[isPANN]

Summary

Moves all four workspace crates from Rust 2021 to Rust 2024, following the official transition guide. Each commit maps to steps of the guide:

Commit	Guide step	What changed
81c2e1be	2–4: cargo fix --edition, set edition, test	edition = "2024" in all crates and rustfmt. Closure patterns rewritten for 2024 binding modes. Nested ifs collapsed into let-chains (the clippy follow-up the guide asks for). Illustrative doc snippet fenced as text, because make test --include-ignored was already failing on main.
16ee25fe	5: cargo fmt (separate commit)	Formatting only: 2024 style edition, mostly version-sorted imports. Safe to add to .git-blame-ignore-revs.
57ebb235	review follow-up	rust-version = "1.88" and one indentation fix.
8368ccaf	codecov follow-up	Tests for error branches the let-chain rewrite reflowed (test code only).
5a02d669	unrelated tooling fix	scripts/pipeline_pr.py wait-ci crashed (AttributeError: ... 'format') because it lacked --format. Added the option and a regression test.
4152d3ec	unrelated tooling fix	test_create_pr_uses_gh_pr_create_and_returns_current_context was stale: b17d57de made create_pr look up the PR by head branch without updating the test. Test-only.

Step 1 (cargo update): Cargo.lock is untracked, so this step produces no commit. make check passes on freshly updated dependencies.

Manual review the guide calls for:

• Macros: the tool rewrote expr to expr_2021; this PR reverts that. In 2024, expr also accepts _ and const {}, but in none of the 27 :expr fragments could that input fall into a different macro arm.
• Drop order: the migration flagged 6 sites. The workspace has no Drop impls and no lock or RefCell guards, so the only effect is when memory is freed.
• Doctests: they pass under 2024.
• Platform-specific code: there is no cfg-gated platform code.
• unsafe: the existing unsafe already uses explicit blocks.

MSRV: 1.88. The let-chains this migration introduces need 1.88, and so does the dev-dependency proptest 1.12. Verified with cargo +1.88 check --workspace --all-targets --all-features. On 1.88, the MSRV-aware resolver picks uuid 1.26 for rmcp, since uuid 1.28 needs 1.89.

Test plan

• make check (fmt-check, clippy --all-features -D warnings, full test suite, --include-ignored) after each commit
• make paper
• cargo +1.88 check --workspace --all-targets --all-features

🤖 Generated with Claude Code

[isPANN]

Agentic Review Report

PR #1207: Migrate workspace to Rust 2024 edition (head 81c2e1be). This is a generic, toolchain-only PR: 54 files, +255/−266, with no models or rules added or removed.

Structural and semantic

• Hard fails: none. The diff has no generated exports, no unrelated edits, and no trait or macro behavior changes.
• Edition bump: all four crates (problemreductions, -cli, -expr, -macros) and rustfmt.toml are set to 2024. style_edition = "2021" keeps formatting unchanged, and make fmt-check passes.
• Closure patterns (|(_, &v)| → |&(_, &v)|): these are the mechanical cargo fix rewrites required by the 2024 match-ergonomics rules. Every call site is .iter().enumerate().filter(...) over Copy items (bool, i64, tuples of usize), so the binding types stay the same.
• Let-chain collapses (all 39 hunks checked by hand):
  • Each rewrite merges an outer if / if let that has no else with an inner if / if let that also has no else. The new if A && let B = … && C { body } evaluates A, then B, then C, left to right with short-circuiting. That is the same order as the nested form. No else branch was added, removed or moved.
  • Statements after the inner block stayed outside the chain: for example, return; in cli.rs:392ff and normalized.push(...) after continue in ilp.rs:585ff/625ff.
  • ? inside a condition (rules/graph.rs:1581 aggregate_witness_evaluation(..)?, and solvers/decision_search.rs:39,79 is_satisfiable(..)?) runs at the same point and under the same guard as before. The mutable borrow from normalized.last_mut() is still scoped to the body.
  • The parenthesized || in mcp/tools.rs:254 keeps the original grouping.
• Edition-2024 drop order (if_let_rescope, tail_expr_drop_order), including sites the lint did not flag:
  • I searched the whole workspace (src, the three sub-crates, tests, examples). It has no Drop impls, no RefCell / Mutex / RwLock / .lock() / .borrow_mut() guards, and no ManuallyDrop / mem::forget or tempfile guards.
  • The only interior mutability is Cell counters (unit_tests/solvers/brute_force.rs:238, unit_tests/rules/traits.rs:513), the OnceLock registries, and the allocation-counting GlobalAlloc test (unit_tests/models/graph/bounded_component_spanning_forest.rs). That test counts only alloc and realloc, not dealloc, so earlier drops cannot change its count.
  • So the rescoping changes only when memory is freed. No observable behavior changes.
• Reverting expr_2021 to expr: this is safe. In 2024, $x:expr also accepts const { … } and _. That can change behavior only when an earlier arm with expr would now capture input meant for a later arm.
  • I checked all 27 :expr fragments across the 35 macro_rules!. No arm order lets _ or const {} reach a different arm.
  • impl_variant_param! arms are distinguished by , k:. The @declare arms of register_decision_variant! are distinguished by a trailing random. impl_variant_reduction!, impl_sat_to_ksat! and the test macros have one arm each.
  • Fragment semantics follow the edition of the crate that defines the macro, so the exported macros now accept slightly more input downstream. They never accept less.
• Other 2024 changes:
  • Box<[T]>::into_iter: the Box<[Plan]> / Box<[Expr]> fields are only iterated with .iter(), and SerializedNode uses Vec.
  • The gen keyword appears only as a string in the expr reserved-word list.
  • unsafe extern and set_var changes don't apply. The unsafe {} blocks in the global allocator were already there.
  • Merged doctests and never-type fallback compile and pass under -D warnings.
• Doc fence ```ignore → ```text in problemreductions-macros/src/lib.rs:211: this is correct, because the block is illustrative syntax and cannot compile under --include-ignored.
• Issue compliance: there is no linked issue. The PR body matches the diff.

Quality

• No production logic changed, so no new tests are needed. The existing suite exercises every touched model, rule and solver path.
• Local gates on the PR head (rustc 1.99.0):
  • make fmt-check: OK
  • make clippy (--all-targets --all-features -D warnings): OK
  • make test (workspace, --include-ignored): all pass, including 6241 lib unit tests, 344 integration tests and the MCP integration tests, with 0 failures.
• cargo doc prints 4 rustdoc warnings (URL is not a hyperlink, unclosed HTML tag, unresolved link). They are in doc comments this PR does not touch.

Feature test (make cli, scratch directory)

• pred show MIS prints the schema description line (commands/graph.rs let-chain).
• pred create --example MaximumMatching → pred solve returns Max(2) via ILP (maximum_matching.rs let-chains).
• pred create --example MinimumSetCovering → pred solve --solver brute-force returns Min(2) (covered_elements let-chain).
• pred path MIS QUBO → pred reduce --via → pred solve returns Max(4) with QUBO intermediate Min(-394). This recovers through recover_completed_result (rules/graph.rs let-chain).
• pred solve with a missing argument gives the usual clap error and usage text.

Findings

Critical: none.

Important: none.

Minor

1. No declared MSRV (Cargo.toml:12, problemreductions-{cli,expr,macros}/Cargo.toml:4). Edition 2024 requires Rust ≥ 1.85, and the PR body says so, but no crate sets rust-version. Without it, crates.io users on an older toolchain get a less helpful error, and the implicit resolver = "3" that 2024 brings has no MSRV to resolve against. Fix: add rust-version = "1.85" to each [package], or to a [workspace.package] that the crates inherit.
2. Misaligned indentation after the collapse. At problemreductions-cli/src/mcp/tools.rs:261-265, the anyhow::bail! arguments still sit at the old, deeper nesting level. At src/models/algebraic/ilp.rs:597, the format! string in normalize_constraint no longer lines up with its format!(. rustfmt skips these blocks because the string literal is longer than max_width, so fmt-check doesn't catch it. Fix: dedent them by hand, or split the long strings so rustfmt can format the blocks. This is cosmetic only.
3. PR body rationale for reverting expr_2021. The body says "no caller passes const {} or _", but the stronger reason is that no macro has an arm that such input could be redirected to. That also covers future downstream callers of the exported macros. Fix: optionally reword the body; no code change is needed.

Verdict: the change preserves behavior. Every let-chain is equivalent to the nested form, the drop-order changes have no observable effect because the workspace has no guard-type destructors, and reverting to expr is safe. It can be merged once a human approves; adding rust-version is recommended.

[codecov]

Codecov Report

❌ Patch coverage is 95.00000% with 20 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.99%. Comparing base (f11b2a4) to head (4152d3e).
⚠️ Report is 1 commits behind head on main.

Files with missing lines	Patch %	Lines
src/unit_tests/example_db.rs	82.35%	6 Missing ⚠️
...odels/misc/minimum_code_generation_one_register.rs	55.55%	4 Missing ⚠️
src/rules/ksatisfiability_cyclicordering.rs	50.00%	4 Missing ⚠️
.../rules/consistencyofdatabasefrequencytables_ilp.rs	0.00%	2 Missing ⚠️
src/rules/stringtostringcorrection_ilp.rs	0.00%	2 Missing ⚠️
...s/decisionminimumvertexcover_hamiltoniancircuit.rs	0.00%	1 Missing ⚠️
src/rules/hamiltoniancircuit_hamiltonianpath.rs	0.00%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1207      +/-   ##
==========================================
+ Coverage   96.90%   96.99%   +0.08%     
==========================================
  Files        1095     1095              
  Lines      145109   145468     +359     
==========================================
+ Hits       140621   141096     +475     
+ Misses       4488     4372     -116     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[isPANN]

Addressed the Minor findings of the Agentic Review Report in 57ebb23:

1. MSRV: declared rust-version = "1.88", not 1.85. The let-chains this migration introduces need 1.88. Verified with a real 1.88 toolchain across the whole workspace.
2. Indentation: tools.rs was already fixed by the 2024 style reformat in 16ee25f. ilp.rs:597 was realigned by hand.
3. PR body: now gives the stronger reason the expr revert is safe (no macro arm can be re-matched by the new input) and maps each commit to a step of the edition guide.