Skip to content

R-2.12.0 problem on Solaris 10

6 messages · Dr. David Kirkby, Brian Ripley, Zhang,Jun

Original
Dr. David Kirkby
#
On 11/ 9/10 09:22 PM, Zhang,Jun wrote:
Is this a 32-bit or 64-bit build? Is it Sun Studio or gcc?

There are issues with the R libraries when built with Sun Studio - they have 
relocations to read-only segments.

So info at

http://trac.sagemath.org/sage_trac/ticket/9040

However, the issues I have seem to be different to yours, so I expect you are 
either using Sun Studio or you are using a 32-bit build.

Dave
Dr. David Kirkby
#
On 11/ 9/10 10:30 PM, Dr. David Kirkby wrote:
I meant the issues are with gcc - SunStudio is ok.

I would be nice to be able to build R with gcc though. I know the gcc compiler 
is inferior, but some systems don't have SunStudio. For a project like Sage, we 
have 100 packages, and R is the only one that needs SunStudio to build. (I'd 
rather build the whole of Sage with SunStudio, but there are too many GNUisms 
for that). So I might have to use two compilers, which is a real pain in the butt

Dave
Brian Ripley
# 
I don't see enough information to help here, and this is rather 
an R-devel topic (see the posting guide for the difference).

Which compiler and which version?
What level of optimization and other compiler flags?
32- or 64-bit build?

The Sun Studio compiler works (see the R-admin manual and the package 
tests on CRAN), but Matrix does segfault in its tests on 32-bit 
Sparc.

32-bit gcc is less reliable, but it is currently working for me with 
gcc 4.3.3 from OpenCSW.

As a first step, compare the flags you used with those documented in 
the R-admin manual, and perhaps try minimal levels of optimization.
And please follow up to R-devel.
On Tue, 9 Nov 2010, Zhang,Jun wrote:

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Zhang,Jun
    

    

      
      #
    

  


  

      
Thank you Dave for the reply. I used solstudio12.2, and followed the R admin installation instruction for Solaris. When I run R as the compiled result, there is a line "Platform: sparc-sun-solaris2.10 (32-bit)". 

Jun


CC="cc -xc99"
CPPFLAGS="-I/opt/csw/include" 
CFLAGS="-O -xlibmieee"
F77=f95
FFLAGS=-O4
CXX="CC -library=stlport4"
CXXFLAGS=-O
FC=f95
FCFLAGS=$FFLAGS
LDFLAGS="-L/opt/csw/lib" 
export CC CPPFLAGS CFLAGS F77 FFLAGS CXX CXXFLAGS FC FCFLAGS LDFLAGS
./configure --prefix=/apps/R-2.12.0 --with-tcl-config=/opt/csw/lib/tclConfig.sh --with-tk-config=/opt/csw/lib/tkConfig.sh --disable-nls --with-recommended-packages=no

-----Original Message-----
From: Dr. David Kirkby [mailto:david.kirkby at onetel.net] 
Sent: Tuesday, November 09, 2010 5:07 PM
To: Zhang,Jun; r-help at r-project.org
Subject: Re: [R] R-2.12.0 problem on Solaris 10

        
On 11/ 9/10 10:30 PM, Dr. David Kirkby wrote:

            

      
      
      
    

        
I meant the issues are with gcc - SunStudio is ok.

I would be nice to be able to build R with gcc though. I know the gcc compiler 
is inferior, but some systems don't have SunStudio. For a project like Sage, we 
have 100 packages, and R is the only one that needs SunStudio to build. (I'd 
rather build the whole of Sage with SunStudio, but there are too many GNUisms 
for that). So I might have to use two compilers, which is a real pain in the butt

Dave

  
  


  


      

    

    

      
      

        


  

        

      Zhang,Jun
    

    

      
      #
    

  


  

      
I used solstudio12.2 to build 32-bit R. How do you get around the Matrix segfault? 

-----Original Message-----
From: Prof Brian Ripley [mailto:ripley at stats.ox.ac.uk] 
Sent: Wednesday, November 10, 2010 12:46 AM
To: Zhang,Jun
Cc: 'r-help at r-project.org'
Subject: Re: [R] R-2.12.0 problem on Sparc Solaris 10

I don't see enough information to help here, and this is rather 
an R-devel topic (see the posting guide for the difference).

Which compiler and which version?
What level of optimization and other compiler flags?
32- or 64-bit build?

The Sun Studio compiler works (see the R-admin manual and the package 
tests on CRAN), but Matrix does segfault in its tests on 32-bit 
Sparc.

32-bit gcc is less reliable, but it is currently working for me with 
gcc 4.3.3 from OpenCSW.

As a first step, compare the flags you used with those documented in 
the R-admin manual, and perhaps try minimal levels of optimization.
And please follow up to R-devel.

        
On Tue, 9 Nov 2010, Zhang,Jun wrote: