Re: [eigen] gcc 4.2.4 warning on type-punned pointers

[Manoj Rajagopalan <rmanoj@xxxxxxxxx>]

I see this only in release build. The exact error is as follows (just pulled 
in latest sources from hg):

/usr/local/include/eigen3/Eigen/src/Core/arch/SSE/Complex.h: In 
function ‘typename Eigen::ei_unpacket_traits<T>::type Eigen::ei_pfirst(const 
Packet&) [with Packet = Eigen::Packet1cd]’:
/usr/local/include/eigen3/Eigen/src/Core/arch/SSE/Complex.h:311: warning: 
dereferencing type-punned pointer will break strict-aliasing rules


    My command-line flags are (mind the -march option):

    -Wall -Wmissing-include-dirs -Wswitch-default -Wunused-parameter -Wextra -Wconversion
    -march=pentium4 -g0 -O3 -msse2 -mfpmath=sse -ftree-vectorize  -fgcse-sm -fgcse-las
    -fgcse-after-reload -fipa-pta  -ftree-loop-linear -ftree-loop-im -ftree-loop-ivcanon
    -ftracer -funroll-loops -fvariable-expansion-in-unroller -fprefetch-loop-arrays -fno-math-errno
    -ffinite-math-only -fno-trapping-math -fno-signaling-nans -DEIGEN_NO_AUTOMATIC_RESIZING

Thanks,
Manoj


On Friday 13 August 2010 12:23:56 am Benoit Jacob wrote:
> 2010/8/12 Manoj Rajagopalan <rmanoj@xxxxxxxxx>:
> > Hi,
> >
> >  I'm seeing this warning
>
> what warning? please paste it all here. i can't reproduce with gcc 4.4
> and usually gcc only gets more pedantic with time.
>
> Benoit
>
> > from Complex.h, line 311 where a real[2] is cast as
> > complex* in C-style. I am tempted to submit a correction with static_cast
> > or reinterpret_cast instead but I'd first like to ask if this is likely
> > to cause issues related to alignment or otherwise.
> >
> > thanks,
> > Manoj