GitLab

Now it breaks properly again.

Remove exact base 2 logarithms, since their values depend on whether
excess precision is used or not (see also #5165).

Also make T5975b actually load the file, it looks like this was an
oversight when the test was created (it wasn't doing anything useful
at all).

As per the comments on that ticket, *not* inlining rnf is correct, so
make this test pass.

It'll be removed soon anyway

Includes a test for the problem in #6111.

Apparently GHC got faster

Most of these tests were written by Nicolas Frisby.

but this change makes it work again by adding a type signature

I had the test lying around already, just hooking it up