rusty_v8

[mrsurge]

How are you getting past this wrench?

[DioNanos]

Hey! Took me a while to crack this one too :)

The short answer: we reuse the prebuilt V8 musl binaries that OpenAI publishes on their releases and plug them into the Termux build with a few minimal patches.

The key pieces:

• Prebuilt V8 musl binaries from openai/codex releases (tags rusty-v8-v<version>) — no need to compile V8 from source
• Three small patches to make it all work on ARM64:
  • patches/rusty_v8_prebuilt_out_dir.patch — points the build to the prebuilt archives
  • patches/v8_source_portability.patch — minimal source-level portability fixes
  • patches/v8_bazel_rules.patch — adapts Bazel rules for hermetic toolchains
• Cross-compilation via GitHub Actions — see BUILDING.md for the full setup

The V8 binaries stay as-is until an upstream dependency changes and requires an update — no arbitrary recompilation.

TL;DR: prebuilt binaries + light patches + hermetic CI = happy Termux ARM64.