[Clear1nsight]

Change and mathematical scope

Publish research/h25-v02-compiler-input-evidence, a self-contained source archive
of a completed static prerequisite. It exports 471 compiler-input metadata rows,
6,424 lookup names, and baseline/census graphs. Its portable JSON-only replay
checks 354 declared-file and 6,070 first-missing outcomes plus 12 specs walks,
unchanged after the census plan's two added files.

The claim covers conditional finite evidence. V02 completion, compiler enclosure,
installed-view acceptance, measured performance and theorem qualification remain open. Original decisions and omission
cases accepted remain 0/8. The source scan and 56 configuration assumptions are
inherited; effective frontend, complete namespace interval, assembler exclusion,
and original command/output/custody obligations remain open.

Task and overlap

User-assigned publication-only scope, refined from original V02 completion to a
completed prerequisite archive. Base e54523517d0f03e58352e647c403a5af3e525a57.
Reuse the original-role contract from #6397 at
0b3b5d923a54808a13c3102657fc51ef2ec5ba3e and conditional source-root/PCH argument
from #6354 at b3771dbc5f713ba3e90170c1dc3321e653222c65. Review-only #6358 at
84b0e70ba4da44b7904fe8d13903c26a3a8239e1 retains descriptor currentness and
interval limits. The archive adds the finite join to these merged contracts. Exact original artifact hashes and extraction/relocation mapping are in
SOURCE_MAP.json; original source archives remain external provenance.

Validation

python3 -I -B research/h25-v02-compiler-input-evidence/verify.py
python3 -I -B -O research/h25-v02-compiler-input-evidence/verify.py

The author and a separate publication reviewer checked both modes on Python 3.12. Archive integrity, candidate
arithmetic, both graph joins, and four rejection controls pass. Controls cover
PCH file, PCH directory, quote-origin shadow, and missing named gconv row. The
last is a metadata guard for a retained failure, not a trace replay. Export also
checked all 12,872 original-to-relocated literal-walk results against pinned JSON.
The public verifier uses only package bytes and standard-library code. The publication
reviewer checked all eleven original provenance pins and independently reproduced
all 12,872 relocation comparisons using a separate walker.

No original source scan, compiler/native execution, host path observation,
installed namespace, original omission case, full suite or benchmark was run
for this publication. No registered replay hook is added. Prior independent
local graph review does not automatically accept the relocated export.

Credit

Chris R (Clear1nsight) and the V02 workstream's retained observer, source-root
and projection work; predecessor PRs #6397, #6354 and #6358 at the heads above.
Substantial OpenAI Codex assistance for this archive and its checks, with earlier
AI assistance retained as disclosed in the predecessor work. Apache-2.0.

Open next questions

The maintained verification overview
connects this package to the original goal and contributor tasks.

The package README
requests one source-bound review of the 56 frontend/configuration assumptions and
hosted implicit include against the exact original two-role contract. Deliver an
assumption-by-assumption source table, effective frontend/argument/search-chain
identity, independent check, and explicit unresolved prerequisites. This closes
one necessary premise only; it does not replace installed-view or original-case
acceptance.

[Clear1nsight]

Publication and contributor handoff are complete at 1998ae8ee6d46bafac647329e5b41a85941bb2e7. The PR is Ready for review with maintainer edits disabled. All seven public file bodies and Git blob identities matched the reviewed export.

The author and separate publication reviewer passed the public JSON-only verifier in normal and optimized Python modes. Review independently checked eleven provenance pins and all 12,872 path-relocation comparisons. Original V02 remains 0/8.

Hosted run 42919554574 failed in plan at this head. The onboarding suite ran 760 tests, with 24 skips and one failure in test_current_ci_section_matches_selected_record. At pinned base e54523517d0f03e58352e647c403a5af3e525a57, the selected record requires make frontier7000-verify, while docs/ci-verification.md names make frontier6030-verify. The changed package does not edit either file. Package selection and replay were skipped. Local package checks supply the stated static evidence.

Focused contributor request asks for the source-bound 56-assumption/effective-frontend review. The maintained overview links both new archives and preserves the original acceptance counts. Please claim the exact starting head and files on this PR.