Le lun. 08 déc. à 11:17, Dirk Eddelbuettel a écrit :
On 8 December 2008 at 10:54, Vincent Goulet wrote:
| > [EMAIL PROTECTED]:~$ grep blas /etc/R/Makeconf
| > BLAS_LIBS = -lf77blas -latlas
| > [EMAIL PROTECTED]:~$
| >
| > which is what you have, and which may have gotten there by
accident.
|
| Same here (obviously) on the build system. It seems to be because I
| have libatlas-base-dev installed. I rebuilt without the package and
| now /etc/R/Makeconf has "BLAS_LIBS = -lblas". Should I provide an
| updated binary?
Might be good as we otherwise force Atlas down onto everybody
compiling
Fortran. It should really be just -lblas -- which work with and
without
Atlas.
Done. Version 2.8.0-1intrepid1 On its way to CRAN now.
Two questions though:
1) Should we enforce this as Build-Conflicts: in debian/control ?
I'm not sure I'd go as far, especially given it's not a "conflict" per
se. This would be inconvenient for, say, the person building the
packages by hand and who knows full well what he/she is doing.
I'd rather say: let's be careful in the future...
2) Related to 1, I still do not understand who this happened on the
MIPS
machines breaking the Debian builds (currently two Rmetrics packages
with
Fortran fail because of this). There will be a new R package around
Dec 22
so that should shake things up.
Sorry, can't comment on this one.
Vincent
Dirk
--
Three out of two people have difficulties with fractions.
_______________________________________________
R-SIG-Debian mailing list
R-SIG-Debian@r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-debian