Re: [Boost-testing] [XL C/C++][IBM_Canada_Ltd][serialization][multi_index] strange error with boost/serialization/force_include.hpp

Subject: Re: [Boost-testing] [XL C/C++][IBM_Canada_Ltd][serialization][multi_index] strange error with boost/serialization/force_include.hpp
From: Robert Ramey (ramey_at_[hidden])
Date: 2009-03-30 12:18:56

joaquin_at_[hidden] wrote:
> Hi,
>
> Currently, a serialization test for Boost.MultiIndex is failing for XL
> C/C++ at
> IBM_Canada_Ltd regression runner:
>
> Compile [2009-03-29 15:21:08 UTC]: fail
>
>     xlC -c   -DBOOST_ALL_NO_LIB=1 -qcpluscmt -qNOOPTimize
>   -qnoinline -g -qfullpath -qfuncsect -qeh -qrtti -I".."  -o
>
> "/nfs/sparky/home/montana/boost_NEW/boost-CVS/101.aix/results/boost/bin.v2/libs/
>
> multi_index/test/test_serialization.test/vacpp/debug/test_serialization1.o"
> "../libs/
>   multi_index/test/test_serialization1.cpp"
>
> [...]
> "../boost/archive/detail/oserializer.hpp", line 162.52: 1540-0063 (S)
> The text "__attribute__" is unexpected.
>
> (see http://tinyurl.com/ce66ny )
>
> The problem lies in boost/serialization/force_include.hpp, which,
> for reasons unbeknownst to me, defines BOOST_USED as the
> compiler-specific qualifier __attribute__ ((used)). Looking at the
> code for force_include.hpp (http://tinyurl.com/cpk7ux ) seems
> like this definition of BOOST_USED can only take place if
> either __GNUC__ or __INTEL_COMPILER is defined, which
> does not seem to be the case here.

Last time I looked, this attribute was only supported on the compilers
listed.  My understanding was that for the IBM compiler, this might
change and perhaps it has.  Feel free to tweak force_include to
see if that fixes the problem.

Robert Ramey

> Any clues? Thank you in advance,
>
> Joaquín M López Muñoz
> Telefónica, Investigación y Desarrollo