Hello everyone,
This is just a friendly reminder about the importance of test cases and
testing the build process, especially as we approach the Amber 12 release.
The test cases are a great way to check that the code you've added (even a
one-liner) hasn't broken existing functionality.
Given the way that many parts of Amber are designed, (e.g. duplicated
header files, array indexing, roll-your-own depend scripts unique to each
program), even a 1-line change in one part of the code can have a huge
impact on how a seemingly independent part of the code operates.
Therefore, it's really not enough to just make sure that the first couple
tests pass in serial, or that the GB tests pass if you think you only
modified the GB code.
Also, changes in the code should be accompanied by updating test cases for
changed output format (or updating the test values after the differences
have been validated).
Thanks!
Jason
P.S. This is not just a "best practices" comment -- we've actually been
spending non-negligible amounts of time fixing bugs introduced by untested
(or not fully tested) code, and it's always harder to fix code that's not
your own.
--
Jason M. Swails
Quantum Theory Project,
University of Florida
Ph.D. Candidate
352-392-4032
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Sun Feb 12 2012 - 13:00:02 PST
