Hi. > > > > Following issue MATH-838, the "probability(double)" method was moved up to > > the base class "AbstractRealDistribution". Following MATH-839, a new > > "probabibility(double, double)" has been created. That method is overridden > > in "NormalDistribution" and "LogNormalDistribution". > > Clirr now considers that, in those classes, method "probabibility" has > > changed signature, but this not true since both (overloaded) methods (with > > one and two arguments) exist: the one-arg is inherited and the two-args is > > overridden. > > > > Isn't Clirr's assessment (that backwards compatiblity is broken) wrong? > > > > It's easy enough to create a test class to check this. > > Compile against previous version of MATH and see if class still runs > with current MATH.
I've done that. This: --- In method 'public double probability(double)' the number of arguments has changed --- is indicated as an error, but runs fine. This: --- Return type of method 'public org.apache.commons.math3.analysis.DifferentiableUnivariateFunction derivative()' has been changed to org.apache.commons.math3.analysis.UnivariateFunction --- is indicated as an error, but runs fine. This: --- Method 'public double dotProduct(org.apache.commons.math3.linear.OpenMapRealVector)' has been removed --- is indicated as an error, and fails to run (unless the argument is cast to "RealVector"). [Sébastien, I think that you have to add this method back for 3.1.] Is there some procedure to keep track that some of the problems reported by Clirr are not problems actually? Regards, Gilles --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@commons.apache.org For additional commands, e-mail: dev-h...@commons.apache.org
