select_site_header never actually works about http-observatory

[april]

This code will be fixed in the code push that allows scans without any backend. :)

[april]

row.get('public_headers', {}) will return None and not {}, which is what I need, because the value is actually NULL (None).

We actually use these columns inside Mozilla quite a bit (for testing sites that need sessions and the like), but they are currently not exposed because I haven't had the time to think of a safe way to do it.

[jarondl]

You are right about the row.get of-course, I haven't considered the actual None value.

If these columns are in use internally, then the title of this bug is wrong, but I still think that they way this code handles an empty DB result is wrong. I guess it should raise an error. Because currently it returns {}, while the retriever expects headers and cookies to exist.

In the DB-less scenario this code should never have been called (#157), so that's not an issue.

[april]

Oh hey, this was fixed a while back by moving it into scan() instead of being in retrieve_all().