> I agree — the NULL handling seems wrong.
> Here's a quick patch to fix it. What do you think?
LGTM, but maybe the test should be expanded with NULL for the two-column
indexes? I mean to add NULLs at zero column too.
Also, about the tests but not related to NULLs...
Seems like we need to run those tests on 32bit system, the test output should be different?
--
Best regards,
Roman Khapov