Re: [Boost-build] address-model as build requirement broken?

Subject: Re: [Boost-build] address-model as build requirement broken?
From: Vladimir Prus (vladimir.prus_at_[hidden])
Date: 2015-08-19 14:43:02

On 08-Aug-15 3:25 PM, John Maddock wrote:
> There seems to be a recent(?) change to boost.build that means adding <address-model> as a build requirement is now broken.
>
> I have an existing Jamfile that basically builds the same program with a bunch of different compiler options and
> compares the results (see https://github.com/boostorg/math/blob/accuracy/reporting/performance/Jamfile.v2), this used to
> work fine, but now I get:
>
> M:/data/boost/boost/tools/build/src/build\feature.jam:782: in expand-composites from module feature
> error: explicitly-specified values of non-free feature <address-model> conflict
> error: existing values: 64 32
> error: value from expanding  <deduced-address-model>32 : 32

I can reproduce this with 'accuracy' branch of Boost.Math.

> To reproduce just add:
>
> <toolset>msvc:<address-model>64
>
> as a build requirement to any otherwise trivial project and try and build with msvc.

That did reproduce the problem for me, i.e.:

        exe first : first.cpp : <toolset>msvc:<address-model>64 ;

works for me, with msvc.

I'll take a closer look at Boost.Math.

Sorry for delay,
Volodya