On Thu, Oct 05, 2017 at 12:24:08PM +0200, Rafael Sadowski wrote: > On Thu Oct 05, 2017 at 06:42:32AM +0000, Ingo Feinerer wrote: > > update math/R 3.4.1 -> 3.4.2 > > OK? > > I ran into: > > ===> Regression tests for R-3.4.2 > ... > gmake[3]: Entering directory '/usr/ports/pobj/R-3.4.2/build-amd64/tests' > running code in '/usr/ports/pobj/R-3.4.2/R-3.4.2/tests/complex.R' > ...gmake[3]: *** [/usr/ports/pobj/R-3.4.2/R-3.4.2/tests/Makefile.common:100: > complex.Rout] Error 1
The problem is on -current: > (z <- tan(1+1000i)) [1] NaN+1i At first glance it seems like a regression in OpenBSD: 3.4.2/amd64/-current: 3.4.1/amd64/-current: 3.4.0/amd64/-current: 3.3.3/amd64/-current: > (z <- tan(1+1000i)) [1] NaN+1i Warning message: In tan(1 + (0+1000i)) : NaNs produced in function "tan" 3.3.3/i386/6.1 (kindly tested by ratchov@): > (z <- tan(1+1000i)) [1] 0+1i I investigated libm commits (http://cvsweb.openbsd.org/cgi-bin/cvsweb/src/lib/libm/) but found nothing obvious (for me). Any ideas? Best regards, Ingo
