Roadmap for stabilization of vm modules

[targos]

According to @SimenB, here are the issues that should probably be taken care of before we unflag and mark vm modules as stable:

• SourceTextModule memory leak (#33439)
• vm.Script's importModuleDynamically should get a reference to the context used to run it (#35714)
• Segmentation fault with import() instead of calling importModuleDynamically (#35889)
• importModuleDynamically for vm module is cached (#36351)
• VM ESM with dynamic imports resolves promises before linking is complete (#33216)

/cc @devsnek

[SimenB]

That last one might be a bug/race condition in Jest's implementation still. I haven't found the time to dig into it properly since opening up the issue, unfortunately. It feels like it's related to some module caching or context GC thing, but that's pure speculation. I have some hope it's related to whatever needs fixing for #36351 (again, speculating)

[devsnek]

definitely blocked on v8:9968

[SimenB]

The above linked v8 issue has been closed as fixed now (:tada:) - any idea of when we'll know if it's enough for Node's use cases?

[targos]

The commit fixing the above V8 issue is in V8 9.4. I'll probably open a PR upgrading to that version next week.

[devsnek]

it is not fixed quite yet, check v8:10284

[SimenB]

Ah ok, thanks for the link! Seems there's quite a bit of activity recently, so hopefully it's not too far away 🙂

[vjeux]

Just wanted to say that Jest is blocked on those tasks being completed in order to properly support ESM. It would be amazing if this could be prioritized!

[jasonwilliams]

Vitest is also blocked on those tasks too by the looks of vitest-dev/vitest#795 (comment). It seems to segfault on vm.runInThisContext when ran in debug mode on VSCode

[benmccann]

I got another segfault in Vitest: vitest-dev/vitest#317

[piranna]

bytenode is also blocked by this.

[targos]

I think the root V8 bug is now chromium:1238312. There was recent activity on this CL: https://chromium-review.googlesource.com/c/v8/v8/+/3172764

[devsnek]

expansion doesn't work for chromium bugs, correct url is: https://bugs.chromium.org/p/chromium/issues/detail?id=1238312

v8 is working on overhauling host_defined_options to be a single Data instead of a FixedArray, as well as moving it to a common place that behaves better with caching for both modules and scripts (this includes the deprecation of v8::ScriptOrModule!!). i don't know what their planned timelines are but there's also some info here: https://bugs.chromium.org/p/chromium/issues/detail?id=1244145

[targos]

I added support for chromium: expansion.

[SimenB]

Should #44211 be added to the OP?

[silverwind]

vite also would have a use case for VM modules, if they were made stable.

[HarikrishnanBalagopal]

Any update on this? As mentioned earlier, using imports in jest test files needs the experimental flag

NODE_OPTIONS=--experimental-vm-modules jest

jestjs/jest#12990 (comment)

[GeoffreyBooth]

Any update on this?

Pull requests are welcome!

[AlbertMarashi]

cmon.. it's been years. This feature would help w/ so much in the JS ecosystem

[targos]

Pull requests are still welcome.

[SimenB]

After 2.5 years, one of the issues in the OP can finally be checked off 😀

[thernstig]

@joyeecheung can the first item be ticked off since #33439 is closed?

[SimenB]

Not sure if there's an issue for it, but the cache misses in vm.Script that was mitigated in #50137 (released in 21.1.0) will resurface when vm modules are unflagged - so fixing the cache in those cases should probably also be a blocker before unflagging/stabilizing.

[metawrap-dev]

Just to be clear, is this what is blocking VSCode extensions from using ES modules?

[woss]

@metawrap-dev no

[RedYetiDev]

@nodejs/vm
I believe that #36351 may be fixed (See #36351 (comment))