A phytools user reported that phyl.cca breaks if either X or Y contain only one column. Even though we would not normally use CCA when one or the other of our xs or ys contains only one variable (why not use multivariable regression instead?) there is no theoretic reason why it should not work. The main reason seems to be R's different behavior towards vectors & matrices with one column.
The updated code is here. Please let me know if you run into any problems.