This avoids a situation where the score of lang becomes lower or equal to others and then figures out the best font according to other properties and the order of family names.
This typically happens only when our orthography files are the subset of lang
in patterns. i.e.
fc-match :lang=en-us to match on en.orth.
In this case, the score is lower than the exact match (en to en) and the partial
match (en to en-us). thus, the result of
fc-match :lang=en-us isn't necessarily
So 35-lang-normalize.conf contains languages only which is available as orth without countries and tries to update properties to match on orth exactly like:
<match> <test name="lang" compare="contains"> <string>en</string> </test> <edit name="lang" mode="assign" binding="same"> <string>en</string> </edit> </match>
Fixes #155 (closed)