Re: acscd.m failing

[Daniel J Sebald]

John W. Eaton wrote:
> On  8-Feb-2007, David Bateman wrote:
>
> | Daniel J Sebald wrote:
> | > Make check either hangs or takes a very long time on the following:
> | >
> | > liboctave/dMatrix.cc ................................... PASS    6/6
> | > scripts/elfun/acosd.m .................................. PASS    3/3
> | > scripts/elfun/acotd.m .................................. PASS    3/3
> | > scripts/elfun/acscd.m ..................................
> | >
> | > The acosd.m and acotd.m tests take a fraction of a second while the
> | > acscd.m test sits for at least a half hour (I broke out of the
> | > program) with CPU pegged at 100%.  Not sure why cosecant should fail
> | > while cotangent is fine.  I tried following the fntests.m script but
> | > it is somewhat convoluted.
> | >
> | > Dan
> | >
> | Very odd. I don't see this, the check goes straight through this section...
>
> The tests for acscd are just
>
> %!error(acscd())
> %!error(acscd(1,2))
> %!assert(acscd(0:10:90),180/pi*acsc(0:10:90),-10*eps)
>
> so can you try evaluating the expressions
>
>   acscd(0:10:90),
>   180/pi*acsc(0:10:90)
>
> and see which one of them is hanging?  On my system acsc(0) returns
> NaN + NaNi for 0, so maybe that is the problem case?  Should we maybe
> be avoiding this argument value in the tests?

Yes, that is it.  On my system I get:

octave:1> acsc(0)
warning: division by zero
Press Control-C again to abort.
panic: Interrupt -- stopping myself...
attempting to save variables to `octave-core'...
save to `octave-core' complete

address@hidden ]$

Values between 0 and 1 come out as complex, so that is working.

Naturally the result should be something graceful for csc(0), but this is 
probably a math library causing the problem, right?

Dan