Repository navigation
0.4.1: v0.4.0 with the semantic-module nesting fix (#528) - #534
humuhumu33 wants to merge 6 commits into
Conversation
The fix of #529, applied to tag v0.4.0 for consumers that pin it and must keep their artifacts byte-identical. A list nests one JSON level per element, and v0.4.0 read a module's JSON with a fixed 128-level limit, so a list of more than about 122 elements was refused and past about 115 the snapshot panicked writing its own bytes. Both readers now take the project's max_scope_depth on the compile thread's stack, and every later reading of what the linker admitted reads to the same depth on a stack sized for it. The readable type and term forms of the notation gain the depth check the generic form had. No identity input changes: every committed lock passes lock --check and check-golden passes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The SDK read a module's notation to 256 levels and its JSON with serde's 128-level limit on its caller's stack, so it could not read the 1000-cell fixture, and AU-08, which reads every committed module through it, failed in the gate. Consumers read and write application sources through it too. It now reads, converts, and writes a semantic module to 1024 levels, the max_scope_depth a project is created with, on a stack of its own sized like a compile thread's for that depth. Tooling and core modules keep their readers. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The workspace version moves to 0.4.1, and with it the version field of every committed build manifest and the one fixture hash that names a manifest. No compiler-semantics ID, source, semantic, or build ID moves: every committed lock passes lock --check and check-golden passes. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
afflom
left a comment
There was a problem hiding this comment.
Bounded release-contract review only; runtime/snapshot implementation and stopped#431/oldTC03 validation are outside this review.
The patch basee0b74b81 is exactly v0.4.0. Complete53-path inventory contains no changed language/schema/embedded-golden or committed example-lock path. All25 changed build manifests differ only in compiler.version0.4.0→0.4.1, with corresponding CL05 manifest hash. This is concrete byte-inventory credit, not universal identity or consumer acceptance.
Before release, reconcile the proposed patch/v0.4 tag placement with unchanged SPEC§9.4/AGENTS, which require each vX.Y.Z tag from release/vX.Y and the associated forward history. The body reports an approved patch plan, but the normative exception/placement decision is not encoded. Preserve the existing release branches and required actual forward history; this review does not authorize retargeting, merging or tagging.
Also identify existing custody evidence for the25manifest/CL05 version transformation: exact source, official writer command/run and outputs, or the applicable authorized version-only rewrite policy. SPEC§9.2 requires the official expected-output writer; check-golden success alone does not establish writer provenance. Reuse existing evidence rather than repeating a writer to answer this review.
Ready PR CI is eligible despite patch/v0.4 being excluded from push CI. The existing release gate can reuse a successful PR-head run for an actual merge's second parent when their trees match and checkout/run binding is established. Synthetic checkout is not itself disqualifying. Current CI43581315031/honesty43581315136/repro43581315079 are queued; complete final-head acceptance and actual merged-tree equality remain owed. Retained#529 source findings are not waived by backporting. Receipt pr534-release-contract-review/receipt.json; no execution, duplicate checks or release admission.
The verified directory's build-manifest.json carries the compiler version, so its hash in the VR-11 fixture moves with the bump; the new hash is the gate's observed value, and conformance_vr_11 passes with it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The advisory was published while the 0.4.1 gate ran; cargo deny's advisories check refused rustix 1.1.4, reached through fs4 and tempfile. 1.1.6 is the fixed release; no LexLean code calls the xattr functions. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
afflom
left a comment
There was a problem hiding this comment.
Bounded release-scope delta: f4efe2a changes only Cargo.lock, moving rustix1.1.4→1.1.6 and four Windows dependency edges. The commit reports an advisory repair; that diagnosis is owner-reported in this source review. No manifests, fixture hashes, workflows or release-policy text change, so the prior25manifest byte-comparison evidence remains reusable.
Please include this dependency update in the patch release description and evidence scope: the current body still says it carries only the nesting fix and version changes. Preserved manifest bytes and unchanged compiler-semantics inputs do not independently establish equivalence of the updated dependency's behavior.
Existing SPEC§9.4 tag-placement and expected-output writer-custody findings remain unresolved. Prior55c013 CI43587293663 failed its rest leg and has canceled jobs; do not aggregate it into current-head acceptance. Current43599855050 CI/honesty43599855067 remain queued and exact repro43599855058 is in progress. Retain applicable evidence and complete the existing final-head gate/actual-tree acceptance; no new writer or repeated checks requested. Receipt pr534-release-contract-review/delta-f4efe2a1.json. No snapshot/runtime/#431 validation or release admission was performed.
afflom
left a comment
There was a problem hiding this comment.
Release metadata changed materially: annotated v0.4.1 peels to this still-unmerged, single-parent PR head f4efe2a; patch/v0.4 remains e0b74b8. The body promises tagging the merged patch commit, so the current tag does not match that stated history. Release run43626401790 is queued; no v0.4.1 GitHub Release was observed in this refresh.
Current CI43599855050 completed all34 jobs successfully, with honesty and exact reproducibility also successful. Retain this evidence; no completed gate repeat is requested. It does not resolve the existing SPEC9.4 release-placement decision, official writer/version-rewrite custody, omitted Rustix dependency scope, or tested/tagged-tree acceptance obligations.
Existing release owner should disposition those points before asset publication and align the release description with the actual target/history. This is a metadata-only observation, not release approval or permission to delete/move tags or rewrite branches. No runtime/nesting/#431/oldTC03 review or validation was performed. Receipt pr534-release-contract-review/tag-observation-20261114T1830.json.
A patch release on
v0.4.0carrying only the nesting fix of #529, for consumers that pinv0.4.0(PrismPM) and need lists past 128 levels without moving any compiler-semantics ID.patch/v0.4isv0.4.0itself; this PR is how the gate runs on the patch before thev0.4.1tag, which the owner approved.What it carries
max_scope_depthin its notation and data, and everything downstream of the linker (rendering, definition digests, the snapshot and its canonical bytes) reads to the same depth on a sized stack. The authoring API reads, converts, and writes a semantic module to 1024 levels on a stack of its own.v0.4.0accepted.Cargo.toml,Cargo.lock, SPEC §2.3,RP-01's andRP-12's strings, the version field of every committed build manifest, and the one fixture hash (CL-05) that names a manifest.Against v0.4.0
lexlean lock --check;check-goldenpasses. No compiler-semantics, source, semantic, build, or snapshot ID moves. Build manifests differ only in"version".conformance_sm_12passes and fails on five planted defects;conformance_au_08passes (862 s);RP-01,CL-05pass.lexlean verifywith pinned Lean 4.32.1: 256 cells (26 s), 1000 cells (22 s).Merging lands the patch on
patch/v0.4; thev0.4.1tag goes on the merged commit once this gate passes.🤖 Generated with Claude Code