[simon1hofmann]

🤖 AI text below 🤖

Description

Compiling RZZ(0.3) to Clifford+T at optimization levels 2 or 3 can fail because general optimization reintroduces arbitrary gates after synthesis. Run continuous optimization, Solovay–Kitaev synthesis, and discrete cleanup in one staged Qiskit pass manager. Keep Solovay–Kitaev and the Qiskit 2.2.3 minimum unchanged; version-dependent synthesis remains in #1081.

Disable permutation elision for native Clifford+T compilation so wire permutations remain in the circuit and survive QASM export without manual layout repair. After synthesis, apply OptimizeCliffordT recursively for nonzero optimization levels and translate to the target basis. Preserve native target normalization, structured loops, loop jumps, and classical measurement destinations.

Mirrors reuse only the exact cleanup pipeline before terminal measurements. High-level synthesis receives the target and equivalence library to lower inverse gates such as iswap_dg and dcx_dg, without repeating approximate synthesis.

Validation after updating to main: 656 tests passed with four existing expected failures on Qiskit 2.6.0; 103 focused Clifford+T, mirror, and loop tests passed on Qiskit 2.2.3. All lint hooks passed, using Python 3.14 for type checking. Hosted CI is running on the updated branch.

Codex assisted with implementation, Ponytail review, regression tests, validation, 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.

[simon1hofmann]

@burgholzer I tried different fixed to this problem but I'm still not really happy with it.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[burgholzer]

Alright. Played around with this a little. I think the solution now is pretty much fine and relatively clean 😌