[R-sig-Geo] problems with predict() in package raster

[Carsten Neumann]

  Dear all,

after updating package raster_1.8-38 and randomForest_4.6-2 the 
predict() function fails to calculate a resulting rasterLayer. The 
Script always functioned well, before. Here is the problem:

###############################
pred <- predict(b1, model1)
class(b1) -> RasterStack
class(model1) -> randomForest

Warnmeldungen:
1: NAs durch Umwandlung erzeugt
2: NAs durch Umwandlung erzeugt
###############################
The Error tells me that NA's are produced through conversion. But why??? 
What do I have to change in comparison to former trials?
(Input for randomForest: 1. factor variable   2. dataframe)

I am thankfull for any help, because I want to present this function 
students on Friday.

With best regards

Carsten