The patch release, 3.1.0 fixes a problem due to a change in the way the adapt command works (and the adapt() function in rjags). If the model has an adaptive phase, you should normally include a burn in period (with the update command/function) straight after adaptation before trying to monitor any nodes. Failure to do this in JAGS 3.0.0 can result in a “logic_error”. The new version allows you to do this without complaining.
This bug had some knock-on effects on R packages that use JAGS such as R2jags and bamdit.
To leave a comment for the author, please follow the link and comment on their blog: JAGS News » R.