Error bars in logarithmic plots

[jzazo]

Hi! I'm experiencing some issues while plotting error bars and using logarithmic scale on the y-axis.
As I understood from the docs, the ErrorBars(y=[...]) indicates the relative value of the error with respect to the Plot.
Consider the following example (normal scale):

x=[1,2,3,4]`
y=[0.1, 0.01, 0.001, 0.0001]
errY = [0.05, 0.05, 0.05, 0.05]
Axis(Plots.Linear(x,y, errorBars = ErrorBars(y=errY)))

The output is correct, centered on y:

However, if I write

x=[1,2,3,4]
y=[1e-1, 1e-2, 1e-3, 1e-4]
errY = [1e-2, 1e-3, 1e-4, 1e-5]
Axis(Plots.Linear(x,y, errorBars = ErrorBars(y=y-errY)), ymode="log")

(note that I've reduce the errY, so that the outcome is always positive for the logarithmic scale), I get a wrong plot, there are no error bars:

Finally, if I do not assume relative error but absolute, I get what I want:

x=[1,2,3,4]
y=[1e-1, 1e-2, 1e-3, 1e-4]
errY = [1e-2, 1e-3, 1e-4, 1e-5]
Axis(Plots.Linear(x,y, errorBars = ErrorBars(y=y-errY)), ymode="log")

Am I missing something? Is there a bug in the logarithmic scale? Maybe in pgfplots, or the output is processed wrongly by ErrorBars() in the logarithmic case? I just wanted to check, I believe the third plot is correct, but getting there is a bit fishy. Thanks.