Reproducibility archive for "The move-only converse and the ω-infinite island of the complex sin²-algorithm"

Reproducibility archive accompanying the converse-half paper "The move-only converse and the ω-infinite island of the complex sin²-algorithm" (preprint https://doi.org/10.5281/zenodo.23058612 , 228 pages, which pins this archive by its printed SHA-256; its four-page statement document gives the statements and their statuses). The move-only converse is our own formulation; it is not part of Karpenkov's Problem 4. Contents (560 files; 559 entries in the root SHA-256 manifest): finite island fixtures and their certificate programs (island/); registers of the periodic-dichotomy and block-grammar sections (registers/); calculation programs and data (external_records/); manuscript sources and figures (manuscript/); Problem 4 calculation programs (problem4/); supplements; and the execution register (replay/). The manuscript sources omit the printed archive SHA-256, so a PDF recompiled from them lacks that fingerprint and has a different SHA-256 from the deposited PDF. The register records 135 executions: 76 with return code 0 and 59 with a nonzero return code (54 blocked, 5 path); 0 control failures. It covers 133 distinct program paths: 76 have a recorded return code 0 and 57 have none. The Python inventory contains 152 files: 137 in the main archive and 15 in the supplement; 19 are library modules without an autonomous entry point. Each nonzero return code is classified by its observed cause in replay/BLOCKED.tsv and MISSING.md: “blocked” identifies a required input that is unavailable, including bytes with a required fingerprint; “path” identifies a configured location that does not resolve; “control failure” means a reached mathematical check failed. A zero return code alone does not establish that a mathematical check was reached or passed; the scope and outcome must be read in the output and command descriptor. The five recorded path outcomes are distinct from the two successful mapped executions. verify_extension and analyze_long_returns stop because the configured verify_uniform_seven_returns.py location does not resolve; neither has a successful mapped execution. external_records__bc_len20_exact_certificate reaches the output write after its mathematical checks, then stops because external_records/<workspace> does not exist; its separate declared-mapping execution returns 0. Here <workspace>, denoted WORKSPACE_TOKEN in MISSING.md, is the unresolved workspace-location token in a configured path, not a supplied directory or an assembly field. external_records__extract_uniform_ratio_D stops because the configured external_records/verify_gloc2_D.py does not resolve; its separate declared-mapping execution returns 0. manuscript__figs__make_certification_summary stops because /etc/matplotlibrc is outside the declared isolated mounts; it has no successful mapped execution. The executions of external_records/bc_len20_exact_certificate.py and external_records/extract_uniform_ratio_D.py are recorded as “replayed with declared path mapping”. The first mapping provides an initially empty writable directory for the output. The second links the expected external_records/verify_gloc2_D.py to the deposited file registers/audit_ge47_uniform_repair/verify_gloc2_D.py, SHA-256 012a9386043f474567112074e9ab53a8e8f9a8b57b0f00503e61adfef9b34c97. The extraction imports that identified copy; it does not run the imported module's standalone verifier. Neither mapping supplies absent mathematical input or changes any program or input bytes. The command descriptors and replay/COMMAND_PROTOCOL.txt specify how to reproduce these mappings using the archive alone. The 40 finite calculations supporting the computer-assisted construction are indexed in replay/CONSTRUCTION_CALCULATIONS.tsv by manuscript reference, deposited program and SHA-256, recorded output and return code. The programs are in supplements/uniform_gluing_sources.zip. These finite checks do not constitute execution of the infinite construction. The construction has not been executed in its stated domain: transport Lipschitz constant 1, spacing threshold at least 47, and the return, avoidance and spacing conditions of Corollary 3.C49. That execution is missing, not known to fail. replay/NUMBERS.tsv compares manuscript quantities with recorded outputs: 263 equal, 0 different and 90 not produced. “Not produced” means that a quantity is absent from the recorded output; it does not mean disagreement. The table identifies the extraction rule, manuscript location and comparison target. State-fingerprint comparisons are metadata comparisons, not mathematical results. The replay establishes neither the written proofs nor their adequacy to the stated claims, and supplies no formal seal. Commands, environments, resource limits, raw outputs and return codes are deposited. SANITIZATION.tsv identifies source and deposited file fingerprints by content category. Directory and file identifiers are used by the manuscript and programs; they carry no additional mathematical claim. This work was assisted by AI. No independent mathematician has reviewed it to date. Mechanical verification does not address whether the statements express the intended mathematics. The author alone is responsible for the content. Archive SHA256: e46412e3f3a10bbb61cc39448f4d9f4164f4d19375ae587e32aab7a31d59e6d2