GNU test suite compatibility graph seems incorrect

[blackbit42]

I enjoy looking at the graph a lot.
Though, today I started to notice that something is odd with it.
The current statistics are:

Pass: 95.3%
Fail:  3.4%
Skip:  1.2%

The distance of the 'Fail' graph from the zero floor should be 2.8 times the distance of the 'Skip' graph from the zero floor.
Though, that's not the case. Both graphs are very close to each other.
I don't know how this graph is generated. The math should be checked.
Either the numbers or the graphs (or both 😄) are off.

[geldot]

It is indeed a visualization defect. The "Latest Results" percentages box is currently sourced from a different dataset, but the way they're presented makes them look like they are supposed to be a legend. A standard legend representing stats from the plotted data would probably be more suitable. See the impl at graph.py

[sylvestre]

https://github.com/uutils/coreutils-tracking/ it is here