For URLs where there's fewer characters left to the right edge of the
window than the number of escape characters in the line, the next line
would be included in the URL, as if the URL went all the way to the
right edge.
For example with a 40 chars wide terminal, if you run:
echo -e '\e[31m1\e[m https://github.com/kovidgoyal/kitty\ntest'
And launch the hints kitten, you'll see that test on the next line will
be included in the URL.
This happened because the calculation for filling the rest of the line
with NUL characters counted the escape characters as well as the visible
characters, so it filled in too few characters.
This is a regression introduced in commit 91966712.
On the initial commit of this feature, IPs were just matched with a
very simple regex that prioritised simplicity/readability over
accuracy.
This commit adds a postprocessor for ip matches that makes use of
Python's `ipaddress` in the standard library to validate all the IP
matches.
This way we don't need huge and complex regex patterns to match _and_
validate the IPs, and we can just use `ipaddress` to abstract us from
implementing all the validation logic into the regex pattern.
It now uses information about the geometry of the screen.
So URLs that stretch over multiple lines are detected even if there is a
hard line break in between (some programs like mutt generate these).
