Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!cs.utexas.edu!helios!bioch.tamu.edu!antek
From: antek@bioch.tamu.edu
Newsgroups: comp.lang.fortran
Subject: Non-VMS like behaviour of "f77 -vms ..."
Message-ID: <9811@helios.TAMU.EDU>
Date: 3 Nov 90 20:15:04 GMT
Sender: usenet@helios.TAMU.EDU
Reply-To: antek@bioch.tamu.edu
Distribution: usa
Organization: Texas A&M University
Lines: 32

Hi! 

I'm trying to port a Fortran program written for a VAX to a MIPS UNIX BSD4.3
environment. There is quite pure F77 code inside, no VAX system calls etc., 
but two things are strange: 

1) During the execution of such a code: 
	CHARACTER*10 ANSWER
	WRITE (6,'(''$Enter your answer: '')')	! The "$" to suppress <CR>
	READ (5,'(A10)') ANSWER
a user have to hit the <RETURN> key TWICE, regardless of the length of the 
string s/he enters. I tried 3 different positions of the $ sign and several
options (-vms, -systype bsd43, etc.) but the problem remains unchanged or
the "$" is not interpreted at all. 
NUMERICAL input is ok, only the "A" format is giving troubles. 

2) Ordinary divisions by zero and SQRT of a negative doesn't produce any run-
   time errors, instead they set the NaN value to a number, or a zero: 
Pseudocode: 
	r = 0.0;	i = 0;		rneg = -1.0;
	r = 1.0/r;	i = 1/i;	rneg = SQRT(rneg);
	write (6,*) r, i, rneg

Answer: NaN     0     0.00000          (1.0 / NaN gives 0.0 again)

Great, errors don't break the program (what I wanted anyway), but I would
rather like them to create an interrupt which I can catch in a handler. 
And in the debuging phase I believe I AM writing parts of the code wrong,
so nonsense values should be reported, I can't write out *everything* or
go step-by-step with a debuger through 10000 lines of the program.

Any suggestions ?                                  Antek @ Bioch.Tamu.Edu