Some priors are not being evaluated

A flag was introduced in 1.7.3 which meant certain priors could flag that they 
should be evaluated first because they may return a 0 likelihood and thus 
short-circuit the full evaluation. Priors that set this flag are Uniform 
distribution and any positive distribution with an offset (values less than the 
offset give a 0 likelihood). 

When evaluating these, the late evaluation priors overwrote the loglikelihood 
of the early priors (rather than being added). This had the effect of removing 
the priors. 

The consequence is that uniform priors were not being respected and an prior 
with an offset were being ignored (this will include calibration priors and 
poisson priors in BSSVS).

Original issue reported on code.google.com by ramb...@gmail.com on 12 Sep 2012 at 8:01

evolvedmicrobe / beast-mcmc

Some priors are not being evaluated #664