Skip to content

installation of affy under Mac OS X 10.5.6 (Leopard)

12 messages · Massimo Pinto, Sean Davis, Kjell Konis +3 more

Original
Kjell Konis
# 
It looks like you don't have all the development tools installed.  The  
missing library /usr/local/lib/libgfortran.2.dylib is part of the  
Fortran compiler which is included with the full R installer.

Kjell
On 21 avr. 09, at 15:10, Massimo Pinto wrote:

            

      
      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Simon Urbanek
    

    

      
      #
    

  


  

      
Massimo,

        
On Apr 21, 2009, at 9:10 AM, Massimo Pinto wrote:

        

            

      
      
      
    

        
This shows you what kind of R you are running (i.e. where was the R  
compiled), not what your OS is. You are running the Tiger release of  
R, which is ok since it is designed for Mac OS X 10.4 (Tiger) and  
later which includes your 10.5.

            

      
      
      
    

        
The "therefore" is wrong here ;).

            

      
      
      
    

        
The problem above doesn't come from affy, but from the preprocessCore  
package (see the error). The binary for preprocessCore you use is not  
quite correct. It uses local Fortran libraries on the machine that  
built it where it should be using the Fortran library inside R. If  
this is a binary from the BioC repository, please notify the  
maintainers, otherwise re-install the binary of preprocessCore from  
BioC. Note the the correct mailing list for such questions is BioC.

A work-around (that fixes the symptom, not the problem) is to install  
the full R (you seem to have only the mini version installed) which  
includes a local Fortran compiler - or you can install the Fortran  
manually from
http://cran.r-project.org/bin/macosx/tools/

Cheers,
Simon

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      James W. MacDonald
    

    

      
      #
    

  


  

      
And downloading the tarball and installing is a recipe for failure. If 
you want to install source packages, just use the correct arguments.

biocLite("affy", type = "source")

will get you the correct affy version plus all required dependencies.

Best,

Jim

        
Kjell Konis wrote:

            

      
      
      
    

        

      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Simon Urbanek
    

    

      
      #
    

  


  

      
On Apr 21, 2009, at 10:48 , James W. MacDonald wrote:

        

            

      
      
      
    

        
Why? People that don't have direct internet connection do that on a  
regular basis. You only have to make sure you don't unpack the content  
involuntarily.

            

      
      
      
    

        
I don't think Massimo was talking about source packages. That would  
require much more - you'll need Xcode tools etc. and you wouldn't see  
the error he was having since you'd be compiling it yourself. Note  
that on Mac binary packages are also tar balls and R knows what to do  
with them ...

Cheers,
Simon

PS: Normally you don't need development tools for binary packages - as  
I was pointing out there is something wrong with the binary Massimo  
was using.

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      James W. MacDonald
    

    

      
      #
    

  


  

      
Hi Simon,

        
Simon Urbanek wrote:

            

      
      
      
    

        
The recipe for failure is directed towards BioC packages, which tend to 
have complicated dependencies. A significant proportion of the questions 
on the BioC listserv are due to people downloading things directly and 
installing and then wondering why they don't work (because they have the 
wrong version, are missing dependencies, etc).

The affy package is a perfect example. It relies on affyio and 
preprocessCore, so if any one package is out of sync you can get weird 
errors that would have been avoided by using biocLite().

            

      
      
      
    

        
Good call. I somehow translated .tgz to .tar.gz when I read the original 
post. My bad.

Best,

Jim

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

  
    

      
      
    

  


  


      

    

    

      
      

        


  

        

      Patrick Aboyoun
    

    

      
      #
    

  


  

      
Massimo,
I'll break down these issues in two and go through them in turn:

1) stale binaries for R 2.8 at bioconductor.org
You had found that the Mac OS X binary packages at bioconductor.org that 
used fortran, like preprocessCore, had a bad gfortran link in them. I 
updated those packages last week and they are now properly redirected to 
the gfortran library that is provided with the pre-built version of R 
2.8. To fix this issue, please download the troublesome libraries again 
for a fix.

2) Determining which binary packages are appropriate for your Mac
I think you were downloading the Mac OS X 10.4 binary packages from 
bioconductor.org rather than the Mac OS X 10.5 binary packages because 
your pkgType == "mac.binary" instead of "mac.binary.leopard". You and 
either perform a global setting of the pkgType to "mac.binary.leopard" 
using options("pkgType" = "mac.binary.leopard") or specify type = 
"mac.binary.leopard" when using the biocLite / install.packages functions.

I have also modified the underpinnings of biocLite to produce a warning 
message when you specify type = "mac.binary" when you are running on Mac 
OS X 10.5 machine with some suggestions on what you can do to avoid this 
type of problem in the future.


Cheers,
Patrick

        
Simon Urbanek wrote:

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Simon Urbanek
    

    

      
      #
    

  


  

      
Patrick,

        
On Apr 21, 2009, at 14:53 , Patrick Aboyoun wrote:

        

            

      
      
      
    

        
Do NOT do that manually! You can only use Tiger binaries in Tiger  
build of R even if you are running Leopard system, so the default  
setting is there for a reason (at least in CRAN binaries). The OS  
version is entirely irrelevant in this matter.

            

      
      
      
    

        
That will be very bad for the reason above ... Are you sure you mean  
that? Or am I missing something - are you talking about the running OS  
or the R build OS?

Cheers,
Simon

            

      
      
      
    

        

      
      
    

            

      
      
      
    

        

      
      
    

            

      
      
      
    

  
  


  


      

    

    

      
      

        


  

        

      Patrick Aboyoun
    

    

      
      #
    

  


  

      
Simon,
I just tested the scenario of using standard R-2.x.x.dmg from CRAN is 
installed on a Mac OS X 10.5 machine and download the standard set of 
BioC packages and it worked as you predicted using the default type = 
"mac.binary". My bad. The OS version was a red herring.


Patrick

        
Simon Urbanek wrote: