Problem 86

Summary:	floating point exceptions
Product:	Geant4	Reporter:	paetzold
Component:	global/management	Assignee:	Gabriele Cosmo <Gabriele.Cosmo>
Status:	RESOLVED WONTFIX
Severity:	normal
Priority:	P2
Version:	1.0
Hardware:	DEC
OS:	OSF/1

Description paetzold 2000-04-11 04:02:01 CEST

In calculating Interaction length I got floating point exeptions in:
G4VDiscreteProcess::PostStepGetPhysicalInteractionLength
I traced this error and I realized that the highest double number is multiplied
by a number greater 1.
This number was supplied by some GetMeanFreePath function of a process
I solved this problem by setting the DBL_MAX to a smaller value, wich should be
ok.
I am using gcc-2.95.2 on osf1, because cxx give only problems.
I know this is not supported, but as I said with cxx I was even not able to
run the examples. Is there somthing which is against using gcc and making this
change

Comment 1 Gabriele Cosmo 2000-04-11 08:10:59 CEST

Geant4 coding conventions and installation setup explicitly follow the
ANSI/IEEE-754 Standard for the initialization of floating-point arithmetic
hardware and portability. The Standard foresees floating-point arithmetic
to be nonstop and underflows to be gradual.
On DEC platforms the ANSI/IEEE-754 Standard compliance needs to be explicitly
set (since deactivated by default), we use infact the option "-ieee" on
the DEC/cxx native C++ compiler to achieve this.
You should check if your compiler provides compilation options for activating
Standard initialization of FP arithmetic (it can be DEC specific).