Backport fix of Wasm on arm64 (M1) to Node 16

[surma]

Version

16.3

Platform

Darwin SurmBook.local 20.5.0 Darwin Kernel Version 20.5.0: Sat May 8 05:10:31 PDT 2021; root:xnu-7195.121.3~9/RELEASE_ARM64_T8101 arm64

Subsystem

No response

What steps will reproduce the bug?

Detailled reproduction can be found in this crbug: https://crbug.com/1224882

How often does it reproduce? Is there a required condition?

100%

What is the expected behavior?

Should execute the Wasm successfully.

What do you see instead?

E_BAD_ACCESS

Additional information

There are two CLs that landed in V8 that fix this problem:

• https://chromium-review.googlesource.com/c/v8/v8/+/2917612
• https://chromium-review.googlesource.com/c/v8/v8/+/2998582

They will be part of the next v8 release, but it might be worth back-porting them to Node 16 as it is the first release that support arm64 natively (as far as I know?)

[devsnek]

These patches don't apply cleanly but I could try to manually apply the logical changes. Dunno what our policy is on that.

[mcollina]

I would start with the backport and fix when they do not apply.

[targos]

I'm able to cherry-pick cleanly if I add two other commits. Compiling locally...

[targos]

#39337