Skip to content

deps: update v8 to 15.8 - #67070

Open
legendecas wants to merge 14 commits into
nodejs:mainfrom
legendecas:v8-158
Open

legendecas wants to merge 14 commits into
nodejs:mainfrom
legendecas:v8-158

Conversation

@legendecas

@legendecas legendecas commented Oct 27, 2026 •

Copy link
Copy Markdown
Member

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

Review requested:

  • @nodejs/gyp
  • @nodejs/security-wg

@nodejs-github-bot nodejs-github-bot added build Issues and PRs related to Node.js builds or CI infrastructure. dependencies PRs that add, update, or configure Node.js dependencies. needs-ci PRs that need a full CI run. labels Oct 27, 2026
@legendecas legendecas mentioned this pull request Oct 27, 2026
2 of 3 tasks
@nodejs-github-bot

This comment was marked as outdated.

@joyeecheung

joyeecheung commented Oct 28, 2026 •

Copy link
Copy Markdown
Member

The current stable train is still on 15.7 (stable release next week). Shouldn’t we update to that one? https://chromiumdash.appspot.com/schedule

@aduh95

aduh95 commented Oct 28, 2026

Copy link
Copy Markdown
Contributor

My 2cts on this: I would be in favor of skipping 15.7, my reasoning is:

  • for users of Alpha releases, the more grouped are breaking changes, the better.
  • for us maintainers, landing V8 upgrades invalidate most of our binary cache.

If instead we group semver-major changes every few months, e.g. I see that V8 16.0 will be LTS and plan for December, that seems like a good candidate.

@joyeecheung

joyeecheung commented Oct 28, 2026 •

Copy link
Copy Markdown
Member

for us maintainers, landing V8 upgrades invalidate most of our binary cache.

I think the downside of following a non-stable V8 upgrade is that it generally contains bugs/regressions that are not yet caught and will only be fixed later after it is rolled out to Chrome users and noticed. As a result we need to patch it more often to fix the bugs/regressions introduced in the non-stable releases. And because we need to bump the embedder string every time we do it, every one of these incurs a V8 rebuild in the CI, the releases will invalidate previously compiled code cache, etc. If we stick to the stable release there will generally be fewer churn because they are already battle-tested by Chrome users.

@aduh95

aduh95 commented Oct 28, 2026

Copy link
Copy Markdown
Contributor

Hum case in point, V8 15.7 supports ICU 79.1 and V8 15.6 doesn't: #66987 (comment)
That being said, we likely want to backport this to 26.x, so cherry-picking it to main would also make sense

@legendecas

legendecas commented Oct 29, 2026 •

Copy link
Copy Markdown
Member Author

I can also revive #66799 if it's preferred. The CI was green on it, I think I just need to make a rebase. Also when I was about to rebase, V8 had the 15.8 branch cut so I went ahead with 15.8.

@joyeecheung

Copy link
Copy Markdown
Member

We do have

We upgrade the version of V8 in Node.js main whenever a V8 release goes stable upstream, that is, whenever a new release of Chrome comes out.

In https://github.com/nodejs/node/blob/main/doc/contributing/maintaining/maintaining-V8.md. Perhaps we should clarify further. I opened #67106

legendecas and others added 7 commits November 17, 2026 09:16
Major V8 updates are usually API/ABI incompatible with previous
versions. This commit adapts NODE_MODULE_VERSION for V8 15.8.

Refs: https://github.com/nodejs/CTC/blob/master/meetings/2016-09-28.md
Original commit message:

    [sandbox] fix dispatch table access on illumos

    illumos pointers are VA48, can allocate from the top of the 64-bit range
    as well.

    Co-Authored-By: Dan McDonald <danmcd@mnx.io>
    Refs: nodejs#65161
    Change-Id: I7d803ca0da2ab6c7cc206cb964d63683523a13da

Refs: v8/v8@011b2de
Original commit message:

    Add V8_DISABLE_PRESERVE_MOST for embedders to opt out of preserve_most

    preserve_most changes the calling convention, which clang
    encodes in the mangled name (e.g. "U" instead of "A" for cdecl in the
    Microsoft C++ ABI). A consumer whose toolchain does not support
    the attribute would reference a differently mangled symbol.
    This patch adds a V8_DISABLE_PRESERVE_MOST to disable it, so that
    it's possible to link to V8 using different toolchains, as long
    as the value of it is set consistently.

    Refs: nodejs#55784

    Change-Id: I67b49e739942d4026ac060be35bb39ec22256cbd
    Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/8532546
    Commit-Queue: Joyee Cheung <joyee@igalia.com>
    Reviewed-by: Michael Lippautz <mlippautz@chromium.org>
    Cr-Commit-Position: refs/heads/main@{#111381}

Refs: v8/v8@c389861
Original commit message:

    [temporal] Compute zoneinfo64.res length from ResB index table

    ICU4C's offsetTOCLookupFn sets UDataMemory::length = -1 for the last
    entry in a .dat archive (which occurs for zoneinfo64.res in small-icu
    builds where zu.res is excluded), and pointerTOCLookupFn sets length = -1
    for all entries in pointer TOC builds.

    When UDataMemory::length is -1, compute the bundle byte length from the
    ResB header offset (udata_getMemory(memory) - data) and the
    self-describing URES_INDEX_BUNDLE_TOP word in the ResB index table,
    matching ICU4C's res_init() in uresdata.cpp. Also check USING_SYSTEM_ICU
    in addition to V8_ENABLE_TEMPORAL_SYSTEMICU when deciding whether
    internal udatamem.h is available.

    Bug: 576986967
    Change-Id: I7a2c881eb3891a97eafb244f71aa405ff2434b51
    Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/8642579
    Commit-Queue: Manish Goregaokar <manishearth@google.com>
    Reviewed-by: Rezvan Mahdavi Hezaveh <rezvan@chromium.org>
    Cr-Commit-Position: refs/heads/main@{#111181}

Refs: v8/v8@a051772
joyeecheung and others added 3 commits November 17, 2026 09:18
Refs: https://chromium-review.googlesource.com/c/v8/v8/+/8524366
Signed-off-by: Joyee Cheung <joyeec9h3@gmail.com>
Signed-off-by: Chengzhong Wu <cwu631@bloomberg.net>
Signed-off-by: Joyee Cheung <joyeec9h3@gmail.com>
V8 deprecated --harmony-struct.

Refs: https://chromium-review.googlesource.com/c/v8/v8/+/8610686
Signed-off-by: Chengzhong Wu <cwu631@bloomberg.net>
Signed-off-by: Joyee Cheung <joyeec9h3@gmail.com>
@legendecas
legendecas marked this pull request as ready for review November 17, 2026 14:27
@nodejs-github-bot

This comment was marked as outdated.

@legendecas

legendecas commented Nov 17, 2026 •

Copy link
Copy Markdown
Member Author

Updated to V8 15.8.62. m158 is stable on Tue Nov 17 so this is going to supersede the 15.7 update.

Given that 15.7 is skipped. I will try to expedite getting this ready.

@legendecas legendecas added the semver-major PRs that contain breaking changes and should be released in the next major version. label Nov 17, 2026
Co-Authored-By: Joyee Cheung <joyeec9h3@gmail.com>
Signed-Off-By: Michaël Zasso <targos@protonmail.com>
Signed-off-by: Chengzhong Wu <cwu631@bloomberg.net>
@nodejs-github-bot

This comment was marked as outdated.

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

@richardlau

Copy link
Copy Markdown
Member

V8 CI: https://ci.nodejs.org/job/node-test-commit-v8-linux/nodes=rhel8-s390x,v8test=v8test/7514/

05:19:49 ld.lld: error: undefined symbol: gimli::read::abbrev::Attributes::push::h0bb0717b250c25e7
05:19:49 >>> referenced by abbrev.rs:361 (vendor/gimli-0.31.1/src/read/abbrev.rs:361)
05:19:49 >>>               std-199215890f3335f7.std.692f79298df501ce-cgu.0.rcgu.o:(gimli::read::dwarf::Unit$LT$R$GT$::new::h1ac327fd5aa79399) in archive prebuilt_rustc_sysroot/lib/rustlib/s390x-unknown-linux-gnu/lib/libstd.rlib

We might need a newer version of Rust (currently 1.88) for the V8 CI builds? Node.js CI seems to be ok.

libstdc++ before 12.3 passes the predicate of `std::erase_if` on a
`std::map` a const reference, so `KnaMap::EraseIf`, whose predicate
mutates the mapped value, fails to compile. Replace it with an explicit
erase loop.

Refs: https://gcc.gnu.org/PR107850
Signed-off-by: Chengzhong Wu <cwu631@bloomberg.net>
V8 now starts the `v8_internal_headers` source list with `sources = []`.
The GN scraper skips the empty list and matches `v8_compiler_sources = [`
instead, so every dependent of `v8_internal_headers` compiled all of the
compiler sources again. That breaks the shared library link with
duplicate symbols. Match the `sources += [` list that holds the headers.

Signed-off-by: Chengzhong Wu <cwu631@bloomberg.net>
Chromium's build no longer lists the Rust std backtrace crates in
`stdlib_files`, because its own std is built without backtrace support.
The system Rust toolchain used by `make-v8.sh` ships a std built with
it, so linking mksnapshot fails with undefined `gimli` and `addr2line`
symbols. Pass the crates through `added_rust_stdlib_libs`, which
replaces the libadler patch of `build/rust/std/BUILD.gn`.

Signed-off-by: Chengzhong Wu <cwu631@bloomberg.net>
@legendecas

Copy link
Copy Markdown
Member Author

We might need a newer version of Rust (currently 1.88) for the V8 CI builds? Node.js CI seems to be ok.

No, I think it's a make-v8.sh issue. Added commit 2aaf14a.

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

@nodejs-github-bot

Copy link
Copy Markdown
Collaborator

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

build Issues and PRs related to Node.js builds or CI infrastructure. dependencies PRs that add, update, or configure Node.js dependencies. needs-ci PRs that need a full CI run. semver-major PRs that contain breaking changes and should be released in the next major version. v8 engine Issues and PRs related to the V8 dependency.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

8 participants