2.0.5 NaN and inf

[Ted Harding]

Hi Folks,
I'm posting this for a friend of mine who ain't on the lists yet.

He pulled and compiled octave-2.0.5 on a Red Hat 4.1 Linux system.
This has gcc-2.7.2.1, and there is no libieee.

He followed the instructions to modify the gcc specs file as in
readme.linux.

Now he observes that (for instance)

x=NaN -> x=0

NaN -> NaN=0

NaN=1 -> NaN=1

1/0 -> lots of messages and finally "inf"

-1/0 crashes octave ("memory exhausted")

inf can also be redefined (e.g. inf=1 -> inf-1).

Clearly the IEEE stuff for NaN and inf is not being compiled in, NaN and
inf are redefinable, and apparently have dfefault values of 0 when octave
is started up.

Can anyone offer advice? Sorry if the info is scrappy or inadequate to
define the problem: this is based on a telephone conversation.

Best wishes to all,
Ted.                                    (address@hidden)