Subject: Re: perl-package: Faulty test
To: None <hermit@cs.tu-berlin.de>
From: Bill Studenmund <skippy@macro.stanford.edu>
List: current-users
Date: 11/11/1997 10:13:23
On Tue, 11 Nov 1997, Bjoern Labitzke wrote:
>
> Hello...
>
> I am running NetBSD-current on an Amiga 4000/40. Perhaps this information
> helps to find the problem (?).
>
> I made the perl5-package from the package-system and installed it. Both
> went quite well, but "make test" was mentioned. Because I prefer to know
> that errors in results are caused by my program and not by the system
> itself, I ran the tests.
>
> The tests for lib/gdbm and lib/odbm were skipped for my platform, but
> lib/complex reported errors. Running the test by itself showed three
> problems:
>
> 1) cot(acot(z)) reported 1 for z=1.2. 1.2 would be correct.
> 2) tan(atan(z)) reported 1.095032... for z=1.2. 1.2 would be correct.
> 3) another wrong calculation was reported, which included longer numbers,
> so I didn't copy it.
>
>
> Because of the automatic skipping of lib/gdbm and lib/odbm I expect some
> problems with the m68k-architecture. I didn't follow it closely, but IIRC
> there was a architecture related discussion about libm. Could it be that
> both problems are related?
It turns out the error was architecture independant. atan's been broken
for a while. It will be fixed shortly. :-)
Take care,
Bill