Enigma forwarder repair: independent acceptance and source integration

PASS for the bounded malformed-export-forwarder repair at immutable candidate 34a6d1b31d0e3280ca363a3c25200ee07f56afc8. Independent review found no substantive defect in the changed parser/resolver. This accepts static diagnostic tooling; dynamic closure, emulator implementation, startup and gameplay remain unproven.

The executed parser SHA256 is f949185c282300ac3463b5524420464f87883b5ab5cb6e833fdf564cf6d83bbf. Forwarder reads stop at the export directory's exclusive end and the 4096-byte string bound. Empty strings cannot become direct exports: only None denotes a direct export. Both PE decoding and caller-supplied resolver graphs validate forwarder syntax. Provider matching normalizes case and optional .dll once; decimal ordinals normalize before lookup, and recursive binding rejects cycles. System-provider matches remain recorded requirements rather than evidence that the emulator implements those functions. The existing final-dot convention and bounded ASCII/module-name policy remain the diagnostic tool's supported scope.

The unchanged baseline diagnostic is deliberately preserved with SHA256 41050c56ffe274f87f38fc3020efbd06eec29f3e7bd77f188495dd0a99691d20. Its hardcoded parser SHA256 is 9cb012199d5307b63e369f140aea0875aee0b519b136c0033dd3ef735a8a298a. This review reuses the retained matching baseline execution (exit zero, REJECT_PARSER_CORRECTNESS), independently matching the exact source and oracle to Git revision 17cb4edae7d5a20874c7afc560dede1978c92ee4 and rehashing its receipt/log. It was not weakened or run as an acceptance oracle on repaired source. Its historical reproduction requires that exact baseline checkout.

Fresh candidate regression executes exact source bytes and rejects the original four-byte SDL_Init EAT mutation semantically, before immutable seals. File offset 299908 changes RVA 0x1340 to 0x53000; mutant SHA256 remains 04448e081b6a27c4cf02989d5c763cd2cc9f1d1f0cb7e0e3934964fa5a6791c0. The full verifier separately rejects the mutant's hash. No hash bypass is claimed. All 33 rejecting controls and eight valid forwarders pass, as do the unchanged producer's fourteen semantic controls and two guard/sentinel controls.

The new independent test targets gaps in those controls, rather than repeating their syntax matrix. A multihop named/ordinal chain must choose b.dll rather than b.dll.dll, normalize #00012, traverse the terminal provider's imports, and allow repeated consumers. Removing the exact provider rejects; a cycle through an already visited module rejects. Actual owned SDL copies verify a non-unit export ordinal base, reject named EAT holes and out-of-range name ordinal indexes, and distinguish a malformed forwarder on the last directory byte from a legitimate direct mapped data export at the exclusive end. These are three positive groups and five rejecting controls, all passing.

Authentic normal-import counts remain 849 entries, 40 edges and 192 packaged bindings. Six fixed codec initializers request 74 symbols: 50 packaged and 24 from missing providers. No inference of codec startup necessity follows. Generic/register-derived loads, including Enigma VA 0x5ab960, remain unknown. The original helper SHA256 remains e29aa3fb9f182a1a6ca48e8f932dc7e6857b3404c0d945be685f8ad2e47225fe; its original verifier/test and their recorded 2460-file BLOCKED_PACKED_IMPORTS outcome are unchanged. Full original mutation tests were not rerun against read-only standing originals. Only their safe guard tests run.

Source integration uses the exclusive shared integration lock, reconciles actual remote main, and copies only the reviewed Enigma tools, tests and historical notes/evidence absent there, plus this review and its independent test/evidence. It creates a source-only commit on fresh main without importing unrelated donor history. Root tracked files, index and refs, all donors, and standing assets stay read-only. Publication receipts record actual task/main refs after nonforce push.

Reproduce using fresh owned destinations (the producer test mutates fixtures):

python3 -B test/enigma-forwarder-repair.test.py \
  /tmp/enigma-authentic-static-unpack-oct24 /tmp/enigma-final-repair
python3 -B test/enigma-unpacked-closure.test.py /tmp/enigma-final-repair
python3 -B test/enigma-forwarder-independent.test.py \
  /tmp/enigma-authentic-static-unpack-oct24 /tmp/enigma-final-independent
python3 -B test/enigma-original-closure.test.py

Exact final-tree checks and execution provenance are in the evidence JSON beside this note. Raw logs, mutation fixtures, full final diff, resource/admission, publication and cleanup receipts are retained under /tmp/enigma-forwarder-independent-integrate-oct24/; external launcher output and the full session are charged by its one-owner ledger.py. Initial projection was 56,197,120 / 67,108,864 bytes including all ten copied input/output modules, temporary overlap, external history/SQLite future growth and 8 MiB reserve. No installed tool was copied or acquired. All executed Python commands have fresh 805306368-byte MemAvailable admission and 60-second CPU/wall bounds. No Node, compiler, build, unpacking, acquisition, guest, browser, box or deployment runs. Mount argv and box ID/state/expiry are N/A; coordinator owns runtime lifecycle. Receipt log hashes bind retained bytes but do not prove arbitrary supplied logs were produced by their claimed commands. No dynamic closure or runtime claim.