Re: [boost] [type_traits] is_list_constructible regression failures

Subject: Re: [boost] [type_traits] is_list_constructible regression failures
From: Peter Dimov (lists_at_[hidden])
Date: 2018-03-12 19:50:21

John Maddock wrote:
> On 12/03/2018 19:07, Lorenzo Caminiti via Boost wrote:
> > Hello all,
> >
> > I noticed the following failure in Boost.TypeTraits regression tests on 
> > GCC 4.6 C++11. Any idea why this is failing?
>
> There's something different about that tester - there are several gcc-4.6 
> test runners in the matrix and they're all passing
is_list_constructible_test bar that one.  What/why the difference is there I 
have no idea :(

It's possible that this is the only 4.6 tester that uses -std=c++0x. 
is_list_constructible has

#if defined(BOOST_NO_SFINAE_EXPR) || 
defined(BOOST_NO_CXX11_VARIADIC_TEMPLATES) || 
defined(BOOST_NO_CXX11_DECLTYPE) || 
defined(BOOST_NO_CXX11_UNIFIED_INITIALIZATION_SYNTAX) || 
defined(BOOST_NO_CXX11_FUNCTION_TEMPLATE_DEFAULT_ARGS)

but that's apparently not enough to disable it on 4.6. (It's probably the 
derivation from `decltype` that confuses the compiler.)

Travis does have 4.4 and 4.6, so we may want to add these to Type Traits.

https://github.com/boostorg/smart_ptr/commit/77a35856c6f51176dc15c00f0e4ca95097f803ff