Update SpiderMonkey to 153_0esr_RELEASE by sagudev · Pull Request #803 · servo/mozjs

sagudev · 2026-08-24T17:33:41Z

There are some changes in hazard packaging in TC so we repackage to older format to avoid more work.

SM now unconditionally depends on some rust code: vendored&patched icu crates and some glue crates that bridge icu through rust to C. Because recursive cargo invocation are no go we extract rust crates from tar and import them in our tree. This means we will publish all those mozilla and patched icu crates under mozjs- prefix. We also need to provide headers from cbindgen to SM, so we reuse same pattern as we used for zlib for this.

Then there are some boring changes like requiring newer clang, NDK.

We fixed and added some more patches and some glue. The most important parts are changes to JobQueue (for microtask changes) and some additional moduleloading code. For their use see corresponding commits in servo companion PR.

Reviewable per commits. During working on this I squashed many commits otherwise rebase would be hellish.

Testing: Covered by existing tests.
Servo PR: servo/servo#47489
Fixes #687

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

jschwe · 2026-08-24T18:27:43Z

Whats the reason for removing the Linux armv7 builds? Upstream dropped support?

sagudev · 2026-08-24T19:59:31Z

Whats the reason for removing the Linux armv7 builds? Upstream dropped support?

No, we would just need to update cross images, but my rationale was that armv7 as an arch is covered by android and arm64-linux means that it will probably also work on linux. Ideally we should be able to run/compile it on arm64 runners anyway.

atouchet · 2026-08-24T23:08:55Z

https://github.com/servo/mozjs/blob/main/README.md instructions are still for 140. Should that be updated to 153 as well?

jschwe · 2026-08-25T06:34:54Z

I think we might want to still do something about HOST_CC / HOST_CXX before this lands (could be in a seperate PR that we rebase on). My suggestion would be to not just check clang but also search for clang-XX so if a user on ubuntu has a newer (non-default) clang in PATH, then we choose that as the host compiler. Ubuntu 24.04 LTS still ships with clang-18 by default, so I think if we don't do anything it would be more breaking than it needs to be.
What do you think? I could probably make a PR.
I think the code should probably be in our build.rs so also embedders benefit from it (they are still free to override HOST_CC).

sagudev · 2026-08-25T09:04:49Z

I think we might want to still do something about HOST_CC / HOST_CXX before this lands (could be in a seperate PR that we rebase on). My suggestion would be to not just check clang but also search for clang-XX so if a user on ubuntu has a newer (non-default) clang in PATH, then we choose that as the host compiler. Ubuntu 24.04 LTS still ships with clang-18 by default, so I think if we don't do anything it would be more breaking than it needs to be. What do you think? I could probably make a PR. I think the code should probably be in our build.rs so also embedders benefit from it (they are still free to override HOST_CC).

So you propose we populate HOST_CC with newer clang? I am not sure how I feel about this aditional logic. It would be ok if we could reuse something from cc-rs or bindgen, but I am not sure something version oriented like this exists there.

Anyway I consider those stuff non blocking as most user will use artifact anyway, meaning they will get away with older clang.

jdm · 2026-08-27T12:35:56Z

I will have time tomorrow to review this.

sagudev · 2026-08-27T20:16:59Z

I published dummy version of mozjs extracted crates and invited jdm as co-owner and setup trusted publishing, so it should just work on land.

updated filters extract patched icu crates as mozjs-extracted-crates update patches Co-authored-by: Jonathan Schwender <schwenderjonathan@gmail.com> Co-authored-by: Gae24 <96017547+Gae24@users.noreply.github.com> Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

jschwe · 2026-08-28T08:50:47Z

For some reason I can't see jsglue.cpp in the github diff, but in SavedQueue cx is added as a member, and I don't see it being initialized anywhere. (This was reported in an agent review with the claim that it causes ./mach test-devtools to segfault. I didn't check if it crashes, but I looked at the code, and definitly don't see cx being initialized anywhere, so seems legit)

sagudev · 2026-08-28T08:58:25Z

For some reason I can't see jsglue.cpp in the github diff, but in SavedQueue cx is added as a member, and I don't see it being initialized anywhere. (This was reported in an agent review with the claim that it causes ./mach test-devtools to segfault. I didn't check if it crashes, but I looked at the code, and definitly don't see cx being initialized anywhere, so seems legit)

We run devtools test CI and it doesn't crashes, but it sure looks wrong (C++ compilers are useless).

use extracted icu stuff and other glue fixes (mainly microtasks) Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

sagudev · 2026-08-28T10:04:49Z

For some reason I can't see jsglue.cpp in the github diff, but in SavedQueue cx is added as a member, and I don't see it being initialized anywhere. (This was reported in an agent review with the claim that it causes ./mach test-devtools to segfault. I didn't check if it crashes, but I looked at the code, and definitly don't see cx being initialized anywhere, so seems legit)

We run devtools test CI and it doesn't crashes, but it sure looks wrong (C++ compilers are useless).

They did crash, but they work now so good catch.

jdm

The changes in this PR look reasonable to me. I'm going to take a look at the servo one next.

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

@Gae24

Companion PR to servo/mozjs#803 So this has been in the baking for a month at servo/mozjs#687. As in mozjs we need to update clang, NDK. The most interesting changes are microtask as jobqueue now lives in SM which greatly simplifies our logic, but running the queue is somehow harder. There is no need for enqueue promise job hook and we now have special hook to define incubemnt global of task (we do things more correctly now which brings two more passes). The second big change is moduleloading which is now mainly implemented in SM. Moduleloading commit was mainly written by @Gae24. Reviewable per commit. Testing: Covered by WPT. --------- Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com> Signed-off-by: Gae24 <96017547+Gae24@users.noreply.github.com> Co-authored-by: Gae24 <96017547+Gae24@users.noreply.github.com>

We already run publish there. I did this in #803 for testing but when squashing forgot to revert it. Signed-off-by: Sam <16504129+sagudev@users.noreply.github.com>

update dl and gen scripts

9b0c256

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

sagudev changed the title ~~Update SpiderMonkey to ESR7153.0.0~~ Update SpiderMonkey to ESR 153.0.0 Aug 24, 2026

sagudev changed the title ~~Update SpiderMonkey to ESR 153.0.0~~ Update SpiderMonkey to 153_0esr_RELEASE Aug 24, 2026

sagudev mentioned this pull request Aug 24, 2026

Update mozjs to include SpiderMonkey to 153_0esr_RELEASE servo/servo#47489

Merged

sagudev requested a review from jdm August 25, 2026 04:52

jschwe reviewed Aug 25, 2026

View reviewed changes

Comment thread mozjs-extracted-crates/collator_glue/Cargo.toml

jschwe reviewed Aug 25, 2026

View reviewed changes

Comment thread mozjs-sys/etc/update.py Outdated

sagudev force-pushed the 153-off branch from aeb30a6 to f1593e7 Compare August 25, 2026 08:51

sagudev marked this pull request as draft August 25, 2026 09:47

sagudev force-pushed the 153-off branch from 6efbbc6 to 7ecc058 Compare August 27, 2026 17:58

sagudev marked this pull request as ready for review August 27, 2026 20:01

jschwe reviewed Aug 28, 2026

View reviewed changes

Comment thread mozjs-sys/Cargo.toml Outdated

Comment thread mozjs-extracted-crates/icu_collections/.cargo-checksum.json Outdated

Comment thread mozjs-extracted-crates/icu_collections/Cargo.toml.orig Outdated

Comment thread mozjs-extracted-crates/collator_glue/Cargo.toml Outdated

sagudev and others added 3 commits August 28, 2026 10:45

Update SM and apply patches

499fb32

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

use newer clang, bump NDK and macos sdk, bump MSRV, rm armv7 builds

95170e1

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

sagudev force-pushed the 153-off branch from 7ecc058 to 40ddcf5 Compare August 28, 2026 08:50

sagudev added 5 commits August 28, 2026 11:07

mozjs-sys

51d3536

use extracted icu stuff and other glue fixes (mainly microtasks) Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

mozjs - regen wrappers and fixup imports

d9da751

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

handle mozjs-extracted-crates in CI

85dcb41

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

140 -> 153 in docs

290e28b

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

mark mozjs-extracted-crates as generated

b4c6f6b

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

sagudev added 3 commits August 28, 2026 11:07

support proper dry run in publish_crates_io.py

eab7590

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

impl Rootable for Box

612b574

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

borrowed_error_report

3f542f1

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

sagudev force-pushed the 153-off branch from 40ddcf5 to 3f542f1 Compare August 28, 2026 09:07

jschwe reviewed Aug 28, 2026

View reviewed changes

Comment thread mozjs-extracted-crates/icu_normalizer_data/Cargo.toml

jdm reviewed Aug 28, 2026

View reviewed changes

Comment thread mozjs-sys/etc/patches/0032-macos-sdk-min.patch Outdated

bring macos sdk back down

9b381d5

Signed-off-by: sagudev <16504129+sagudev@users.noreply.github.com>

sagudev requested a review from jdm August 29, 2026 06:39

jdm approved these changes Aug 29, 2026

View reviewed changes

sagudev merged commit 29ba719 into servo:main Aug 29, 2026
45 checks passed

sagudev mentioned this pull request Aug 29, 2026

do not run main workflow on push on main #807

Merged

sagudev added a commit that referenced this pull request Aug 29, 2026

do not run main workflow on push on main (#807)

7352d27

We already run publish there. I did this in #803 for testing but when squashing forgot to revert it. Signed-off-by: Sam <16504129+sagudev@users.noreply.github.com>

atouchet mentioned this pull request Sep 4, 2026

Upgrade to icux to 2.2 #751

Closed

Conversation

sagudev commented Aug 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

jschwe commented Aug 24, 2026

Uh oh!

sagudev commented Aug 24, 2026

Uh oh!

atouchet commented Aug 24, 2026

Uh oh!

Uh oh!

Uh oh!

jschwe commented Aug 25, 2026

Uh oh!

sagudev commented Aug 25, 2026

Uh oh!

jdm commented Aug 27, 2026

Uh oh!

sagudev commented Aug 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

jschwe commented Aug 28, 2026

Uh oh!

sagudev commented Aug 28, 2026

Uh oh!

Uh oh!

sagudev commented Aug 28, 2026

Uh oh!

jdm left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

sagudev commented Aug 24, 2026 •

edited

Loading

sagudev commented Aug 27, 2026 •

edited

Loading