I am trying to install some packages from source on a linux (RHEL) server. But whenever the package requires some C++ compilation it fails.
For example, I try to install Ckmeans.1d.dp package. If I call R CMD INSTALL Ckmeans.1.dp_4.2.1.tar.gz I get the following output (truncated) where you can see the compiler is omitted.
...* installing *source* package ‘Ckmeans.1d.dp’ ...
** libs
I/usr/include/R -DNDEBUG -I/usr/local/include -c Ckmeans.1d.dp.cpp -o Ckmeans.1d.dp.o
make: I/usr/include/R: Command not found
...
-shared -L/usr/local/lib64 -o Ckmeans.1d.dp.so Ckmeans.1d.dp.o Ckmeans.1d.dp_main.o dynamic_prog.o fill_SMAWK.o fill_log_linear.o fill_quadratic.o select_levels.o weighted_select_levels.o -L/usr/lib64/R/lib -lR
/bin/sh: line 2: -shared: command not found
make: *** [Ckmeans.1d.dp.so] Error 127
ERROR: compilation failed for package ‘Ckmeans.1d.dp’
I am currently stuck with a server that only has R 3.1.1 and I do not have admin privileges. The $Rhome/etc/Makeconf has CXX and CC defined. I have also verified g++ and gcc are installed.
What could be happening here?
Invariably an environment variable with an 'empty' value gets expanded. Maybe (just guessing here) the package would use C++11, so $(CXX11} -shared gets expanded and ... becomes -shared and an error is triggered.
Look at the R settings, eg via less $(R HOME)/etc/Makevars and see what it expects.
I think you can get by, if you must, installing gcc et al below, say, ~/bin. Some packages will require a compiler...
Related
I am working on an R package (https://github.com/bgbrink/dropClust) and I was testing if it can be installed, since it depends on a number of packages both from CRAN and from Bioconductor. I specified the three dependencies from Bioconductor in the DESCRIPTION:
Remotes:
bioc::flowDensity,
bioc::SamSPECTRAL,
bioc::flowPeaks
However, when I try to install the package, the installation of flowPeaks fails, because the script downloades the source version of the package, which requires GSL to be present in order to be compiled.
* installing *source* package ‘flowPeaks’ ...
** libs
clang++ -I/Library/Frameworks/R.framework/Resources/include -DNDEBUG -I/usr/local/include `gsl-config --cflags` -fPIC -Wall -g -O2 -c Rpack.cpp -o Rpack.o
/bin/sh: gsl-config: command not found
In file included from Rpack.cpp:16:
./gvector_gmatrix.h:24:10: fatal error: 'gsl/gsl_math.h' file not found
#include <gsl/gsl_math.h>
^
1 error generated.
make: *** [Rpack.o] Error 1
ERROR: compilation failed for package ‘flowPeaks’
When I download the package manually from Bioconductor, everything works fine, because the script downloades the binary version of the package. Can I specify somewhere that this is the default behaviour? I tried options(pkgType = "binary") without success.
Edit: No new ideas, so I'm bumping it once.
In order to learn how to create a R packages that uses Fortran I created a simple one to calculate Pi with a monte carlo simulation. Now I'm trying to make the Fortran code parallel with MPI.
When I add use MPI to my fortran code and click the Build and Reload buton on Rstudio I get the following error.
==> R CMD INSTALL --no-multiarch --with-keep.source MyPi
* installing to library ‘/home/ignacio/R/x86_64-pc-linux-gnu-library/3.2’
* installing *source* package ‘MyPi’ ...
gfortran -fpic -g -O2 -fstack-protector-strong -c Fpi.f90 -o Fpi.o
/usr/lib/R/etc/Makeconf:161: recipe for target 'Fpi.o' failed
** libs
Fpi.f90:2.4:
use mpi
1
Fatal Error: Can't open module file 'mpi.mod' for reading at (1): No such file or directory
make: *** [Fpi.o] Error 1
ERROR: compilation failed for package ‘MyPi’
* removing ‘/home/ignacio/R/x86_64-pc-linux-gnu-library/3.2/MyPi’
* restoring previous ‘/home/ignacio/R/x86_64-pc-linux-gnu-library/3.2/MyPi’
Exited with status 1.
I think the problem is that Rstudio is trying to compile using gfortran instead of mpi90. I'm able to follow this post to do a very basic hello world, but there you compile the code with mpif90 -fpic -shared -o test.so test.f90
How can I tell Rstudio to use mpif90 -fpic -shared instead of gfortran?
I added this to src/Makevars
FC=mpif90
This is what I get:
==> R CMD INSTALL --no-multiarch --with-keep.source MyPi
* installing to library ‘/home/ignacio/R/x86_64-pc-linux-gnu-library/3.2’
* installing *source* package ‘MyPi’ ...
gfortran -fpic -g -O2 -fstack-protector-strong -c Fpi.f90 -o Fpi.o
/usr/lib/R/etc/Makeconf:161: recipe for target 'Fpi.o' failed
** libs
Fpi.f90:2.4:
USE MPI
1
Fatal Error: Can't open module file 'mpi.mod' for reading at (1): No such file or directory
make: *** [Fpi.o] Error 1
ERROR: compilation failed for package ‘MyPi’
* removing ‘/home/ignacio/R/x86_64-pc-linux-gnu-library/3.2/MyPi’
* restoring previous ‘/home/ignacio/R/x86_64-pc-linux-gnu-library/3.2/MyPi’
Exited with status 1.
You just need to set the FC flag in your Makevars file to indicate the fortran compiler you wish to use.
FC=mpif90
EDIT - The above doesn't work if included locally in your package
The FC macro will be overwritten when specified within a package Makevars because the src/Makevars file is read before $RHOME/etc/Makeconf which overwrites the FC macro. In order to change the default compiler is to create a Makevars file in a .R/ directory in your $HOME directory.
Then just create the file with the same line as above. Restart R and try to compile again (this worked on my Ubuntu system).
I am trying to create my first package using the instructions on the hadley devtools wiki. I am using Windows 7 Professional, I have loaded R-tools 3.1, and I am using R 3.0.2. I get the error below when I run the has_devel() command and I do no know why. Does anyone know how I can successfully install devtools given the error below.
has_devel()
"C:/PROGRA~1/R/R-30~1.2/bin/x64/R" --vanilla CMD SHLIB foo.c
Error: Command failed (1)
> traceback()
6: stop("Command failed (", status, ")", call. = FALSE)
5: system_check(r_path, options, c(r_env_vars(), env_vars), ...)
4: force(code)
3: in_dir(path, system_check(r_path, options, c(r_env_vars(), env_vars),
...))
2: R("CMD SHLIB foo.c", tempdir())
1: has_devel()
There is an existing related question, except in that example R is run on MAC OSX (The check for successful devtools load (has_devel) fails).
My Rtools installation was not complete.
I re-installed the most recent version of Rtools from CRAN (http://cran.rstudio.com/). I re-installed devtools using the instructions supplied by Hadley Wickham (http://adv-r.had.co.nz/Philosophy.html) and the error was fixed, has_devel() returned TRUE.
has_devel()
"C:/PROGRA~1/R/R-30~1.2/bin/x64/R" --vanilla CMD SHLIB foo.c
gcc -m64 -I"C:/PROGRA~1/R/R-30~1.2/include" -DNDEBUG -I"d:/RCompile/CRANpkg/extralibs64/local/include" -O2 -Wall -std=gnu99 -mtune=core2 -c foo.c -o foo.o
gcc -m64 -shared -s -static-libgcc -o foo.dll tmp.def foo.o -Ld:/RCompile/CRANpkg/extralibs64/local/lib/x64 -Ld:/RCompile/CRANpkg/extralibs64/local/lib -LC:/PROGRA~1/R/R-30~1.2/bin/x64 -lR
[1] TRUE
see here for the same issue https://github.com/hadley/devtools/issues/234
(solution provided by hadley, running the script in this gist https://gist.github.com/hadley/4506250)
I have installed first time to a directory C:\Apps\Rtools without checking the environmental variable add step during installation. But, I have manually added the path into the PATH variable. Then, I had the same problem.
Then I reinstalled to the default directory C:\Rtools and this time I checked the environmental variable add step during installation. Then it worked.
I've successfully build my own package with Rcpp in R-Studio. However, when building the package in the Windows console, there are some error messages complaining file missing of R.h.
I set the path and R_Home environments with
SET PATH=D:\RTools\gcc-4.6.3\bin;D:\R3\bin;D:\RTools\bin;
SET R_HOME=D:\R3\
And the build command is
R CMD INSTALL --byte-compile --build mypkg
Below are the compiling error messages:
g++ -m32 -I"/include" -DNDEBUG -I"D:/R3/library/Rcpp/include" -I"d:
/RCompile/CRANpkg/extralibs64/local/include" -O2 -Wall -mtune=core2 -c Rcpp
Exports.cpp -o RcppExports.o
In file included from D:/R3/library/Rcpp/include/Rcpp.h:27:0,
from RcppExports.cpp:4:
D:/R3/library/Rcpp/include/RcppCommon.h:35:15: fatal error: R.h: No su
ch file or directory
compilation terminated.
You are missing an include for R.h, as the error says.
And if you look at your compile line, the statement
-I"/include"
is wrong relative to your stated R_HOME in D:/R3
You need to check your setup, somehow you confused R from using the correct include directory.
Rcpp builds just fine on Windows, you can even check by submitting your package to the win-builder service.
I have modified a C function in the phmm package. Now I want to rebuild the package, but the following error occurred. How can I fix the error?
$ cd ~/Desktop/phmmw
$ R CMD build phmm
* checking for file ‘phmm/DESCRIPTION’ ... OK
* preparing ‘phmm’:
* checking DESCRIPTION meta-information ... OK
* cleaning src
* installing the package to re-build vignettes
-----------------------------------
* installing *source* package ‘phmm’ ...
** libs
*** arch - x86_64
gcc-4.2 -arch x86_64 -std=gnu99 -I/Library/Frameworks/R.framework/Resources/include -I/Library/Frameworks/R.framework/Resources/include/x86_64 -I/usr/local/include -fPIC -g -O2 -c arms.c -o arms.o
make: gcc-4.2: No such file or directory
make: *** [arms.o] Error 1
ERROR: compilation failed for package ‘phmm’
* removing ‘/private/var/folders/r0/9qv15p6j59j6jtfq775k6j380000gn/T/RtmpIPgvDE/Rinst48317ea7/phmm’
-----------------------------------
ERROR: package installation failed
Your R installation was configured with gcc in such a (suboptimal) was that the gcc-4.2 command was hardwired.
So you need to
either provide a command gcc-4.2, most commonly done via a symlink from gcc to gcc-4.2,
or change the R setting, presumably in the file $RHOME/etc/Makeconf where you need to look for CC
or change it on the fly by doing CC=gcc R CMD build phmm
You may need to modify more than just one CC setting, possibly CXX, LD and more. It may be as easy as always removing the -4.2 part.