iraf.net

---
Date: Thu, 10 Dec 92 12:56:36 PST
From: irwin@dao.nrc.ca (Alan Irwin--UVic Guest)
To: tody@noao.edu
Subject: Re: some iraf problemsHi Doug,Thanks for your quick response. 3) is answered. 2) I will have to
try your solution. 1) is still a problem. I agree x or y = 1.e30
works. However, the exact sequence below does not.x = -1.e30
y = -xThe second statement bombs, but y = 0 - x works after the first statement.Alan
Date: Thu, 10 Dec 92 14:02:08 MST
From: tody@lepus (Doug Tody)
To: irwin@dao.nrc.ca
Subject: bug?Alan, I still can't reproduce your problem. cl> x = -1.e30
cl> y = -x
cl>
cl> print (x, y)
-1.0000000000000E30 1.0000000000000E30
Date: Thu, 10 Dec 92 13:08:05 PST
From: irwin@dao.nrc.ca (Alan Irwin--UVic Guest)
To: tody@noao.edu
Subject: Re: bug?interesting.... Do you have a guest account here? I assure you that
our present release of iraf (2.10, I believe) and the identical
system at University of Victoria have this problem.I will try and put together a log file for you after lunch.Alan
Date: Thu, 10 Dec 92 14:15:28 MST
From: tody@lepus (Doug Tody)
To: irwin@dao.nrc.ca
Subject: CL problemNever mind, I tried your example under iraf on one of our sun servers and it
failed with a floating operand error. The test earlier was done on the
Mac-A/UX IRAF system I use in my home office. Odd. I will take a look at
it. This is not a syntax error, it looks like an arithmetic error occurring
during computation of the expression (for no good reason that I can see just
looking at the input).
Date: Thu, 10 Dec 92 14:38:03 MST
From: tody@lepus (Doug Tody)
To: irwin@dao.nrc.ca
Subject: CL bugOk - what was happening was that an integer overflow was occuring when the
CL attempted to convert the large floating value to an integer (the integer
value is not used in this computation but the CL code used here generates it
internally anyway). The CL checks for a large floating value before doing
this conversion, but the check would fail for large negative values. This
will be fixed in V2.10.3. Positive values or negative values smaller than
-2.e9 avoid the problem. Integer overflow is not detected on all
architectures, e.g. my 68030 Mac system, but we have it enabled for IRAF
running on sparc systems. This explains why the problem only shows up on
some systems. - Doug(We don't normally fix bugs this quickly by the way; this time it was easier
than logging the problem).