Node.js 8.0.0 Tracking Issue

[jasnell]

It's still very early in the process, but this will be the issue that I use to track the issues and todo's around the Node.js 8.0.0 release that will land in April 2017. The current plan is to cut the v8.x branch off master in the first couple of weeks in March 2017 (just about 3 months from now). This issue will be updated periodically with the relevant todo's.

[targos]

We could also try to use a Github project?

[jasnell]

Perhaps. I'm not sure how that would be any different than the milestone, however.

…

On Sun, Dec 4, 2016 at 9:29 PM Michaël Zasso ***@***.***> wrote: We could also try to use a Github project? — You are receiving this because you were assigned. Reply to this email directly, view it on GitHub <#10117 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAa2eSLEy7YB4K-hfxJuKkc93AWUMx_Tks5rE6EjgaJpZM4LDy9d> .

[Trott]

Probably need to add an issue for deprecating the current CLI debugger.

[jasnell]

Ooh... Yeah, good idea. We will need the new stuff to come out of experimental at the same time the old debugger deprecates.

[targos]

Yeah this is very important because we will probably have V8 5.8 or 5.9 by the time of the release and the old debugging protocol is removed as of 5.7! See: #9789

[Fishrock123]

and the old debugging protocol is removed as of 5.7!

It would be nice if they could have, you know, given us a heads up specifically about that. Thanks V8 team. 😠

Guess we need to issue another semver-minor deprecation in node 7 and maybe 6 that everyone will hate.

[Trott]

It would be nice if they could have, you know, given us a heads up specifically about that. Thanks V8 team. 😠

The protocol has been deprecated in V8 for at least one year, and I think longer than that.

@ofrobots opened an issue in June to discuss getting off the deprecated protocol. I'm pretty sure he surfaced the issue before that too, but certainly at least since then.

[Trott]

By the way, the debugger thing is on the CTC agenda this week, at least indirectly, as nodejs/CTC#40.

[klimashkin]

@targos Is there a date when 8.0.0 is going to be released?
V5 5.7 will be around the beginning of March, 5.8 - middle of April

[chicoxyzzy]

Node.js 8.0.0 release that will land in April 2017

[klimashkin]

So there is hope that nodejs release will wait for V8 5.8, as 7.0.0 did with 5.4

[Fishrock123]

Only if V8 5.8 comes out in April.

[ofrobots]

Extrapolating from the schedule, my non-official estimat is that V8 5.8 will go stable during the last week of April. My guesstimate is that Node 8.0.0 will ship on April 25. If we hold things, V8 5.8 may be possible in 8.0.0, but the schedule is too tight for comfort. I don't see a way to get V8 5.9 in though. @targos: perhaps you meant as a ABI compatible upgrade sometime after 8.0?

Re: the old debug protocol. This has been deprecated for at least 2 years – we just haven't done anything about it. I guess the fault is partially mine that I didn't make even more noise about the imminent deprecation. After chatting with the V8 team, they have agreed to delay the removal of the old debug protocol from V8 5.7 to V8 5.8. This should give us some deprecation breathing room if 8.0.0 ships with V8 5.7 instead of V8 5.8. Coming up with an API/ABI compatible patch to upgrade to 5.8/5.9 during the life of 8.x would be a lot more work though.

Regardless, as discussed in the collaborator summit, we will need to immediately start working on figuring out a path forward for the old debug protocol going away.

[jasnell]

The timing is going to depend largely on what kinds of breakage is expected in 5.8. Once the branch is cut in early March, the ability to land semver-major PRs will be cut significantly and will require CTC sign off (that's not an absolute rule, just the way that I'll be handling this release). If there are any breaking changes going into 5.8, they will be need to be reviewed carefully.

[targos]

I don't see a way to get V8 5.9 in though. @targos: perhaps you meant as a ABI compatible upgrade sometime after 8.0?

No you're right. I just went too fast and looked at the Branch points schedule.

[hashseed]

For the record: the legacy debugging JSON protocol has been deprecated for a very long time already. The reason it stayed functional until now is because there was no reasonable migration path away from it (aside from using the not-well-documented debug context). This has changed since v8-inspector finally landed in V8.

[targos]

I'm trying to integrate V8 5.8 and it seems that vm.runInDebugContext is also compromised.

[klimashkin]

There is only 2 month left for Node 8.
Not sure that 5.5 for Node 7 still makes sense, anyway it's not LTS.
Maybe it's better to spend more effort on 5.8 and its new inspect api, community will happily wait for it : )
Imo

Anyway async in 5.5 is not so fast:

[hashseed]

@targos what do you mean by "compromised"? v8::Debug::GetDebugContext is deprecated but should still work.

[targos]

@hashseed see https://github.com/nodejs/node/blob/master/src/node_contextify.cc#L262-L275. We are using v8::Debug::GetMirror.
I don't know if this is still needed or can be replaced by something else.
/cc @ofrobots @bnoordhuis

[hashseed]

Hehehe... check out this: v8@293fc84

[Trott]

Three things that need to be worked out, at least two of which will likely come down to the CTC in my opinion:

• what's the CLI debugger story in v8.0.0 (I think Diagnostics has that worked out already)
• what to do about Buffer constructor and when (paging @ChALkeR)
• whether or not to go with the new error system as-is or if it needs modifications or what. My primary motivation at this time is realizing that all error messages are going to change and not being certain of the cost to the ecosystem vs. the benefit. See errors: add internal/errors.js #11220 (comment) and discussion that follows

[jasnell]

The v8.x and v8.x-staging branches have been cut.
I'm running the first test release build now (https://ci-release.nodejs.org/job/iojs+release/1575/). (Note: This test build does not have the release bit set... it is essentially a cut directly off master).

I will generate a new test build each week and will begin working on the release candidate process starting April 4th.

@nodejs/collaborators @nodejs/ctc ... please work on landing all semver-major commits that should go into Node.js 8.0.0 by April 4th.

[jasnell]

New test build: https://ci-release.nodejs.org/job/iojs+release/1576
(the first run errored out because I'd forgotten that the REPLACEME tags needed to be replaced before running a release build (even for tests... that seems odd)

[jasnell]

Test build #2: https://ci-release.nodejs.org/job/iojs+release/1593/

[targos]

@jasnell this page is not public? I get "Access Denied".

[jasnell]

It's on the release ci so only release team members have access.

[jasnell]

ping @nodejs/build @nodejs/release ... see https://ci-release.nodejs.org/job/iojs+release/1593/nodes=vs2015-x86/console (/cc @jbergstroem)

[jasnell]

Test build #2: https://nodejs.org/download/test/v8.0.0-test20170327db9805885fb817c5b36e5b2bb53b82c0765e438f/

[joaocgreis]

@jasnell that looks like nodejs/v8#4, did that run with all the commits from #11752? I see 9c9e2d7 in v8.x-staging now, if it was not when you run the job then that is expected.

[jasnell]

@joaocgreis ... ok. I'll be running another test build tomorrow and will be syncing v8.x-staging and v8.x with master before I run.

[joaocgreis]

@jasnell there's a real issue here, tracking in nodejs/build#669 and nodejs/v8#4.

[gordysc]

Both tickets in the above comment have been closed. Is there still outstanding issues?

[iyobo]

@gordysc None! Release the Kraken! :D

[kyrylkov]

@gordysc @iyobo #12220 (comment)

[vjeux]

Sorry if this is the wrong place but is node 8 going to have support for trailing commas for function calls? Here's the commit on v8 that implements it: https://bugs.chromium.org/p/v8/issues/detail?id=5051

[klimashkin]

@vjeux Yes, it was shipped in v8 5.8

[ConAntonakos]

https://v8project.blogspot.com/2017/05/launching-ignition-and-turbofan.html

[bnoordhuis]

One for the release notes: new Date(...) date parsing follows ES2015 behavior now:

When the time zone offset is absent, date-only forms are interpreted as a UTC time and date-time forms are interpreted as a local time.

Refs #13210 and others.

[jasnell]

Landed. Thank you all.