Dear Simon,
Thank you for your answer which is really helpful for me.
Best regards
Christian
Simon Urbanek wrote:
On Jan 29, 2010, at 13:49 , cstrato wrote:
Dear Simon,
As a subscriber to RootTalk I have just received the following mail:
http://root.cern.ch/root/roottalk/roottalk10/0122.html
As you see there are other Mac users of ROOT who do not want to use
fink but another gfortran compiler.
I seems that besides gfortran from your website, the following
gfortran can also be used:
http://www.macresearch.org/gfortran-leopard
My question now is:
Can gfortran from macresearch also be used to compile R 32 bit and 64
bit?
Pretty much any compiler including fink can be used to compile R so
the answer specifically to your question is yes. However it will be
incompatible with any binaries we provide.
The compilers from macresearch.org seem to be the same that used to be
on the HPC site and are still incomplete and incompatible with Xcode
and the usual Mac configuration so they cannot be used to replace our
compilers. AFAIK ours are the only complete gfortran compilers for the
Mac with Apple driver (unfortunately - I'd love to give that
responsibility to someone else :)).
Cheers,
Simon
cstrato wrote:
Dear Kasper, dear Simon,
Thank you both for your help, this was easier than I thought.
Since I had already installed "R-2.10.1.pkg" all I needed to do was:
$ R --arch=x86_64 --no-multiarch CMD INSTALL -l
~/Library/R/2.10/library xps_1.7.4.tar.gz
* installing *source* package 'xps' ...
checking for gcc... gcc-4.2 -arch x86_64 -std=gnu99
checking for C compiler default output file name... a.out
checking whether the C compiler works... yes
checking whether we are cross compiling... no
checking for suffix of executables...
checking for suffix of object files... o
checking whether we are using the GNU C compiler... yes
checking whether gcc-4.2 -arch x86_64 -std=gnu99 accepts -g... yes
checking for gcc-4.2 -arch x86_64 -std=gnu99 option to accept ANSI
C... none needed
checking how to run the C preprocessor... gcc-4.2 -arch x86_64
-std=gnu99 -E
checking for gcc... (cached) gcc-4.2 -arch x86_64 -std=gnu99
checking whether we are using the GNU C compiler... (cached) yes
checking whether gcc-4.2 -arch x86_64 -std=gnu99 accepts -g...
(cached) yes
checking for gcc-4.2 -arch x86_64 -std=gnu99 option to accept ANSI
C... (cached) none needed
found ROOT version 5.24/00 in directory /Users/rabbitus/ROOT/root
** libs
** arch - x86_64
Now xps compiles for 64 bit and starting R with:
$ R --arch=x86_64
I am able to load my package:
> library(xps)
Welcome to xps version 1.7.4
an R wrapper for XPS - eXpression Profiling System
(c) Copyright 2001-2009 by Christian Stratowa
Now one final question remains, which is a question of a user of xps
who has started the whole discussion since he has compiled ROOT as
64 bit and wanted to install xps using:
> biocLite("xps", type = "source")
I assume that when I start R using "R --arch=x86_64" then every
package is compiled as 64 bit automatically.
Best regards
Christian
Simon Urbanek wrote:
On Jan 26, 2010, at 15:24 , cstrato wrote:
Dear Simon,
Meanwhile I have tried to install xps:
$ R CMD INSTALL -l ~/Library/R/2.10/library xps_1.7.4.tar.gz
* installing *source* package 'xps' ...
checking for gcc... gcc-4.2 -arch i386 -std=gnu99
checking for C compiler default output file name... a.out
checking whether the C compiler works... yes
checking whether we are cross compiling... no
checking for suffix of executables...
checking for suffix of object files... o
checking whether we are using the GNU C compiler... yes
checking whether gcc-4.2 -arch i386 -std=gnu99 accepts -g... yes
checking for gcc-4.2 -arch i386 -std=gnu99 option to accept ANSI
C... none needed
checking how to run the C preprocessor... gcc-4.2 -arch i386
-std=gnu99 -E
checking for gcc... (cached) gcc-4.2 -arch i386 -std=gnu99
checking whether we are using the GNU C compiler... (cached) yes
checking whether gcc-4.2 -arch i386 -std=gnu99 accepts -g...
(cached) yes
checking for gcc-4.2 -arch i386 -std=gnu99 option to accept ANSI
C... (cached) none needed
found ROOT version 5.24/00 in directory /Users/rabbitus/ROOT/root
** libs
** arch - i386
As you can see it compiles using the wrong architecture:
Well, you're running the wrong architecture ... As I said when you
want to link, you have to use 64-bit.
> library(xps)
Error in dyn.load(file, DLLpath = DLLpath, ...) :
unable to load shared library
'/Users/rabbitus/Library/R/2.10/library/xps/libs/i386/xps.so':
dlopen(/Users/rabbitus/Library/R/2.10/library/xps/libs/i386/xps.so,
6): no suitable image found. Did find:
/Users/rabbitus/Library/R/2.10/library/xps/libs/i386/xps.so:
mach-o, but wrong architecture
Error: package/namespace load failed for 'xps'
>
In my file "config.in" I have the following lines:
# See "Writing R Extensions"
: ${R_HOME=`R RHOME`}
if test -z "${R_HOME}"; then
echo "could not determine R_HOME"
exit 1
fi
CC=`"${R_HOME}/bin/R" CMD config CC`
CFLAGS=`"${R_HOME}/bin/R" CMD config CFLAGS`
AC_PROG_CPP
AC_PROG_CC
Thus I assume that I need to use 64 bit R so that the flags are
set correctly. Is this correct?
You have to use 64-bit R if you want to link to a 64-bit program -
as I said.
If yes, can I use:
http://r.research.att.com/R-GUI-5548-2.10-leopard-Leopard64.dmg
or do I need to build 64 bit R from source?
Neither - just pick it from CRAN - it's the first download in the
list. Note that you still have to run 64-bit R -- e.g. R --arch=x86_64
Cheers,
Simon
Simon Urbanek wrote:
On Jan 25, 2010, at 5:34 PM, cstrato wrote:
Dear Simon,
Thank you, with this version I could finally compile ROOT as 64
bit.
Now I have one more question:
When I do "R CMD Install xps.tar.gz" which now depends on 64 bit
ROOT, do I need to install the 64 bit version of R or can I use
the current R version?
I have no idea since I don't know your package. If you *link* to
ROOT then yes, you have to use 64-bit R to load 64-bit code.
(Linking in itself can be quite interesting given the flags ROOT
uses, so your journey may not end there ;)).
Cheers,
Simon
Best regards
Christian
Simon Urbanek wrote:
On Jan 25, 2010, at 15:03 , cstrato wrote:
Dear Simon,
Thank you for your suggestions, which I have just tried. Sadly
the error remains the same.
The problem seems to be that for some reason
"gfortran-4.2-5564-darwin9.tar.gz" does not have a
"/usr/lib/gcc/i686-apple-darwin9/4.2.1/x86_64" directory
containing libgfortranbegin.a. For this reason the 64 bit
version of libgfortranbegin.a is not installed in the
corresponding directory x86_64 which contains only libgcc.a etc.
Oh ... that's a surprise - it should be there, sorry. Ok, so
please use the more recent 5566 build:
http://r.research.att.com/tools/gfortran-4.2-5566-darwin9.tar.gz
that one has the x86_64 files - I checked:
ginaz:~$ curl -s -S
http://r.research.att.com/tools/gfortran-4.2-5566-darwin9.tar.gz |
tar fvtz - | grep gfortran.a
-rw-r--r-- root/wheel 3592608 2008-12-24 14:51:33
usr/lib/gcc/i686-apple-darwin9/4.2.1/libgfortran.a
-rw-r--r-- root/wheel 3870032 2008-12-24 14:51:33
usr/lib/gcc/powerpc-apple-darwin9/4.2.1/libgfortran.a
-rw-r--r-- root/wheel 5002152 2008-12-24 14:51:33
usr/lib/gcc/i686-apple-darwin9/4.2.1/x86_64/libgfortran.a
(Ideally you should install Xcode 3.1.2 for this since that is
where the 5566 build comes from, but it will probably work even
if the builds don't match perfectly).
Cheers,
Simon
You can see this by double-clicking on
"gfortran-4.2-5564-darwin9.tar.gz" when located in some
directory.
In contrast, "gfortran-4.2.3.dmg" did install
libgfortranbegin.a in the
"/usr/local/lib/gcc/i686-apple-darwin9/4.2.1/x86_64" directory.
Could you please check or tell me what the reason might be?
I have installed it using:
$ gunzip gfortran-4.2-5564-darwin9.tar.gz
$ sudo tar -xvf gfortran-4.2-5564-darwin9.tar -C /
Best regards
Christian
Simon Urbanek wrote:
On Jan 24, 2010, at 3:28 PM, cstrato wrote:
Dear Kasper,
You are right. Since ROOT supports the CC, CXX, F77 and LD
flags I did:
./configure macosx64 --with-cxx="/usr/bin/g++-4.2 -arch
x86_64" --with-f77="/usr/bin/gfortran-4.2 -arch x86_64"
You don't really need the override - it works without and
they add -m64 anyway so the -arch gets essentially ignored.
The error was still the same:
i686-apple-darwin9-g++-4.2.1: libgfortran.dylib: No such
file or directory
However, when I replaced in the Makefile.macosx64:
F77LIBS := $(shell $(F77) -m64
-print-file-name=libgfortran.dylib)
F77LIBS += $(shell $(F77) -m64
-print-file-name=libgfortranbegin.a)
with:
F77LIBS := $(shell $(F77) -m64
-print-file-name=libgfortranbegin.a)
You found the right spot but your fix is wrong - just remove
the two lines, they are superfluous and only cause trouble.
the error disappeared, and I could finish the compilation.
However, now I got a link-error:
ld warning: in
/usr/lib/gcc/i686-apple-darwin9/4.2.1/libgfortranbegin.a,
file is not of required architecture
Undefined symbols:
"_main", referenced from:
start in crt1.10.5.o
Your'e getting the 32-bit version (of the wrong file - see
above). It should really be ../4.2.1/x86_64/lib...
I am not sure what this error means since Makefile.macosx64
contains:
F77FLAGS = -m64 -std=legacy
Could it be that libgfortranbegin.a can only be used for
32bit compilation?
No, you're picking the wrong one - there is one for each
architecture.
Furthermore, I still have the question (and hopefully Simon
will answer it):
Why does gfortran-4.2-5564-darwin9.tar.gz not contain any
libgfortran.dylib files?
Because it's using a static library.
Interestingly, when googling "/usr/lib/libgfortran.dylib" I
get the answer that NO results were found, while googling
"/usr/local/lib/libgfortran.dylib" gives many results. Thus
it seems that there will never be any libgfortran.dylib in
"/usr/lib/"!?
Nope - OS X has no Fortran by default.
Cheers,
Simon
I must admit that I do not understand this issue.
Best regards
Christian
Kasper Daniel Hansen wrote:
On Jan 23, 2010, at 8:37 AM, cstrato wrote:
Dear Kasper,
Thank you for your suggestion, I did indeed use gcc 4.0.
Since gcc_select does no longer exist on Leopard I had to do:
cd /usr/bin
sudo ln -s gcc-4.2 gcc
sudo ln -s g++-4.2 g++
sudo ln -s gcov-4.2 gcov
I am not sure this is advisable in general. When I compile
stuff with gcc-4.2 I usually explicitly set
CC=/usr/bin/gcc-4.2
when I run configure. For example, my entire configure
line for R on my macbook is
../${SRCDIR}/configure SHELL='/bin/bash' \
CC="/usr/bin/gcc-4.2 -arch x86_64 -std=gnu99" \
CFLAGS="-g -O2 -std=gnu99 -march=nocona" \
CXX="/usr/bin/g++-4.2 -arch x86_64" \
CXXFLAGS="-g -O2 -march=nocona" \
OBJC="/usr/bin/gcc-4.2 -arch x86_64" \
F77="/usr/bin/gfortran-4.2 -arch x86_64" \
FFLAGS="-g -O2 -march=nocona" \
FC="/usr/bin/gfortran-4.2 -arch x86_64" \
FCFLAGS="-g -O2 -march=nocona" \
--x-includes=/usr/X11/include
--x-libraries=/usr/X11/lib \
--with-system-zlib \
--with-blas='-framework vecLib' --with-lapack
Of course this requires that ROOT uses the CC/CFLAGS
macros, but that should be pretty standard if it uses
autoconf.
Sadly, compiling ROOT resulted still in the same error:
i686-apple-darwin9-g++-4.2.1: libgfortran.dylib: No such
file or directory
When I check the location for gfortran-4.2.3.dmg I get:
$ cd
/usr/local/lib/gcc/i686-apple-darwin8/4.2.3/../../../x86_64
$ ls -al
total 32
drwxrwxr-x 6 root wheel 204 Jan 21 20:26 .
drwxr-xr-x 35 root wheel 1190 Jan 21 20:26 ..
lrwxr-xr-x 1 root wheel 26 Jan 21 20:26
libgfortran.2.0.0.dylib -> ../libgfortran.2.0.0.dylib
lrwxr-xr-x 1 root wheel 23 Jan 21 20:26
libgfortran.2.dylib -> libgfortran.2.0.0.dylib
lrwxr-xr-x 1 root wheel 16 Jan 21 20:26
libgfortran.a -> ../libgfortran.a
lrwxr-xr-x 1 root wheel 23 Jan 21 20:26
libgfortran.dylib -> libgfortran.2.0.0.dylib
This means that /usr/local/bin does contain
libgfortran.dylib whereas /usr/bin does not contain any of
these files or links.
Does this mean that gfortran-4.2.3.dmg does contain these
file for libgfortran.dylib while
gfortran-4.2-5564-darwin9.tar.gz does not contain these
files?
How do I get libgfortran.dylib?
Simon might be able to answer this better, but the general
idea is that the
gfortran-4.2-5564-darwin9.tar.gz
installs into /usr/bin
gfortran-4.2.3.dmg
installs into /usr/local/bin. There is some logic here,
the way I think about it is that the former is a direct
extension of the Apple build.
I would start from scratch, which means remove all traces
of both gfortrans, install the 5564-tarball and then set
the various macros as I have set it above for R (your exact
values may of course be different depending on what machine
etc. you are using).
Kasper
BTW, I must apologize to Simon. On his website he lists
the commands how to install gfortran.tar.gz, but in the
Tiger section, so I did miss it. Nevertheless it would
still be helpful for Mac users to know how to uninstall it:
$ tar -tf gfortran.tar | sort -r | (cd /; xargs -p -n 1 rm
-d)
Best regards
Christian
Kasper Daniel Hansen wrote:
Remember that Apple has two version of GCC on Leopard:
4.0 and 4.2. You are using 4.0, you might want to switch
to 4.2. In Tiger there used to be gcc_select (or
select_gcc) that let you choose between 3.x and 4.0, I
don't remember if that is still around on Leopard.
The error seems to indicate that you need 4.2 in order to
use Simon's gfortran.
Kasper
On Jan 22, 2010, at 16:04 PM, cstrato wrote:
Dear Simon,
Thank you for this information.
Now I have installed gfortran-4.2-5564-darwin9.tar.gz on
Leopard and tried to compile ROOT again, but this time I
got the following error:
g++ -dynamiclib -single_module -undefined dynamic_lookup
-install_name @rpath/libminicern.so -O2 -m64 -o
lib/libminicern.so misc/minicern/src/cernlib.o -ldl
misc/minicern/src/hbook.o misc/minicern/src/kernlib.o
misc/minicern/src/zebra.o libgfortran.dylib
/usr/lib/gcc/i686-apple-darwin9/4.2.1/libgfortranbegin.a
i686-apple-darwin9-g++-4.0.1: libgfortran.dylib: No such
file or directory
I know that this is not necessarily an R-related
question but I would greatly appreciate your help if
possible (at least it is a Mac- and somehow a
BioC-related question).
When I have tried to compile ROOT with
gfortran-4.2.3.dmg I got the following error:
gfortran -O2 -m64 -o bin/g2root main/src/g2root.o -Llib
lib/libminicern.so
/usr/local/lib/gcc/i686-apple-darwin8/4.2.3/../../../x86_64/libgfortran.dylib
/usr/local/lib/gcc/i686-apple-darwin8/4.2.3/x86_64/libgfortranbegin.a
-lm -Wl,-rpath,@loader_path/../lib -pthread -ldl
ld: -rpath can only be used when targeting Mac OS X 10.5
or later
As you can see gfortran-4.2.3 could find
libgfortran.dylib at:
/usr/local/lib/gcc/i686-apple-darwin8/4.2.3/../../../x86_64/libgfortran.dylib
while gfortran-4.2.4 could not find libgfortran.dylib.
Thus maybe I did make a mistake when installing
gfortran-4.2.4. Here is what I did:
$ gunzip gfortran-4.2-5564-darwin9.tar.gz
$ sudo tar -xvf gfortran-4.2-5564-darwin9.tar -C /
$ sudo ln -s /usr/bin/gfortran-4.2 /usr/bin/gfortran
Do you know what I need to do so that libgfortran.dylib
can be found?
P.S.:Since many Mac users are no Unix experts please
allow me to suggest to place on your website the
commands to install/uninstall gfortran. I have found the
following website, which told me how to install gfortran
and how I could uninstall it:
http://www.webmo.net/support/fortran_osx.html
Thank you in advance.
Best regards
Christian
Simon Urbanek wrote:
On Jan 21, 2010, at 15:56 , cstrato wrote:
Dear Simon,
One more question:
I have just unzipped gfortran-4.2-5564-darwin9.tar.gz.
The version gives:
Using built-in specs.
Target: i686-apple-darwin9
Configured with:
/Builds/gcc/gcc_42-5564/build/obj/src/configure
--disable-checking --enable-werror --prefix=/usr
--mandir=/usr/share/man
--enable-languages=c,objc,c++,obj-c++,fortran
--program-transform-name=/^[cg][^.-]*$/s/$/-4.2/
--with-gxx-include-dir=/usr/include/c++/4.0.0
--with-slibdir=/usr/lib --build=i686-apple-darwin9
--host=i686-apple-darwin9 --target=i686-apple-darwin9
Thread model: posix
gcc version 4.2.1 (Apple Inc. build 5564)
This means that this version seems to be older than
the one I am using now?
No, that is GNU Fortran 4.2.4 that I was referring to.
The "gcc version 4.2.1" comes from Apple's Xcode 3.1
gcc-4.2 branch since that is what they define - it just
adds Fortran support to the existing gcc 4.2.1 branch
of Apple that you installed as Xcode. However, the
Fortran part is merged from the GNU Fortran 4.2.4
release - hence I refer to it as 4.2.4.
Cheers,
Simon
Simon Urbanek wrote:
On Jan 21, 2010, at 14:47 , cstrato wrote:
Dear all,
I have the following problem:
My Bioconductor package xps requires both R and
ROOT, whereby I compile ROOT usually from source.
Until now this has not been a problem since I do all
my development on Tiger. However, now I wanted to
compile ROOT on Leopard but failed since my gfortran
compiler is too old, see:
http://root.cern.ch/phpBB2/viewtopic.php?t=9721&sid=003b01ddb64f33dbe0c74651ee089c34 To
solve the problem I should upgrade to the latest
version coming with gcc-4.4 in Fink.
I would advise against that - using gcc from Fink
usually calls for a disaster. The real solution is to
fix their flags, but that's another story ...
Thus my question is:
Can I also use this version for R or do I need to
use the version downloaded from att?
I would recommend using the Leopard gfortran 4.2.4
for Xcode 3.1.x from the Mac devel page
http://r.research.att.com/tools/
which I assume is what you meant by the latter.
Cheers,
Simon
_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-mac
_______________________________________________
R-SIG-Mac mailing list
R-SIG-Mac@stat.math.ethz.ch
https://stat.ethz.ch/mailman/listinfo/r-sig-mac
