core: single-own compiled observation schemas - #460
Closed
lemone112 wants to merge 2 commits into
Closed
Conversation
|
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: 📒 Files selected for processing (11)
WalkthroughКаноническая схема наблюдений больше не хранится отдельно в сессиях и планах: она извлекается через ChangesВладение схемой и admission
Estimated code review effort: 4 (Complex) | ~45 minutes Sequence Diagram(s)sequenceDiagram
participant Session
participant OwnerLease
participant SessionPlanV1
participant ObservationAdmission
Session->>OwnerLease: try_acquire_owner()
Session->>SessionPlanV1: observation_schema(&owner)
SessionPlanV1-->>Session: canonical schema reference
Session->>ObservationAdmission: prepare_observation(schema)
ObservationAdmission->>ObservationAdmission: share_for_observation()
Possibly related issues
Possibly related PRs
🚥 Pre-merge checks | ✅ 5✅ Passed checks (5 passed)
✨ Finishing Touches📝 Generate docstrings
🧪 Generate unit tests (beta)
Comment |
Collaborator
Author
|
@coderabbitai review Please review this stacked schema-SSOT slice for exact-owner/schema coupling, lifetime and Rc-count regressions, hidden allocations, equal-content authority leaks, proof-rebind honesty, and Python verifier portability. It must not be assessed as independently mergeable from #459. |
|
✅ Action performedReview finished.
|
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@scripts/verify_point_support_surplus.py`:
- Around line 388-392: Добавьте короткий комментарий перед вызовом zip в участке
с assert len(left_monomial) == len(right_monomial), объясняющий, что явная
проверка равенства длин гарантирует строгий zip, а zip(..., strict=True) не
используется для сохранения совместимости с Python 3.9.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro Plus
Run ID: f508d0ca-464e-44f3-86dc-c76187356c6e
📒 Files selected for processing (11)
crates/labcolors-core/contracts/point-support-reference-surplus-q55-bps-proof-v1.jsoncrates/labcolors-core/src/generic_boundary_tests.rscrates/labcolors-core/src/observation.rscrates/labcolors-core/src/observation_tests.rscrates/labcolors-core/src/point_support.rscrates/labcolors-core/src/point_support_tests.rscrates/labcolors-core/src/program_session.rscrates/labcolors-core/src/program_session_tests.rscrates/labcolors-core/src/session.rscrates/labcolors-core/src/session_tests.rsscripts/verify_point_support_surplus.py
lemone112
added a commit
that referenced
this pull request
Jul 25, 2026
…fact The exact-length gate still pinned 376830B from `canonical-authored-program-lowerer` (#456, run 29971399220). This slice grew the runtime artifact to 376985B and did not carry the re-pin that every earlier stacked slice performed (#450, #452, #454, #456), so the gate has rejected this head and every head above it. The +155B is attributable to this slice alone: runs 30115821523 (#457), 30124467410 (#458), 30125634830 (#459), 30129537515 (#460) and 30136346868 (#461) all measure exactly 376985B, so #458-#461 contribute zero bytes to the artifact and were failing only on the inherited pin. The new measurement is the CI run for this exact head (39ee0a7), not a local build: the canonical platform is linux-x64 and a local arm64 build only produces a DIAGNOSTIC result. The budget file's own SHA-256 is re-pinned in the checker so the drift gate keeps rejecting unattributed edits. Co-Authored-By: Claude <noreply@anthropic.com>
lemone112
force-pushed
the
agent/owner-bound-program-projection
branch
from
July 25, 2026 17:01
d00c63e to
6d9044b
Compare
lemone112
force-pushed
the
agent/session-schema-ssot
branch
from
July 25, 2026 17:01
f574140 to
cdba3fb
Compare
lemone112
added a commit
that referenced
this pull request
Jul 26, 2026
This slice moves files inside the point-support semantic cone, so the capsule digest and the committed surplus proof move with it. Both are now regenerated in the same commit that causes the drift, matching the convention the rest of the stack follows; previously the re-bind was batched at #460, which left #457-#459 fail-closed on their own heads and made the stack unmergeable in order. Numerical review: every proof field is unchanged. Only the source-binding identities move -- the file hashes of the cone files this slice edits, the resulting closure digest, the verifier hash and the rolled-up payload hash. The surplus mathematics is byte-identical. Co-Authored-By: Claude <noreply@anthropic.com>
lemone112
added a commit
that referenced
this pull request
Jul 26, 2026
This slice moves files inside the point-support semantic cone, so the capsule digest and the committed surplus proof move with it. Both are now regenerated in the same commit that causes the drift, matching the convention the rest of the stack follows; previously the re-bind was batched at #460, which left #457-#459 fail-closed on their own heads and made the stack unmergeable in order. Numerical review: every proof field is unchanged. Only the source-binding identities move -- the file hashes of the cone files this slice edits, the resulting closure digest, the verifier hash and the rolled-up payload hash. The surplus mathematics is byte-identical. Co-Authored-By: Claude <noreply@anthropic.com>
lemone112
added a commit
that referenced
this pull request
Jul 26, 2026
This slice moves files inside the point-support semantic cone, so the capsule digest and the committed surplus proof move with it. Both are now regenerated in the same commit that causes the drift, matching the convention the rest of the stack follows; previously the re-bind was batched at #460, which left #457-#459 fail-closed on their own heads and made the stack unmergeable in order. Numerical review: every proof field is unchanged. Only the source-binding identities move -- the file hashes of the cone files this slice edits, the resulting closure digest, the verifier hash and the rolled-up payload hash. The surplus mathematics is byte-identical. Co-Authored-By: Claude <noreply@anthropic.com>
lemone112
force-pushed
the
agent/owner-bound-program-projection
branch
from
July 26, 2026 04:18
6d9044b to
8b5fb46
Compare
lemone112
force-pushed
the
agent/session-schema-ssot
branch
from
July 26, 2026 04:18
cdba3fb to
35472cb
Compare
Collaborator
Author
|
@coderabbitai review |
✅ Action performedReview finished.
|
lemone112
added a commit
that referenced
this pull request
Jul 26, 2026
* feat(core): compile generic point-render sessions * fix(core): satisfy format and all-target lint gates * perf(core): recycle point-render session buffers * style(core): format reusable session buffers * feat(lcs): execute sealed sRGB8 tristimulus derivation * style(lcs): format sealed tristimulus slice * feat(lcs): derive versioned appearance views * style(lcs): canonicalize Oklab coefficient * feat(core): expose typed terminal program path * fix(core): satisfy terminal Program quality gates * feat(lcs): bind private output projection registry * style(core): canonicalize release module order * fix(core): scope private registry lint firewall * fix(core): harden terminal Program admission * style(core): match pinned Rust formatter * feat(core): bind F0 release descriptors * perf(core): add linear canonical surface ingestion * refactor(core): unify terminal physical identities * perf(core): prebind terminal evaluation slots * perf(core): make present ingestion lazy and borrowed * refactor(core): remove per-port surface mutation * feat(core): evaluate typed constraints and emit terminal paints * feat(core): bind observation groups to runtime streams * refactor(core): share one encoded point Paint value * refactor(core): share revision-bound observation state - admit correlated scenario sets once behind shared immutable backing - make joint selection domain-safe, linear, and ownership-retryable - refresh source-bound release proof gates * fix(ci): bind V2a budget to measured artifact - keep allocation tests ownership-preserving without a large Result closure - pin the zero-headroom WASM ratchet to the reproducible V2a artifact * refactor(core): remove superseded Program runtime Keep only the private Program compiler/lowering payload for the direct sole-Session bridge. Delete the duplicate owner, lifecycle, output materialization, and test evaluator scaffolding. Pin the hard cut with a negative facade gate and refresh the exact point-support source receipt. * feat(core): execute compiled plans through sole Session Replace the point-support-specific lifecycle with one sealed monomorphized Session, attach reusable CompiledProgram epochs through strong ownership, and retain complete case-by-constraint reports. Bind every decision to the exact admitted observation before atomic commit and refresh release proof pins. * feat(core): bind Program assessments to context-bound LCS * Hard-delete Pair taxonomy from shipping surfaces * Compile typed finite targets into Program sessions * quarantine unsupported Display P3 promises * core: make compiled programs the sole session-generation owner * core: close the evaluator union and package session bridge * core: derive a typed CAM16-UCS occurrence view * core: lower authored physical declarations into the canonical Program * core: bind Program content identity * core: name versioned identity discriminants * fix(ci): re-pin the runtime WASM budget to this slice's measured artifact The exact-length gate still pinned 376830B from `canonical-authored-program-lowerer` (#456, run 29971399220). This slice grew the runtime artifact to 376985B and did not carry the re-pin that every earlier stacked slice performed (#450, #452, #454, #456), so the gate has rejected this head and every head above it. The +155B is attributable to this slice alone: runs 30115821523 (#457), 30124467410 (#458), 30125634830 (#459), 30129537515 (#460) and 30136346868 (#461) all measure exactly 376985B, so #458-#461 contribute zero bytes to the artifact and were failing only on the inherited pin. The new measurement is the CI run for this exact head (39ee0a7), not a local build: the canonical platform is linux-x64 and a local arm64 build only produces a DIAGNOSTIC result. The budget file's own SHA-256 is re-pinned in the checker so the drift gate keeps rejecting unattributed edits. Co-Authored-By: Claude <noreply@anthropic.com> * fix(ci): re-bind the point-support source capsule to this slice's cone This slice moves files inside the point-support semantic cone, so the capsule digest and the committed surplus proof move with it. Both are now regenerated in the same commit that causes the drift, matching the convention the rest of the stack follows; previously the re-bind was batched at #460, which left #457-#459 fail-closed on their own heads and made the stack unmergeable in order. Numerical review: every proof field is unchanged. Only the source-binding identities move -- the file hashes of the cone files this slice edits, the resulting closure digest, the verifier hash and the rolled-up payload hash. The surplus mathematics is byte-identical. Co-Authored-By: Claude <noreply@anthropic.com> * core: project complete program certificate evidence * fix(ci): re-bind the point-support source capsule to this slice's cone This slice moves files inside the point-support semantic cone, so the capsule digest and the committed surplus proof move with it. Both are now regenerated in the same commit that causes the drift, matching the convention the rest of the stack follows; previously the re-bind was batched at #460, which left #457-#459 fail-closed on their own heads and made the stack unmergeable in order. Numerical review: every proof field is unchanged. Only the source-binding identities move -- the file hashes of the cone files this slice edits, the resulting closure digest, the verifier hash and the rolled-up payload hash. The surplus mathematics is byte-identical. Co-Authored-By: Claude <noreply@anthropic.com> * core: bind Program operations to exact owner snapshots * fix(ci): re-bind the point-support source capsule to this slice's cone This slice moves files inside the point-support semantic cone, so the capsule digest and the committed surplus proof move with it. Both are now regenerated in the same commit that causes the drift, matching the convention the rest of the stack follows; previously the re-bind was batched at #460, which left #457-#459 fail-closed on their own heads and made the stack unmergeable in order. Numerical review: every proof field is unchanged. Only the source-binding identities move -- the file hashes of the cone files this slice edits, the resulting closure digest, the verifier hash and the rolled-up payload hash. The surplus mathematics is byte-identical. Co-Authored-By: Claude <noreply@anthropic.com> * core: single-own compiled observation schemas * chore(proof): document Python 3.9 zip invariant * core: define the public Program API * core: address Program API review * fix(ci): re-bind public Program proof capsule * test: close compile-fail sentinel escape * core: make joint order non-empty by construction and bound evidence cells `AdmittedFiniteJointOrderV1` stored one flat tuple slice, so an empty admitted order was representable and the evaluation path carried a runtime `state_count == 0 -> InternalInvariant` guard to reject it. The guard proved nothing about the type; it only re-checked a property the constructor already enforced. Split the order into `first + rest`. Non-emptiness becomes structural, `state_count()` is total, and the `InternalInvariant` branch in `prepare_program_evaluation_buffers` is deleted rather than left dead. The `joint_state_count: Option<usize>` parameter disappears with it: cell counts are now derived from the epoch itself. The same pass stops reserving an exhaustive-conflict buffer that no constraint can ever fill. `can_conflict` is false when every compiled constraint is report-only, so a report-only program no longer reserves `cases x constraints x states` cells and no longer reports `ResourceExhausted` for a conflict it cannot produce. `OwnerV1::evidence_cell_bounds` exposes the same arithmetic as a pure preflight, with `EvidenceBoundsErrorV1::CardinalityOverflow` as the only closed failure. It creates no Session and mutates no state. Verified locally on the CI-pinned toolchain: full workspace tests green, `cargo fmt --all --check` and `cargo clippy --workspace --all-targets -- -D warnings` clean. Co-Authored-By: Claude <noreply@anthropic.com> * docs: explain packed proof invariants * core: fail closed without current Program evidence * core: revoke outputs on unknown session handoff * core: keep incomplete Program crate-private * perf: bind smaller private Program wasm * core: make staged Program unexportable * ci: explain rustdoc surface mismatch --------- Co-authored-by: Claude <noreply@anthropic.com>
Collaborator
Author
|
Superseded by the reviewed cumulative squash merge #465 ( |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Stack
Depends on #459. This is a stacked draft and must not be merged independently.
Law
The canonical observation schema is compiled-contract state, not generic runtime lifecycle state.
Sessionfirst pins the exact owner, then derives the schema from that same lease for both keyed and packed/schema-ordered admission;Rchandle;General production
Clonewas removed from the schema type. The sole sharing capability is private to observation admission.Proof
-D warnings: passedRcmutant: both Program and point-support hostile tests REDSessionPlanV1implementationsProof rebind
The source closure changed, so the existing exact point-support proof was replayed and regenerated. Human/numerical claims, domains, exclusions, algebra and envelopes are unchanged; only binding/digest metadata changed. The verifier now preserves the same strict-length invariant on Python 3.9.
Next stacked slice
Hard-cut hidden
package_bridgeto an intentional documented publicprogramRust API, without aliases or root re-exports. That rename will remain explicitly distinct from the later Program→WASM/browser runtime cutover.Summary by CodeRabbit
Надёжность
Проверки