R-1.7.0 'make check' fails on reg-tests-1

You appear to have a problem with the accuracy of the arithmetic on that
machine. Please run the test by hand and tell us what N and max(dd) are
when the test fails (or is it the mvftt test?)

The test is not new, but the change in random-number generator will affect 
what values are tested.

On Tue, 20 May 2003, James F. Reid wrote:

Dear all,

I am trying to upgrade R-1.6.2 to R-1.7.0 on an SGI running IRIX6.5.

Everything compiles but I get an error in the testing phase:

[snip]
running regression tests
make[3]: Entering directory `/usr/local/src/R-1.7.0/tests'
running code in 'reg-tests-1.R' ...make[3]: *** [reg-tests-1.Rout] Error 1
make[3]: Leaving directory `/usr/local/src/R-1.7.0/tests'
make[2]: *** [test-Reg] Error 2
make[2]: Leaving directory `/usr/local/src/R-1.7.0/tests'
make[1]: *** [test-all-basics] Error 1
make[1]: Leaving directory `/usr/local/src/R-1.7.0/tests'
make: *** [check] Error 2

this is the part in reg-tests-1.Rout.fail which fails:

## fft
set.seed(123)
eps <- 1e-11
for(N in 1:130) {

+     x <- rnorm(N)
+     if(N %% 5 == 0) {
+       m5 <- matrix(x,ncol=5)
+       stopifnot(apply(m5,2,fft) == mvfft(m5))
+     }
+     dd <- Mod(1 - (f2 <- fft(fft(x), inverse=TRUE)/(x*length(x))))
+     stopifnot(dd < eps)
+ }


Any help concerning this matter would be very much appreciated.

Best regards,
James Reid.