On Wed, Jan 02, 2013, Kirill Nuzhdin wrote:
> to my opinion a program should not crash under any
> circumstances (even when used in unintended way), and if it does - there
> is something seriously wrong with it.
By this definition, there is indeed something "seriously wrong" with most
or all of the Amber programs. For example, almost all of the test cases
use good inputs, and relatively little effort is spent checking inputs
for consistency and correctness. So, all Amber users should take care in
examining input and output files, should run simple calculations before trying
complex ones, and so on.
What keeps me awake some nights are not crashes (where it is at least obvious
that things have gone wrong), but outputs that look plausible but are actually
incorrect.
....dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jan 02 2013 - 13:00:02 PST
