Re: [Boost-users] Bug in boost::phoenix ?

Subject: Re: [Boost-users] Bug in boost::phoenix ?
From: Nicolas Savoire (nicos_at_[hidden])
Date: 2011-02-25 05:09:23

On Friday, February 25, 2011 01:12:51 am Joel de Guzman wrote:
> On 2/24/2011 11:05 PM, Nicolas Savoire wrote:
> > I am using boost::phoenix lambda functions and I am observing some weird
> > behaviours when compiling with optimizations (-O3).
> > 
> > The first issue is that compiling the following piece of code
> > 
> > Compiling:
> >    const float x = 1.0f;
> >    const float y = 1.0f;
> >    std::cout<<  (arg1+arg2)(x,y);
> > 
> > with g++ -Wall gives the warning:
> > boost_1_46_0/boost/spirit/home/phoenix/operator/arithmetic.hpp:74:5:
> > warning: returning reference to temporary
> > 
> > And I suspect a bug in the project I am working on is a consequence of
> > this. It is as if a boost::for_each( rng, lambda_fun ) was optimized
> > out. I have opened a ticket on this issue:
> > https://svn.boost.org/trac/boost/ticket/5190
> > I suspect the issue comes from the type deduction inside phoenix.
> > I first discovered the issue with boost 1.45.0, but it still here in
> > 1.46.0.
> > 
> > The other issue I am facing is similar. Consider the following code:
> > 
> > namespace bp = boost::phoenix;
> > using boost::phoenix::arg_names::arg1;
> > 
> > int i=0;
> > int a[] = {0};
> > 
> > template<class UnaryFunction>
> > inline void foo( UnaryFunction fun )
> > {
> > 
> >    fun(i);
> > 
> > }
> > 
> > int main()
> > {
> > 
> >    foo( bp::ref(a)[arg1]++ );
> >    std::cout<<  a[0]<<  " ";
> >    
> >    (bp::ref(a)[arg1]++)(i);
> >    std::cout<<  a[0]<<  "\n";
> > 
> > }
> > 
> > Compiled without optimizations, it corectly prints: 1 2
> > Whereas compiled with g++-4.5.1 -O3 with boost 1.45.0, it prints: 0 1,
> > 
> >   as if the first lambda invocation was removed.
> > 
> > The issue disappears if I use g++-4.4 or boost 1.46.0. I tracked the
> > changes between boost 1.45.0 and 1.46.0, and it seems related to
> > https://svn.boost.org/trac/boost/changeset/66411
> > applying the part of the changeset concerning:
> > fusion/support/detail/is_mpl_sequence.hpp
> > fusion/support/is_sequence.hpp
> > fusion/support/sequence_base.hpp
> > to boost 1.45.0 fixes the issue, but I can't see how this is related to
> > the bug I observed.
> > Is it a compiler bug, a phoenix bug or a mistake in my code ?
> 
> Could you try with Phoenix3 (on going mini-review) and see if this
> is an issue? I can try and investigate with Phoenix2 if you can
> provide a minimal cpp test case that I can run.
> 
> Regards,

I attached to this email two minimal test cases. Both fail with boost 1.45.0 
and g++-4.5.1 with compilation flags -O3.

My gcc version is :
Using built-in specs.
COLLECT_GCC=g++-4.5
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.5.1/lto-wrapper
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro 
4.5.1-7ubuntu2' --with-bugurl=file:///usr/share/doc/gcc-4.5/README.Bugs --
enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.5 
--enable-shared --enable-multiarch --enable-linker-build-id --with-system-zlib 
--libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --
with-gxx-include-dir=/usr/include/c++/4.5 --libdir=/usr/lib --enable-nls --
with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-
libstdcxx-time=yes --enable-plugin --enable-gold --with-plugin-ld=ld.gold --
enable-objc-gc --disable-werror --with-arch-32=i686 --with-tune=generic --
enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --
target=x86_64-linux-gnu
Thread model: posix
gcc version 4.5.1 (Ubuntu/Linaro 4.5.1-7ubuntu2) 

They both pass if I use g++-4.4.
Second test case pass with g++-4.5.1 and boost 1.46.0, but as I said in my 
previous mail, I don't understand how the changes in boost 1.46.0 contributed 
to solve the issue.
First test case still fails with g++-4.5.1 and boost 1.46.0. I was able to fix 
the issue by adding float to the handled types in 
boost/spirit/home/phoenix/detail/type_deduction.hpp, in a manner similar to 
bool,int,uint and double.

The good news is that both test cases pass with phoenix v3 ! But the inclusion 
of phoenix v3 in boost is still months away...

Thanks for your help,
nicolas