[R] trouble compiling on Windows

[Prof Brian Ripley]

On 31/03/2012 10:38, Joshua Wiley wrote:

> I may be putting my foot in my mouth, but I want to say that Rtools
> changed at 2.15 from mingw to gcc.  I would check that you have the
> most recent version of Rtools installed, make sure you have the right
> directories in your path environment variable (which also changed, I
> believe), and make sure you use the default tool chain in
> Mkrules.local if you set it.

I think the 'giving credit where credit is due' dept requires an update 
there.

All builds of R under Windows in recent years have used GCC ('GNU 
Compiler Collection') and GNU binutils (for the assembler, linker, ...). 
   All 64-bit Windows builds have used the ports developed for the 
MinGW-w64 project, as well as their runtime ('libc', 'libm' and more, 
including headers).

Prior to R 2.14.2 we used for 32-bit Windows the ports from the 
Mingw.org project and their runtime.  We now use the 32-bit versions 
from MinGW-w64.

Erin had libgomp.spec missing.  There are MinGW-w64 toolschains out 
there with it missing, but the master sources for the toolchain I 
compiled does have it.

The OPENMP setting in src/gnuwin32/MkRules.dist should not need changing 
if the correct toolchain is used (and that the binary distributions 
manage to get built with the default setting is some confirmation of this).

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595