[R-SIG-Mac] Building package from source error "file is not of required architecture"

[Peter Cowan]

On Fri, Sep 11, 2009 at 6:32 AM, Simon Urbanek
<simon.urbanek at r-project.org> wrote:
> Peter,
>
> On Sep 11, 2009, at 2:05 AM, Peter Cowan wrote:
>
>> Bouncing to the list after failing to reply-all
>>
>> On Thu, Sep 10, 2009 at 5:25 PM, Simon Urbanek
>> <simon.urbanek at r-project.org> wrote:
>>>
>>> Peter,
>>>
>>> you have probably compiled ncl in 64-bit (default on SL) but you're using
>>> 32-bit R, so it won't work. Either use 64-bit R or compile ncl for
>>> 32-bit.
>>
>> I suspected something obvious of the like.  Unfortunately, I've not been
>> able to get 64 bit R working for me.  When I try the following for the CRAN
>> version:
>>
>> Birch$ R --arch x86_64
>> /usr/bin/R: line 207:
>> /Library/Frameworks/R.framework/Resources/etc/x86_64/ldpaths: No such
>> file or directory
>>
>
> Yes, CRAN R is 32-bit only.

Yes, this makes sense, but I was a bit confused by this comment on
r.research.att.com

"Leopard builds, 64-bit binaries and packages are becoming available
from CRAN as of R 2.8.0."
Does that mean r.research.att.com ?  I don't find a link on CRAN.

>> So I got the 2.9.2 patched version distributed as a tar.gz from
>> <http://r.research.att.com/> (thanks for providing this Simon!), but that
>> doesn't work with the 64bit GUI build on the same site (it won't even
>> launch, because it need 2.10.0).
>>
>
> Oh, true, the 2.9 GUI is simply not there. I'll have to look to see what
> went wrong.

Thanks!

>
>> So I installed the 32/64 Universal 2.9.0,
>
> [BTW: you can simply overwrite the installation with the 2.9.2 patched
> binary to upgrade it - the only part you really needed from the universal
> package was the GUI]

Okay, I was concerned there might be a version mismatch between 2.9.2
and the 2.9.0 gui so I didn't do that.  BTW is the repository for the
GUI public?  I see the .tar on CRAN.

>
>> but now building any package gives errors [1] perhaps because SL doesn't
>> install ppc64 libraries?
>
> Yes, unfortunately. The hot fix is
> rm -rf /Library/Frameworks/R.framework/Resources/bin/exec/ppc*

Brilliant!  This works perfectly.  Thank you for not only contributing
software but patiently helping me understand it.

Peter

>> I the same (or similar warning) using the R-devel nightly.
>>
>> I apologize if this to is obvious.  When I install xcode, do I need to
>> install the tools for prior OSs to get ppc compilers?
>>
>
> The compilers are there all right, it's the libraries that are missing. They
> are now only part of the 10.5 SDK in Xcode which is not used by R.
>
> Cheers,
> Simon
>
>
>> [1]
>> ======
>> (WARNING: partial output only, ask package author to use Rprintf instead!)
>> newick.c: In function ‘readNewickString’:
>> newick.c:233: warning: format not a string literal and no format arguments
>> ld: warning: in /usr/lib/dylib1.10.5.o, missing required architecture
>> ppc64 in file
>> ld: warning: in /usr/lib/libSystemStubs.a, missing required
>> architecture ppc64 in file
>> ld: warning: in
>> /System/Library/Frameworks//CoreFoundation.framework/CoreFoundation,
>> missing required architecture ppc64 in file
>> ld: warning: in /usr/lib/libSystem.dylib, missing required
>> architecture ppc64 in file
>> ld: in /usr/lib/libSystem.B.dylib, missing required architecture ppc64 in
>> file
>> collect2: ld returned 1 exit status
>> make: *** [ape.so] Error 1
>> ERROR: compilation failed for package ‘ape’
>> * Removing
>> ‘/Library/Frameworks/R.framework/Versions/2.9/Resources/library/ape’
>> * Restoring previous
>> ‘/Library/Frameworks/R.framework/Versions/2.9/Resources/library/ape’
>> ========
>>
>>
>>> Cheers,
>>> Simon
>>>
>>>
>>> On Sep 10, 2009, at 7:58 PM, Peter Cowan wrote:
>>>
>>>> I'm running 10.6 from clean install, with the supplied Xcode, and R
>>>> 2.9.2.  I am unable to build a package (specifically the phylobase
>>>> package [1]) which built fine under 10.5.8.  Even more confusingly it
>>>> seems to build fine for others using 10.6.   The end of the install
>>>> log (appended below) contains the these potentially informative
>>>> errors:
>>>>
>>>> ld: warning: in ncl/ncl/.libs/libncl.a, file is not of required
>>>> architecture
>>>> ld: warning: duplicate dylib /usr/local/lib/libgcc_s.1.dylib
>>>>
>>>> The latter error I get with any source package, but doesn't seem to
>>>> prevent them from being installed.
>>>>
>>>> The package also makes use of the Rcpp interface to use an included
>>>> library (NCL).  It is this included library that seems to not build
>>>> correctly, which could very well be related to the old version of
>>>> Rcpp, or the makefiles responsible for building it.
>>>>
>>>> I've tried all manner of tweaking, upgrading, and list searching, but
>>>> the (probably obvious) solution still eludes me.  Any pointers in the
>>>> right direction are greatly appreciated.
>>>>
>>>> Peter
>>>>
>>>> [1] https://r-forge.r-project.org/projects/phylobase/
>>>> =========================
>>>>>
>>>>> sessionInfo()
>>>>
>>>> R version 2.9.2 (2009-08-24)
>>>> i386-apple-darwin8.11.1
>>>>
>>>> locale:
>>>> en_US.UTF-8/en_US.UTF-8/C/C/en_US.UTF-8/en_US.UTF-8
>>>>
>>>> attached base packages:
>>>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>>>
>>>> =========================
>>>> birch$ which gfortran
>>>> /usr/local/bin/gfortran
>>>>
>>>> birch$ gfortran --version
>>>> GNU Fortran (GCC) 4.2.3
>>>> Copyright (C) 2007 Free Software Foundation, Inc.
>>>>
>>>> GNU Fortran comes with NO WARRANTY, to the extent permitted by law.
>>>> You may redistribute copies of GNU Fortran
>>>> under the terms of the GNU General Public License.
>>>> For more information about these matters, see the file named COPYING
>>>>
>>>> ============================
>>>>
>> [original error snipped]
>>
>> _______________________________________________
>> R-SIG-Mac mailing list
>> R-SIG-Mac at stat.math.ethz.ch
>> https://stat.ethz.ch/mailman/listinfo/r-sig-mac
>>
>>
>
>