[burgholzer]

🤖 AI text below 🤖

Description

Extract independently useful fixes and simplifications from #2545 onto current main. Each change has its own commit; this PR does not depend on the exception replacement.

• Preserve bare and uppercase DD imaginary-unit coefficients.
• Reject zero allocation capacities in MemoryManager.
• Handle terminal measurement fidelity and validate widths/permutation indices.
• Report buffered DD file-write failures after closing the file.
• Reuse ParsedBenchmark for JSON evaluation, preserving manifest validation and diagnostic order.
• Check typed gate-parameter arity at compile time.
• Remove the unused register-width argument from the MLIR DD adapter.
• Bind JIT constructors/destructors to the session runtime and retain DDSIM output through teardown.
• Query the existing functionality DD in the exhaustive Shor test instead of constructing and multiplying dense basis vectors.
• Correct the documented benchmark-generator result type.
• Execute DD documentation examples in a temporary directory.

Core's native throwing APIs remain. Source-tree MLIR callers use std::array for typed gate parameters and omit the unused adapter width. Result types, diagnostic adapters and exception policy remain in #2545.

CI investigation

The reported Windows failure occurred on main: SamplingRetainsStateWithoutChangingSamples reached 600 seconds. Successful previous main and #2545 suites took 76.73 and 78.29 seconds, respectively. These snapshots do not show a sustained exception-replacement regression. No root cause for the isolated hang was reproduced; host and worker stacks from a recurrence are still needed.

The Shor change preserves every modulus, multiplier, input and control case, plus the separate coherence, inverse and workspace checks. Matched Linux arm64/GCC 13.3 Release binaries, three runs each pinned to CPU 0, gave medians of 13.770 → 13.053 seconds. Ranges overlap (13.666–14.401 versus 12.524–14.211 seconds), so this is not a reliable CI speedup claim. Most duplicate tests previously removed from #2545 were never on main; no further deletion was justified without losing distinct coverage.

Validation

• Release build and native suite: 4,014 tests, one existing skip.
• Focused Python DD, benchmark, compiler and built-in driver tests: 455 passed.
• Executable documentation and internal links passed.
• Repository lint and whole-file C++ lint passed after fixing and rechecking the reported test-style findings.
• Windows/macOS execution was not run locally. The buffered-write regression uses /dev/full where available.

Implemented with Codex assistance; human review remains required.

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.
• 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

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!