svn commit: r452962 - head/math/libRmath

[Joseph Mingrone]

Tijl Coosemans <tijl at FreeBSD.org> writes:
> So I'd like to propose the following compromise:
> We keep USES=fortran:gcc and fortran:flang but also introduce
> DEFAULT_VERSIONS=fortran and let plain USES=fortran choose the Fortran
> compiler based on that.  Then math/R can provide a flang option that
> sets USES=fortran:flang when on, and USES=fortran when off.  So the
> description of the option would be something like "Use flang instead
> of the default Fortran compiler".  The default for this option would
> be off, also on amd64.

Sounds good [1].  I am even fine with your original proposal as long as
R works for users without tinkering [2]. 

Yuri, could you share your thoughts?  Based on the comments on bug
220359, I assume these proposed changes will not have a negative impact
on devel/rstudio.  Is that true?

Thanks,

J.

[1] Assuming that after testing with gfortran everything is again
    working fine.  A poudriere build is still running.

[2] Unfortunately switching fortran compilers probably means users will
    have to reinstall R packages installed with install.packages().
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 962 bytes
Desc: not available
URL: <http://lists.freebsd.org/pipermail/svn-ports-all/attachments/20171114/4554b76e/attachment.sig>