Skip to content

Problems installing packages on Lion

8 messages · R. Michael Weylandt, Colstat, David Winsemius +2 more

#
On Mar 9, 2012, at 6:30 PM, Colstat wrote:

            
Your version of R is out-of-date. Binaries are only reliably synced to  
the current major version (and you are asked not to post problems  
until you _have_ upgraded.)
David Winsemius, MD
West Hartford, CT
#
Following on what David said,

Are you equipped to build the package locally? That is, do you have
gcc and g++ installed?

I'm almost certain that a local build (or an upgrade to 2.14 or 2.15
in just over two weeks time) will be the easiest way to fix things.

Michael
On Fri, Mar 9, 2012 at 6:30 PM, Colstat <colstat at gmail.com> wrote:
#
On Mar 9, 2012, at 6:42 PM, R. Michael Weylandt wrote:

            
I'm not sure why you say that. He seems to have a fairly standard / 
R.framework/ location for his /Resources/ and the fact that he still  
posting in HTML suggests he is not likely to have compiled his R from  
source.  A few years ago I would often solve problems by building from  
source, but in recent years when there is a binary package available,  
it generally installs without problems.

(I do see that there is a ../bin/../2.13/  node in the CRAN mirror at  
ku, which is not typical, so perhaps we are both wrong here. Now that  
I check, both the Berkeley and the CMU CRAN mirrors seem to keep their  
Mac binaries a lot longer than is required, so KU is just following US- 
local tradition. The CMU mirror has binaries back to R 2.8.)

Is your experience different?
#
Oh -- perhaps I was ambiguous and I apologize for any confusion that
caused: I meant to just build the package himself from source so that
it would match his architecture, hence why gfortran wouldn't be needed
as it would to build R.

Incidentally, doesn't CRAN usually support one minor version back? It
seems that it checks for "R-oldrel" (which appears to currently be
2.13.2) -- I don't know if the patch releases are ever incompatible,
though I doubt so since things seem to divide into R x.y folders,
rather than R x.y.z in the CRAN servers.

Michael Weylandt
On Fri, Mar 9, 2012 at 7:25 PM, David Winsemius <dwinsemius at comcast.net> wrote:
#
On 10/03/2012 00:53, R. Michael Weylandt wrote:
Yes, that's pretty much the policy.  But with 2.15.0 alpha, 2.14.x is 
now 'one minor release back', and building for 2.13.x has been switched 
off (it seems to have been for Macs on Feb 10, a little earlier than I 
would have expected -- the Windows last updates were on Feb 28).

  
    
#
On Mar 10, 2012, at 12:59 AM, Prof Brian Ripley wrote:

            
On the Mac side we never build more than two releases so whenever we start building R-devel packages on daily basis is the time that we stop building R-oldrel. In fact it is very common that after a release packages are coming in requiring the new release so it becomes increasingly pointless to run oldrel builds -- so historically they have been switched off even much earlier. This may be the case even more so in the future because of the new R release schedule and the fact that we may move up OS X target version which will make building old releases even harder.

Cheers,
Simon