forecast.gts silently produces NAN for all levels when at least one base level is all zeros

[dashaub]

hts version is 4.4 from CRAN
See the reproducible example below

library(hts)
myhts <- hts(cbind(wineind, 0))

# The total is fine here
aggts(myhts)

# The forecast produces NAN at all levels, 
# including Total and wineind
forecast.gts(myhts)

[robjhyndman]

This issue does not occur with v4.5 on github.