fix: give an actionable error when moonbitlang/core is not bundled

[mvanhorn]

• Related issues: Fixes Windows: Cannot inject the standard library moonbitlang/core after fresh install (v0.1.20260522) #1718
• PR kind: Bugfix (diagnostics)

Summary

moon build on a fresh install where ~/.moon/lib/core exists but was never bundled now fails with an actionable message that names the core directory and the exact remediation (moon -C <core_dir> bundle --target all), instead of the generic Cannot load the core file. A missing core directory gets its own message pointing at reinstalling the toolchain or $MOON_CORE_OVERRIDE.

Why this matters

#1718 hit this on Windows: the old error reads like a corruption/path bug and sends users down the wrong debugging path, while peter-jerry-ye's confirmed fix is just running the bundle command. The change is confined to the .context(...) on read_module_desc_file_in_dir in inject_std (crates/mooncake/src/resolver.rs); the underlying error stays in the source chain and the CannotInjectCore wrapper still prefixes the standard-library context.

Output after this change (exists-but-unbundled path):

Cannot inject the standard library `moonbitlang/core`: Cannot load the core file. The standard library at `/tmp/.../unbundled_core` does not appear to be bundled. Run `moon -C /tmp/.../unbundled_core bundle --target all` to bundle it

Metadata

• Tests added/updated for bug fixes or new features - N/A for this diagnostics-only change; both failure paths verified by running the built moon binary against a minimal project with MOON_CORE_OVERRIDE pointing at an empty dir and a nonexistent dir (outputs above)
• Compatible with Windows/Linux/macOS - message formatting only, no platform-specific APIs; cargo check -p mooncake and cargo fmt --check pass

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
crates/mooncake/src/resolver.rs	22% smaller

[peter-jerry-ye]

Although I think the fix is fine, I don't think such situation should happen often enough to add a fix. Either it's because the user is trying to develop moonbitlang/core without reading CONTRIBUTING or there's a bug.

Such a fix would hide away such unexpected situation. I would need a stronger reason for this fix.

[peter-jerry-ye]

Also I'm still waiting for the clarification for that issue because it claimed to be a "fresh installation"

[mvanhorn]

That's fair - no pressure on this one. If the reporter's clarification shows it really was a fresh install hitting a missing core, this is ready; if it turns out to be a broken dev setup, happy to close it as not worth the surface area. I'll leave it to your call once they respond.