This is a bit hacky, because comparing size doesn't really work like difference, but I've squeezed it into a shape where it seems to work okay?
It basically works like this:
for each table produced by both compilers, we find the ratio of their sizes
if the ratio exceeds a threshold (currently 1/10) then we add it to the the report
in the actual report, we just display these size differences as a delta, in bytes. Negative numbers mean that fontc is smaller than fontmake, and positive numbers mean the inverse.
This is a bit hacky, because comparing size doesn't really work like difference, but I've squeezed it into a shape where it seems to work okay?
It basically works like this: