Re: Crash for Armidillo

From: Alan Anderson (aranders_at_insightbb.com)
Date: 08/13/04 

Date: Fri, 13 Aug 2004 01:58:29 GMT

> sib101@gmail.com (Jack Sprat) wrote:
> > Honestly, that's simply an absurd thing to post. If I'm developing
> > medical software where lives are at stake, and I get a compile error,
> > should I wring my hands in despair that I'm not taking it seriously
> > enough?
> >
> > No, obviously, because I'm in development. At that stage of the
> > process, errors are expected. It's the end-product that matters, and
> > late-stage testing is different from early-stage testing.

I'd have to say that your analogy is flawed. A compile error is akin to
finding that a part does not fit correctly, an electrical connector is
miswired, or some other relatively "static" problem. This was more like a
run-time error, a missing bounds check on an array reference or a division
by zero.

A compile error catches mistakes before the program can do anything
dangerous, and is rarely a cause for worry. A run-time error during
development of something like a medical monitoring device is often a minor
disaster, since if you *were* taking it more seriously, you'd already have
written the code to detect and deal with anomalous events. But it's
almost never a *major* disaster, because you can usually get enough
information about the error to ensure that it won't happen again.

jamcguir@yahoo.com (Jake McGuire) wrote:
> If you're developing medical device code, and the attempt to build an
> image for QA fails with a compilation error, you should be *very*
> concerned, as there's no excuse for a developer to check code into
> your version control system that doesn't compile. If he didn't
> compile his changes, he certainly didn't test them.

*ahem*

If YOU are the one working on the code, you're obviously making changes
that aren't yet checked in. A compile error is usually due to something
as simple as a typo, and isn't something I would get worked up over.

(Unless I were writing in C, that is, because a simple typo in a line of C
often results in a perfectly valid line of code with a meaning quite
different from the intent.)

> And it's very hard to retrofit reliability into a design. It's much
> easier to design it in from the beginning.

If you don't have a good collection of data on what parts of a design are
likely to fail, you don't have a good idea where to concentrate your
efforts. Finding the weak links (by breaking them) and redesigning them
to be more robust is definitely more effective than analyzing the whole
thing to death but never getting around to building it.

> What data did he gather, other than "rocket engines don't work so well
> without propellant", "my vehicle is indeed aerodynamically unstable",
> and "our vehicle will not survive an unpowered fall from 1000 feet"?

I saw lots of interesting stuff about how engine performance is affected
when the catalytic elements are distributed badly, and how the temperature
sensing system can give invalid readings, and how rough acceleration can
fool the control system. All can contribute to a more reliable vehicle
in the next iteration.

> And was that data worth five weeks of downtime?

Arguably only three. The known flaw was already being addressed, with a
part that takes a couple of weeks to fabricate.

Relevant Pages

  • Re: xml documentation
    ... I was creating my first custom control, ... The first compile would be fine. ... would throw a compile error with the "Could Not Copy Temporary Files ...
    (microsoft.public.dotnet.framework.aspnet)
  • Re: Cant find old post - Command Button Visible
    ... it created a compile error. ... to do with the Library 3.6 compile problems I've had? ... It has been a learning experience on this database ... "Joan Wild" wrote: ...
    (microsoft.public.access.security)
  • Problems complieing apache_1.3.31 and SSL
    ... > Final apache compile error message: ... not be in the mod_ssl source tree. ... around it by including a version of ssl_expr_scan.c that does compile. ...
    (comp.unix.solaris)
  • Re: How to Find the Date of Last Change in Inventory
    ... your shoulder and give you a hand. ... I do not get the compile error when viewing the code ... > I should tell you that the only VBA code I know how to write is your ...
    (microsoft.public.access.modulesdaovba)
  • Re: Sorry, but not happy with D2007
    ... that until you save or compile, ... This was a specific design decision since once the uses ... Because you 'automatically' remove all empty events, it forces me to take the EXTRA time to always had a comment line in each event, which I wouldn't have to do otherwise. ...
    (borland.public.delphi.non-technical)