[unpaidmoon]

Change and mathematical scope

Audit, not a construction and not a κ change. The selected frontier stays 363005817/100000000000.

The public replay of the PR CrocSwap#7136 V02 export authenticates only the count of original_source_condition_assumptions. That finite program mentions the object once, in len(...) == 56. Replaying it unchanged returns PASS with original_commands_executed = 0. Changing __GNUC_MINOR__ from 3 to 4, and updating the manifest hash, leaves that PASS object unchanged. Deleting HAVE_IEEEFP_H is rejected with assumption count differs.

An exhaustive census of the 56 recorded names against the frozen CrocSwap#6397 commands (pyinit_failure, then exec_failure) finds no assumption name used as an argv element. The only token agreements are __GNUC__ = 13 with the basename x86_64-linux-gnu-gcc-13, and __STDC_VERSION__ = 201112 with the element -std=c11. Those are spellings in the driver, not an installed dump. /usr/include/stdc-predef.h is a regular graph node the pinned walker resolves to itself; the public checker never names it.

This is a proof about that finite checker, plus an exhaustive census of one 56-key object and two argvs. It is not a proof that any recorded value is false, and it is not a statement about every compiler.

Task and overlap

Base: e54523517d0f03e58352e647c403a5af3e525a57.

Task card: formal-interface-lemma, scope changed. The live request is the CrocSwap#1166 handoff on the 56 frontend facts (discussion comment 19530668), not a Lean lemma and not one of the five selected assumptions. The parent-origin handoff on the same thread is a different request and is not taken.

Dependencies: Clear1nsight commit 1998ae8ee6d46bafac647329e5b41a85941bb2e7 (PR CrocSwap#7136), vendored under research/v02-assumption-binding-gap/pinned-export/ because that commit is not an ancestor of this branch, and checked by Git blob id. Caller contract: research/h25-c05-criterion-consumer-contract/ROLE-PROOF.json and variant_driver_v1.py.txt from PR CrocSwap#6397.

Nearest overlap is CrocSwap#7136 itself, which publishes the export and leaves the binding open, and the v03 constructor model that already records automatic_preinclude.effective = False. This package does not rerun that derivation. No open PR states this count-only predicate or the argv census.

Validation

python3 -B research/v02-assumption-binding-gap/verify.py
python3 -B scripts/research_package.py --package research/v02-assumption-binding-gap

Python 3.12.12 was used for the author run. Both commands passed: verifier status PASS_V02_ASSUMPTION_VALUES_UNBOUND, package status PASS_REGISTERED_RESEARCH_PACKAGE with outcome=negative-result. No compiler, network fetch, or selected-replay was run. The public checker was executed only on temporary copies of the vendored bytes.

Credit

The export and the handoff are Chris R / Clear1nsight, PR CrocSwap#7136. The two-command contract is PR CrocSwap#6397. Prepared by unpaidmoon with substantial Cursor Grok assistance. Apache-2.0.

Open next questions

1. Can a transcript of these two frozen argvs, with the sample phase named, bind each of the 56 values? First step: archive one stdout whose argv equals the frozen command. Blocked by the absence of that transcript and by the export's exclusion of host file contents.
2. Do the two hashed pyconfig.h rows define HAVE_STAT_H, HAVE_IEEEFP_H, HAVE_SYS_TERMIO_H, MS_WINDOWS, MS_WINDOWS_DESKTOP, MS_WINDOWS_GAMES, or NT_THREADS? First step: compare sha256 dcda0cfa3f4971db195817881c58025db6c76f19ae45f26d3dda0a43001253e8 and 23931f53bc7ee512c6bd3162828747494eaa5e21f23dd63b31170ec1adfbe65e to file text. Blocked because the fixture stores hashes only.

Workflow feedback (optional)

The handoff names a commit that is not reachable from current main, so the audit vendors the seven public blobs and checks Git blob ids instead of a head: pin.