degrib2 tests are failing on Intel and Windows because of differences in calculated averages

[edwardhartnett]

degrib2 tests are failing on Intel and Windows because of differences in calculated averages

For example:

399c399
< ( PARM= SWDIR ) :  MIN=               0.09999999 AVE=             219.31314087 MAX=             359.94000244
---
> ( PARM= SWDIR ) :  MIN=               0.09999999 AVE=             219.31278992 MAX=             359.94000244

Testing degrib2 is fraught with difficulties when comparing text output. What we need is something (snippet of sed? python?) to get rid of the 3rd decimal onward.

(Or, better still, add a flag to degrib2 which causes output to be rounded, then use that for testing.)

[edwardhartnett]

OK, turns out I already have a utility function to compare two degrib2 outputs and check within tolerances. I just have to remove the two old tests that still use a diff for this, but those tests are reproduced already using the new method.