Invalid Floating Point Number exception

I have a customer who told me that an 'Invalid Floating Point Number' had been raised, and the error box disappeared and the program halted.

I am looking for either a way to flag where the error occurred, or to acknowledge the error by substituting a different value (valid, of course) and flagging the error and continuing.

Any insights would be helpful.

It is possible that a variable does not get initialized properly; without a location given, it will be impossible. The error does not appear to be repeatable; however, I will ask my user to re-run his data a few time, JIC.

'Invalid Floating Point Number' is runtime error number 27 in the FTN95 library but I can't see any reference to it .

I put in the call to trap FP exceptions (Win32). This is a random occurrence so far as I can tell; likely my problem with some FP I don't properly initialize. Still...

It would be helpful to know how to force an invalid FP number (and other exceptions) to see if it now ignores the error, or still crashes.

One possibility is to use the IEE_ARITHMETIC module and a call to

x = ieee_value(x,IEEE_QUIET_NAN)

Using x might give the desired result.

Otherwise the TRANSFER function can be used to to set an integer hexadecimal value and convert it to REAL.

For example, a 32 bit NAN has 255 in the exponent part and non-zero in the fractional part.

Thanks, Paul. I'm going to experiment with this a bit!

Bill