Re: [boost] [ANN] BOOST_FOREACH committed to main CVS

From: David Abrahams (dave_at_[hidden])
Date: 2005-08-26 14:49:55

Stefan Slapeta <stefan_at_[hidden]> writes:

> Eric Niebler wrote:
>> 
>> BTW, is there a guide for adding new libraries to Boost main CVS? I 
>> think I got all the bases covered as far as integration with the main 
>> docs and the regression test, but I'd like to be sure.
>> 
>
> Hi,
>
> http://tinyurl.com/dl8ze
>
> [
>    because of:
>
>    #ifdef BOOST_FOREACH_NO_CONST_RVALUE_DETECTION
>    # error Expected failure : const rvalues disallowed
>    #else
> ]
>
> I think it's not quite correct to fail the test case in this situation. 
> "Failure" means that some functionality of this library doesn't work 
> correctly in this environment, which is obviously not the case because, 
> according to the documentation, FOREACH isn't intended to handle const 
> rvalues on VC (btw, what, if any, is the intended behaviour on a full 
> standard compliant compiler?)
>
> IMO, this test case should do nothing and simply return ok for the 
> affected compilers, but others may have a different opinion.

IMO if the test is going to pass, it should test *something*.  For
example, the parameter library SFINAE tests pass where SFINAE isn't
supported in order to prove that if you use the SFINAE syntax, your
code will still compile and work on that compiler with reduced
functionality.

Technically, though, we should have a separate sfinae_syntax test for
that, I think.  So I guess I agree with what Eric did.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com