Boost logo

Boost :

Subject: Re: [boost] [typeof][scope_exit] VS2017 and option /permissive-
From: Daniela Engert (dani_at_[hidden])
Date: 2017-04-15 14:07:32


Am 15.04.2017 um 15:05 schrieb Peter Dimov via Boost:
> Daniela Engert wrote:
> ...
>> I've implemented a possible solution and created two pull-requests:
>> - https://github.com/boostorg/typeof/pull/7
>
> I think that a more principled solution here would be to check
> !defined(BOOST_NO_CXX11_DECLTYPE) at the very start of typeof.hpp (if
> BOOST_TYPEOF_EMULATION is not defined) and then use a generic
> decltype-based implementation. Something like
...
> where decltype.hpp is more or less the same as the suggested
> msvc/typeof_impl_decltype.hpp, but without ensure_obj (shouldn't be
> needed with decltype) and with remove_cv_ref instead of remove_reference.
>
> Except that we don't have remove_cv_ref in TypeTraits, so
> remove_cv<remove_reference>.

Thanks for this suggestion, Peter, I was thinking along the same lines
as well. But for a start, I rather went for the low-risk path affecting
only one compiler. During my experiments to find a fix for my problem I
noticed that the decltype-based implemention required a template type
alias to work thoughout all Boost libraries. Otherwise the compiler
would complain about either a missing 'typename' keyword or a
superfluous one. Thus, as long as there is no better way to deal with
that the first test would have to check for
!defined(BOOST_NO_CXX11_TEMPLATE_ALIASES) as well. I will implement this
in the next couple of minutes and start-off another run of the full
Boost test-suite to see what's happening...

Any comments on that?

Ciao
   Dani

-- 
PGP/GPG: 2CCB 3ECB 0954 5CD3 B0DB 6AA0 BA03 56A1 2C4638C5

Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk