[Rd] Problems with 64bit dll compile in R-2.15.0

[Duncan Murdoch]

On 12-05-10 2:07 PM, Marian K Talbert wrote:
> I built my package under the R version 2.14.1 on windows without any
> problems by first checking for issues using R CMD check (no warnings) and
> then R CMD build to build the tar.gz.  I can install this on version
> 2.14.1 using install.packages(...).  I next tested to see if I could also
> install on version 2.15.0 (also on my computer) using install.packages()
> with the version 2.14.1 tar.gz this failed  with the error message
>
> sorry, unimplemented: 64-bit mode not compiled in
> make: *** [InternalFcts.o] Error 1
> ERROR: compilation failed for package 'Blossom'

The make files have been changed to be compatible with a newer version 
of gcc.  You should update your Rtools installation to get the new one.

Duncan Murdoch

> I then tried to check the package using the 2.15.0 version (after changing
> my path to specify this version).  The R CMD check  reports the following:
> warning:  Clock skew detected.  Your build may be incomplete.
> installing to H:/Desktop/PkgBuild/BlossomPkg.Rcheck/Blossom/libs/i386
>
> *** arch - x64
> make: Warning: File `InternalFcts.f90' has modification time 0.52 s in the
> future
> gfortran -m64   -O2  -mtune=core2 -c  InternalFcts.f90 -o InternalFcts.o
> f951.exe: sorry, unimplemented: 64-bit mode not compiled in
> make: *** [InternalFcts.o] Error 1
> gfortran -m64   -O2  -mtune=core2 -c  InternalFcts.f90 -o InternalFcts.o
> f951.exe: sorry, unimplemented: 64-bit mode not compiled in
> make: *** [InternalFcts.o] Error 1
> ERROR: compilation failed for package 'Blossom'
> * removing
>
> I'm not sure why this issue would come up only on the newest version of R.
>   I'm working on Windows 7 64bit.  My package has some internal code
> written in Fortran and C.  I don't have a makefile in my src but this
> didn't affect anything before so I'm not sure why it would under the new
> version.  Other thoughts I've had are a possible mismatch between some of
> the Rtools and the newer version or R or might there possibly be more
> rigorous checks on the dll build for newer versions?  Anyway, if anyone
> has any insight into what might cause this to break under the new version
> or R, it would be greatly appreciated.   Also when I do build the package
> (using 2.14.1)  it won't install using install.packages() on another
> computer.  Might this be because the other compute lacks the tools
> required to build the dll?  If so, what's the best way to build a tarball
> that can be installed by anyone on their computer which might not have the
> tools require for building packages on windows.
>
> Thanks,
>
> Marian Talbert
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel