[AMBER-Developers] Error counting in test scripts

From: Ben Roberts <roberts.qtp.ufl.edu>
Date: Thu, 1 Apr 2010 19:55:56 -0400


I've just had a look at the test failures and the error count, particularly in the parallel tests, is much higher than it ought to be.

On inspection, this is because the test script greps for "error" and "Error". There are a lot of times one of those two strings (which need not even be a whole word) appears. For example, in a test that fails because there are more groups than processors, the total error count would be 3: One for the error message thrown by Sander, one for the error message thrown by MPI_ABORT, and one for the "Program error".

The obvious solution is to use "Program error" instead of "[Ee]rror" in the grep string. But in my experience, obvious solutions often turn out to be incorrect. Does anyone know any reason why this solution wouldn't work? If not, I'm happy to write a patch (if no-one else beats me to it).

AMBER-Developers mailing list
Received on Thu Apr 01 2010 - 17:00:03 PDT
Custom Search