R help - Feb 2008 - R-2.6.2 installation (64bit) falling over with the grid package

It's a long time since I had a problem compiling R, but I've now
encountered one with R-2.6.2 on Fedora Core 6 (64bit).

The problem arises when the installation process gets to the grid
package, the unit.c part in particular.  This is the output I get:


gcc -std=gnu99 -I../../../../include -I../../../../include  -I/usr/local/include
-fvisibility=hidden -fpic  -g -O2 -c unit.c -o unit.o
unit.c: In function ??pureNullUnit??:
unit.c:332: error: stray ??#?? in program
unit.c:332: error: ??R_fcall?? undeclared (first use in this function)
unit.c:332: error: (Each undeclared identifier is reported only once
unit.c:332: error: for each function it appears in.)
make[5]: *** [unit.o] Error 1
make[5]: Leaving directory `/usr/local/R-2.6.2/src/library/grid/src'
make[4]: *** [all] Error 2
make[4]: Leaving directory `/usr/local/R-2.6.2/src/library/grid/src'
make[3]: *** [all] Error 1
make[3]: Leaving directory `/usr/local/R-2.6.2/src/library/grid'
make[2]: *** [R] Error 1
make[2]: Leaving directory `/usr/local/R-2.6.2/src/library'
make[1]: *** [R] Error 1
make[1]: Leaving directory `/usr/local/R-2.6.2/src'
make: *** [R] Error 1


I installed the same tgz file on a 32-bit Mepis machine and on a
CentOS5 machine without such any problem .  All previous 64bit
compilations have also been without a problem.

There has been a problem with one PCI slot on the machine in question
after an electrical misadventure.  If nobody else has had a problem on
64 bit machines, could it be that I have a hardware problem?  Seems a
bit unlikely it would get as far as it did were that the case, but I
could be convinced otherwise.

Ideas? Other experience?

-- 
~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.   
   ___    Patrick Connolly   
 {~._.~}          		 Great minds discuss ideas    
 _( Y )_  	  	        Middle minds discuss events 
(:_~*~_:) 	       		 Small minds discuss people  
 (_)-(_)  	                           ..... Anon
	  
~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.~.

It's your system: try unpacking the tarball again.  That line is

             eval(R_fcall3, R_gridEvalEnv);

and has no # there.

I suspect FC6 systems are thin on the ground (it is end-of-life, I 
believe), but R-2.6.2.tar.gz has been installed on many, many 64-bit Linux 
systems (e.g. CRAN tests on one), including an FC5 system (which we still, 
just, have for testing compatibility with an old OS).

People seem to fail to appreciate the effort that goes into checking that 
the release tarballs are complete and uncorrupted.

On Sun, 24 Feb 2008, Patrick Connolly wrote:
> It's a long time since I had a problem compiling R, but I've now
> encountered one with R-2.6.2 on Fedora Core 6 (64bit).
>
> The problem arises when the installation process gets to the grid
> package, the unit.c part in particular.  This is the output I get:
>
>
> gcc -std=gnu99 -I../../../../include -I../../../../include 
-I/usr/local/include   -fvisibility=hidden -fpic  -g -O2 -c unit.c -o unit.o
> unit.c: In function ????pureNullUnit????:
> unit.c:332: error: stray ????#???? in program
> unit.c:332: error: ???R_fcall???? undeclared (first use in this function)
> unit.c:332: error: (Each undeclared identifier is reported only once
> unit.c:332: error: for each function it appears in.)
> make[5]: *** [unit.o] Error 1
> make[5]: Leaving directory `/usr/local/R-2.6.2/src/library/grid/src'
> make[4]: *** [all] Error 2
> make[4]: Leaving directory `/usr/local/R-2.6.2/src/library/grid/src'
> make[3]: *** [all] Error 1
> make[3]: Leaving directory `/usr/local/R-2.6.2/src/library/grid'
> make[2]: *** [R] Error 1
> make[2]: Leaving directory `/usr/local/R-2.6.2/src/library'
> make[1]: *** [R] Error 1
> make[1]: Leaving directory `/usr/local/R-2.6.2/src'
> make: *** [R] Error 1
>
>
> I installed the same tgz file on a 32-bit Mepis machine and on a
> CentOS5 machine without such any problem .  All previous 64bit
> compilations have also been without a problem.
>
> There has been a problem with one PCI slot on the machine in question
> after an electrical misadventure.  If nobody else has had a problem on
> 64 bit machines, could it be that I have a hardware problem?  Seems a
> bit unlikely it would get as far as it did were that the case, but I
> could be convinced otherwise.
>
> Ideas? Other experience?
>
>
-- 
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