gnuplot can't handle extra large denominators
When plotting functions that are mathematically non-negative, but that reach 0 the plot
tactic sometimes generates a rational number whose denominator is larger than floating point infinity. gnuplot
will be unable to handle the situation, and this will result in a failure at display time.
Example plot_fail.v
It seems the failure of gnuplot happens when the denominator reaches 2 ^ 1024 (the highest number representable as a 64-bit floating point number?)
An unsatisfactory workaround is to post-process the generated gnuplot file to replace inverses of very large numbers by 0, as in the following sed
command.
sed -E 's!\(-1. / [1-9][0-9]{307,}\.\)!0.!'
The value 307
here was chosen to reflect approximately the decimal size of 2 ^ 1024.