llvm cache

[zeriyoshi]

No description provided.

[Copilot]

🟡 Changes recommended

The LLVM prefetch script can fail on a full cache miss due to adding the native dpkg architecture, and the new documentation overstates “Linux build jobs” coverage relative to the unchanged emulated workflow.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR introduces an LLVM apt package prefetch/cache flow intended to eliminate repeated downloads from apt.llvm.org during CI container builds by restoring a local llvm-debs/<codename>-<arch>/ apt repository and teaching the Dockerfile to consume it when present.

Changes:

• Add LLVM_PACKAGES build arg and Dockerfile logic to install LLVM either from a local llvm-debs/... file-based apt repo or from apt.llvm.org.
• Extend CI “Prepare” to prefetch LLVM .deb packages per architecture, cache them, and pass cache keys into native Linux jobs for restore.
• Document the LLVM install/caching behavior in both English and Japanese READMEs and ignore llvm-debs/ in git.

File summaries

File	Description
README.md	Documents LLVM installation and CI cache behavior.
README_ja.md	Japanese documentation for LLVM installation and CI cache behavior.
Dockerfile	Adds LLVM package list arg and local-apt-repo install path via bind-mounted build context.
.gitignore	Ignores llvm-debs/ directory.
.github/workflows/ci.yaml	Wires LLVM cache key outputs from Prepare into native Linux workflow call.
.github/workflows/ci_prepare.yaml	Prefetches/caches LLVM .deb packages and outputs cache keys per arch.
.github/workflows/ci_native_linux.yaml	Restores the LLVM .deb cache for native Debian builds before container build.

Review details

• Files reviewed: 6/7 changed files
• Comments generated: 4
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.