Re: [boost] metactrl header [compile time asserts]

From: Beman Dawes (beman_at_[hidden])
Date: 2000-07-26 16:00:45

John Maddock wrote:

 >... once I had figured out that
 >"sizeof(incomplete_type)" gave useful error messages, I had to find a way
 >to wrap this up, originally I just assigned this to an enum, but Steve
 >wasn't so happy with this and prefered a typedef, so we added the
 >ct_assert_test just as a way of converting an 
integral-constant-expression
 >to a type.

Ah! Well, it really does need documenting, the rationale for the code 
choice isn't at all obvious (at least to me)>

I tend to like the enum idea.  Whether a typedef is fully evaluated is a 
bit tricky; when I was testing last fall I kept hitting cases where at 
least some compilers didn't evaluate an unreferenced typedef.  Of course 
now I can't find the particular wording in the standard.

 >BTW there is some difference of opinion on names here:  I tend to prefer
 >something with "assert" in the title, Steve (and others I think) have 
gone
 >for "postulate", any preferences?

"postulate" sounds a bit pompous.  How about "require"?

--Beman