R SIG Debian - Jun 2012 - Multiple sub-architecture: linking issue

Hi again,

following my previous post on RCurl, I cannot install Rcpp either:

g++ -m32 -I/home/renaud/bin/R/2.15/lib64/R/include 
-I/home/renaud/bin/R/2.15/lib64/R/include/i386 -DNDEBUG 
-I../inst/include/ -I/usr/local/include    -fpic  -g -O2  -c r_cast.cpp 
-o r_cast.o
g++ -m32 -shared -L/usr/local/lib64 -o Rcpp.so Date.o DateVector.o 
Datetime.o DatetimeVector.o Dimension.o DottedPair.o Environment.o 
Evaluator.o Formula.o Function.o Language.o Module.o Pairlist.o 
Promise.o RObject.o RcppCommon.o Rcpp_init.o Reference.o Rostream.o 
Rstreambuf.o S4.o Symbol.o WeakReference.o barrier.o cache.o coerce.o 
complex.o debugging.o exceptions.o posixt.o r_cast.o
/usr/bin/ld: skipping incompatible 
/usr/lib/x86_64-linux-gnu/gcc/x86_64-linux-gnu/4.5.2/libstdc++.so when 
searching for -lstdc++
/usr/bin/ld: skipping incompatible 
/usr/lib/x86_64-linux-gnu/gcc/x86_64-linux-gnu/4.5.2/libstdc++.a when 
searching for -lstdc++
/usr/bin/ld: cannot find -lstdc++
collect2: ld returned 1 exit status

There is definitely some conflict or missing libraries going on...
Any help is appreciated.

Thank you.

PS: Dirk do you want me to post this on the Rcpp list for record?

-- 
Renaud Gaujoux
Computational Biology - University of Cape Town
South Africa

Make sure you have installed multilib gcc and the ia32 dev packages - it seems
you don't have them. I'm a

Sent from my iPhone

On Jun 8, 2012, at 6:27 AM, Renaud Gaujoux <renaud at
mancala.cbio.uct.ac.za> wrote:
> Hi again,
> 
> following my previous post on RCurl, I cannot install Rcpp either:
> 
> g++ -m32 -I/home/renaud/bin/R/2.15/lib64/R/include
-I/home/renaud/bin/R/2.15/lib64/R/include/i386 -DNDEBUG -I../inst/include/
-I/usr/local/include    -fpic  -g -O2  -c r_cast.cpp -o r_cast.o
> g++ -m32 -shared -L/usr/local/lib64 -o Rcpp.so Date.o DateVector.o
Datetime.o DatetimeVector.o Dimension.o DottedPair.o Environment.o Evaluator.o
Formula.o Function.o Language.o Module.o Pairlist.o Promise.o RObject.o
RcppCommon.o Rcpp_init.o Reference.o Rostream.o Rstreambuf.o S4.o Symbol.o
WeakReference.o barrier.o cache.o coerce.o complex.o debugging.o exceptions.o
posixt.o r_cast.o
> /usr/bin/ld: skipping incompatible
/usr/lib/x86_64-linux-gnu/gcc/x86_64-linux-gnu/4.5.2/libstdc++.so when searching
for -lstdc++
> /usr/bin/ld: skipping incompatible
/usr/lib/x86_64-linux-gnu/gcc/x86_64-linux-gnu/4.5.2/libstdc++.a when searching
for -lstdc++
> /usr/bin/ld: cannot find -lstdc++
> collect2: ld returned 1 exit status
> 
> There is definitely some conflict or missing libraries going on...
> Any help is appreciated.
> 
> Thank you.
> 
> PS: Dirk do you want me to post this on the Rcpp list for record?
> 
> -- 
> Renaud Gaujoux
> Computational Biology - University of Cape Town
> South Africa
> 
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
> 
>

On 8 June 2012 at 12:27, Renaud Gaujoux wrote:
| PS: Dirk do you want me to post this on the Rcpp list for record?

Yes, that generally is where Rcpp questions / comments / hints should go. 

As for multi-arch builds, I am not sure we even thought about supporting this so
if
it breaks your use of Rcpp and related packages, you get to keep the pieces.
That said, I'd be interested in supporting it eventually but I guess I want
to first understand better how/if it is supported (on Linux) by R itself.

Dirk

-- 
Dirk Eddelbuettel | edd at debian.org | http://dirk.eddelbuettel.com

On 12/06/2012 15:35, Simon Urbanek wrote:
> On Jun 12, 2012, at 7:01 AM, Renaud Gaujoux wrote:
> Any package that has configure on Makefile has to be installed in multiple
steps - one for each architecture, e.g.:
>
> R --arch i386 CMD INSTALL foo_1.2-3.tar.gz
> R --arch amd64 CMD INSTALL --libs-only foo_1.2-3.tar.gz
>
> see R-admin 2.6 Sub-architectures
> http://r.research.att.com/man/R-admin.html#Sub_002darchitectures
>
> Cheers,
> Simon
Worked perfectly, thank you!

I am sure this is a stupid question, but why aren't these packages 
installed like this by default?

Renaud

On 12/06/2012 16:03, Simon Urbanek wrote:
> Because INSTALL can be used on directories and that won't work, because
the installation will modify the content so you can't run it twice.
Technically, it would work on tar balls, and I think there is now
--merge-multiarch that will do it. I suspect that the reason to not make it
default is that it takes twice as long so it is a bit annoying if you are
installing in the development process (but that's just a guess).
>
> Cheers,
> Simon
>
>
Understood. I guess it would be possible (nice) to detect such special 
case where INSTALL is a directory.

I used install.packages('RCurl', libs_only=TRUE) to install RCurl from 
CRAN from the 64 bit R session after installing it on the 32 bit, and it 
worked perfectly.

I would say that when a user/admin installs a package she expects it to 
be available in all its R installations.
So to me it looks like package installations from anything except 
directories (repositories, tar balls) could be rather assumed to be 
multiarch installs by default, with the option of choosing a single 
architecture. This "install all" behaviour could possible be enabled 
with an option, for backward compatibility.

Renaud