#weak-learners

Boosting runs on 10 training points with these weak classifiers, each listed with the points it misclassifies:

Weights start at $1/10$. Each round picks the classifier with the smallest weighted error $\varepsilon$ (ties go to the earlier letter); then the misclassified points are rescaled to total weight $1/2$ and the correct ones to total weight $1/2$. Which classifier is picked in round 2?