warnings don't seem to be warnings

[oharar]

I've been trying to use suppressWarnings() to suppress warnings, and it doesn't work (e.g. on models with bad inits, when the MCMC finds the right part of the parameter space). I assume this means that Nimble warnings are being passed to R as something other than a warning class.

[paciorek]

@oharar Are you talking about warnings emitted while the MCMC runs?

[oharar]

Yes. Or possibly at initialisation (which I guess counts).

[paciorek]

Whoops, lost track of this. Indeed we don't have a way of suppressing warnings printed out from compiled C++.

For messages from code run by NIMBLE uncompiled in R, I've done some work recently so that most of our messages are now controlled by nimbleOptions('verbose').

@perrydv Control over messages from C++ might be something to consider in our nCompiler revamp of the compilation system.