src: reduce number of exported symbols

[addaleax]

Use static definitions and anonymous namespaces to reduce the
number of symbols that are exported from the node binary.

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• commit message follows commit guidelines

Affected core subsystem(s)

src

CI: https://ci.nodejs.org/job/node-test-commit/9057/
CITGM: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/712/

[addaleax]

Any chance this will break a "rouge" embedder?

Only if they were relying on these symbols on purpose – so far we’ve always been taking the liberty of changing code that’s guarded by NODE_WANT_INTERNALS or completely local to our source files and didn’t hear anybody complain.

[addaleax]

CI: https://ci.nodejs.org/job/node-test-commit/9082/

[refack]

Only if they were relying on these symbols on purpose

That's why I dubbed them "rogue"... I've been know to use undocumented entry points on occasion 😞... You know then might disappear but then you're sad...
Although it's obviously semver-minor if at all, is there a way to put a big warning in the release notes, or land this only in node8?

[addaleax]

I've been know to use undocumented entry points on occasion

Everything I have seen so far indicates that our embedders and addon developers avoid doing that. (Also, I strongly doubt anything in here would be useful for outside users.)

Although it's obviously semver-minor if at all, is there a way to put a big warning in the release notes, or land this only in node8?

This doesn’t touch our public API, so it’s semver-nothing, and I really don’t think this is worth an explicit mention in the release notes. If you feel really strongly about it, we can only land it in Node 8, I don’t care too much.

(edited a bit later)

[refack]

This doesn’t touch our public API, so it’s semver-nothing, and I really don’t think this is worth an explicit mention in the release notes. If you feel really strongly about it, we can only land it in Node 8, I don’t care too much.

I agree it's semver-nothing, and I don't feel too strongly one way or another.
But I do think it's worth a special mention, writing comments in the release notes is easier than fielding issues from frustrated (even if misguided) users.

Just an example in the description of #8749 is says "Any non-null value of NODE_PRESERVE_SYMLINKS will enable symlinks." even tough the code acts differently and is documented correctly, that caused me a few hours of frustration, with no-one to blame but myself...

[benjamingr]

LGTM if citgm is green. I would prefer a semver major because if embedders but honestly I've never heard of anyone doing that before and I won't block this.

[addaleax]

Landed in 9d52222

I would prefer a semver major because if embedders but honestly I've never heard of anyone doing that before and I won't block this.

@benjamingr Like I’ve said, we’ve been quite liberal with internals so far, so it would be very surprising if this broke some actual code. I’ve added the dont-land labels because it shouldn’t really matter, but I would like to avoid getting in a habit of considering these actual breaking changes.

(edit: if this makes it into a release < 8.x it should probably come together with #12432)