[R] Installing on 64-bit Dec or SGI

[George White]

I've "compiled" R on an OSF1 alpha and on Irix 6.2 and 6.5 (R4000):

$ uname -a
OSF1 atlantic.bio.dfo.ca V4.0 878 alpha
./configure --prefix=$HOME
[...]
R is now configured for alpha-dec-osf4.0

  Source directory:         .
  Installation directory:   /usr2/gwhite
  C compiler:               cc -ieee_with_inexact -g
  FORTRAN compiler:         f77 

$ uname -a
IRIX newton 6.5 05190003 IP22 mips
$ ./configure --prefix=$HOME
[...]
R is now configured for mips-sgi-irix6.5

  Source directory:         .
  Installation directory:   /usr/people/gwhite
  C compiler:               cc -g
  FORTRAN compiler:         f90

$ uname -a
IRIX caligo 6.2 03131015 IP22
./configure 
[...]
R is now configured for mips-sgi-irix6.2

  Source directory:         .
  Installation directory:   /usr/local
  C compiler:               gcc -O2 -I/usr/local/include.bugs
  FORTRAN compiler:         g77 -O2

The Irix 6.2 system is a mess due to the number of patches that have
been applied, reversed, etc.  Some headers are just missing and had to
be copied from another system into "/usr/local/include.bugs".

On the Irix 6.5 system, there were some name clashes which required
additions to the "SGI nonsense" section of src/include/Mathlib.h:

...3.2/src/include $ diff -C 3 Mathlib.h.000 Mathlib.h
*** Mathlib.h.000       Wed Dec 23 05:54:25 1998
--- Mathlib.h   Tue Feb  2 14:18:26 1999
***************
*** 154,159 ****
--- 154,165 ----
  #ifdef qexp
  #undef qexp
  #endif
+ #ifdef qgamma
+ #undef qgamma
+ #undef qlgamma
+ #undef qsigngam
+ #endif
+ 
  
        /* Name Hiding to Avoid Clashes with Fortran */
---------------------------
  
There are problems with both the OSF1 and Irix 6.5 versions.  In both
cases I used the vendor's compilers.  Both these systems are very fresh,
out of the box installations with no patches applied.  SGI provides perl
5, but OSF1 (at least with the options installed on the system available
to me) does not, so I don't get documentation on OSF1.  Furthermore, tick
marks and some other lines (e.g., in boxplots) are missing from plots when
I use the x11 device, but present using the PS device.

The SGI version gives:

> demo(lm.glm)
[...]
> anova(lmwg <- lm(weight ~ group))
Analysis of Variance Table

Response: weight
Warning: NAs introduced by coercion
Error: names attribute must be the same length as the vector

I made more headway using Irix 6.2 and GNU compilers, but had to use GNU
rx instead of the SGI regcomp routines, which gave:

> gsub("([ab])", "\\1_\\1_", "abc and ABC")
Error: invalid backreference in regular expression

Finally, I want to use the cluster package, but the "daisy" function
fails on a simple example for:

> codes(as.ordered(NULL))
Error: no applicable method for "codes"

Is this a bug or a change in the definition of "codes"?

--
George White <aa056 at chebucto.ns.ca>  tel: 902.426.8509
  Bedford Inst. of Oceanography, Nova Scotia, Canada.

On Tue, 9 Feb 1999, Gregory R. Warnes wrote:

> 
> Hi all, 
> 
> The systems guys here in the stat dept don't seem to be able to
> compile R on the Dec Alphas or on the SGIs.  Can anyone give them a hand?
> 
> -Greg
> 
> -------------------------------------------------------------------------------
>     Gregory R. Warnes          | It is high time that the ideal of success
> warnes at biostat.washington.edu  |  be replaced by the ideal of service.
>                                |                       Albert Einstein
> -------------------------------------------------------------------------------
> 
> ---------- Forwarded message ----------
> Date: Tue, 9 Feb 1999 10:06:57 -0800 (PST)
> From: Richard Fairfield <rcf at ms.washington.edu>
> To: "Adrian E. Raftery" <raftery at stat.washington.edu>
> Cc: Statistics Help <help at stat.washington.edu>
> Subject: Re: installing R on monto
> 
> 
> Adrian:
> 
> We are unable to compile this software on either platform. I don't think
> there is anything more we can do at this time. 
> 
> Here are several options for the future:
> 
> 1) If you know of anyone who has successfully compiled this software on
>    either DEC Alpha (running Dunix 4.0) or on a 64 bit SGI (running Irix 6.2)
>    then maybe they would share their work with us.
> 
> 2) We can watch for future versions of the software. A newer version may
>    fix the comilation problems.
> 
> 3) Maybe you have a grad student who knows Fortran who could work on the
>    code?
> 
> that's all for now,
> rcf
> 
> On Fri, 5 Feb 1999, Adrian E. Raftery wrote:
> 
> > Thanks, Rick. - Adrian
> > 
> > On Fri, 5 Feb 1999, Richard Fairfield wrote:
> > 
> > > Today I tried to build "R" on both a DEC Alpha and on an SGI. It failed
> > > in both cases. There errors looked a bit less formidable on the Alpha than
> > > on the SGI. We are going to have one of our grad students employees look
> > > at the DEC Alpha version to see if he can isolate the error.

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._