[m-rev.] for post-commit review: float NULL

[Julien Fischer]

On Thu, 7 Jun 2018, Zoltan Somogyi wrote:

> The simpler fix is to make each field of the structure an MR_Box, but that makes
> the structure bigger. The better fix would be to generalize MR_hl_field to non-MR_Box
> field types, and to teach the compiler to use that generality when needed, as here.
> Unfortunately, I am pretty sure it is a significantly bigger job.
>
> Does anyone depend on the performance of the hl.gc grade?

The hl.gc grade is no longer publicly documented, so no-one should
be relying on it.

Julien.