[ianmccul]

Uni20 PR #118 makes log_gamma_options::max_terms optional. An unset limit compares below an integer, so both Bethe Barnes coefficient guards reject default construction with the new dependency: native contexts throw “Barnes requires sixteen native coefficients” and MPFR contexts throw “Barnes precision exceeds coefficient budget”.

Normalize the legacy integer/new optional field to an optional budget and enforce it only when set. Explicit limits remain strict, the existing coefficient-count formulas are unchanged, and Bethe's Uni20 pin remains at ca350254.

Added focused numerical regressions for default and explicit budgets in double, long double, and MPFR at 128/256 bits, checking the independent identity G(4)=2 and end-to-end rejection of insufficient explicit caps. In MPFR, Gamma is constructed first and adds its own 64 guard bits to Bethe's working precision: it requires floor((p+128)/6)+8 coefficients versus Bethe's floor((p+64)/6)+8. Gamma therefore rejects insufficient explicit caps before Bethe's defensive guard. That assertion does not independently cover the Bethe MPFR guard; default construction is the discriminating regression for the unset-budget bug. Native explicit-cap tests exercise Bethe's own guard.

Validation at 2579ad0: all three tests fail before the fix with Uni20 697e49e7; all three pass afterward with both 697e49e7 and pinned ca350254 (GCC 13, Release, MPFR enabled). CTest discovers all three cases. Follow-up 643f6d3 changes comments only; non-comment source and whitespace were checked without rerunning numerical tests. Existing pinned production is unaffected; this is a consumer compatibility fix for newer dependency overrides.

[chatgpt-codex-connector]

Codex Review Summary: ✅ Completed

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-11-08T11:00:47.260790Z	643f6d3	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[Copilot]

🟢 Approval recommended

The review identifies no unresolved issues, and the changes preserve explicit-limit validation while covering default and supported-precision Barnes behavior with focused regressions.

No open issues 🎉

What changed in this PR

Updates Bethe’s spinon Barnes coefficient-budget handling for compatibility with Uni20’s optional max_terms, while preserving strict validation for explicitly configured limits. Adds focused numerical regressions across native and MPFR configurations.

Changes:

• Normalize legacy integer and optional coefficient budgets before applying Barnes guards.
• Allow default construction when no budget is set while rejecting insufficient explicit budgets.
• Add Barnes identity and budget regression tests for native and MPFR precisions.
• Register the new regression executable with CTest.

File	Description
include/​bethe/​detail/​spinon_barnes.hpp	Applies native and MPFR coefficient-budget checks only when an explicit budget is set.
tests/​test_spinon_barnes.cpp	Tests default construction, Barnes identity evaluation, and sufficient or insufficient budgets across supported precisions.
tests/​CMakeLists.txt	Builds and discovers the new Barnes regression test executable.

_{Review effort: Lite}

[ian-review-bot]

Claude review of exact head 2579ad07409b519e7d71b6f1efe413cab0928b53

Found 1 issue, in the test rather than the fix; details in the inline comment on tests/test_spinon_barnes.cpp:46. The MPFR "insufficient explicit cap" assertion is satisfied by Uni20's own guard. Bethe's MPFR guard cannot fire for any explicit budget.

What the PR does.

• SpinonBarnesContext in include/bethe/detail/spinon_barnes.hpp now copies log_gamma_options::max_terms into a std::optional<std::size_t>. Both coefficient guards (MPFR and native) run only when a budget is set.
• Without this, Uni20 697e49e7 (where max_terms became optional with {} as default) makes nullopt compare below any count, so a default-constructed context throws.
• The PR adds tests/test_spinon_barnes.cpp: a G(4)=2 identity check plus checks for default, explicit-OK and explicit-too-small budgets. It covers the native types and MPFR at 128 and 256 bits.
• The Uni20 pin stays at ca350254.

What I reviewed.

• The header change and the new test.
• log_gamma_options and log_gamma_context in Uni20 ca350254 and 697e49e7: default budget, guard order, guard bits.
• All Bethe uses of log_gamma_options/max_terms (git grep). The only log_gamma_options::max_terms reader is this header; the xxz_thermodynamic_structure_factor max_terms is a different options type.
• Consumers: XXXSpinonDensity (via xxx_barnes_transition_rate.hpp) and the XXX thermo apps.

Verified. All runs: GCC 13.3, Release, x86_64 (80-bit long double), float128 and MPFR on, MPLAPACK off. I configured with the build recipe from REVIEW.md and -DFETCHCONTENT_SOURCE_DIR_UNI20 pointing at Uni20 ca350254 or 697e49e7.

• bethe_spinon_barnes_tests at head: 4/4 pass with both Uni20 revisions, both directly and via ctest -R SpinonBarnes. CTest discovers 4 cases (fp64, long_double, fp128, MPFR), not the 3 the description says, because float128 is enabled here.
• Same test compiled against the base (17656b06) header: all 4 fail on 697e49e7 ("Barnes requires sixteen native coefficients" ×3, "Barnes precision exceeds coefficient budget" ×1) and all 4 pass on ca350254. This matches the description.
• Neighbouring inputs, using a probe that checks G(4) and G(7)=34560 against the returned bounds:
  • Native: budgets 0 and 1 are rejected by Uni20; 2 and 15 by Bethe; 16 and 17 succeed. Results are identical on both Uni20 revisions.
  • MPFR defaults at 24, 53, 113, 1000 and 2900 bits are inside the bounds on both revisions.
  • At 2960, 3000 and 6000 bits, ca350254's default 512-term cap rejects the request, while 697e49e7 now accepts it and agrees with G(4) and G(7). That is a deliberate behaviour change from Uni20's precision-aware default.
• On 697e49e7: bethe_xxx_spinon_density_tests (24 cases), bethe-xxx-structure-factor-thermo and bethe-xxx-moments-thermo all build. The 24 gtest cases pass, and XXXSpinonDensityPrecision.cli passes (94 s). The 24 density cases also pass on ca350254.

Not checked.

• Clang 19 or the ASan/UBSan preset.
• The ci-wide MPLAPACK build.
• aarch64 and MSVC, where long double differs.
• The rest of the CTest suite and the other apps, with either Uni20 revision.
• CI for this head: no failed-job logs were present when I reviewed.

---

_{Review tier: full (code or new pull request) · model: claude-opus-5-5 · effort: high}