5+3 council: real-corpus proof + auto-resolves + LAND verdict (Q1-Q9 queue)#26
Conversation
…+ deprecation hygiene The 5+3 review council's two directly-fixable findings, resolved: R5 (completeness, OVERCLAIMED->fixed): compile_source had only ever seen inline snippets. Now: (1) examples/real_corpus_probe.rs sweeps a real addon tree — measured on /home/user/odoo addons/account/models: 55 files -> 71 models / 642 attrs / 293 assocs / 1496 ActionDefs / 347 kausal / 14 classid-resolved; (2) the verbatim 7380-line account_move.py is vendored (data/account_move_real.py, the form-view fixture precedent) and tests/real_source_compile.rs pins its compile: 1 model, classid 0x0202_0002, 104 attrs / 38 assocs / 93 computed / 354 ActionDefs / 92 kausal — exact drift-fuses, plus the _compute_amount spot identity. The V3 pipeline is now proven on REAL Odoo source, committed and fused. R1 (toolchain honesty, HOLD->resolved): (a) the 8 call sites using the deprecated corpus witness without #[allow(deprecated)] now carry it with justification (tests-mod-level + probe site) — in-build deprecation warnings 8 -> 0; (b) the deprecation note no longer promises a removal the CLI can't survive: corpus_to_actions is named as the implementation behind od-codegen --actions (ndjson corpora, not source) — do NOT remove until that mode migrates; (c) stale corpus size 2 739 -> 3 065 (verified wc -l) in README x3 + probe comment. Gate: 14/14 test binaries green (+1 = real_source_compile), clippy 0 errors, 0 deprecation warnings.
All five lenses reported; 3-meta synthesis consolidated: R1 HOLD->RESOLVED (5b45170) · R2 CATCH-LATENT x2 (queued) · R3 TARGETS-CORE + triple.rs VIOLATION (Q1, THE next refactor) · R4 3 OPPORTUNITY / 3 WORTH-EXPLORING / 0 DROP · R5 OVERCLAIMED -> core gap RESOLVED same-hour (real 55-file account addon compiled + verbatim account_move.py fused). One reviewer premise corrected by synthesis: L6 SPOG IS ruled (operator 2026-07-06); what is missing is the lance-graph ledger paperwork + the attribute->byte-position assignor, not the ruling. Queue Q1-Q9 with sizes + owners recorded in docs/COUNCIL-5+3-2026-07-07.md: triple.rs upstream migration, group_functions EmittedBy arm (smallest highest-leverage ruff diff), ledger appends x3 repos, od-posting re-architecture off SurrealQL vocabulary, mint relocation + odoo_actions authority table, guard-classifier convergence, 5.1%-tail ownership, carrier asks (#11/#15/#17 + stored verify).
Bugbot couldn't run - usage limit reachedBugbot is counted against Cursor usage for this user or team, and this run hit a usage or spend limit. A user or team admin can review and increase usage limits in the Cursor dashboard. (requestId: serverGenReqId_69317610-8d39-4248-9d56-150b576721a1) |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: a10c0612ac
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| let dir = std::env::var("ODOO_MODELS") | ||
| .unwrap_or_else(|_| "/home/user/odoo/addons/account/models".into()); | ||
| let Ok(entries) = std::fs::read_dir(&dir) else { | ||
| eprintln!("real_corpus_probe: {dir} not present — run next to an odoo checkout"); | ||
| return; |
There was a problem hiding this comment.
When ODOO_MODELS is unset outside the author's machine, this hard-coded /home/user/odoo default misses the documented sibling checkout and then returns exit 0. That makes cargo run -p od-ontology --example real_corpus_probe look successful while compiling zero real files, so the real-corpus proof can be silently skipped in local or CI verification runs; derive the default relative to the repo or exit with an error.
Useful? React with 👍 / 👎.
2 participants
What
The 5+3 review council ran against the "transpile complete" claim at
7b5b5f2(5 Sonnet lenses: toolchain honesty / baton audit / consumer doctrine / convergence / completeness critic + Opus 3-meta synthesis). Verdict: LAND with named tail — after two same-session auto-resolves carried in this PR.The overclaim, caught and killed (R5 →
5b45170)compile_sourcehad only ever seen inline snippets. Now:examples/real_corpus_probe.rs— the V3 pipeline over the REALaccountaddon: 55 files → 71 models / 642 attrs / 293 assocs / 1 496 ActionDefs / 347 kausal / 14 classid-resolved (measured, first try, no panic).data/account_move_real.py— the verbatim 7 380-lineaccount_move.pyvendored (form-view fixture precedent) +tests/real_source_compile.rspinning its compile with exact drift-fuses: 1 model, classid0x0202_0002, 104 attrs / 38 assocs / 93 computed / 354 ActionDefs / 92 kausal,_compute_amountspot identity.Toolchain hygiene (R1 →
5b45170)#[allow(deprecated)]now carry it with justification — in-build deprecation warnings 8 → 0.corpus_to_actionsis named as the implementation behindod-codegen --actions).wc -l) in README ×3 + probe comment.The verdict doc (
a10c061)docs/COUNCIL-5+3-2026-07-07.md— lens scoreboard, the synthesis correction (L6=SPOG IS ruled; the lance-graph paperwork + byte-position assignor are what's missing), and the Q1–Q9 queue with sizes/owners (triple.rs upstream migration,group_functionsEmittedBy arm, ledger appends, od-posting re-architecture off SurrealQL vocabulary, mint relocation +odoo_actionstable, guard-classifier convergence, 5.1%-tail ownership, carrier asks #11/#15/#17).Verification
cargo test -p od-ontology --features cli,fieldmask— 14/14 test binaries green (+1 =real_source_compile); clippy 0 errors, 0 deprecation warnings. F15/F16/F17 fuses re-run live by R1 and match the docs verbatim.🤖 Generated with Claude Code
https://claude.ai/code/session_01Fy8D6Boh4ZBaHSBpadSENN
Generated by Claude Code