feat(codegen): resume completed LLVM units after interrupted builds

[proggeramlug]

Change

Atomically checkpoint completed native units under the driver's full object-cache identity and a fingerprint of each frozen LLVM input. Treat missing, truncated, corrupt or unwritable records as cache misses.

Independent regression

Unit tests cover input isolation, corruption, nesting, unwinding and best-effort writes. Bounded CLI regression removes one of four completed units and the final module object, then requires 3 reused/1 rebuilt and byte-identical cold/resumed/uncached objects.

The branch starts directly from main and contains its own synthetic fixtures. No proprietary application sources, credentials, account or investigation artifacts are needed.

Validation completed

All four checkpoint unit tests passed. The bounded native driver proved 3 reused units/1 regenerated unit and identical cold/resumed/uncached objects. On unchanged main it found 0 of the required 4 checkpoints.

For transparency: the local before/after native and unit validation used main 8d88e481c and an integration checkout containing the nine audited patches together. Compiler and all nine runtime/extension archives had matching source identities. Each published branch is separate and passes its own CI warnings/check compilation jobs.

Local gate run: all applicable script gates passed except the already-stale public benchmark artifact; product/host warnings and Clippy passed. The all-in-one run reached its 10-minute cap in the API-doc rebuild; both API outputs were then independently verified byte-for-byte with the matching built compiler (no drift).

Existing CI failures / landing note

• Main already has the same public benchmark freshness failure and Linux custom thread-stack test failure. Neither is changed or suppressed here.
• The existing parallel typed_feedback test race encountered by the scoped codegen suites is fixed separately in test(codegen): isolate the default feedback lane #9999. Please include that test-only fix in the landing batch.
• Scoped CI may still be running or show those recorded failures; this is not a claim that every CI job is green.

Ready for the maintainer landing workflow. The full application will be rebuilt only after the required production changes are verified on main.

[coderabbitai]

Warning

Review limit reached

Next included review available in 59 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 8 included reviews currently available.

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Learn how review limits work.

Review configuration:

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: f63d572f-a9c8-461e-bc82-62b11c26a2cc

📥 Commits

Reviewing files that changed from the base of the PR and between 8d88e48 and 72f1bb1.

📒 Files selected for processing (8)

• changelog.d/9997-llvm-unit-checkpoints.md
• crates/perry-codegen/src/lib.rs
• crates/perry-codegen/src/native_emit.rs
• crates/perry-codegen/src/unit_cache.rs
• crates/perry/src/commands/compile/run_pipeline.rs
• crates/perry/tests/llvm_unit_checkpoints.rs
• scripts/test-codegen-unit-resume.mjs
• tests/modules/codegen_unit_resume/main.ts

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[proggeramlug]

Landed via #10001 (rebase train; author commits retained). Audited head: 72f1bb166da0ef4ff1884824033fbf18fd2c085e. Post-merge main 6b92bb2ef5ffe277e5f2773bea1a2e244a58347e is byte-identical to the validated tree. Full audit, validation, and conflict/test-harness repair details are recorded in the train PR.