[0.5.0] Release 0.5.0

Plan item L6.1 of #106.

Item. The 0.5.0 release per the amended §2.3 and §30 (#113), after every item of #106 and #15's definition of done.

Accepted when. release-check passes with no named unmet criterion; the §30.3 artifacts are published from the tagged commit.

Required v0.5 tooling integration. Complete #468: use PrismPM's Prism model and factory for LexLean development, CI and release tooling, with qualified input-bound evidence reuse, measured removal of duplicate work, and retirement of replaced orchestration. An installed dependency, descriptive model or standalone regeneration pilot is insufficient. Existing proof/CI guarantees and all release criteria above remain mandatory.

Generated by Claude Code

Release lines, by the owner's direction of 2026-10-20: a release X.Y lives on the branch release/vX.Y, cut from main when its work begins; the milestone's pull requests target it; tags are made from it; it is merged forward into the next line after each landing; main is the newest released line. release/v0.4 and release/v0.5 are cut at main 17f7e46 now, so v0.5 work starts without waiting for v0.4 to close. PR #226 states the rule in SPEC.md §9.4, runs the three workflows on a push to a release line under the same plan, and extends RP-14 to read the push triggers. The open v0.4 pull requests are retargeted to release/v0.4 when #226 lands; v0.5 pull requests target release/v0.5. Identity-moving landings stay one at a time per line, one regeneration and one full gate each.

Generated by Claude Code

Status of #128, 2026-10-22: `release/v0.5` @ `2959a4db2`

What stands between the line and a v0.5.0 tag, criterion by criterion. Facts only; scope and order are the owner's.

How it was read. A fresh clone of release/v0.5 at 2959a4db21ff0c84a052a7f0daddda352aacb348, WSL Ubuntu x86-64 (not the normative host). cargo xtask release-check was run twice: first on the bare checkout, then after cargo xtask release-artifacts staged what one machine can stage (10 files under release/). The rest is read from CI. The push run on 2959a4db2 (39542757483) ran only plan, under the §9.4 rule. 2959a4db2 has tree d2cbd13d5, the same tree as #246's head e49dcbc05, whose PR run 39536444144 passed all 28 jobs: the rest, examples and seven bdd-* legs, the five hosts, and the Atlas shards and join.

release-check after release-artifacts (exit 1):

gate failed: RP-12: the release is refused; unmet criteria:
  source-tag: the workspace version is not 1.0.0; §2.3 fixes the first complete release at 1.0.0
  source-tag: CHANGELOG.md has no `## 1.0.0` entry
  host-binaries: release/bin/x86_64-unknown-linux-gnu/lexlean is missing or empty
  host-binaries: release/bin/aarch64-unknown-linux-gnu/lexlean is missing or empty
  host-binaries: release/bin/x86_64-apple-darwin/lexlean is missing or empty
  host-binaries: release/bin/aarch64-apple-darwin/lexlean is missing or empty
  host-binaries: release/bin/x86_64-pc-windows-msvc/lexlean.exe is missing or empty
  version-output: release/version-output.txt must be exactly:
lexlean 1.0.0
language 1.0
compiler-semantics 8554008034f6e9284382619a7aa71e0f0071ad404ea9d821068294d2d8804851
lean-toolchain leanprover/lean4:v4.32.1
found:
lexlean 0.3.0
language 1.0
compiler-semantics 8554008034f6e9284382619a7aa71e0f0071ad404ea9d821068294d2d8804851
lean-toolchain leanprover/lean4:v4.32.1

  sbom: release/sbom.json: No such file or directory (os error 2)
  ci-evidence: release/vv-evidence.txt: No such file or directory (os error 2)

On the bare checkout it also named checksums, crate-package and semantics-id (files absent); release-artifacts produced all three, and the packaged crate built standalone with the same four-line identity. It did not refuse conformance-doc, errors-doc, spec, licenses, no-ignored-test, gate-evidence or schemas-exercised.

§30.3 artifacts and §30.4 criterion

Criterion	Status	Evidence	Owner item
§30.3 source tag	Unmet	`source-tag` refused (above); `Cargo.toml` `version = "0.3.0"`; no `## 0.5.0` or `## 1.0.0` in `CHANGELOG.md`; only tag is `v0.1.0`. #288 makes the accepted version one §2.3 names (`0.5.0` or `1.0.0`); no open PR bumps the version	#128 (L6.1); text in #113 / #288
§30.3 checksums	Met as derived	`release-artifacts` writes `release/checksums.txt`, and `release-check` accepts it over the 10 staged files; the published one is written in `release.yml` `publish`	—
§30.3 supported-host binaries	Unmet	`host-binaries` refused for all five targets; `release.yml` `binaries` stages them only in a `release.yml` run; five-host `build and non-Lean tests` green in 39536444144	#128 (L6.1)
§30.3 crate package	Met	`release/lexlean.crate` produced; `release-check: the packaged crate builds standalone with the same identity`	—
§30.3 compiler-semantics ID	Met for language 1.0; per-language form pending	one 64-hex line `8554…4851` accepted; the amended §30.3 (one ID per supported language) is in #288	#113 / #288
§30.3 `CONFORMANCE.md`	Met	`conformance-doc` not refused	—
§30.3 `ERRORS.md`	Met	`errors-doc` not refused	—
§30.3 `SPEC.md`	Met	`spec` not refused	—
§30.3 license files	Met	`licenses` not refused	—
§30.3 SBOM	Unmet	`sbom` refused; produced only by `cargo cyclonedx` in `release.yml` `publish`	#128 (L6.1)
§30.3 CI evidence of `just vv` on the tagged commit	Unmet	`ci-evidence` refused; produced only by `release.yml` `gate`; no tagged commit exists	#128 (L6.1)
§30.3 version report	Unmet	`version-output` refused (above). #288 replaces the four-line report with eight lines (languages 1.0, 1.1, 1.2, each with its ID); the version line still needs the bump	#113 / #288, then #128
§30.4 every §31 ID at level `build`	Met	`model/ids.toml` at `2959a4db2`: 443 rows, 443 `level = "build"`; `validate-model` green in 39536444144. Open PRs add rows (#287 AU-14, #288 RP-17, #289 PC-01/PC-02), and items #115, #116, #119–#123, #125 and #127 enter as new rows (R3)	—
§30.4 no stub or ignored test	Met	`no-ignored-test` not refused; `audit-deferral` (in `validate-model`) green in 39536444144	—
§30.4 schemas committed and exercised	Met	`schemas-exercised` not refused (32 files under `schemas/`)	—
§30.4 core, Nat, Int, UOR Atlas lexicons validate	Met in CI	every `just vv` leg and `atlas (join)` green in 39536444144	—
§30.4 literal example verifies	Met in CI	`examples` leg green in 39536444144	—
§30.4 negative fixtures fail for the prescribed reason	Met in CI	every `just vv` leg green in 39536444144	—
§30.4 every gate has falsifiability evidence	Met	`gate-evidence` not refused (33 gates, `### <gate> can fail` each)	—
§30.4 `just vv` from a clean checkout	Met in CI for this tree	39536444144 on `e49dcbc05` (tree equal to `2959a4db2`); not run whole here	—
§30.4 default branch has no `open` row	Met	`main` @ `6bcc8d598`: 430 rows, all `build`; `release/v0.5`: 443, all `build`	—
§30.4 README claims tied to model IDs	Met in CI	`RP-11` is part of `just vv`, green in 39536444144	—
§30.4 no normative behavior dependent on unspecified implementation choice	Not evaluated	no `release-check` criterion and no `vv` recipe reads it	—
#128: `release-check` passes with no named unmet criterion	Unmet	5 criteria named after `release-artifacts` (above)	#128
#128: §30.3 artifacts published from the tagged commit	Unmet	no `v0.5.0` tag; `release.yml` last succeeded on `v0.1.0` (below)	#128

#128's precondition: every item of #106 and #15's definition of done

Item	Status	Evidence	Owner item
L1.3 #111	Met on the line, issue open	stages #231, #232, #233 merged; #233 at `e5281f062`	#111
L1.1 #112	Pending #286	ready, base `release/v0.5`	#112
L1.2 #113	Pending #288	ready, base `release/v0.5`; CI full gate in progress at `edb686bb4`	#113
L3.2 #115	Pending afflom's drafts	#228, #230, #238, #239 (PRs 1–4 of 5)	#115
L3.3 #116	In progress, no PR yet		#116
L4.6 #124	Pending #289	ready, base `release/v0.5`	#124
#241, #242	Pending #287, #285	both ready, base `release/v0.5`	#241, #242
L4.1 #119, L4.2 #120, L4.3 #121, L4.4 #122, L4.5 #123, L4.7 #125	Unmet, no PR	behind decisions #107 (D2), #108 (D3), #109 (D4), all open	#119–#123, #125
L5.2 #127	Unmet, no PR	behind decision #110 (D7), open	#127
#179, #181	Unmet, no PR	open in the v0.5 milestone	#179, #181
#15 DoD; L2.1 #27, L2.2 #30, L2.5 #34	Unmet	open on `release/v0.4`; state in the #34 table	#15, #27, #30, #34

Release mechanics

Version strings. Cargo.toml line 6 version = "0.3.0" (and the path pins on lines 20–22), the SPEC §2.3 metadata fence version = "0.3.0", and CHANGELOG.md's newest release heading ## 0.3.0. release.yml identity compares the tag only with the first version = line of Cargo.toml and fails on a mismatch. release-check at the head wants 1.0.0 (RELEASE_VERSION); with [0.5.0] #113: release identity, §2.3 completion release 0.5.0 and §30 for languages 1.0-1.2 (RP-17, CL-18) #288 it wants a version §2.3 names and a ## <version> heading.
[0.5.0] #113: release identity, §2.3 completion release 0.5.0 and §30 for languages 1.0-1.2 (RP-17, CL-18) #288 changes what release-check reads. It drops RELEASE_VERSION and reads §2.3's sentence beginning "The completion release" and the §30.3 text fence beginning lexlean <version>. semantics-id becomes one <language> <ID> line per language, and release.yml's allowance before the completion release becomes source-tag only.
[0.5.0] #113: release identity, §2.3 completion release 0.5.0 and §30 for languages 1.0-1.2 (RP-17, CL-18) #288 and release.yml binaries. At [0.5.0] #113: release identity, §2.3 completion release 0.5.0 and §30 for languages 1.0-1.2 (RP-17, CL-18) #288's head edb686bb4, .github/workflows/release.yml lines 177–179 still fail any --version output that is not exactly 4 lines (--version printed $lines lines, not the four of §30.3). Under amended CL-18 the output is 8 lines, so each binary (<target>) job would fail, and publish, which needs binaries, would not run.
§30.4 and the [0.5.0] Completion plan: optimal compilation for every production root #106 objective. Neither the head nor [0.5.0] #113: release identity, §2.3 completion release 0.5.0 and §30 for languages 1.0-1.2 (RP-17, CL-18) #288 adds certified optimality of production roots to §30.4 ([0.5.0] #113: release identity, §2.3 completion release 0.5.0 and §30 for languages 1.0-1.2 (RP-17, CL-18) #288 changes only its lead line). None of ci: a toolchain fetch is retried three times and no more (#242) #285, [0.5.0] #112: SPEC §3 and §4 state production compilation and its optimality guarantee #286, #241 AU-14: locate finds the environment the structural parser opens #287, #124: the general performance corpus (PC-01, PC-02, L4.6 of #106) #289, #115 PR 1/5: R3 skeleton (TC-08, RB-18, SP-24, MD-14) #228, #115 PR 2/5: TC-08, a calculus literal is charged its weight once per read #230, #115 PR 3/5: RB-18, a heap literal read from a function of no parameters is a once-per-thread constant #238 or #115 PR 4/5: SP-24, MD-14, an artifact's decoded value is one literal of the lowered program #239 touches Chapter30, Chapter02, release.yml, release.rs or Cargo.toml. release-check names no criterion for the L4 items.
release.yml history (corrects the [v0.4] End-to-end release closure, examples, documentation, packaging, and clean-checkout reproducibility #34 table, which said it never ran on a tag). 17 runs, all push events. Two were on tag v0.1.0: run 32081554137 at 0d66c6cfd (2026-08-17) passed 10 jobs and failed publish the release; run 32085962099 at 16046ebdd (2026-08-18) succeeded. The other 15 failed with zero jobs, the last on 2026-09-22. None was a workflow_dispatch. Since v0.1.0, release.yml and crates/model/src/release.rs changed by 83 insertions and 8 deletions (git diff --stat v0.1.0 2959a4db2).
Rehearsal already in the file (refines [v0.4] End-to-end release closure, examples, documentation, packaging, and clean-checkout reproducibility #34 proposal item 2). workflow_dispatch is a trigger. A manual run sets tagged=false, pushes images tagged sha-<7> to ghcr.io, and skips publish (if: needs.identity.outputs.tagged == 'true'). So a rehearsal never runs cargo cyclonedx, release-artifacts, the release-check allowance, or the crates.io step.
§9.4 release lines. Each vX.Y.Z tag is made from release/vX.Y; a line is merged into main when its release is tagged, and merged forward into the next line after each landing. main @ 6bcc8d598 is an ancestor of release/v0.5 (89 commits ahead, 0 behind). release/v0.4 @ f057b8308 is not an ancestor: 140 of its commits are not on release/v0.5. The last forward merge, 79d11693e, took release/v0.4 at 998927881.
Inherited unchanged from the [v0.4] End-to-end release closure, examples, documentation, packaging, and clean-checkout reproducibility #34 proposal (comment):
- The version bump precedes the tag, because identity refuses a mismatch.
- publish still runs cargo publish before gh release create.
- crates.io answers 404 for lexlean today (GET /api/v1/crates/lexlean).
- The repository's environment list has no crates.io environment (only copilot), which publish names.
- Whether CARGO_REGISTRY_TOKEN is set could not be read with this account.

Generated by Claude Code

release.yml: check the host binary's report against the §30.3 fence, …

Maintenance checkpoint, 2026-11-02 (America/Denver): the October 22 status above predates the completed v0.4.0 publication recorded on #15. It is not the current release blocker list.

Current next landings are:

Integrate Host request types and deep package export/replay #389: reviewed Host effects/types slice; prior full gate green, now needs Host+Order regeneration against release/v0.4 after #276 member 2 (SL-10): sorting and order in the standard library (LexLeanStd.Order) — draft #328. Then Integrate certificate fixes, GNAF request validation and nested induction #393 can finish its certificate-fix identity wave (Certificate A is rejected by the kernel for a recursive function with a list parameter before its recursive argument #314/Certificate B is rejected by the kernel for a match arm that returns a binder of the matched recursive type #310).
BS-02: the lexer's scans run only in the branch that reads them (#373) #394: lexer improvement retains its intended semantics, but BS-02, bootstrap regeneration and measured performance rows are unfinished. The Intel macOS host leg has crept from ~60 to 100+ minutes; bring its cost down #235 separately addresses measured redundant host-gate work.
Add pinned Lean-library imports, glossary generation, wrappers and axiom auditing #362: moved to draft and release/v0.5 per the existing review; schema parity, numeral-capture coverage and outstanding CI fixes must be completed. A disconnected replay-cache library does not close Mathlib import: content-addressed replay cache (DigestSet Merkle DAG) #338's warm-verification acceptance goal.
Port Host to v0.5 and make closed byte and string proofs practical #396: v0.5 Host interface/kernel evaluation is still draft. Its final-head draft run is not a full release gate. Coordinate its identical Host types with the eventual Integrate Host request types and deep package export/replay #389 forward merge.

I have also claimed implementation of #374's parser crash and #235's CI cost correction through issue-closing PRs, and #386's Lake-build documentation is being reproduced. #391's single-kernel-check principle is accepted for project/certificate modules, with runtime modules retaining their in-process check initially.

The larger #92 bootstrap fixed point, #106 optimality work and #343 proof-block goals remain implementation work. They are not satisfied by component ports or infrastructure-only PRs. We will preserve those acceptance criteria and close each item through its completed PR.

Release checkpoint, 2026-11-03 16:48 UTC:

release/v0.5 is now a8ad0588155881d6a5437f0bb18c5e05438d1bea after kappa: the four other hash axes and the compositions on all five axes (#127, KA-05, KA-06) #407: KA-05/06 passed all 34 CI jobs plus honesty/reproducibility, and the actual merge tree equals the tested tree. Scoped κ-addressing slice: five hash axes and compositions on every axis (KA-05/KA-06) #410 is closed; [0.5.0] κ-addressing in LexLean, validated against uor-addr #127 remains open for GGUF/ONNX.
release/v0.4 is 0a01df682 after BS-20: the LaTeX backend of the tooling modules as language-1.2 source (first slice of #270) #364/[v0.4] BS-20: tooling LaTeX backend as LexLean source #399. Combined maintenance Fix compiler diagnostics and release-evidence validation #403 is reviewed at 215a9d591; 41/43 final-head jobs pass, with Intel and bootstrap-package still running. Its five issues remain open until a passing landing. Gate expensive CI fan-out on existing honesty checks #413's same-run honesty optimization follows that landing and still requires its own complete final-head gate.
Host+Order and reviewed Fix deep package manifest export and program replay readers (PX-07) #402 are in one maintainer-owned regeneration wave under Integrate Host request types and deep package export/replay #389. Official exports and build goldens passed; fixtures and normalized verification remain in progress. Integrate Host request types and deep package export/replay #389 is draft until the integrated source/artifacts, corpus measurements and final gate are ready. Integrate certificate fixes, GNAF request validation and nested induction #393 follows it; no competing identity wave is running.
Forward-merge release/v0.4 backlog into release/v0.5 (integration blocked) #405 remains the required forward-integration vehicle. Source preparation now includes KA-05/06 and the v0.4 BS-20 additions with a static 493-claim union; generated conflicts still require official writers. Add pinned Lean-library imports, glossary generation, wrappers and axiom auditing #362's reviewed import/schema follow-ups and accepted Move the Lean pin from v4.32.1 to v4.33.1 #420 Lean 4.33.1 upgrade remain coordinated v0.5 work, not completed compatibility or release evidence.

The next release is not ready to tag. #106 optimality, #92 self-hosting and #343 proof-block acceptance remain open for their actual product deliveries. The measured #400 CI reduction and #391's passing real false-declaration regression do not substitute for those goals; #391 still owes certificate/shard validation, measurements and its coordinated identity/full-gate work.

Release checkpoint, 2026-11-03 19:45 UTC: #403 has landed on v0.4 as 5a44fe49e after all 43 CI jobs plus honesty/repro passed, with actual/tested/reviewed tree equality. #374/#375/#377/#235/#408 are explicitly closed. v0.5 remains a8ad0588.

#413 is now ready on the landed v0.4 base. Its same-run honesty job passed before expensive fan-out, and final-head CI is progressing without failures; complete acceptance is still required before #411 can close. No additional local gate is competing with the Host writer.

Host/#402 fixture regeneration passed with all negative diagnostic codes preserved. The durable verification writer has completed 20 of its 21 selected projects and is still verifying bootstrap. Final corpus/artifact review and CI remain. #405's source preparation includes the #403 landing but remains draft pending real generated integration.

#396 has advanced to 84e38cb0 with SL-12 and refreshed outputs; independent delta review and its new full gate are underway. Its closure remains #409 only. The #362/#433 and forward-source integration requirements, #393's following identity wave, and all broad product/release trackers remain open.

Maintenance landing: #413 merged into release/v0.4 as5393f5528ec46a3fb0f2884a81c0c7087f1de95b after all44 final-head CI jobs and reproducibility passed. Actual/tested/synthetic tree5868cd66 matches; #411 is explicitly closed. The existing honesty checks now precede expensive fan-out, with all812 tests retained and RP13/14/18 passing. No aggregate runtime saving is asserted.

v0.5 remains a8ad058; #405/#404 owns the required forward integration, and its source preparation is being refreshed for this landing before any generated wave. Host/#402 official writers are complete; focused PD10/BS05/BS06/BS13 have passed, BS14 and the remaining checks are running before final publication/CI. #396 remains draft because its full gate confirmed the known fixture/README failures; its reviewed source correction and official fixture writer remain outstanding. #434 is the scoped native formatting regression preparation under #381, not completed implementation. Broad release/self-hosting/Mathlib trackers remain open.

Release/v0.5 advanced to2b0395ef1cf020509826e3a2cd8a0cb9add6ddd2 through reviewed PR #396. All34 final-head CI jobs plus honesty/reproducibility passed, and the actual merge tree exactly equals the reviewed/tested tree7f85850a6f94f540183e26b9540741a790a46920. Scoped #409 is explicitly closed with evidence in #409 (comment).

This delivers the closed kernel decisions, bounded byte slices and reducible joins slice. #376/#301 retain their remaining primitive/law work; #382/#271 retain Host publication and driver scope. The active #389/#402 identity wave is still regenerating its final artifacts on v0.4. Existing draft #405 source preparation is being refreshed against the newly landed v0.5 runtime/Host changes; no competing regeneration or duplicate integration PR was started. The release tracker remains open.

Delivery queue update: #389 has completed its coordinated Host/Order/#402 identity regeneration, final source/artifact review and publication at88b078cc; fullCI41781980948 is running, with same-run honesty and reproducibility passed. #284's reviewed BS15 disk correction is published at708f86dea; its fullCI41780038830 is also running with honesty/repro passed. Neither is merged or issue-complete yet, and current release refs remain5393f552(v0.4)/2b0395ef(v0.5).

Existing #390 was explicitly restacked as draft7ca5a7ff on Host88b078 plus reviewed30171790. Its new head is outside Host ancestry, preserving incomplete BS16 delivery rather than allowing GitHub to mark it merged through the older contributor history. Its official generation/count reconciliation/full acceptance remains in the next coordinated #393 wave. #405 must retain the landed #396 runtime/Host source and incorporate actual later release landings before the v0.5 wave; no competing writer is active.

While these gates run, #434 gained an actual native UInt8 formatting-policy LLV7005 witness using the admitted #396 CLI. That is baseline evidence only; implementation and complete acceptance remain open. Broad release, self-hosting, optimality and Mathlib trackers remain open.

Maintenance delivery update: #284 has landed as1a690b51 on release/v0.4 after all44 final CI jobs, same-run honesty and reproducibility passed; actual/tested/reviewed trees matched. DF-19 (#428) and the generated-workspace BS-15 ENOSPC fix (#436) are explicitly closed. Parent #278 remains open for #292/#429 nested induction.

Host #389's final-pin PX-07 now passes with real Lean, but its full gate remains incomplete and the newly advanced release base must be integrated. It is draft during that integration; existing CI will finish untouched, and the combined final tree still needs complete acceptance. #390's separate BS-16 redelivery stays draft/open.

New draft #437 contains the independently reviewed four-part formatting-isolation source for scoped #434. The named pre-fix LLV7005 was reproduced before implementation; post-fix identity/artifact generation and complete positive/negative acceptance remain in the coordinated v0.5 wave with #405/#362/#433. #434 and parent #381 remain open. No broad roadmap completion is claimed.

Current delivery queue: Host #389 is ready at943bba379a after integrating the actual #284 landing, with its complete new-tree gate41802893306 running and reproducibility passed. Its prior88b gate completed44/44, but does not accept the integrated tree. #390 remains separate draft7ca; live GitHub ancestry comparison confirms it is outside Host ancestry. No new release landing or broad completion is claimed.

Useful work during the gate advanced two independently reviewed source preparations: #405 now preserves current v0.5/#396 plus Host and landed DF-19 in a124-path,496-claim source union; #292's accepted SM-37 was reconciled with current Host/DF-19 source and retargeted to release/v0.4, while remaining draft with generated products and final acceptance owed. SM-37 joins the next #393/#390 source-union review before a single coordinated regeneration.

For #391, the existing width and whole/shard/join checks now pass with explicit artifact-binding assertions and five real refusal cases. Follow-up104f755ae9 is published with independent evidence review. Atlas/gateway measurements, coordinated identities/artifacts and complete CI still remain; no speedup or overall optimization completion is asserted.

Maintenance landing: Host/deep-package delivery #389 merged into release/v0.4 as c613290 after all44 final CI jobs plus honesty/repro passed; actual/tested/reviewed tree equality verified. #382/#378 are explicitly closed. #402 closed by ancestry through the integrated delivery; #390 remains an open draft, so incomplete BS16 was not accidentally marked delivered. #271 and other broad trackers stay open.

Required #405 forward integration now records actual c613/1a and landed #396/2b ancestry. Its initial Mathlib + formatting + active Lean-pin source composition is independently reviewed (511 claims); Atlas repair, actual dependency/probe compatibility, final identities/products, controlled upgrade evidence and full CI still block acceptance. Unused replay-cache infrastructure remains explicitly deferred with its complete future obligations; streaming SHA is not an initial-release prerequisite.

The next v0.4 source union for #393/#390/#292 is independently reviewed at475 claims, with all strict cases retained. No new identity wave has started. Separately, #391's matched baseline CLI and immutable public gateway corpus are prepared; its single controlled verification pair is now running. No speedup or overall release-completion claim is made before actual results and remaining acceptance.

Current delivery state: accepted Host/#389 and earlier landings remain complete; release/v0.4 is c613290 and release/v0.5 is 2b0395e. No completed gate was rerun.

The #393/#390/#292 source union has now passed its official canonical-identity, initial model/calculus and lock/input phases, with independent reviews. The single all-five export writer is active; golden products, fixtures, verification, measured combined corpus checks and final-head CI still remain. Scoped #314/#310/#432/#429 and broad trackers remain open.

Separately, #391's new controlled public-corpus comparison completed and was independently reviewed: the same 58 modules, four roots and 2,732 declaration-policy rows verified in 1863.117s at baseline versus 1243.365s for the candidate (33.264% lower wall time in this one pair). Baseline-first order and unchanged global caches limit the conclusion; this is not a general, flag-only or historical WSL claim. Evidence is recorded in #391 and documentation-only follow-up494fc28a. Atlas, coordinated identities/products and full acceptance remain outstanding; #391 stays in v0.4.1, separate from the current v0.4.2 wave.

The v0.5 source preparation remains reviewed but unaccepted at runtime; #433's existing owner still owns the bounded Atlas compatibility repair. No broad release or optimality completion is claimed.

The #393/#390/#292 wave has completed its export, golden/calculus and fixture producers with independent actual artifact reviews. The fixture run preserved all old outcomes and inputs across 309 cases; the new nested-type refusal is LLT4001 with no artifacts. Its one 22-project non-Atlas verification writer is now active, with final records, combined corpus/focused acceptance and final-head CI still pending. Scoped issues remain open.

New draft #439 proposes native Host command execution on v0.5 and overlaps existing #419/#271. It is under bounded source/ownership review; the author has been asked to report any active writer and preserve its outputs while coordinating with the existing #405/#362/#433/#437 source union before another wave. Its reported partial BS-09 agreement and larger resource limits do not establish final acceptance. No draft or broad tracker has been closed, and release refs remain v0.4 c613290 / v0.5 2b0395e.

05:40 UTC live release update: #466 is merged as bc71b33 on release/v0.4. Independent post-landing review5854389516 confirms all54 jobs in exact-head CI42946701043, same-run honesty and reproducibility42946701086 passed before the05:03:25 merge. Actual ordered parents are700196+c89; tree339f2df5 equals the reviewed head and recorded GitHub synthetic. Prior failed/cancelled runs supply no coverage, and no producer, gate or log was repeated.

The compiler repair is delivered, but #465/#454 remain open for their SDK/downstream scope. The owner has already recorded actualbc71 for #405 (6787548956). SDK source admission is actively owned at UOR-Foundation/PrismPM#594 on fix/lexlean-466-sdk-admission, currently524dd61b. This is one preparation transaction, not a qualified immutable SDK or downstream acceptance. Preserve that owner and the existing #194 candidate route.

release/v0.5 remains2b0395ef1cf020509826e3a2cd8a0cb9add6ddd2;26 open PRs and91 open issues. #316's published source/frozen positive and omission evidence remain intact while it waits for actual accepted forward history and a reviewed integration/product-impact decision. #459's priorc66-tree qualification requires owner reconciliation with actualbc71 before further acceptance. Existing count findings and broad parents remain open. Monitoring remains enabled; accepted backlog is incomplete.

06:49 UTC live checkpoint: release/v0.4 remains actualbc71b33f9a4ce7f5582c337b5752a7ba9eb49706 and release/v0.5 remains2b0395ef1cf020509826e3a2cd8a0cb9add6ddd2. All26 open PR heads are unchanged;91 issues remain open. Completed #466 compiler qualification/landing evidence is retained without repeating its gate or producers.

New delivery status: PrismPM#594's sole source-admission preparation run42964759470 on524dd61b completed failure at05:47:58 UTC: https://github.com/UOR-Foundation/PrismPM/actions/runs/42964759470 . This refresh read run metadata only; no failure cause or generated-output availability is inferred. Branch/ownership are unchanged, with no associated PR or new owner handoff. Existing SDK owner retains diagnosis, source admission and later immutable SDK qualification; no second writer or retry was started. #465/#454 remain open for downstream acceptance.

#405 still owns actualbc71 forward integration; #316 continues to preserve its source/frozen positive and omission evidence while waiting for actual accepted forward history. #459 requires current-release reconciliation. A bounded #115/#116 ownership check also confirms the existing #239/#295 drafts and recorded remaining work; no duplicate patch or historic test rerun. Monitoring remains enabled; accepted backlog is incomplete.

2026-11-11 13:34 UTC delivery update: release/v0.4 remains bc71b33; release/v0.5 remains 2b0395e. Live inventory is 27 open PRs / 91 open issues.

New contributor-owned draft ci: regenerate.yml, the write recipes on runners #467 moves official write recipes onto runners. Bounded review at a66bd0c found that sequential patch application does not enforce its promised rejection of shared paths; exact-head COMMENT ci: regenerate.yml, the write recipes on runners #467 (review) records the isolated reproduction. The existing owner retains repair and smoke-run/output qualification; no competing writer was launched.
Draft BS-27: the notation of a tooling module as language-1.2 source (#269) #452 advanced to 6b8100d and now includes actual bc71 release ancestry. Its owner regeneration is active. The three BS-27 packages and print_tooling_module stack row remain absent at this head; resulting products, count reconciliation and complete final acceptance remain owed. Artifact presence alone is insufficient because its temporary workflow can upload partial output after failure. Self-hosting: the front end after the primitive scanner (grammar, surface, lexicon, link, elaboration) as language-1.2 source #269 remains open.
Both draft CI runs currently skip required host/Atlas groups, so their status is not full acceptance. No rerun, cancellation, readiness change or merge was initiated.

#316 remains preserved pending the owner’s accepted #405 forward landing and reviewed composition/product impact. #465/#454 still need qualified SDK/downstream delivery; the existing PrismPM #594 admission run remains failed and owner diagnosis remains separate. Completed producer receipts stay complete, monitoring remains enabled, and the accepted backlog remains unfinished.

Maintenance update: live release refs remain bc71b33 (v0.4) and 2b0395e (v0.5), with 28 open PRs / 92 open issues.

New ready SM-38: declaration IDs and statement IDs #469 (05e70af) has a source-reviewed P2 declaration-ID invalidation gap: generated constructor references omit their owning inductive dependency. Exact-head COMMENT SM-38: declaration IDs and statement IDs #469 (review) records the source trace. Existing owner retains repair; full CI43039894969 is active, with 7/36 currently materialized jobs successful and exact-head repro passed. No final acceptance.
ci: regenerate.yml, the write recipes on runners #467's existing without-Atlas regeneration43027788154 completed all four jobs successfully; owner reports an empty patch. This is scoped producer evidence, not PR acceptance: unchanged draft head retains the shared-path review finding and draft CI43027793300 failed (2 success,1 failure,12 skipped). No logs re-fetched, producer repeated or cause inferred.
BS-27: the notation of a tooling module as language-1.2 source (#269) #452's draft CI43026064948 likewise failed (2 success,1 failure,12 skipped); its owner regeneration43026106897 remains active. No partial output was adopted.
Required factory integration [v0.5] Adopt PrismPM's Prism factory for development, CI and release tooling #468 now has a 62-obligation inventory and reviewed initial source/test design. Unnecessary exact-scope gate reruns were removed from the proposal; portable host-policy coverage still needs resolution before projection or execution. This is preparation, not delivery.

#405 forward integration and #316's resulting composition remain separately owned/pending. PrismPM's published v0.4 factory is the pinned starting point; its broader #586 scheduler remains unmerged, and publication does not close #465/#454. Monitoring remains enabled. Completed gates/producers and all frozen evidence are preserved; no new execution phase or merge was admitted.

Live release refs remain bc71b33 (v0.4) and 2b0395e (v0.5); inventory is 29 open PRs / 92 open issues.

SM-38: declaration IDs and statement IDs #469 advanced to 3f8d53d. The constructor-owner invalidation finding is resolved at source; bounded repair review: SM-38: declaration IDs and statement IDs #469 (review) . Its exact-head reproducibility check passed. Full CI43050132653 remains incomplete (36 currently materialized jobs at this observation); earlier canceled runs are excluded.
New contributor-owned draft conformance: the BS evaluator differential runs on every processor #470 parallelizes bootstrap evaluator work. Review conformance: the BS evaluator differential runs on every processor #470 (review) identifies a local concurrent-caller budget gap: N−1 helpers plus C callers can exceed N processors. Single-case hosted parts are unaffected by that schedule. The owner retains repair, measurements and its existing manual qualification run43046966603; draft PR CI remains separate.
BS-27: the notation of a tooling module as language-1.2 source (#269) #452's prior owner regeneration43026106897 failed. The owner started43050104167 on temporary1fc92054, whose sole delta from6b8100 replaces the temporary workflow with regenerate.yml; it is explicitly not for merge. No logs, partial products or artifacts were adopted.
Required Prism factory work [v0.5] Adopt PrismPM's Prism factory for development, CI and release tooling #468 now has a concrete portable RP-19 source proposal. Review is resolving the executable/build lifecycle before adoption: adding a redundant all-host tool build would undermine the efficiency goal. No projection, named failure, implementation or new build has occurred.

#467's shared-path finding and #405/SDK owner qualifications remain pending. #316's frozen evidence and stale-output designation are preserved. Monitoring stays enabled; no completed producer/gate was repeated, and the accepted backlog remains incomplete.

Maintenance source progress: #471 now has reviewed resolver-produced commit 4b2d0ff on actual bc71. It updates rustix to1.1.6 for RUSTSEC-2026-0486 and includes four independently reviewed, compatible Windows dependency-edge changes. Only Cargo.lock changed; published tree53bdea equals the reviewed candidate. #469/#470 owners received the commit for one coordinated resulting-head qualification. No separate PR/full gate or advisory suppression was started. #471 stays open pending accepted issue-closing landing and merged-tree equality.

#470's draft CI failed; its manual full run retains one failed rest job while other jobs continue. The rustix diagnosis was already supplied in comment6801273300, so no logs were fetched again. #469's full run remains active with exact-head reproducibility passed. #452's owner regeneration continues; no partial output is adopted.

#468's genuine local R3 source sequence and disabled staged driver are prepared. Final portable policy uses the already-linked evaluator, avoiding an extra all-host xtask build. Resource review cannot support build admission: its source+output planning scenario exceeds1GB before scratch/freezes, versus roughly640MB currently free. This is an estimate, not a peak/minimum/hard limit. Formatter applicability and exact driver review also remain prerequisites; no468 checkout, projection, build, named failure or implementation has run.

Release refs remain bc71(v0.4)/2b039(v0.5). #405 remains the forward owner; #316's frozen evidence and seven stale mutable outputs are preserved. Monitoring is enabled, completed producers/gates were not repeated, and the accepted backlog remains incomplete.

Live inventory is 30 open PRs / 93 open issues; release refs remain bc71b33 (v0.4) and 2b0395e (v0.5).

Existing-owner deps: rustix 1.1.6 for RUSTSEC-2026-0486 #472 is now the [maintenance] Update rustix to resolve RUSTSEC-2026-0486 #471 advisory-repair delivery route. Independent review5862857039 confirmed its Cargo.lock is byte-identical to the already reviewed repair, plus a CHANGELOG addition. Its full CI43067237659 is active; same-run honesty and repro passed. Root will not create a duplicate maintenance PR or qualification run. [maintenance] Update rustix to resolve RUSTSEC-2026-0486 #471 stays open until complete acceptance and actual merged-tree equality.
SM-38: declaration IDs and statement IDs #469's new970c66 head changes only the README capability row to include SM-38; bounded review5862846158 retains the prior implementation review. New fullCI43064053425 is active; old3f8d run was canceled and is excluded.
BS-27: the notation of a tooling module as language-1.2 source (#269) #452's scoped owner regeneration43050104167 finished4/4 successful jobs. Output review/adoption remains with the owner; workflow success alone does not qualify its PR. conformance: the BS evaluator differential runs on every processor #470's separate manual43046966603 is terminal53/54 with the previously diagnosed rest failure; its draft gate remains separate. No logs or producers were repeated.
Required Prism factory integration [v0.5] Adopt PrismPM's Prism factory for development, CI and release tooling #468 remains source preparation. Driver review repaired root coverage and artifact-graph checks, then caught a preserved-copy alias flaw before execution; the bounded fix is underway. Capacity and applicable formatter evidence still block admission. The portable linked-evaluator design avoids an additional all-host xtask build; no factory checkout/build/projection/runtime or identity wave has started.

#405 ownership and actual forward-history requirements remain; #316's frozen evidence and stale mutable-output designation are preserved. Monitoring is enabled. Broad unfinished parents and the accepted backlog remain open.

Current release refs remain bc71b33 / 2b0395e; inventory remains30 open PRs /93 open issues.

#469's new970c66 full gate has a failed rest job. One focused log retrieval confirmed deny / rustix1.1.4 / RUSTSEC-2026-0486; diagnosis6804229667 preserves the log hash. Existing-owner #472 already carries the exact reviewed repair and remains under full-gate watch (44 of53 currently materialized jobs successful at18:44, no failures then). Neither PR is accepted yet; no canceled runs or separate manual runs are combined, and no new rerun/cancel was initiated.

Required factory work #468 advanced to an independently reviewed same-root combined native-tool provisioning plan. One ordinary Cargo build supplies formatter and xtask; exact artifact authentication occurs before either is admitted as a producer. The bounded source-only driver revision retains the R3 sequence and all17 expectations while avoiding a separate discovery build or unnecessary CLI rebuild. This is preparation, not implemented factory behavior or runtime evidence.

Capacity remains a concrete blocker:639MB available; one conditionally eligible completed-root conversion could recover at most475MB, still below even the1.170GB source/output planning estimate. With frozen final tools and illustrative scratch, the estimate is2.098GB before unknown costs. Independent review confirmed the arithmetic and excluded other carriers without exact completion evidence. No cleanup is admitted without a viable phase allocation; budgets remain estimates, not hard limits.

#405 ownership/actual forward ancestry and all #316 evidence remain intact. No completed producer was repeated. Monitoring stays enabled, broad parents remain open, and the accepted backlog remains incomplete.

The current #452 head034efd now includes all three previously missing BS27 packages and committed build/verify products. Bounded review5864468843 supersedes that missing-product finding. Count and documentation findings remain: the source-derived BS15 run total is3262 while the mirrored claim still says3190; the new evidence paragraph misnames one package and leaves its ≤300-node accounting unexplained. Its new CI is a limited draft event, not full acceptance. Existing owner retains repairs and qualification; no regeneration was repeated.

#472 remains the single advisory-repair delivery route: current exact-head CI43067237659 has51 successful jobs and two running (package and Intel moves); exact repro passed. Full acceptance, owner landing and actual merged-tree equality remain prerequisites to #471 closure. #469 remains blocked by its already diagnosed advisory failure; no logs or gates were repeated.

For required factory work #468, a bounded independent review identified an unnecessary first-build constraint: predicting Cargo-generated paths and native compiler probe results before their first observation. A source-only amendment is underway to admit reviewed finite source/tool/effect possibilities, then authenticate the exact observed graph, script outputs and both executables before use. The mandatory post-build stop, R3 ordering, all17 expectations and final proof/CI guarantees remain. This does not admit execution: concrete effect policies and fresh custody/resource controls remain unresolved, with635,826,176B free at the latest resource observation and workspace expansion still pending.

Release refs remain bc71b33 /2b0395ef. #405 ownership and actual forward-history requirements, all #316 frozen evidence and stopped scopes remain preserved. Monitoring remains enabled; broad unfinished parents and the accepted backlog stay open.

Delivery blockers and completed reviews:

deps: rustix 1.1.6 for RUSTSEC-2026-0486 #472 attempt1 completed with53/54 jobs successful. The sole failure was artifact publication after the package normative gate and coverage assertion had passed: CreateArtifact exhausted retries on an HTML response parsed as JSON. Diagnosis6806996206 retains the single log hash. The existing owner already queued failed-job attempt2 at the same head; no root rerun/cancel/source patch was made. [maintenance] Update rustix to resolve RUSTSEC-2026-0486 #471 remains open pending complete acceptance and actual merged-tree equality.
fix(canonical-json): compare duplicate keys after escape decoding #473's small decoded-duplicate-key fix passed bounded source review5865272948. The bug also exists on current release/v0.4; new Reject escaped-equivalent duplicate JSON keys before parsing #475 tracks delivery through this same contributor, with release targeting, issue linkage and CHANGELOG as the next owner actions. No extra test matrix or duplicate implementation requested.
New draft Decision regions: the region_check proof form (PF-24, PrismPM#605) #474's decision-region proof feature received bounded triage5865263518. No kernel bypass was demonstrated in the inspected boundary, but the author-admitted stale Atlas records and contradictory blanket regeneration wording block acceptance. This is not a full proof/product approval; the owner retains the missing output and final qualification work.
Forward-merge release/v0.4 backlog into release/v0.5 (integration blocked) #405 received concrete checkpoint request6806938590: disposition of the already-produced outputs, the actual remaining blocker and next owner action, without repeating its completed producer. Glossary document denotations cannot name semantic-module declarations #316 remains preserved pending that accepted forward lane.

The #472 upload failure is a concrete efficiency case for #468: a successful package gate ran181m19s before publication failed. Preserve that distinction when designing factory reuse; an uploaded artifact or accepted complete gate cannot be invented from partial evidence. No measured savings or recoverable payload is claimed. The reviewed factory driver remains disabled pending concrete effects/custody/resources; workspace expansion is still unresolved (634MB free observed).

Current release refs remain bc71b33 /2b0395ef. There are32 open PRs and94 open issues after #475. Monitoring remains enabled; unfinished parents stay open.

Actual release state changed: #472 was externally merged as7eb816284bd8668e6835c3892ac541d279e5b5bf before its retry finished. The landed tree equals the reviewed/tested tree, but complete final acceptance is still unestablished; #471 remains open (audit6807836339). Native attempt2 now exposes the full54-job official inventory with53 retained successes and package qualification active. The earlier transient enumeration concern needs no collector patch or replay of those53 jobs. Root initiated no rerun, cancellation or merge.

Completed owner follow-through:

fix(canonical-json): compare duplicate keys after escape decoding #473 now targets release/v0.4 on actual7eb, links Reject escaped-equivalent duplicate JSON keys before parsing #475 and has its changelog/reproducer corrected. Review5865741365 confirmed both fix/test blobs unchanged and reused the prior review. No CI/check runs were visible at the new8cc04d head, so the next owner action is the single intended exact-head qualification path.
Decision regions: the region_check proof form (PF-24, PrismPM#605) #474's regeneration wording is corrected; review5865729927 confirms its latest9f9d928 release integration changes only the maintenance lock/changelog. Atlas records still remain missing; it stays draft and has no full proof/product acceptance.

Bounded delta review confirms #469/#470/#452 carried only the maintenance base, preserving previous source review and existing findings. #469's new full gate is active; #470's caller-budget issue and #452's count/evidence wording remain with their owners. #467 adds an ordered finish phase, but its previously reported path-collision check is still absent; the earlier scoped run cannot qualify that added phase. No competing patch or producer was started.

Actual v0.4 is7eb816; the latest fully qualified baseline remainsbc71 until the outstanding acceptance is established. v0.5 remains2b039. #405 owns actual forward history; #316's frozen evidence is preserved. #468 execution remains capacity-blocked. Monitoring stays enabled;31 PRs/94 issues are open.

Two concrete delivery assists completed:

fix(canonical-json): compare duplicate keys after escape decoding #473's missing qualification now has a verified upstream ref, ci/473-8cc04d22, pointing to exactly its reviewed current commit. No source, PR or readiness change was made, and no run was started. Comment6809071690 gives the authorized CI operator the existing ci.yml + reproducibility.yml dispatch path. The connector cannot dispatch workflows. Draft→ready would omit independent repro, so it was not used. The manual route is valid full-gate evidence but does not promise to suppress the later merge-push gate; no acceptance condition is relaxed.
Bootstrap count cases: report every moved count in one pass, and write them #442/Fix ambiguous count rewrites (stacked on #442) #462's current combined source is clean. GitHub's synthetic e050a49 has parents[current44218def,462head810cf], and changes only counts.rs to the exact already reviewed repair blob. Comment6808994594 withdraws the transient conflict concern: the owner needs no speculative rebase or second implementation. Existing native evidence retains its original scope; adoption and resulting release qualification remain with the owner.

The latest #474 delta only corrects the six-runtime-module test expectation and clarifies the Atlas identity exception. It supplies no missing Atlas records. #471 still awaits complete qualification despite #472's external landing; #316 still waits for the existing #405 forward lane. No completed producer, test or gate was repeated. Monitoring remains enabled.

Completed: #471 is explicitly closed through merged #472. Exact final-head CI43067237659 attempt2 passed all54 jobs and600 steps, same-run honesty and independent repro passed, and actual7eb816/treeea7955 equals the reviewed/tested tree. Final review5866955298 preserves that acceptance was completed after the external merge. The same-tree contract permits this completed evidence; no duplicate gate replay or cancellation was requested. #405 has the accepted actual-history handoff6810186760 and retains forward ownership.

Removed an unnecessary producer requirement from #474: Atlas is language1.0 and outside the changed1.2 identity partition. Review5866949431 explicitly retracts our earlier stale-output inference. The minimum remaining correction is REGIONS.md's outdated claim, plus the owner's normal affected1.2 output review and complete final qualification. No Atlas regeneration is requested on the withdrawn premise.

Current release/v0.4 is now fully qualified at7eb816; v0.5 remains2b039. #469's existing full run is still active (51/53 materialized jobs passed at refresh). #470/#467/#452/#442 draft runs completed3 successful jobs with12 required groups skipped; these are development evidence, not full acceptance. Existing source findings and owners remain.

#473's exact upstream qualification ref is prepared, but no current-head workflow is registered yet; an authorized CI operator must dispatch the existing full CI and independent repro as recorded in6809071690. #468 factory execution remains capacity/effect-policy blocked, and #316 waits for the existing #405 lane. No completed producers or checks were repeated. Monitoring stays enabled;31 PRs and93 issues remain open.

Delivered: #469 landed after final review5867323276 and full54/54 CI, same-run honesty and repro. Actual5dccf9e5ba20d91d81f4f702f46a79e8b363bfb6 has ordered parents[7eb816,685aa104]; its0c7b1612 tree equals the reviewed/tested tree. SM-38 declaration/statement IDs are now on v0.4. #405 received actual-history handoff6811129084. #354/#343 remain open because kernel block-binding is broader than these source IDs.

Two owner CI-efficiency changes are under review, with no competing implementation:

ci: export the bootstrap project once per run (SPEC.md §9.4) #476 shares one bootstrap export per run. Review5867372298 identifies a narrow environment-lock race in the new consumer; the owner needs that repair and real transfer/final qualification. Historical export cost is not measured savings yet.
ci: run the Windows host's tests as five lanes (SPEC.md §8.3) #477 partitions Windows into five lanes. Bounded review5867407708 found no concrete source regression; its first full run must establish native execution, receipt join and actual timings. No test or gate was repeated by review.

#474's remaining documentation finding is resolved (5867314141); the unnecessary Atlas-regeneration requirement remains withdrawn. Affected1.2 product acceptance and final qualification remain with its owner. All pending branches must assess actual5dcc integration; old-tree runs retain their original scope.

For #473, operator instructions6809071690 are now explicitly on hold pending the contributor's current-release reconciliation. The old8cc ref still has no registered run, and dispatching it now would spend a full gate on the previous composition. Existing implementation review can be reused where unchanged.

#468 gained one disabled, source-reviewed quote build-script effect fragment; tool provenance, remaining policies and capacity remain unresolved. #316 stays preserved behind the existing #405 lane. Monitoring is enabled;32 PRs/93 issues remain open.

Delivery checkpoint (2026-11-13 08:28 UTC): release/v0.4 is actualb8b6e86a94533d87dd89455d5a5bf9400bda96ba; v0.5 remains2b0395ef1cf020509826e3a2cd8a0cb9add6ddd2. Inventory46 open PRs/96 open issues. Existing owners retain implementation and qualification.

#419 landed externally as b8b6 after full CI43329099701 passed70/70 jobs and772/772 steps, same-run honesty and exact repro43329037236. Actual ordered parents26e5+0b77 and treecb61de9e equal the qualified synthetic checkout. Postlanding review5886731674 credits registered driver behavior, provenance/refusals/plants, current products and the complete728-vector/49-rule/8-plant route. The historical BS09 aggregate failure is retired as a current blocker; no historical forensic gate or completed producer replay is owed. No final gate/source acceptance submission before the external merge was found; the new review is explicitly postlanding.

#426 remains OPEN. The full-body audit corrected the earlier conditional mapping: source/shipped/embedded RuntimeHost byte equality and Rust-data decoding/validation, plus real Lean verification of the command-root fixture, do not establish direct compiler/pinned-Lean qualification of RuntimeHost itself. The inspected compiler Main/manifest closure excludes it. Comment6845268645 asks the existing owner first for any retained exact-source-bound RuntimeHost receipt and mapping of the two originally missing twins; if absent, the smallest missing qualification. The PR body now records current adoption43311807929 and this remaining clause honestly. #271/#269 remain open.

The IR stack now integrates actual #419 through #50493d727. Review5886737547 identifies a new concrete R1 regression: current CONFORMANCE.md is inherited release data and omits all nine BS65–73 claims still present in the authoritative register. Restore the official projection in the existing repair batch. Sampled manifest5d59/compiler06639 binds a1598-byte release config, not the current1849-byte IR config; retained term_text still binds64ebb/compiler635a. Exact inventory still has85 roots versus67 packages. These new bytes are not completed IR finish adoption. Finish43361968064 atd084 ended failure, separate from current PR CI/repro failures. Owner owes failed-output disposition, retained source-contract repairs and the smallest necessary product delta; no blanket rerun requested. BS73 reader/witness source credit remains, while BS67 coverage/fallback/memory attribution/budget disposition and BS69 coverage findings persist.

#507d9f532 is a bot product commit atop147; regeneration43352312619 ended success. Sample manifests/locks agree on compiler043cd7e5, but exact workflow still masks writer exits and always publishes. Review5886740216 requests the owner-promised per-writer/verify exit disposition and workflow removal (owner6842324403); workflow success alone is not completion. Current branch conflicts with actualb8b6. Preserve valid outputs and review resulting identity effects before further work; PF26/Laws source fixes and unchanged export reuse remain credited.

#500ef8cca merges exact25a32+b8b6, preserving source history. LexiconSchema authored bytes are unchanged; current release manifest/compiler06639 and retained oldd2c3/compiler635a packages are pending reconciliation. Owner regeneration43373252951 completed successfully; adoption is still pending. Prior25a32 product credit remains valid for its original input scope, not blanket current acceptance. #501b361/#502ee762 likewise merge their exact prior heads with b8b6 and explicitly retain release generated files pending regeneration. Samples show current06639/5d59 manifests alongside retained old specialized maps/packages. Earlier product credit remains scoped to the original inputs; both remain owner drafts.

New #5142cff GrammarMath and #51509e6 PropositionTerms received bounded source review5886740317 with no demonstrated semantic/plant bug. #514 is independent on26e5; #515 stacks on #505. Their selected corpora and attributed runtime evidence remain distinct from final acceptance; BS43 evaluator150 means runs, not150 sentences. #505/#509–#513 changes inspected this turn are child-output32→256MiB/docs/lock/config amendments with unchanged implementation/cases. Prior corpus findings persist; configured output bounds are not measured memory peaks or proof acceptance. Existing canonical batch ownership remains; no canceled producer evidence is aggregated.

Current bounded reviews: #504 comment6845958314 now tracks separate finish preparation0ac1b85/run43392517878 ACTIVE. The sole workflow delta narrows measurement toBS70 and adds core/process-memory diagnostics. The pipeline-status P2 remains: commands after cargo-test can mask its failure. Capture and return the original status; disposition the actual result before consuming rows. Prior da80043377086335 is cancelled and is not aggregated. No source/projection/config/product change, current R1 repair, actual failure or OOM cause is inferred.

#500/#501/#502 completed exact-source regeneration43373252951/43373638120/43374086737: each5/5 jobs and54/54 steps successful, five nonexpired artifacts including a final source-bound patch. Their workflows return artifacts with read-only repository permissions and do not publish commits; all three PR heads remain unchanged. Updated comment6846026516 assigns the existing owner artifact disposition, canonical composition and reviewed adoption/publication or a concrete blocker. No successful writer needs repeating merely because publication is pending. Fast count sampling is not full conformance acceptance; pre-adoption PR CI failures/repro passes remain separate.

The prior bounded duplication finding remains: fold-laws/compiler project trees, complete manifest inputs and compiler/runtime/toolchain bindings match accepted #419. Retain those products and scoped evidence. Distinct bootstrap inputs prohibit interchangeable products; no only:bootstrap selector exists. Final resulting-head qualification remains required, with no #508 cross-run policy adoption. Existing owners retain implementation; no new local producer or gate was started.

#493's prior26e5-scope70/70+repro remains credited; SM40 coordination6843396283, allocation wording and retained initialR3 receipt link remain owner actions, plus actual current-base applicability. #4826a's SM41/VR29 coordination6843417805 and retained13-failure diagnosis remain pending. #499 and other prior source findings remain; successful historical gates do not clear them. #508 verification memo is still an unaccepted policy proposal, with no cross-run proof reuse admitted.

Forward comment6845271604 supplies actualb8b6 ancestry and retained follow-ups to #405's existing owner. The completed6521 output disposition, real blocker and publication step remain outstanding; the unchanged477a placeholder still blocks #316/v0.5 integration. #468 factory guard remains disabled source preparation with actual tools/config/output/libm/native/capacity STOP. All28 #316 fields/frozen905/BOTH/errors/sevenSTALE outputs remain preserved.517,545,984 bytes free observed; budgets are planning assumptions. No local build/probe/test/writer/cleanup/adoption or competing patch ran. R25, #431/oldTC03 STOP and future #338/#438 scope remain unchanged. Hourly monitoring enabled; accepted backlog incomplete.

Navigation Menu

Repository navigation

[0.5.0] Release 0.5.0 #128

Status of #128, 2026-10-22: `release/v0.5` @ `2959a4db2`

§30.3 artifacts and §30.4 criterion

#128's precondition: every item of #106 and #15's definition of done

Release mechanics

41 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

[0.5.0] Release 0.5.0 #128

Description

Activity

afflom commented on Oct 20, 2026

humuhumu33 commented on Oct 22, 2026

Status of #128, 2026-10-22: release/v0.5 @ 2959a4db2

§30.3 artifacts and §30.4 criterion

#128's precondition: every item of #106 and #15's definition of done

Release mechanics

afflom commented on Nov 3, 2026

afflom commented on Nov 3, 2026

afflom commented on Nov 3, 2026

afflom commented on Nov 3, 2026

afflom commented on Nov 4, 2026

afflom commented on Nov 4, 2026

afflom commented on Nov 4, 2026

afflom commented on Nov 4, 2026

afflom commented on Nov 4, 2026

afflom commented on Nov 4, 2026

afflom commented on Nov 4, 2026

41 remaining items

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 11, 2026

afflom commented on Nov 12, 2026

afflom commented on Nov 12, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions

Status of #128, 2026-10-22: `release/v0.5` @ `2959a4db2`