j
k
j a
j l
The proposed NaN handling sounds like it's hiding another bug. Are we sure that these NaNs aren't coming from an earlier failure?
Back to the thread
Back to the list