VMS upgrade changed program behavior, what could cause this?

From: Lawrence Bleau (bleau_at_umtof.umd.edu)
Date: 02/25/04 

Date: Tue, 24 Feb 2004 23:28:26 +0000 (UTC)

I have a very puzzling problem, folks, that shakes my group's confidence in
VMS. I hope I've missed a step in my logic, but I don't think I have.

First, the configuration: I'm running OpenVMS AXP V7.3-1, which I upgraded
to on Dec 18, 2003; I was running V7.1-2. I did not upgrade the Fortran
compiler; we're running Compaq Fortran V7.2-780, which I installed back in
1999.

Now, the problem statement, brief version: I have a program, HOUR_AVERAGE,
that was last compiled and linked in 2000. It takes as input an ascii data
file that has columns of numbers, and writes as its output another ascii
file in which several rows from a given column are averaged according to
their grouping by their time, with a new averaging being produced for each
hour; hence its name. This averaging is done for each column, so that the
output file has the same number of columns, but only one row for each hour.
This is just straight addition and division, folks, nothing fancy. All
arithmetic is single precision.

HOUR_AVERAGE ran before the upgrade last Dec. I ran it again today, giving
it the same input file. The output files are not the same. More precisely,
the average values produced by the program differ in the lowest decimal place;
sometimes 3rd, sometimes 5th.

Emphasis: HOUR_AVERAGE did not change! A DIR/DATE of the .EXE file still
shows sometime in 2000.

So, what did change? Obviously, the operating system did. What about the
upgrade, though, could possibly have caused this change of behavior? I suspect
some roundoff behavior has changed, but the release notes say nothing about
this; it's just a guess anyways.

I have a lot of details I could post, but for now, does anyone have a hint
as to what could case this, how widespread it could be, etc.?

Lawrence Bleau
University of Maryland
Physics Dept., Space Physics Group
301-405-6223
bleau@umtof.umd.edu

Relevant Pages

  • Re: Terminal Services after Migration to SBS 2003
    ... It's funny, but there are a fair number of folks that I can say "yes, I ... remember we talked about this last time we did that upgrade". ... Kit reference because at least it's a standardized ... emails I get start with "Jeff, I'm in Phase 2, Step B, Item 5 and I'm stuck. ...
    (microsoft.public.backoffice.smallbiz2000)
  • Re: A tale of two macro libraries LIB vs STARLET
    ... The point about help is we told folks they can put stuff in there we take ... >> on an upgrade to replace it. ... >> to be unique and not comflict with anything put in there. ... That's why we have the lovely Product Registry isn't it? ...
    (comp.os.vms)
  • Re: Upgrading laptop processor
    ... 50% of the folks who buy PC parts from these stores have no idea how to ... The gentleman in question has a newish notebook with a Sempron CPU... ... developing expertise and imparting it freely - encouraging consumerism by ... telling someone who wants to upgrade his laptop he must be joking. ...
    (microsoft.public.windowsxp.hardware)
  • Re: [fw-wiz] Variations of firewall ruleset bypass via FTP
    ... "Check the specific OS line, or your version number, or upgrade." ... folks using IPF don't get the wrong idea. ... If I were still admining NetBSD systems in production, ... entry well before I'd look at the NetBSD entry because I'd expect you to ...
    (Firewall-Wizards)