From: Roy S. <roy...@ic...> - 2005-03-18 18:24:50
|
On Fri, 18 Mar 2005, KIRK, BENJAMIN (JSC-EG) (NASA) wrote: > Some day I'd really like to get better, automated regression testing... We could probably start by setting up something in the examples Makefile to produce deterministic output, to run individual examples repeatedly with multiple inputs, and to produce more numerical output (i.e. solution norms) even for examples where we can't easily produce error norms. As it is, things like perflog output prevent me from doing any regression testing with just "diff", even if I go through example outputs by hand there are important code paths that go untested, and there are examples where some failures can be silent. Just a few changes there would probably be a good start. --- Roy |