[mesa-users] test suite cases

[Bill Paxton]

On Dec 4, 2012, at 10:10 AM, Falk Herwig wrote:

> Bill, we are in the process of preparing some test suite cases. Can you point me to a place where you specify what exactly is needs to be provided for a valid test suite case?

Hi Falk,

I don't have a writeup yet defining what needs to be in a test suite case.   Here's a sketch to get you started:

the test case should automatically check results (i.e., no human looking at a plot to see if it is okay!)

the test should output something to the terminal that indicates success.  
in many cases, this output can simply come from the run having reached 
some stopping criterion (e.g., "stop because star_mass >= star_mass_max_limit").
but for more complex tests, you can add code to extras_after_evolve in your run_star_extras
(e.g., see test_suite/axion_cooling).

the tests should be precise enough to catch cases where mesa has changed in a way that 
screws up your results.  but the tests shouldn't be so exacting that small numerical changes
will trigger false alarms.  

keep the runtime of the test as short as possible while still testing what you need to test.
make use of a saved starting model if that will eliminate early parts of evolution that don't
really matter for the test.  stop the run as soon as it has done enough to make the test useful.


does that give you enough to get going?  

Cheers,
Bill