Unrealistic expectation on size of err_u

[henrikt-ma]

It appears problematic that ModelicaTest.Media.TestAllProperties.PartialMediumFunctions has a single eps constant used to define absolute error tolerance different media properties.

ModelicaStandardLibrary/ModelicaTest/Media.mo

Line 104 in 7a2a5ed

assert(err_u <= eps, "Error: abs(medium.u - u) > eps " + "(err_u = " +

Take ModelicaTest.Media.TestAllProperties.CO2 for example. The expected value of medium.u is 156518.651861815, and the value of eps is 1e-10. This means that the assertion that err_u < eps requires medium.u to have a relative error smaller than 6e-16. Considering that double precision is the de facto standard (the specification only has a recommendation regarding the range, not the precision of Real), this does not seem realistic.