Skip to content

fix: qualify generated-first manuscripts and native publication boundaries - #289

Draft
docxology wants to merge 47 commits into
codex/release-5-1-candidate-20261105from
codex/ageint-renderer-0-2-20261107
Draft

docxology wants to merge 47 commits into
codex/release-5-1-candidate-20261105from
codex/ageint-renderer-0-2-20261107

Conversation

@docxology

@docxology docxology commented Nov 7, 2026 •

Copy link
Copy Markdown
Owner

Current owned release work — exact native policy installed; publication/runtime gates pending

Owned branch 271337baae4f622450114d77234d86a3c809accc, tree 57310b20a72a54ff4570f46a7227105eadd2bf7b. Ordinary parentcad901d; exactly six literal native policy files committed. Source is frozen for the complete original publication candidate. No protected-main, foreign Source/readiness, tag, release, archival or approval transition.

  • Complete actual25 coverage metadata: original24 ordinary measurements from 42451800355 and fresh post-repair Active 42456787792 retain their original0f603/f22 producer identities. The old aggregate/Active failure stays failed. Fresh Active records91.03%/inputc55d7d9c; original artifact13608712878 SHA256bc494579cb7955e2d1a63337d4be03dbd50c9354d125457cacce8c8472879b36 and full raw log authenticate. Native17546 completes all25/current hashes/support and exposes three real old-literal drifts; after only those native literals and the real2026-11-08 cohort date,17552 verifies25 complete/0 failure/0 drift and canonically renders provenance/COUNTS from genuine live facts. Actual full CLI17556 passes, Source unchanged/reaped/remaining[]. Exact four metadata/review files are ordinarily committed as cad901d. Native17725 then compares all25 inputs/support after the policy-only cutover at271337b: no changed project input, no changed Active support, every current literal still equals its original genuine measurement. No hash-only numerical reuse, hand-adjusted counters or coverage-database recomputation claim.
  • Actual repair, not restamped certificates: 42455181907 completes13 declared Active producers, seven regeneration/strict-audit commands and200 canonical Boolean checks. Ordinaryf22cfef0 integrates the unchanged complete36-file output patch with exact original native tree8b58eb88. No coverage rebuild flag, scientific change, compiler/census bypass or direct workflow push.
  • All4 real reviewed author originals: 42459932001 succeeds at exactcad901d and committed review SHA26ee331712b0add404966af502125f4ec4567395c731b5721bc838c7a978f79d. All8 original archives and four CLI exits0 authenticate. All Linux/macOS ×3.14/3.15 complete995-module/13080-family/17261-member inventories and explicit decisions agree. Review preserves981 unrelated records, every prior default/exception and all31 obligations; exactly33 old witnesses are rebound,12 real family anchors/two Pandoc/one uv exceptions added, and two obsolete families retired. Original Linux3.14 draft8597428 bytes SHA25620d3a0fc85718ff442227238ff4bb13b01bf035896a29c000d9e1490a00146b1; other targets review. No missing case is defaulted.
  • Canonical native encoding and literal installation: actual native17698 independently binds all4 original API/source/runtime/process/nonces/bytecode records and4124 current Source hashes, parses the original draft and performs exact encode/decode/literal roundtrip. Logical SHA 52e01e3b09253220cda1c6dbcb983bb860f872fe39016bceab3534cafe1c19d2,7001980 decoded bytes,five native parts/31 obligations. All six complete native files are guarded and installed byte-for-byte, then ordinarily committed/pushed as271337b; no hand-edited policy or obsolete alias/part. Actual installed SDK3.14.10 load_policy17707 verifies13080 families/31 declarations/current Source witnesses. Loading is not admit_plan, runtime prerequisite availability, case execution or release acceptance. Original erroneous/refused harness carriers remain preserved; only own verified temporary seed copies are removed after reaping.
  • Current exact-Source dependencies: all4 post-install census42461995066 and complete genuine publication refresh42461998287 now bind271337b. The existing publication workflow keeps its complete actual refresh/dependent-Pitch/nine-project rebind complement, strict original gates, one worker and unchanged budgets. Neither result is assumed; integrate only the complete authenticated original native candidate tree. All hosted watches are lease-free.
  • Retained scoped native/reader proof: actual complete main-consumer151 cases per supported minor and renderer/PDF/navigation/partition130 per minor pass without skips. Inert-template/frontmatter correction preserves live-ID/container refusals and authored bytes. Actual native browser target/back/unique-TOC,widths,panels/images/captions/math and local-only assets are retained; HTML paint/scroll helpers remain bounded unavailable, so no painted/deep-wheel claim is fabricated. Earlier actual PDF paint stays scoped/historical. Own tabs are released and read-only viewers exit0 normally; no shared browser restart or process kill.

Still required: original post-install metadata and actual runtime admission; genuine complete publication integration, recompare measured source/support identities and remeasure any real affected Source; full unchanged-output rebind/current hardening/CI; independent owner readiness and ordinary protected-main integration; exact-final-main authenticated installed52-profile acceptance and accepted manual SDK5.1.0 release; then stable-SDK AGEINT0.2 production/readiness/promotion. No metadata, component, load or input-shape success substitutes for those gates.

Coordination: finite own manager consumers reaped and released at their normal safe points; no repository context spans a hosted wait. HumOS and every other pane's running work/Source, locks, policy and approvals—including the separate blocked release/Zenodo approval—remain untouched. No force push, manual process kill, lock clearing, policy/runtime override, guard/floor weakening or approval impersonation.

Clean Source cutover

Generated-first manuscript selection, native publication boundaries, lazy provider discovery, explicit offline assets and bounded large-book math remain shared SDK contracts. Compiler freshness remains the producer's obligation; authored-source fields stay truthfully null. Ordinary merges preserve carrier 780748e99e6a36846c8d11d65db4346ab782913c and its original work. The separate shared checkout and approvals are untouched.

Current qualified Source

Published review Source: 71f3b52bf0bef7f2395766367bad0e69c53d6d34.

  • Functional CI passed at code Source 8ea46a3d2f95ec7a78a79628c315a5c49b61eb35: 42394855443. It includes actual complete Active own-release coverage, 91.03%, for its changed input hash.
  • Fresh reviewed authoring passed on all four Linux/Darwin × Python 3.14/3.15 targets at the published review Source: 42399024255. Every target reviews 996 modules / 13,081 families. Original policy DATA still requires authenticated transport, native encoding and Source admission; collection/review is not execution or acceptance.
  • All 25 genuine own-profile measurements are retained. Native current-input comparison admits 24 unchanged C inputs plus fresh Active at R0; the native complete snapshot reports zero percentage drift. Canonical COUNTS and provenance were regenerated through the actual producer, not hand-restamped.
  • Both selected native Python minors passed 259 cases in eight complete integrity modules. Each retained one strict final-census OS PermissionError refusal; no cleanup guard or policy was weakened. Both minors separately passed eight decoder controls, 13 real Active replay cases and fresh offline/hash-required cache consumers.
  • The actual full-book browser loaded all 12 original font faces with the bounded decoder, 177 intact figures and corrected unnumbered navigation. Development reader proof is not offline or release acceptance.

Source repairs

  • Preserve complete generated rosters, genuine BibTeX success and prior-PDF refusal behavior.
  • Keep explicit offline mode and bounded MathJax discovery/reflow; no CDN fallback.
  • Discover lazy Core exports without heavy cold imports.
  • Use reviewed Noble runners and Active's existing group in the actual selected interpreter; preserve locked fresh offline consumers.
  • Distinguish actual pytest workers from inherited environment; retain typed native stream observations and strict artifact closure.
  • Copy independent source bytes without privileged BSD flags; use current module bytecode locations, Python-floor syntax checks and fresh optional distribution metadata.
  • Preserve portable native LZMA streams, exact finite decoder budgets, original font bodies/hashes/licenses. Retire incidental wording/wiring pins rather than repin them.

Remaining release gates

Install only the authenticated native reviewed policy; qualify current-source hosted hardening, including the strict macOS census; regenerate coordinated publication outputs, compare all own/support input hashes, remeasure changes and obtain all-25 unchanged-output rebind. Then ordinary #289 → #281 → main integration, complete exact-final-main native/installed/Root/shard-union acceptance, authenticated original handoff and tag-bound release audits.

No main/tag/release acceptance is claimed. Historical tags, original failures and all unrelated work remain unchanged. No force, bypass, policy override, PyPI or Zenodo publication.

Reproduced the native failing-before boundary: BibTeX exit 2 was reported successful and replaced a valid PDF without its bibliography. Add real bibliography resolution and prior-publication preservation regressions; remove import-wiring checks. After-fix native and hosted release qualification remain pending.
@docxology
docxology marked this pull request as draft November 7, 2026 04:28
@docxology docxology changed the title fix: AGEINT figure identity and fail-closed BibTeX rendering fix: qualify generated-first manuscripts and native publication boundaries Nov 7, 2026
@docxology

Copy link
Copy Markdown
Owner Author

Fresh exact-Source CI passed on aee13dca1c406deccdcf979a71d580fbb7a7acdf: https://github.com/docxology/template/actions/runs/42360591322. The CI Gate retained the complete functional-member receipts; all 25 selected project smokes and infrastructure contracts completed. Requested all-project own coverage was not part of this run and is not claimed. Four-target installed-policy review and downstream full-book verification remain separate running gates.

@docxology

Copy link
Copy Markdown
Owner Author

Native coordinated publication refresh is running on frozen aee13dca1c406deccdcf979a71d580fbb7a7acdf: https://github.com/docxology/template/actions/runs/42361191560. It uses the existing fresh unprivileged Linux execution account, one worker, original phase exits and both strict committed-tree gates. Generation overlaps the independent fresh census; no integration or acceptance is inferred from a launch. Only authenticated successful review artifacts will be considered for ordinary integration.

@docxology

Copy link
Copy Markdown
Owner Author

Fresh final-Source census/review passed on all four Linux/Darwin × Python 3.14/3.15 targets: https://github.com/docxology/template/actions/runs/42360514514. All eight original API ZIPs are retained and authenticated against actual API sizes/SHA256; all four original CLI exits are zero with empty stderr. Every target reviews 996 modules / 13,082 families. The newly authored Linux 3.14 draft is byte-for-byte identical to the already installed genuine payload (draft SHA256 4113f09ab1193fe4b521a50dcc6d3ce17000bfeaea77c3f9078fee0085194b64), with no restamping or policy changes. Together with current-Source CI this closes the code/policy review prerequisite, not runtime/release acceptance. Native publication generation 42361191560 remains running.

@docxology
docxology marked this pull request as ready for review November 7, 2026 10:03
@docxology
docxology marked this pull request as draft November 7, 2026 10:45
@docxology

Copy link
Copy Markdown
Owner Author

A real native Stage03 run reproduced a pre-existing release-blocking offline defect: render.html.asset_mode: local was discarded while reconstructing RenderingConfig, so successful HTML silently referenced CDN MathJax. A native emitted-output regression fails before the one-field correction; full after-fix qualification and actual pinned local-asset hashes are now running. This PR is draft again while that required revision is qualified. Existing publication run 42361191560 at the older P head remains running and untouched; its artifacts will not be presented as proof for changed code. No approvals, processes or policies were overridden.

Real local proofs retain six fresh offline no-build/hash consumers, full regression selection and native protocol/Markdown controls on both minors. Source policy renewal, current coverage/count metadata, fresh publication and exact-main acceptance remain mandatory; this is not release acceptance.
The complete genuine H5 reader reproduced blank References at y=4739856 on a fresh 390px load. Removing four root overflow clips restores actual paint with scrollWidth390, all177 images and13math containers. Retire four CSS token/wiring/bare-not-throw cases instead of repinning them, and remove root clipping from the real math-reflow fixture. This local inspection proof is not runtime or release acceptance.
Review all four authenticated B0 native collection graphs; retire six incidental families and explicitly classify the real two-member Pandoc boundary. Bind changed evidence to actual authenticated old/current AST definitions. Canonical counts follow native all25 own-profile input equivalence with the genuine fresh B0 Active metric. Current-Source native authoring, installed policy admission and full runtime/publication/main qualification remain required.
All four original B1 native censuses bind4123 actual Source hashes,994 modules and13076 families. Actual selected-runtime policy parsing, canonical encoding, byte-exact installed readback and load_policy admission pass with31 prerequisites; logical hash2f66d64a8b91b1668db3abcc6072528126a4001740412e418b7d3108bdd66fd8. Requalify all25 own-profile inputs before canonical metadata refresh. No runtime, publication, main or release acceptance is claimed.
Integrate only the 454 generated paths in original run 42416683937, source 4db5cf6. Original review artifact 13593461123 SHA256 744db4709fab86a3b5f1c64f2c5f6fd2990dd7c0f38b8bac361d6d6166248c8d. All 80 original preparation, generation and committed-tree gate exits are zero with authentic stdout/stderr hashes. Staged tree equals native tested candidate fe5f03d. This is publication preparation, not final required acceptance or a public release; preserve the independent immutable hardening run.
Reconcile the canonical release boundary with the genuine fifteen publication-copy refreshes, actual Pitch analysis/refresh and nine unchanged rebinds. Preserve C/D/Q/E input reconciliation and explicitly authorize automatic GitHub-release/Zenodo archival without consuming another operator approval. Existing native doc auditor passes all three operator documents; native post-Q input comparison has zero drift across all25 and unchanged Active support. No software, policy, hook or measured value changed.
…egrity-20261108

Fix native marker selection and split-page navigation
…f45f9

fix(rendering): global references and complete mobile math/EPUB access
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant