[Issue Backlog] Backlog Analysis - 2026-03-17

[github-actions[bot]]

Issue Backlog Analysis - 2026-03-17

Note: This is an incremental run. Previous analysis was 2026-03-15 (138 issues). Today's run covers 137 open issues (1 closed since last run). Issues already commented on within the last 4 days are skipped to avoid spam.

Executive Summary

• Total open issues analyzed: 137
• Issues recommended for closure: 2
• Potential duplicates / merge candidates: 9 (in 4 clusters; see below)
• Issues with suggested fixes or workarounds: 8
• Issues commented on this run: 3 (new comments; 10 comments placed in prior runs within window)

---

Issues Recommended for Closure

These issues appear to be resolved, no longer relevant, or fully answered.

Issue	Title	Reason for Closure
#7958	Z3 4.15.3 is not on opam	Packaging issue for a version that has since been superseded by 4.15.4, 4.16, etc. If the opam package has been updated for a newer version, this can be closed. Reporter should confirm.
#8940	Is there a way to combine QF_BV and QF_IDL?	A direct answer was provided by NikolajBjorner: use smt.arith.solver=3 (Floyd-Warshall) or smt.arith.solver=2 (Simplex). The follow-up discussion confirmed a benchmark showing arith.solver=3 is 2× faster than the default. Remaining open as a performance tracking issue — consider closing if maintainers prefer.

---

Potential Duplicates / Merge Candidates

These issues overlap with other open or closed issues and could be tracked together.

Cluster 1: String / Seq Invalid Model (Canonical: #8023)

The canonical issue #8023 tracks ongoing soundness violations in the string/sequence solver. These issues all produce invalid models or wrong sat/unsat verdicts involving string or sequence operations:

Issue	Title	Notes
#8023	Soundness issue in string logic (canonical)	9 comments; multiple test cases still trigger as of 2026-03-16
#7841	Incorrect model for QF_SLIA	seq.++ / length constraints
#7664	invalid model (string API)	Python API string soundness
#7593	Invalid model with sequences: foldl, nth and ++	seq.foldl unsound model
#6982	invalid model on incremental QF_SLIA	Incremental mode
#6346	invalid model on incremental string formula	Earlier incremental string bug

Cluster 2: Float / FP Invalid Model (Canonical: #7162)

Issue	Title	Notes
#7162	Invalid model on float formula (canonical)	Consolidated issue
#7431	Invalid model issue on float formula	Reproduced 2024-10
#7321	Invalid model issue on floats	Reproduced 2024-08
#7842	Incorrect model (FP + mixed)	Recently confirmed still present
#8185	Incorrect model in mixed FP/Real + string constraint	FP cluster + string interaction

Cluster 3: JS/Node.js Thread Non-Termination (Canonical: #6512)

Issue	Title	Notes
#6512	NodeJs process won't stop after using solve (canonical)	WASM threading
#6701	Javascript & Typescript API Runs Forever	Same root cause
#7070	NodeJS Z3 doesn't clean up threads	Same root cause

Cluster 4: arith.solver=6 Performance Regressions (Canonical: #4934)

Issue	Title	Notes
#4934	[Consolidated] Performance regressions of arith.solver=6 (canonical)	Known regression tracker
#5614	Performance regression from arith.solver=2 to arith.solver=6	Confirmed regression
#7632	Performance regression Fixedpoint 4.13.2→4.14.1	Fixedpoint + arith
#8940	QF_BV + QF_IDL performance (arith.solver=6 2× slower)	New data point: BV+IDL workloads

---

Issues with Suggested Fixes

These issues have identifiable root causes or affected components.

#7730 — Spacer returns sat instead of expected unsat

• Component: Spacer / Model-Based Projection (fp.spacer.native_mbp)
• Relevant source files: src/muz/spacer/spacer_mev_array.cpp, src/muz/spacer/spacer_manager.cpp
• Suggested fix direction: The native_mbp code path produces a spurious inductive invariant. Setting fp.spacer.native_mbp=false restores correct behavior on 4.13.3; the workaround status on current HEAD is unknown. Assigned to @agurfinkel.

#7593 — Invalid model with sequences: foldl, nth and ++

• Component: Sequence theory solver
• Relevant source files: src/smt/theory_seq.cpp, src/smt/seq_axioms.cpp
• Suggested fix direction: The seq.foldl encoding produces a satisfying assignment for which the model evaluator gives an inconsistent value (the model states all = (seq.unit 4) ++ (seq.unit 0) but final = 6, which contradicts foldl semantics). This is part of the broader string/seq invalid-model cluster (Soundness issue in string logic #8023, invalid model issue on an incremental QF_SLIA formula #6982).

#7592 — PartialOrder hangs when reasoning about bottom element

• Component: PartialOrder Python API + MBQI quantifier instantiation
• Relevant source files: src/api/python/z3/z3.py (PartialOrder function), src/smt/smt_quantifier.cpp
• Suggested fix direction: PartialOrder generates quantified axioms (transitivity, antisymmetry, reflexivity) without explicit :pattern hints. MBQI can loop on ground instances that need multi-step chaining. Explicit pattern annotations or a LinearOrder / manual encoding avoids the loop (as shown in the issue). A fix could add patterns to the PartialOrder-generated axioms.

#7538 — produce-unsat-cores affects convergence

• Component: MBQI / E-matching interaction with unsat-core tracking
• Relevant source files: src/smt/smt_quantifier.cpp, src/smt/smt_relevancy.cpp
• Suggested fix direction: Enabling produce-unsat-cores alters which clauses are marked relevant, changing which quantifier triggers fire. This is a side-effect of relevancy propagation (see also Unexpected proof failure due to minor input tweak #7444 for a detailed analysis by @CanCebeci). Not a correctness issue per se — when cores are enabled, Z3 produces unsat; without them it returns unknown.

#8023 — New segfault in latest comment (2026-03-16)

• Component: Bit-blast tactic / quantified arithmetic
• Relevant source files: src/tactic/core/blast_term_ite_tactic.cpp, src/ast/rewriter/
• Suggested fix direction: The latest comment (2026-03-16) by @wintered reports a segfault on a purely arithmetic formula with :fp.xform.bit_blast true and nested quantifiers — unrelated to the string soundness bugs in the same issue. This segfault deserves its own issue for tracking.

#7417 — Z3 returning wrong sat-answer (Horn + Arrays)

• Component: Spacer proof checker (spacer_proof_utils.cpp)
• Relevant source files: src/muz/spacer/spacer_proof_utils.cpp line 358
• Suggested fix direction: NikolajBjorner confirmed that release mode now returns unsat (correct), but a debug-mode assertion violation remains in spacer_proof_utils.cpp:358 ("Arithmetic proof check failed"). The arithmetic Farkas coefficient check is failing on a valid proof. Likely the proof checker's Farkas lemma reconstruction is too strict. Assigned to @hgvk94, @agurfinkel.

#7403 — Non-termination with mod0/div0 quantifiers

• Component: Quantifier instantiation / E-matching
• Relevant source files: src/smt/smt_quantifier.cpp
• Suggested fix direction: The pattern (mod (mod x y) y) triggers an instantiation that produces (mod (mod x y) y) again, creating a loop. Z3 does not detect this specific idempotency. The query can be solved without those quantifiers since Z3's arithmetic solver handles (mod (+ a b) b) natively. Removing the quantified mod axioms and relying on the built-in theory solver should work.

---

Issues Needing More Information

These issues lack sufficient detail to investigate or reproduce.

Issue	Title	Missing Information
#8052	Unsat only with unnecessary asserts, else unknown	Test case is in a file attachment only — inline SMT-LIB2 snippet would enable analysis
#7730	Spacer returns sat instead of unsat	Confirmation of whether fp.spacer.native_mbp=false still works as workaround in Z3 ≥ 4.15
#7113	Completeness regression on LIRA	No follow-up; no confirmation of whether still present in current versions

---

Notable Issues Deserving Attention

Issues that are particularly impactful, long-standing, or actively accumulating new evidence.

Issue	Title	Age	Notes
#8023	Soundness issue in string logic	~4 months	Critical; now 9 test cases; latest adds unrelated segfault
#7162	Invalid model on float formula	~2 years	Consolidated float soundness; 5 related issues still open
#7730	Spacer returns sat instead of unsat	~8 months	Soundness; assigned to @agurfinkel; no resolution yet
#7417	z3 returning wrong sat-answer (HORN + Arrays)	~5 months	Debug assertion remains; may indicate proof checker bug
#7684	OCaml bindings don't install with cmake	~9 months	18 comments; affects cmake-based OCaml users
#7991	Infinite loop / perf regression 4.15.3 (BV rewriter)	~5 months	Has workaround; 14 comments; BV rewriter regression
#6512	NodeJs process won't stop	~3 years	Affects all WASM/thread users; 3 related issues
#4934	arith.solver=6 performance regressions	~5 years	Consolidated; new data in #8940 (2026-03-11)
#5648	Regex performance cliff with InRe	~4 years	Related to 8893 (4.16 regex regression)
#3962	Spacer: error with uninterpreted mod in Horn	~6 years	Long-standing Horn/Spacer limitation

---

Automated by Issue Backlog Processor — runs every 2 days. Previous run: 2026-03-15.

AI generated by Issue Backlog Processor · history

• expires on Mar 24, 2026, 12:22 AM UTC

[github-actions[bot]]

This discussion has been marked as outdated by Issue Backlog Processor.

A newer discussion is available at Discussion #9039.