Custom limits for dynamic workers

[git-bruh]

No description provided.

[github-actions]

All contributors have signed the CLA ✍️ ✅
_{Posted by the CLA Assistant Lite bot.}

[git-bruh]

I have read the CLA Document and I hereby sign the CLA

[ask-bonk]

I'm Bonk, and I've done a quick review of your PR.

This PR adds a ResourceLimits struct (with cpuMs and subRequests) and plumbs it through the dynamic worker loading and entrypoint/actor-class APIs. I found three issues, two of which are high severity.

1. [HIGH] Build break: WorkerStubImpl in server.c++:4107-4114 does not accept the new limits parameter, so the override keyword will fail to compile against the updated pure virtual interface in WorkerStubChannel.
2. [HIGH] Bug in DynamicWorkerSource::clone(): The new .limits field is not propagated by the clone() method (io-channels.h:361-372), so cloned sources silently lose their limits. This code path is exercised by WorkerLoader::load(), which calls source.clone() inside a closure that may be invoked multiple times.
3. [LOW] No tests: There are no tests covering the new limits plumbing — neither for WorkerCode.limits flowing into DynamicWorkerSource, nor for EntrypointOptions.limits flowing into getEntrypoint()/getActorClass().

---

This review was generated by an AI assistant and may contain inaccuracies.

[github-actions]

The generated output of @cloudflare/workers-types matches the snapshot in types/generated-snapshot 🎉

[ask-bonk]

Review posted successfully on PR #6399. Here's a summary of the findings:

1. [HIGH] Build break: WorkerStubImpl in server.c++ overrides getEntrypoint and getActorClass with the old two-parameter signatures, but the base class WorkerStubChannel now requires a third kj::Maybe<ResourceLimits> limits parameter. The override keyword will cause a compilation failure.

2. [HIGH] DynamicWorkerSource::clone() drops limits: The clone() method does not include the new .limits field in its designated-initializer list, so cloned sources silently lose their resource limits. This affects the WorkerLoader::load() path.

3. [LOW] No tests for the new limits plumbing.

github run

[codspeed-hq]

Merging this PR will not alter performance

✅ 70 untouched benchmarks
⏩ 129 skipped benchmarks¹

---

_{Comparing pkhanna/custom-dynamic-worker-limits (d1dbd0f) with main (0f907f5)}

Footnotes

1. 129 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[kentonv]

We probably don't need a clone() method since the struct is trivially copyable anyway.

[git-bruh]

Yeah, I'd actually shifted the struct out of edgeworker and this method was being used in quite a few places so I let it be as-is for consistency