Peter Trifonov

Ethan, When do you expect version 5.2.5 to be released?

Clipped line from in-range to out-range point is wrong for nonlinear axis

Changing yrange still affects the slope of curve segments, even if there are no invalid values in the input file. For example, in the attached file y=1E-5 is achieved either on x=1.8 or x=1.9 depending on yrange, although both plots are obtained from the same data. This bug leads to misleading plots.

In older gnuplot versions, having observed a vertical line on a plot, I would just set the low yrangeboundary to the minimal valid value, so that this vertical line disappears. Now even with appropriate yrange setting I cannot be sure if the plot is valid.

Thanks for suggesting the workaround. However, I would highly appreciate reverting to the behaviour of previous gnuplot versions, which produced vertical lines in such cases. This is mathematically correct, since log(0)=-infinity. Anyway, some warning message is necessary. Using the ?: operator is quite tedious, and one can easily forget to do this. This may lead to invalid plots, which may go unnoticed and cause a lot of confusion.

Such values are rather common in my datasets (these are error frequencies observed in simulations, which may be quite low). Previous versions of gnuplot produced vertical lines in such cases. This provided a clear indication of the problem. Now given some strange behaviour of the curve I have to guess if these are true data values, or some spurious gnuplot output. Even worse, this may be unnoticed and go into some papers/reports.

See the attachment

The slope of the last segment of the curve depends on yrange