some work on supporting piscina

[alii]

What does this PR do?

Fixes #19924
Fixes #13218

• Documentation or TypeScript types (it's okay to leave the rest blank in this case)
• Code changes

How did you verify your code works?

third party test (but will likely need to do more since there are worker related changes)

[robobun]

^{Updated 4:21 PM PT - Oct 20th, 2025}

❌ @alii, your commit a45c263 has 1 failures in Build #29763 (All Failures):

• src/bun.js/event_loop.zig - zig error on 🐧 x64 - build-zig
• src/bun.js/event_loop.zig - zig error on 🐧 aarch64 - build-zig
• src/bun.js/event_loop.zig - zig error on 🪟 x64-baseline - build-zig
• src/bun.js/event_loop.zig - zig error on 🐧 x64-musl-baseline - build-zig
• src/bun.js/event_loop.zig - zig error on 🍎 x64 - build-zig
• src/bun.js/event_loop.zig - zig error on 🐧 aarch64-musl - build-zig
• src/bun.js/event_loop.zig - zig error on 🐧 x64-musl - build-zig
• src/bun.js/event_loop.zig - zig error on 🐧 x64-baseline - build-zig
• src/bun.js/event_loop.zig - zig error on 🪟 x64 - build-zig
• src/bun.js/event_loop.zig - zig error on 🍎 aarch64 - build-zig
• src/bun.js/event_loop.zig - zig error on 🐧 x64-asan - build-zig

---

🧪   To try this PR locally:

bunx bun-pr 19940

That installs a local version of the PR into your bun-19940 executable, so you can run:

bun-19940 --bun

[heimskr]

Most of my comments are nitpicks, but the use of reinterpret_cast instead of static_cast for casting JSC::JSGlobalObject* to Zig::GlobalObject* is a safety issue. If there's any doubt whatsoever about whether a JSC::JSGlobalObject is also an instance of Zig::GlobalObject, use jsCast for an assertion failure (including in release builds) or jsDynamicCast for returning nullptr if the cast fails. Otherwise, static_cast is fine.

[190n]

This assertion failure in test-worker-message-port-transfer-terminate is interesting: https://buildkite.com/bun/bun/builds/18199#01974255-d25e-4706-8891-290500bfd29b/2013-2014

Trying to repro

[190n]

Comments + I'll figure out the assertion failure

[190n]

we should revert this before merge

[alii]

sg

[alii]

Closing as:

• this PR grew too big
• many of the fixes it implemented already merged elsewhere, or were fixed in other ways
• conflicts