What happened to the errorbar function in R2014b?

As far as I can tell both the ERRORBAR function and its documentation are identical in R2014a and R2014b. The beahviour, however, is drastically different. Consider:

errorbar(1:5, ones(1, 5), -2:2, zeros(1, 5))

In R2014a you get

while in R2014b you get

Both errorbars have length L(i)+U(i) like the documentation says and I am not sure what a negative distance below a point means, but I wasn't expecting the change. Did TMW tell us about this? Is this a bug or a bug fix? Is it documented?

Answer 1

[Scott Webster]

It looks like maybe this was a bug before, and that now it has been fixed in R2014b? If a negative error bar is drawn in the opposite direction then the new behaviour seems correct. I could see an argument for not allowing negative values, but taking the absolute value as seems to have been the previous method could mask errors in code?