Graduate experimental N-API feature napi_threadsafe_function

[pipobscure]

I reworked fsevents to use N-API. For that I had to use the napi_threadsafe_function family of features, which are currently behind experimental. They work like a charm.

Can we graduate these out of experimental, backport them to v8.x and bump the N-API version?

See: fsevents/fsevents#237

(Came out of a discussion with @mhdawson)

[pipobscure]

Interestingly enough, I found that the napi_threadsafe_function family of features are actually entirely unnecessary. That is because the entire 1.x branch of libuv is API/ABI stable to begin with.

Now that being said, it's a nice wrapper that may be easier to use than interacting directly with libuv, so I'm not saying it should be dropped.

[targos]

@nodejs/n-api

[mhdawson]

I'd like to correct that I still don't believe that we can depend on libuv stability. When Node.js moves to a new major version things will break. If you want ABI stability you can only depend on the functions which are part of N-API. It may be possible to implement checks/binaries for different Node.js versions with different major versions of libuv but nothing in our existing versioning will help with that (might be something to look at if we believe updates to the libuv major version are going to be far enough apart).

Based on this real-world use/validation as well as prior use I'd suggest we graduation napi_threadsafe_function from experimental

[mhdawson]

@gabrielschulhof

[gabrielschulhof]

@mhdawson this means we should create version 4 of N-API. What else can we bundle into it?

[gabrielschulhof]

We concluded at today's N-API meeting that napi_threadsafe_function is the only API we could include in version 4, and that it would require a backport to Node.js v8.x before we could consider it stable.

[pipobscure]

Created a new PR for fsevents using these: fsevents/fsevents#247
So now I'm on hold/waiting for the backport for v8.x in #25002

[pipobscure]

This is the latest version of fsevents based on napi. fsevents/fsevents#247

However to be able to release somewhat smoothly #25002 needs to be in.

[mhdawson]

We talked today in the N-API team meeting and @gabrielschulhof is going to going to create a PR to move it out of experimental on master. The goal is to that that backported for a potential SemVer minor 8.x release being planned by the Release WG.