Sorry to anyone who gets this twice, but I sent it yesterday to the old
address (mailbase.ac.uk) and although I got a message to say it had been
sent out, I didn't see it myself.
---------- Forwarded message ----------
On Fri, 24 Nov 2000, Bertrand Meltz wrote:
>
> There has been a few comments, recently, from vendor compiler engineers,
> complaining that bug reports should be sent to the proper support
> provided by the vendor, and not to comp-fortran-90.
I agree with your comments. There is another factor: often a problem
manifests itself in very complex code. At first you think you have made a
mistake, then later the suspicion dawns that perhaps it is a compiler
problem (I think this is the mark of an experienced programmer, beginners
tend to suspect the compiler at a much earlier stage...). But to make a
successful bug report you really have to produce a very small program
which shows the same problem, and this can take a very large amount of
effort. So often, what I would like to do was to submit a question saying
something like "I think there is a problem with feature X in compiler
version Y, could I be right?". If it's something they already know
about, then I am saved the days of effort producing a watertight example
of the bug. It may even be already fixed in version (Y+1).
Unfortunately, vendors bug-reporting services are not set up to deal with
such questions, and I have rarely if ever got a sensible answer to them.
They just say: give us your code and tell us the problem, and we'll get
back to you. That is another way in which this mailing list (and the
Usenet groups) are valuable first steps in reporting a suspected bug, and
discovering whether it is already a well-known problem. If vendors' help
desks were more helpful, this would not be necessary.
--
Clive Page,
Dept of Physics & Astronomy,
University of Leicester.
|