[v8.x] deps: backport V8 update to 6.1

[targos]

We still need to revert some changes for ABI compatibility.
@addaleax would you like to help me with that? I'm not good enough with C++ to determine which changes could break addons or not.

Output of git diff v8.5.0 deps/v8/include/ > v8-61.diff: https://gist.github.com/targos/d3a1bf91592cb2e9257af94476274f27

/cc @nodejs/v8

Closes: #14220

Checklist

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

Affected core subsystem(s)

V8

[addaleax]

I’ll take a look and I’m glad to help in whatever way I can, but just so you know, today I heard that @hashseed might already have tried something out?

[addaleax]

From a quick look, the API/ABI breakage is:

• The removed ForceSet() overload
• The ArrayBuffer::Contents changes (this one is tricky and we had some prev discussion on this, don’t remember where but I think it was in the previous ABI compat updates)
• initial_error_prototype being added as a v8::Intrinsic in alphabetical order, not at the end of the list
• The max_semi_space_size_ field type change
• AllowCodeGenerationFromStringsCallback now refers to a different kind of function type
• The AdjustAmountOfExternalAllocatedMemory changes (at least the added CheckMemoryPressure symbol)

Also, some WASM and Modules changes that I think we don’t need to care about…?

[hashseed]

There was a list of things that Node has to float to make 6.1 ABI compatible with 6.0 somewhere in a node issue by @natorion.

[MylesBorins]

@nodejs/tsc if you could chime in on this it would be great

[addaleax]

@hashseed @natorion was the person who mentioned that you might have worked on sth ;)

I’ll try to work a few things off that list.

[hashseed]

#14220

I think the ArrayBuffer::Contents class layout issue has already been solved by reordering fields, which already landed in V8 long ago on 6.1. But you better verify :)

[addaleax]

Right, thanks for looking that up. I mentioned in that thread why just adding the fields at the end isn’t enough, too. ;)

[hashseed]

Right... In that case it might be better to rip out these fields and essentially revert https://chromium-review.googlesource.com/c/523271/

[addaleax]

Yeah, possibly more items in that feature set too – I don’t think we really got around to figuring out how to deal with segfaults-that-V8-might-handle

[MylesBorins]

/cc @addaleax @bnoordhuis @hashseed do any of you want to own working on ABI compat for this?

[hashseed]

@psmarshall

[addaleax]

I pushed an initial set of ABI compat changes to this PR, which should take care of a few items

[psmarshall]

@addaleax I had a go at solving the Contents issues.
It lives here: psmarshall@0e14e89

[addaleax]

@psmarshall Thanks, lgtm!
@MylesBorins Included your patch from #15431 in here.

From talking with a few V8 people today, we should probably include/wait for #15431 to land. The changes in there should be low-risk.

CI: https://ci.nodejs.org/job/node-test-commit/12382/
V8 CI: https://ci.nodejs.org/job/node-test-commit-v8-linux/906/ (edit: at least locally it works … the V8 CI is not exactly in a wonderful state right now)
CITGM: https://ci.nodejs.org/view/Node.js-citgm/job/citgm-smoker/985/

[xymeng2017]

gogogo! v8 6.2

[vsemozhetbyt]

@xymeng2017 It is unlikely: #15362 (comment)

[gibfahn]

V8 CI should be working better now (see nodejs/build#886), and #15431 has landed. If there are any more infra issues let me know and I'll take a look.

[addaleax]

@xymeng2017 I think V8 6.2 on Node 8 is out of the question, sorry.

[addaleax]

Looks good imho, although ofc I can only sign off on the commits that are not from me

(also, to be explicit: +1 to landing V8 6.1 on Node 8)