[Rd] cross-compiling + expm1

Prof Brian D Ripley ripley at stats.ox.ac.uk
Sun Jun 27 22:40:26 CEST 2004


Another thought: you are not by any chance using sources that have been
configured for the host?  configure on the host would give you
inappropriate files in src/include and then the Windows make would not
create the correct ones.

src/include/Rmath.h should be dated almost immediately after you first ran
make in src/gnuwin32.  To be sure, I would run make distclean and start
again.

Another way to check include paths is to look at the *.d files.

On Sun, 27 Jun 2004, Prof Brian Ripley wrote:

> If set up correctly, this _does_ work.  There is no known bug in the R
> files for cross-compiling, and it _has_ been tested.
>
> Something is seriously wrong with your cross-compiling setup, which since
> you do not credit I presume you made yourself.  See below for some
> guesses as to what might be wrong.
>
> On Sun, 27 Jun 2004, Jonathan Taylor wrote:
>
> > Hello all,
> > 	Just joined this mailing list -- not sure if this is the right list to
> > send this question, but I have a question about cross-compiling R. I am
> > working with R-1.9.1.tgz.
> > 	It may just be with my version of mingw32, but it seems that expm1 is
> > not defined, so I tried to ensure that HAVE_EXPM1 was #undef'ed before
> > cross-compiling. The problem is that, in <include/Rmath.h> if it is
> > #undef'ed it is then #define'd as 1, meaning that the line where expm1
> > is declared automatically not included.
>
> That's not right.  First, the file is called Rmath.h but it is in
> src/include not include.  Second, the Windows Rmath.h has
>
> #ifndef HAVE_EXPM1
> #endif
>
> so I guess your cross-compiling is set up to use the Rmath.h from the host
> and not the target.  Incorrectly built cross-compilers do get the include
> paths wrong, and I have seen several such.  I suggest you use gcc -v and
> check your include paths.
>
> > I had similar problems with
> > logpq, but after changing Rmath.h cross-compiling was straightforward.
> > My question is: is this a "bug" in Rmath.h with respect to expm1 and
> > logp1? If so, how would you cross-compile, or even compile, if you
> > don't have expm1 or logp1?
>
> Not relevant, as the correct Rmath.h will define expm1/log1p (sic, not
> logpq or logp1) iff it is not present in the system's runtime files.
>
> You should have log1p in your mingw32 files, and if you do not your
> system is too old (and the file src/gnuwin32/INSTALL does have warnings
> about 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
>
> ______________________________________________
> R-devel at stat.math.ethz.ch mailing list
> https://www.stat.math.ethz.ch/mailman/listinfo/r-devel
>
>

-- 
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 272860 (secr)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595



More information about the R-devel mailing list