At the point of the line mentioned above the x and y variables have been modified so that they include the right data, i.e. the data to be used to fit the nth cv fold. The same thing happens to the offset variable, but not to the weights variable. Therefore the weights which are passed through are pretty much in a random order.
This is an easy fix AFAIK, you just need to add in a line w <- w[i.train][i] and everyone is happy.
https://github.com/harrysouthworth/gbm/blob/9c11f642fec7268e42837b4abc7fa34e10137651/R/gbmDoFold.R#L41
At the point of the line mentioned above the x and y variables have been modified so that they include the right data, i.e. the data to be used to fit the nth cv fold. The same thing happens to the offset variable, but not to the weights variable. Therefore the weights which are passed through are pretty much in a random order.
This is an easy fix AFAIK, you just need to add in a line
w <- w[i.train][i]and everyone is happy.