Re: [boost] [config] [detail] Macro BOOST_WORKAROUND

Subject: Re: [boost] [config] [detail] Macro BOOST_WORKAROUND
From: Peter Dimov (lists_at_[hidden])
Date: 2015-01-28 10:16:45

Andrzej Krzemienski wrote:
> Hi Everyone,
> I have a couple of concerns regarding macro BOOST_WORKAROUND.

Hi Andrzej,

> 1. Is usage thereof recommended?

Yes, I think so.

> 2. Why is it documented in Boost.Config but defined in Boost.Detail?
> 3. Should it not belong to Boost.Config?

The header is in boost/detail for historical reasons, but it's part of 
Boost.Config, not Boost.Detail.

> 4. Why should I use it rather than manual if (BOOST_INTEL < 551) ?

BOOST_WORKAROUNDs can be deactivated globally with BOOST_STRICT_CONFIG. This 
is useful, f.ex. when testing whether workarounds are necessary on a newer 
compiler version. This doesn't apply to BOOST_INTEL < 551, but in many cases 
before the introduction of BOOST_WORKAROUND, the workarounds did not test 
against a maximum version and were, again for example, activated for all 
MSVC versions, present and future.

Since the path without the BOOST_WORKAROUND is supposed to be the conforming 
one, compiler vendors can use Boost with BOOST_STRICT_CONFIG as a 
conformance test.

> Now, that it is in Boost.Detail, I have to make a dependency on 
> Boost.Detail, and indirectly on Boost.MPL because other components in 
> Boost.Detail do.

Including boost/detail/workaround.hpp is a dependency on Boost.Config, not 
Boost.Detail. That's because the header is actually in 
libs/config/boost/detail/workaround.hpp, not in libs/detail. So it's not a 
problem. See for example

http://www.pdimov.com/tmp/report-develop-c3bb6eb/core.html#config