PUBLISHED_FROM=da57dfe52e315dc1901d761289f0445be6e93cea

PUBLISHED_FROM=da5f7f80a9876ac4f1889a1176a3048f47980659

PUBLISHED_FROM=da57dfe52e315dc1901d761289f0445be6e93cea

PUBLISHED_FROM=dbf75bfba087f1b0aa0531e5003ba3e69ed1a6ab

PUBLISHED_FROM=293a54ab09e6879b6f0c36975a137eeaeff92098

On my way, fixed a couple of cases where we had `()` in the header, and
non-empty argument list in the source file.

PUBLISHED_FROM=5519526cf84e2bbd425a726fcc112fea1a95cbf1

PUBLISHED_FROM=0d71533fd475c152c29fe304ca95559512a6059d

PUBLISHED_FROM=e9a4e5c7b4a1d03b93a2a79e29de19e60e919929

A way to serve a single, specific file as opposed to the hairy monster
that is mg_serve_http.

PUBLISHED_FROM=6ed1bc487b9d41d212a5907337182f23b040aecf

PUBLISHED_FROM=542e7c82fa494bef52222495301ab53dce43cd8f

We haven't actually _fixed_ evaluation of precompiled bcode from
non-mmapped memory: we just never free it. For sane use case (do not
evaluate the same file more than once) it should be acceptable.

PUBLISHED_FROM=a9ca2d8fa72e9c40c93415d060ee8db697caa225

PUBLISHED_FROM=5cd78790a03580aa23dd3d2d8b825ad449083cc2

PUBLISHED_FROM=11c57fc5ece01686e9cecab5e3bfce54a2825836

PUBLISHED_FROM=8b7fcfc1bc32ff9ff38e2904ddb730c83bf9fae4