On 2020/02/02 18:51:08, Dan Eble wrote: > LGTM > > https://codereview.appspot.com/547560044/diff/555240043/lily/spring.cc > File lily/spring.cc (right): > > https://codereview.appspot.com/547560044/diff/555240043/lily/spring.cc#newcode119 > lily/spring.cc:119: avg_stretch /= static_cast<Real> (springs.size ()); > This is not an issue with this change, but should this function do something > other than divide by zero when springs is empty?
Possibly. We have had a few reports of combinations of forced breaks leading to failed assertions. Dividing by Real zero (which was done here before as well due to conversion rules) does not throw divide-by-zero but results in ±inf.0 (unless the numerator is also zero, then you get ±nan.0). So I won't rule out that some of the triggering cases might have come through this code path. https://codereview.appspot.com/547560044/