This walks back a few changes in checkLogProb to allow us to see test results before changes are made that affect the RNG state and therefore the precise samples returned in various MCMC tests. This is because there's some change in behavior in terms of whether a logProb of NaN or -Inf leads to a random sample being drawn and then rejection or immediately to rejection.
If this passes and goldfile values do not change, I'll proceed with updating goldfiles and such in the tests results for checkLogProb.
This is not intended to be merged into devel ever.
This walks back a few changes in
checkLogProbto allow us to see test results before changes are made that affect the RNG state and therefore the precise samples returned in various MCMC tests. This is because there's some change in behavior in terms of whether a logProb of NaN or -Inf leads to a random sample being drawn and then rejection or immediately to rejection.If this passes and goldfile values do not change, I'll proceed with updating goldfiles and such in the tests results for
checkLogProb.This is not intended to be merged into devel ever.