Re: How to get ORCA/C bugs fixed.

[toddpw@pride.ugcs.caltech.edu (Todd P. Whitesel)]

lvirden@cas.org writes:

>So bugs which can not be narrowed down to simple test cases should just
>be bore silently.  Grit your teeth and say "There's no place like source;
>there's no place like source...".

Any reproducible test case (even a large one) is still much easier to fix
than a description of what looks like the problem.

Very consistently at work we have found that the customers who insist that
they know what the bug is without a test case and output to prove it are
always wrong, and 90% of the time they aren't even close. Normally we will
go through at least three or four theories about the bug ourselves, while
narrowing it down to the program or phase of the program responsible.

Blatant bugs are the norm with application code (if you're lucky). Subtle
bugs are the norm with compilers, because of the inherent tendency towards
much more complex algorithms, all of which have to interact with each other.
The style of bug report you would give for something like AudioZap is not
going to be worth much for ORCA/C.

Todd Whitesel
toddpw @ ugcs.caltech.edu