[R-SIG-Mac] R CMD check blind and deaf

[roger koenker]

This may repeat prior advice given here, or elsewhere, if so I wasn't
able to find it.  But I thought it might be worthwhile to report that a
possible consequence of using gfortran with libgfortran*dylibs is
that R CMD check is essentially  disabled.  I learned this last
May  from Simon  (see message below) had forgotten it, and then
experienced the same thing on my office G5 today, had a mild
sensation of deja vu, and fortunately was
able to dig out the prior email.  Removing  libgfortran*dylib
in /usr/local/lib and recompiling R has restored the functionality
of R CMD check.  Except for this behavior, the prior version
of R seemed fine....

Thanks again to Simon for this advice and all his other efforts!

Roger


url:    www.econ.uiuc.edu/~roger            Roger Koenker
email    rkoenker at uiuc.edu            Department of Economics
vox:     217-333-4558                University of Illinois
fax:       217-244-6678                Champaign, IL 61820

On May 22, 2005, at 5:48 PM, Simon Urbanek  wrote:

On May 22, 2005, at 5:48 PM, roger koenker wrote:


>>> PS.  I have another peculiar problem that I thought you might have
>>> encountered or have some idea about.  I upgraded my laptop G4 to  
>>> OS X
>>> 10.4 (tiger) and managed to install R 2.1.0 using Simon's gcc  
>>> bundle.
>>> Everything looked quite fine for a while but then I found running R
>>> CMD check on quantreg it seems that example checking and vignette
>>> checking are somehow disabled.  What I mean is that they appear  
>>> to be
>>> done in the sense that they appear in the standard output as "OK"  
>>> but
>>> when you look in the quantreg-Ex.Rout file all you see is the R
>>> startup header and a single prompt.
>>>

My first guess is that you're using gfortran with dynamic libgfortran  
library. You may look for Brian's post about this. My work-around is  
to use static library (i.e. remove libgfortran*dylib before compiling  
R), but I think Brian added a flag that allows you to compile R with  
static gfortran libs even if dynamic ones are present. A simple test  
is as follows: create some dummy R script (let's say foo.R) that  
prints something and run R --vanilla < foo.R then try cat foo.R | R -- 
vanilla . If the latter works but the first doesn't then you have  
that gfortran problem. It basically prevents R from running any tests.

Cheers,
Simon