[R-pkg-devel] Linking an R package to the Python library
Bruce Hoff
bruce.hoff at sagebase.org
Mon Feb 29 22:01:06 CET 2016
I am trying to create a self-contained R package wrapping Python that works
cross platform and have not been able to do this on Windows. There is
already an R package which calls Python, requiring Python to be installed
on the target machine: https://github.com/cjgb/rPython-win. When the R
package is built it successfully links to the resident Python dll.
I retrieved the Python source code plus associated Visual Studio projects
as described here: https://docs.python.org/devguide/setup.html I ran the
Python 3.6 build and verified that rPython-win can link to the generated
dll, as expected. In order to make a portable, self-contained R package, I
changed the build to create a _static_ Python library as described in the
file PCbuild\readme.txt. This creates a 31MB file "python36.lib" but when
I try to build the rPython-win package under R i386 (and using the latest
RTools) I get:
gcc -m32 -shared -s -static-libgcc -o rPython.dll tmp.def pycall.o
-LC:/cpython/PCbuild/win32 -lpython3stub -lpython36
-Ld:/RCompile/r-compiling/local/local323/lib/i386
-Ld:/RCompile/r-compiling/local/local323/lib
-LC:/PROGRA~1/R/R-32~1.3/bin/i386 -lR
C:/cpython/PCbuild/win32/python36.lib: *error adding symbols: File format
not recognized*
collect2.exe: error: ld returned 1 exit status
no DLL was created
ERROR: compilation failed for package 'rPython'
I see some mention of the link error here:
https://sourceware.org/bugzilla/show_bug.cgi?id=17910
It looks like a related bug was in binutils 2.25 on 2/27/2015. It seems
like the fix would have made it into Rtools 3.3 since:
1) this page suggests that the aforementioned was resolved before the
latest release of binutils, 2.25.1, 9/15/2015:
https://sourceforge.net/projects/mingw/files/MinGW/Base/binutils/
2) Installing Rtools 3.3, one can see that the binaries which are part of
binutils (like ld.exe) are dated 9/22/2015.
I don't know if MinGW is 'buggy', if I'm doing something wrong, and/or what
to do to fix the link error.
Is there anyone on this list with experience linking an R package to a
static library built on Windows using Visual Studio?
Any suggestions are appreciated!
[[alternative HTML version deleted]]
More information about the R-package-devel
mailing list