boost_at_[hidden]
> Subject: Re: [boost] alignment?
> 
> 
> John Max Skaller wrote:
> 
> >         Since the calculation isn't definite, I'd prefer to see
> > a constant
> [...]
> >         boost::max_align
> > 
> > included in the library somewhere.
> 
> I agree with that.  For computing that value at run-time (and thus
> possibly validating in the regression tests if this stands a chance
> of being correct, could we possibly exploit section 5.3.4p10 of the
> standard?  Implementations have to keep the size of an array around
> somewhere, so it's likely that the "integral multiple" the paragraph
> talks about is actually not 0.

Sorry if I'm dense, but I can't see how 5.4.3p10 can be used to get a
max_align at all -- I can see how one could use it to get the alignment of a
particular type, but we already have that in alignment_of.  Am I overlooking
something?  (It wouldn't be the first time! :)

        -Steve