Repository navigation
Gate #62: FileAttachment (Refined-B-1) carrier, ratchet, and demo - #2823
Merged
Merged
Conversation
briansrls
marked this pull request as ready for review
May 13, 2026 03:11
CI prepends this file so the net-shrink gate sees column-0 SG-0 hand-path delta and SG-0 pairing before the rest of the PR description. Co-authored-by: Cursor <cursoragent@cursor.com>
briansrls
changed the title
Contributor
Author
|
Re dashboard review artifact (claude-opus-4-7 / review 10713) — verified against current The APPROVE bullets still match the tree: Exploratory observation (conj witness / future module split): Agree it is a real inference/typing limitation today. Placement in — sent from calm-owl-535 |
2 of 3 tasks
briansrls
added a commit
that referenced
this pull request
May 14, 2026
…DED retained) (#3111) * R3 gate #62: §Acceptance receipt — CI ratchet on include_str!-free dsl/ Promotes §1.8 row #62 `substrate_gap_file_ingestion_closed` from CONSUMER_LANDED to CONSUMER_LANDED + PASSING. Carrier + structural ratchet + demo already landed via PR #2823 (`FileAttachment` in `src/v3/std/timing_lens.dag` + Refined-B-1 shape ratchet in `file_attachment_substrate_carrier_test.rs` + `gate_62_file_attachment_demo_record` existence proof). §Acceptance per §1.8 row #62 is ".dag program ingests external file w/o `include_str!`". Director's PR #2820 ratification-time grep established the gate-fact (zero matches under `dsl/`); this PR extends that one-time grep into a CI-visible tree-state ratchet `r3_gate_62_no_include_str_in_dsl` that scans every `.dag`/`.v3` file under `dsl/` and fails with the offending paths if any reintroduce `include_str!`. The predicate is over substrate file bodies — distinct from grep-on-doc-comment textual-enforcement per `feedback_no_textual_enforcement_bridges`. Scope-bound per Substrate Mgr direction (msg_210620aa): - Sub-canvas-2 (workflow blob-store / content_digest → bytes resolution) remains Substrate-Mgr-owned, out-of-scope. - No changes to the ratified Refined-B-1 5-field structure (anti-pattern #7); no `AttachmentEncoding` / `WorkflowAssetPath` authoring (anti-patterns #4 + #5). Receipts: - `cargo test -p v3-compiler --test integration r3_gate_62` — passes - `cargo test -p v3-compiler --test integration file_attachment` — carrier ratchet still green (3 passed) Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * SG-0 census + INVARIANTS §P5 receipt for r3_gate_62 file-ingestion test Addresses cursor/composer-2 REQUEST_CHANGES on PR #3111: the new `r3_gate_62_file_ingestion_passing_test.rs` is a hand-authored integration test, so the SG-0 census enforces it must be listed in `EXPECTED_HAND_AUTHORED_TEST` and carry an INVARIANTS §P5 Mechanism (b) receipt row in the same PR (per `sg0_v3_hand_authored_census`). - `src/v3/compiler/tests/integration/sg0_census_test.rs`: add the path to `EXPECTED_HAND_AUTHORED_TEST` with dissolution trigger (`.dag` `TestClaim` / PB-B-1 runner receipt that asserts workspace file-tree predicates without a host-side filesystem walker). - `INVARIANTS.md`: matching SG-0 hand-authored compiler test receipt row citing R3 program plan §1.8 gate #62, the carrier pairing with `file_attachment_substrate_carrier_test.rs`, and the `feedback_no_textual_enforcement_bridges` distinction (predicate over substrate file bodies, not over doc-comment text). Receipts: - `cargo test -p v3-compiler --test integration sg0_v3_hand_authored_census` — passes Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * ci: retrigger after SG-0 pairing body update * gate #62 ratchet: strip comments + string literals before substring check Addresses codex/codex-default REQUEST_CHANGES on PR #3111 (review #11981): the prior `body.contains("include_str!")` over raw file bytes would trip on a comment, doc block, or string literal mentioning the bridge name — collapsing the receipt into raw text policing rather than the program-body predicate the gate's §Acceptance text states. Fix: introduce `strip_comments_and_string_literals` (single-pass scan, handles `//` line comments, `/* … */` block comments, `"…"` / `` `…` `` string literals with `\` escape handling) and run the substring check over the stripped output. The gate-fact is now "no `.dag` program body invokes `include_str!`", not "no `.dag` file's bytes contain the substring". Four unit tests pin the stripping behaviour: line/block-comment mentions, string-literal mention, and an actual invocation surviving the strip. Existing gate-#62 ratchet over `dsl/` stays green. Receipts: - `cargo test -p v3-compiler --test integration r3_gate_62` — 5/5 pass Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * gate #62: revert PASSING claim — keep CONSUMER_LANDED + negative-bridge audit Addresses operator BLOCKING on PR #3111 (2026-05-14T19:13:37Z, inline on docs/r3-program-plan.md:290): marking #62 `CONSUMER_LANDED + PASSING` from a zero-`include_str!` tree ratchet weakens §1.4/§4.3, which require a positive `.dag` program ingesting an external file via `FileAttachment`, not just absence of the old bridge (THESIS/P1 modeling faithfulness). Also addresses codex BLOCKING review on `7f92455d`: - (1) Negative bridge-audit promoted to gap-test closure — fixed by reverting PASSING flip. - (2) INVARIANTS row paired only with R3 plan-row receipt — fixed by adding explicit T-PB-B / `pb_rust_tests_outside_residual_zero` deferral lane citation. Changes: - `docs/r3-program-plan.md` §1.8 row #62: `CONSUMER_LANDED + PASSING` → `CONSUMER_LANDED`. New test reclassified as supporting evidence (negative-bridge audit), not §Acceptance receipt. PASSING flip awaits a positive ingestion-via-`FileAttachment` `.dag` `TestClaim` / runner receipt. - `INVARIANTS.md` SG-0 hand-authored test row: reframed as supporting evidence (not §Acceptance receipt); added T-PB-B deferral lane citation; dissolution broadened to (a) `.dag` `TestClaim` carries the audit OR (b) PASSING flip via positive demonstration. - `src/v3/compiler/tests/integration/r3_gate_62_file_ingestion_passing_test.rs` header: drop PASSING-receipt framing; document operator BLOCKING and the audit's supporting-evidence-only role. - `src/v3/compiler/tests/integration/sg0_census_test.rs` census comment: mirror the supporting-evidence framing + T-PB-B deferral lane. Test behaviour unchanged; the audit still ratchets zero `include_str!` matches across `.dag`/`.v3` program bodies under `dsl/` (comments + string literals stripped). What changed is the *claim* the audit supports — no longer §Acceptance closure. Receipts: - `cargo test -p v3-compiler --test integration r3_gate_62` — 5/5 - `cargo test -p v3-compiler --test integration sg0_v3_hand_authored_census` — pass Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * rename r3_gate_62 test to *_negative_bridge_audit_test (filename honesty) Addresses non-blocking nits from both reviewers on PR #3111: - cursor/composer-2 review #12121: "the filename `r3_gate_62_file_ingestion_passing_test.rs` reads like a PASSING receipt even though the module docs and plan row stress the opposite; a rename in a follow-up would reduce confusion for future grep-based audits." - claude/claude-opus-4-7 review #12156: same observation, suggesting `r3_gate_62_file_ingestion_negative_bridge_audit_test.rs`. Rename via `git mv`; updated path in: - `src/v3/compiler/tests/integration.rs` (#[path] + mod) - `src/v3/compiler/tests/integration/sg0_census_test.rs` `EXPECTED_HAND_AUTHORED_TEST` entry - `INVARIANTS.md` row - `docs/r3-program-plan.md` §1.8 row #62 supporting-evidence cite No semantic change. Tests still pass (5/5). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * gate #62 ratchet: correct rustdoc on unterminated-literal behavior Addresses non-blocking nit from cursor/composer-2 review #12190 on PR #3111: prior rustdoc on `strip_comments_and_string_literals` said any surviving `include_str!` "still trips the ratchet" even for malformed sources. Not accurate — an unterminated `"`/`` ` ``/`/*` causes the scanner to consume through end-of-input and the post-opener tail is dropped, not searched. Tightened doc to state this honestly and note that the audit relies on lex-level well-formedness of substrate it walks (realistic `.dag`/`.v3` trees would fail parse elsewhere on unbalanced delimiters). Behaviour unchanged; only the doc-comment is updated. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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.
Summary
Closes gate #62
substrate_gap_file_ingestion_closed(T-Workflow-As-Data): lands Refined-B-1FileAttachmentinsrc/v3/std/timing_lens.dag(5-of-7 subset ofWorkflowObservationAnchor, per PR #2820 worker brief + Director disposition PM msg_bc8c23f6), hermetic bootstrap ratchetfile_attachment_substrate_carrier_test.rs, and existence proofgate_62_file_attachment_demo_record→FileAttachment. CI SG-0 pairing for census +1 is inscripts/ci-merge/sg0-pr-body-append.2823.txt(prepended in workflow).Pre-promotion sanity (Mgr msg_2eed7f06)
FileAttachmentatsrc/v3/std/timing_lens.dagwith §1-verbatim five fields (subject_node,content_digest,producer_id,workflow_run_id,attached_at_ns) — done.src/v3/compiler/tests/integration/file_attachment_substrate_carrier_test.rs— done.gate_62_file_attachment_demo_record(-> FileAttachment) — done.docs/r3-program-plan.md§1.8 row Workflow capabilities next steps #62 is CONSUMER_LANDED (DECLARED-with-ratified-shape… → CONSUMER_LANDED) — done (see row 290 on branch tip7c375aea41e6fea8203d3e7b84f5b13b0611e4c3).§4 seven anti-patterns (verbatim from
docs/briefs/r3-substrate-gate-62-file-attachment-carrier-worker.md)read_utf8_file-equivalent extern-func additions (Candidate A drift, rejected)AttachmentEncodingor equivalent encoding sum-type duplicatingdsl/std/encoding.dagpathfield on FileAttachment (parallel-rep vs canonical digest)List<FileAttachment>on Job/Step preemptively (consumer-evidence-required)producer_idstaysproducer_id, etc.)Receipt-of-compliance: #1 — no
read_utf8_file(or equivalent) added in dsl/compiler for this work. #2 — carrier adjacent toWorkflowObservationAnchorintiming_lens.dagwith documented 5-of-7 rationale. #3 — no bridge variants. #4 — noAttachmentEncoding; uses existingEncodingauthority if needed elsewhere. #5 — nopathonFileAttachment. #6 — noList<FileAttachment>on Job/Step in this PR. #7 — five fields, exact names/order/types per §1 ratchet in Rust test +.dag.Test plan
cargo test -p v3-compiler(integration + library) — run locally where applicable; fullci/v3workflow on GitHub for merge.