🐛 Reject nonlinear QCO programs at checked boundaries

[burgholzer]

🤖 AI text below 🤖

Description

Reject QCO programs that violate the linear quantum-value invariant at every checked ownership boundary.

Scalar qubits and rank-one qubit tensors and vectors are linear. Every quantum operation result and block argument must therefore have exactly one use. Transformation passes may rely on that invariant and do not need recovery paths for malformed IR.

This pull request:

• adds the shared qco::verifyLinearity(Operation*) semantic verifier;
• checks every nested quantum result and block argument and reports the actual use count;
• routes textual and owned-module construction through checked QCOProgram factories;
• validates direct QCO and converted QC/jeff input at the C++ pipeline and mqt-cc boundaries;
• rejects a QCOProgram that a caller invalidated through mutable low-level MLIR access before any public transform runs; and
• keeps QTensor transforms built on the linearity invariant and removes unreachable recovery branches.

This PR is based directly on main.

Validation:

• complete compiler unit suite: 135/135 passed;
• QTensor transform CTest: 1/1 passed;
• mqt-cc CTests: 2/2 passed;
• C++ lint: passed; and
• full repository lint: passed.

Hosted CI is pending for the rebased revision.

AI assistance: OpenAI Codex assisted with implementation, review, testing, and this description.

Checklist

• The pull request only contains commits that are focused and relevant to this change.
• I have added appropriate tests that cover the new/changed functionality.
• I have updated the documentation to reflect these changes.
• I have added entries to the changelog for any noteworthy additions, changes, fixes, or removals.
• I have added migration instructions to the upgrade guide (if needed).
• The changes follow the project's style guidelines and introduce no new warnings.
• The changes are fully tested and pass the CI checks.
• I have reviewed my own code changes.

If PR contains AI-assisted content:

• Any agent that created, edited, or submitted GitHub content was explicitly authorized for that scope, as required by our AI Usage Guidelines.
• Every agent-authored or agent-edited public text body begins with the visible disclosure 🤖 *AI text below* 🤖 (titles are exempt).
• I have disclosed AI assistance in the PR description.
• I confirm that I have personally reviewed and understood all AI-generated content, and accept full responsibility for it.

[codecov]

Codecov Report

❌ Patch coverage is 94.62366% with 5 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
mlir/lib/Compiler/Programs.cpp	92.6%	5 Missing ⚠️

📢 Thoughts on this report? Let us know!

[burgholzer]

This one looks good already.

[simon1hofmann]

Also went through this and looks good to me!

[denialhaag]

I also had a brief look at this. Apart from my three (non-blocking) nitpicks below, this LGTM! 😎

[denialhaag]

If I'm not missing anything right now, this could even be pulled out of the stack and merged immediately. 🤔

[burgholzer]

If I'm not missing anything right now, this could even be pulled out of the stack and merged immediately. 🤔

Yeah. Most likely! And it feels independent enough, that we could just do it. Let me try to do that in the next iteration. This feels stable enough.

[coderabbitai]

Important

Approval pending

CodeRabbit has no unresolved comments, but it has not reviewed the latest commit.

Use the checkbox below to review the latest commit. CodeRabbit will approve the changes if it finds no blocking issues.

• 🔍 Trigger review

---

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.}

[burgholzer]

@denialhaag @simon1hofmann pulled this one out of the PR chain and rebased it on current main. Just waiting for CI to spin up and finish!