On Sat, 2 Apr 2005, Peter Dalgaard wrote:
[EMAIL PROTECTED] writes:
Full_Name: Teru KAMOGASHRIA Version: 2.0.1 OS: SunOS sun 5.10 Generic i86pc i386 i86pc Submission from: (NULL) (219.5.176.24)
rbinom.c cannot be compiled because of the casting problem.
/opt/SUNWspro/bin/cc -I. -I../../src/include -I../../src/include -I/usr/local/include -DHAVE_CONFIG_H -D__NO_MATH_INLINES -g -c rbinom.c -o rbinom.o "rbinom.c", line 60: operand must have real floating type: op "isfinite" cc: acomp failed for rbinom.c
Following patch may solve this problem.
*** R-2.0.1.orig/src/nmath/rbinom.c Mon Nov 15 21:33:01 2004 --- R-2.0.1/src/nmath/rbinom.c Sun Apr 3 00:19:52 2005 *************** *** 57,63 **** n = floor(nin + 0.5); if (n != nin) ML_ERR_return_NAN;
! if (!R_FINITE(n) || !R_FINITE(pp) || /* n=0, p=0, p=1 are not errors <TSL>*/ n < 0 || pp < 0. || pp > 1.) ML_ERR_return_NAN;
--- 57,63 ---- n = floor(nin + 0.5); if (n != nin) ML_ERR_return_NAN;
! if (!R_FINITE((double)n) || !R_FINITE(pp) || /* n=0, p=0, p=1 are not errors <TSL>*/ n < 0 || pp < 0. || pp > 1.) ML_ERR_return_NAN;
Hmm... But does it at all make sense to pass an integer to R_FINITE?
Dunno. It was in the original (r574 code), with finite rather than R_FINITE. It would make sense to test nin (rather than n).
and why isn't there a prototype causing automatic casting anyway?
Because isfinite is a macro. (Solaris 10 is rather new, which is probably why we have not seen this before. We received our CDs last week.)
-- Brian D. Ripley, [EMAIL PROTECTED] 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@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-devel
