> We try to keep the tests so that they pass under many locales. But some
> are just so different that it's too cumbersome to maintain. This is one
> such case. Sorry.
The select_views test has also confused me while I was working on the
geometric types [1]. It is hard to figure out if the failure is
positive or a collation issue. I don't think this test is responsible
from testing collations. Can we change this to ORDER BY name COLLATE
"C"?