[R] Cross-compilation of R and ld bug ?

[Prof Brian Ripley]

On Tue, 9 Jan 2007, Stéphane Dray wrote:

> Not sure to really understand 1) . The makefile (available at 
> http://cran.r-project.org/doc/contrib/Makefile-rcb) use  wget --passive-ftp 
> http://www.stats.ox.ac.uk/pub/Rtools/mingw-cross5.tar.bz2;  This version 
> seems to be the current, I am correct ? Note that :

Well, those tools built 2.4.0 for me, but earlier versions of 
mingw-cross5.tar.bz2 did not because bintools was too old.  So assuming 
you didn't get a cached older file, I don't understand.


> stephane at pcdray:~/Rdev/CrossCompileBuild$ cross-tools/bin/i586-mingw32-ld -v
> GNU ld version 2.16.91 20050827
> stephane at pcdray:~/Rdev/CrossCompileBuild$ cross-tools/i586-mingw32/bin/ld -v
> GNU ld version 2.16.91 20050827
>
>
> However, I have use R-2.4.1 and everything seems ok.
>
> Thanks.
>
>
> Prof Brian Ripley wrote:
>> There are two problems here:
>> 
>> 1) your binutils is not current.
>> 2) your R is not current.
>> 
>> Updating either will solve this, but please update both.
>> Almost up-to-date cross-compilers are (as ever) available from
>> 
>> http://www.stats.ox.ac.uk/pub/Rtools/
>> 
>> (If you want to compile R-devel you will need to update the mingw, which 
>> can be done from the i386 distribution.  I will rebuild the cross-compilers 
>> in due course.)
>> 
>> On Mon, 8 Jan 2007, Stéphane Dray wrote:
>> 
>>> Hello list,
>>> 
>>> I would like to cross-compile R packages using R 2.4.0. I am working on
>>> Linux Debian and cross-compiled (windows binaries) without problems with
>>> older R version.
>>> I have used the doc of Yan and Rossini in the contributed section of the
>>> R documentation (same version of MinGW...).
>>> When I try to cross-compile R (make R), the procedure stopped and returns 
>>> :
>>> ************************************
>>> i586-mingw32-windres  --include-dir ../include -i dllversion.rc -o
>>> dllversion.o
>>> i586-mingw32-gcc  -shared -s -mwindows -o R.dll R.def console.o
>>> dataentry.o dynl oad.o edit.o editor.o embeddedR.o extra.o opt.o pager.o
>>> preferences.o psignal.o rhome.o rui.o run.o shext.o sys-win32.o system.o
>>> e_pow.o malloc.o ../main/libmai n.a ../appl/libappl.a
>>> ../nmath/libnmath.a graphapp/ga.a getline/gl.a ../extra/xd r/libxdr.a
>>> ../extra/zlib/libz.a ../extra/pcre/libpcre.a ../extra/bzip2/libbz2.a
>>> ../extra/intl/libintl.a ../extra/trio/libtrio.a dllversion.o -L. -lg2c
>>> -lRblas - lcomctl32 -lversion
>>> ../extra/pcre/libpcre.a(pcre_dfa_exec.o):pcre_dfa_exec.c:(.text+0x1703):
>>> undefin ed reference to `__pcre_ucp_findprop'
>>> ../extra/pcre/libpcre.a(pcre_dfa_exec.o):pcre_dfa_exec.c:(.text+0x1740):
>>> undefin ed reference to `__pcre_ucp_findprop'
>>> ../extra/pcre/libpcre.a(pcre_dfa_exec.o):pcre_dfa_exec.c:(.text+0x1848):
>>> undefin ed reference to `__pcre_ucp_findprop'
>>> ../extra/pcre/libpcre.a(pcre_dfa_exec.o):pcre_dfa_exec.c:(.text+0x187f):
>>> undefin ed reference to `__pcre_ucp_findprop'
>>> ../extra/pcre/libpcre.a(pcre_dfa_exec.o):pcre_dfa_exec.c:(.text+0x1966):
>>> undefin ed reference to `__pcre_ucp_findprop'
>>> ../extra/pcre/libpcre.a(pcre_dfa_exec.o):pcre_dfa_exec.c:(.text+0x1f2a):
>>> more un defined references to `__pcre_ucp_findprop' follow
>>> collect2: ld returned 1 exit status
>>> make[4]: *** [R.dll] Erreur 1
>>> make[3]: *** [../../bin/R.dll] Erreur 2
>>> make[2]: *** [rbuild] Erreur 2
>>> make[1]: *** [all] Erreur 2
>>> make[1]: quittant le répertoire «
>>> /home/stephane/Rdev/CrossCompileBuild/WinR/R-2 .4.0/src/gnuwin32 »
>>> ****************************************
>>> I have read (http://www.murdoch-sutherland.com/Rtools/) that a bug
>>> exists for ld version 2.16.91 20050827 and that Prof Ripley produced a
>>> patched version. It seems that my problem is also related to ld (and I
>>> have the same version). So I wonder if the same bug could be responsible
>>> of the error in the cross-compiler.
>>> Two questions:
>>> - Are they other people which have the same problem when cross-compiling
>>> R on Linux
>>> - Is it possible that the problem is related to ld. If yes, is it
>>> possible to obtain a patched version ?
>>> 
>>> Thanks in advance.
>>> Sincerely,
>>> 
>>> 
>> 
>
>
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595