[R-SIG-Win] New Rtools sent to CRAN

[Dan Tenenbaum]

On Tue, Sep 22 2015, Duncan Murdoch wrote:

> I have just uploaded a new build of the Rtools33.exe installer to CRAN.  
> It should appear on the mirrors within a day or two.
> 
> This installer includes the current build of gcc 4.9.3 from 
> http://www.stat.ucla.edu/~jeroen/mingw-w64. It also includes
> Jeroen's build of libicu55, which is necessary to build R.
> 
> The current R-devel build on CRAN should have the necessary changes to 
> build R using this toolchain, 

Can you point to the specific changes (svn revision numbers) that were made to R-devel to facilitate this?

> but the default is still the previous gcc 
> 4.6.3 toolchain; users should follow the instructions on 
> https://github.com/rwinlib/r-base#readme to set up their systems to use it.
> 
> Please post any problems here or to me or Jeroen, as appropriate. We're 
> hoping to be able to use this toolchain in R 3.3.x.

I installed this toolchain and the installer suggested adding the following to the beginning of my PATH:

c:\Rtools\bin;
c:\Rtools\gcc-4.6.3\bin;

Is that correct? Should it (also?) have suggested 

C:\Rtools\mingw_32\bin and C:\Rtools\mingw_64\bin?
Or does the new R know where to look for those? If so, how does it know?

Are the R-devel binaries available from CRAN being built with this toolchain?

I myself am not interested in building R from source with this toolchain but rather using it to build packages that have C/C++/Fortran code. 

Is there a way to tell which compiler is being used when I compile such a package?

R CMD config CC
just tells me "gcc -m64" it does not give the path to, or version of, the gcc that's being used.

I ran the Bioconductor builds overnight using this toolchain and recent R-devel and did not notice anything odd with packages containing native code, which means either the new toolchain is working perfectly (which is wonderful) or gcc-4.6.3 is still being used and maybe I need to do something to faciliate the use of gcc-4.9.3. Can you advise?

Thanks,
Dan