phase12: PLAN approved — milestone 750 bodies (+148), stretch 800
Approved by the developer 2026-09-24: milestone 750, stretch 800; both harness repairs authorised; Goal B (the 92-unclassified negatives) in scope and bounded; 4 workers — two on the fresh band, one on the 101 named negatives, one on the third fresh slice plus the 92 unclassified as the Goal B worker; compaction-first. THE ONE STRUCTURAL FACT THIS PLAN IS BUILT ON, measured at the open: 0 of the 193 near_match_negatives rows appear in the 1001-row worklist. The negatives index is held OUT of the worklist (sf3_triage plan --negatives) and is not a queue — it is an index, and a worker dispatched only from the worklist can never reach it. So the negatives pool becomes a first-class partition with its own worker, and the phase's structural change is that it stops being an appendix. Two pools, measured: 1001 fresh rows (410 of them <=244 B) plus 193 attempted-but-unresolved negatives, of which 101 carry a NAMED mechanism and are overturnable by cookbook 182's rule, and 92 carry no class at all and cannot be attempted until someone re-reads them. That is what Goal B is for. The plan states plainly that +148 is LARGER THAN ANY PHASE SO FAR (Phase 11 closed +118) and that it cannot be reached from the fresh worklist alone — it depends on the negatives-overturn programme working at scale, which is a measured route but only a ONE-ROW demonstration so far (worker F's 0x8002622C). Two safeguards are built in rather than discovered late: the cycle-1 checkpoint must include at least 8 bodies from the negatives pool, so a low overturn rate is measured in cycle 1 and not at the close; and the leading-indicator rule stands unchanged — report the projection and ask, never redefine or self-certify the milestone. Also recorded: the fresh-band size distribution, the negatives class distribution, and the baseline revalidation table including the failed precondition that T0 repaired.
This commit is contained in:
Christopher Williams
@@ -0,0 +1,341 @@
|
||||
# Phase 12 Plan — Overturning the Named Negatives
|
||||
|
||||
**Status:** **APPROVED** (2026-09-24).
|
||||
**Prepared:** 2026-09-24, immediately after Phase 11 closure (602 bodies / 611 regions).
|
||||
**Baseline:** commit `9f2543b` on `main`, working tree clean, 722 tracked files.
|
||||
**Approved decisions (developer, 2026-09-24):** 1. milestone **750**, stretch **800**;
|
||||
2. **both harness repairs authorised** (§10.1 the stale maspsx patch, §10.2 `sf3_free`'s ledger path);
|
||||
3. **Goal B in scope, bounded** (the 92-unclassified classification only);
|
||||
4. **4 workers** — two on the fresh band, one on the 101 named negatives, one on the third fresh slice
|
||||
plus the 92 unclassified as the Goal B worker; orchestrator discretion to scale or retire;
|
||||
5. **compaction-first** context policy.
|
||||
|
||||
**Milestone:** **750 distinct matched bodies**, from 602 — i.e. **+148**.
|
||||
**Stretch:** 800 bodies.
|
||||
|
||||
> ### ⚠ That milestone is larger than any phase so far, and the plan says so plainly
|
||||
>
|
||||
> Phase 11's record was **+118** with 4–6 workers. **+148 is 25 % more in one phase**, and it cannot be
|
||||
> reached from the fresh worklist alone (the ≤244 B band is only 410 rows and a third of it yields
|
||||
> +98 on Phase 11's rate). **The milestone therefore depends on the negatives-overturn programme
|
||||
> working at scale**, which is a measured route but only a *one-row* demonstration so far.
|
||||
>
|
||||
> Two honest safeguards are built in rather than discovered late:
|
||||
> 1. **Only one of the 101 named-mechanism negatives has ever been overturned** (worker F's
|
||||
> `0x8002622C`). The first checkpoint (P12-T2, +30 bodies) must include **at least 8 A2 bodies from
|
||||
> the negatives pool**, so that a low overturn rate is measured in cycle 1 instead of at the close.
|
||||
> 2. **The leading-indicator rule** (workflow §9) applies unchanged: a per-batch cluster of sub-8-byte
|
||||
> residuals with no NEW mechanism means the band is exhausted — report the numbers and re-rank, do
|
||||
> not grind. If cycle 2's measured rate projects below 750, the orchestrator reports the projection
|
||||
> and asks; it does not redefine the milestone and does not self-certify it.
|
||||
|
||||
---
|
||||
|
||||
## 0. The honest starting point
|
||||
|
||||
Phase 11 closed at 602 bodies / 611 regions with two headline results that define this phase:
|
||||
|
||||
1. **The 244-byte ceiling was a dispatch artefact, not a toolchain limit.** 28 regions now exist
|
||||
above 244 B, the largest 1232 B. Nothing above 244 B had ever been *attempted*, and once rows were
|
||||
attempted the ceiling evaporated.
|
||||
2. **The dispatch criterion is tie-break density, not size** (cookbook 79/88/99, now implemented in
|
||||
the tracked `tools/sf3_rank`): a 548 B body matched first attempt while a 176 B body took nine
|
||||
spellings and never closed.
|
||||
3. **A recorded negative with a NAMED mechanism is overturnable; one without is not**
|
||||
(cookbook 182). Worker F took `0x8002622C` out of the negatives index — a row four earlier
|
||||
spellings had written off — and matched it inside the hour by making an intermediate live past the
|
||||
second operation, proving the mechanism with a `cc1 -da` dump.
|
||||
|
||||
Result 3 is the measured route to 700, and it is larger than the Phase 11 coordinator first reported.
|
||||
|
||||
### The two dispatch pools, measured at this baseline
|
||||
|
||||
| pool | rows | note |
|
||||
|---|---|---|
|
||||
| `config/match_worklist.tsv` (fresh rows) | **1001** | regenerated 2026-09-24, byte-identical to the tracked file, `excluded_already_registered=611` |
|
||||
| `config/near_match_negatives.tsv` (attempted, unresolved) | **193** | **held OUT of the worklist** — `sf3_triage plan --negatives` excludes them |
|
||||
| **total unregistered candidates** | **1194** | |
|
||||
|
||||
**The negatives index is a separate pool, and this is the single most important fact for Phase 12.**
|
||||
Verified at this baseline: **0 of the 193 negatives rows appear in the 1001-row worklist.** A worker
|
||||
dispatched only from the worklist can never reach them. They are not a queue; they are an index.
|
||||
|
||||
### What is actually in the negatives index (measured)
|
||||
|
||||
| class column | rows |
|
||||
|---|---|
|
||||
| named class or substantive note | **95** |
|
||||
| `LENGTH-MISMATCH ...` prose note | 6 |
|
||||
| **`-` (no class at all)** | **92** |
|
||||
| **total** | **193** |
|
||||
|
||||
So **101 rows carry a named mechanism** and are overturnable by cookbook 182's rule; 92 carry no
|
||||
mechanism and must be re-read before they can be attempted at all. Mean stated size of the 116 rows
|
||||
that have one is 119 B — this is a pool of *small* rows, and the milestone counts bodies.
|
||||
|
||||
### Fresh-worklist size bands (measured)
|
||||
|
||||
| band | rows |
|
||||
|---|---|
|
||||
| ≤120 B | 68 |
|
||||
| 121–200 B | 243 |
|
||||
| 201–244 B | 99 |
|
||||
| **≤244 B subtotal** | **410** |
|
||||
| 245–400 B | 225 |
|
||||
| 401–800 B | 221 |
|
||||
| >800 B | 145 |
|
||||
|
||||
Grade: 999 `exact`, 2 `fallthrough`. Tier: 138 tier-1 (exact leaf), 861 tier-2, 2 tier-3, 0 tier-0
|
||||
(the duplicate-group representatives are all consumed).
|
||||
|
||||
### The arithmetic of +148
|
||||
|
||||
| source | rows available | basis |
|
||||
|---|---|---|
|
||||
| fresh worklist ≤244 B | 410 | the proven band; Phase 11 consumed a comparable slice |
|
||||
| fresh worklist 245–400 B | 225 | **partially proven** — Phase 11 put 28 regions above 244 B |
|
||||
| negatives with a NAMED mechanism | 101 | overturnable by cookbook 182; **1 / 1 demonstrated** |
|
||||
| negatives with NO mechanism | 92 | unattemptable until Goal B classifies them |
|
||||
| fresh worklist 401–800 B / >800 B | 221 + 145 | proven tractable but expensive in context |
|
||||
|
||||
**The milestone mixes three rates, and only one of them is proven at scale:**
|
||||
|
||||
- the ≤244 B band is proven, and the fresh-band workers will consume it at Phase 11's per-worker rate;
|
||||
- the 245–400 B band is 225 rows and Phase 11's 28 regions above 244 B (max 1232 B) show it is
|
||||
reachable — this is where A1 must go once ≤244 B thins, and it is why `sf3_rank` (tie-break density)
|
||||
rather than size drives dispatch;
|
||||
- **the negatives are the swing factor.** 101 rows at a high overturn rate gives the milestone its
|
||||
margin; at a low rate the phase lands nearer Phase 11's own +118 rate and the shortfall is reported
|
||||
with its measured cause.
|
||||
|
||||
**This phase's structural change is therefore: the negatives index stops being an appendix and becomes
|
||||
a first-class partition with its own worker.**
|
||||
|
||||
---
|
||||
|
||||
## 1. Goal
|
||||
|
||||
**Goal A (the milestone): 750 distinct bodies.** Two partitions run in parallel:
|
||||
|
||||
- **A1 — the fresh band**, dispatched by `sf3_rank` (tie-break density) with `--max-size`, not by size.
|
||||
- **A2 — the named negatives**, 101 rows, dispatched from `config/near_match_negatives.tsv` directly.
|
||||
Each row's named mechanism is the hypothesis; cookbook 182's rule is the licence to attempt it. The
|
||||
rule is: **name the pass that does the damage and the dimension that was exhausted**, then try the
|
||||
ONE lever that dimension implies.
|
||||
|
||||
**Goal B (strategic, evidence deliverable): a counted answer to why 92 negatives rows carry no class.**
|
||||
Deliverable is evidence, not bodies. Any of these outcomes is a success: (a) the 92 are classified into
|
||||
named mechanisms and become attemptable; (b) they are shown to collapse into a small number of classes
|
||||
already known; (c) they are characterised as out of reach with the class counted. **A Goal B result
|
||||
that adds zero bodies is still a met Goal B.**
|
||||
|
||||
**Goal C (carry-forward): the GTE class.** `include/gtemac.h` carries the `lwc2`/`swc2` macros, the
|
||||
command-field caveat and the register-binding lever (cookbook 159/160/161); one row matched and four
|
||||
more are fully derived. Bounded: harvest the derived rows, then report.
|
||||
|
||||
---
|
||||
|
||||
## 2. Structure — orchestrator + 4 workers, spawned by the orchestrator
|
||||
|
||||
Unchanged from Phase 11, and **the roster must be spawned from scratch** (`CURRENT_PHASE.md`,
|
||||
workflow §4.0 — Phase 11 tore the roster down; there is nothing to reconnect to).
|
||||
|
||||
- **Topology:** one tab, four agent panes, plus the orchestrator's own pane. 2×2 grid; no four-column
|
||||
or four-row layouts.
|
||||
- **Assignment by goal, not preference:**
|
||||
| worker | partition | goal |
|
||||
|---|---|---|
|
||||
| A | fresh band, `sf3_rank` top slice | A1 |
|
||||
| B | fresh band, `sf3_rank` second slice | A1 |
|
||||
| C | **the 101 named negatives** | A2 |
|
||||
| D | fresh band third slice + **the 92 unclassified negatives** | A1 + **B** |
|
||||
- **D is the Goal B worker** and its artifact is a running investigation record, not a body count.
|
||||
- Roster may scale to 5 or shrink, orchestrator's call, based on measured attempts-per-match.
|
||||
- **Adjacency is the strongest dispatch heuristic found** (4-for-4) — but it crosses partition
|
||||
boundaries, so the in-flight ledger is mandatory (§5 of the workflow).
|
||||
|
||||
---
|
||||
|
||||
## 3. Context policy — compaction first, rotation second
|
||||
|
||||
`pi-context-tools` is installed globally. Unchanged from Phase 11:
|
||||
|
||||
- Every report quotes the exact `context_info` figure. **Never estimate.**
|
||||
- Worker at 70–80% with work left in its band → `compact_context`, keep working.
|
||||
- Band exhausted or staging stale → rotate (verify handoff first).
|
||||
- Orchestrator compacts rather than stopping at the cap.
|
||||
- **Keeping the ledger and staging current is a hard requirement**, because compaction summarises the
|
||||
conversation away; only the files survive.
|
||||
|
||||
---
|
||||
|
||||
## 4. The carried machinery (verify before relying, do not re-derive)
|
||||
|
||||
- **Toolchain:** PsyQ 4.0 `CC1PSX` (GCC 2.7.2.SN32.3.7.0002); flags `-quiet -O2 -G0`; maspsx (ASPSX
|
||||
2.56 emulator) then GNU `as`.
|
||||
- **Harness:** `sf3_match` (per-region `cc1=`, `as=`, `gp=-NAME`,
|
||||
`maspsx=off|noreordernop|regread|moves|epilogue|nopmarker`), `sf3_diff`, `sf3_merge`, `sf3_triage`,
|
||||
`sf3_cc`, `sf3_free`, `sf3_family`, `sf3_rank`.
|
||||
- **The tracked maspsx patch** `tools/patches/maspsx-phase10-r1r2.patch` — **STALE, see §10.1. This is
|
||||
a phase blocker, not a footnote.**
|
||||
- **Evidence tables:** `function_inventory`, `function_extents`, `duplicate_bodies`, `match_worklist`,
|
||||
`near_match_negatives`.
|
||||
- **The hardened merge flow** (workflow §6): verify → merge to CANDIDATE → gate whole-binary → promote
|
||||
only on `result=MATCH` → reconcile negatives → regenerate worklist + refilter partitions → commit →
|
||||
push. Use the gate's **exit code**, never a `grep` for `result=`.
|
||||
|
||||
---
|
||||
|
||||
## 5. Rules already in force (carried into every charter, with provenance in workflow §9)
|
||||
|
||||
`git ls-files --error-unmatch` before any `rm` under `src/`; the dependent-claim rule; exact symbol-name
|
||||
spelling (the gp marker is name-keyed); compute-don't-eyeball; the three-encoding-trap habit; the
|
||||
amended register-field rule (verify the dataflow before calling it a tie-break); the two-step
|
||||
two-arm-branch rule; the paired `&&`/named-boolean rule; **verify the file you are about to stage, at
|
||||
the path you are about to claim**; **run the `src/` cleanup audit BEFORE the final report**; **read start
|
||||
and end from the worklist row, never from a hand dump**; **one attempt on a named lever, then classify**;
|
||||
**record the source md5 per claim**; **a rule every worker must follow belongs in a TRACKED tool**
|
||||
(cookbook 179); **the `maspsx=epilogue` token is decided from the CANDIDATE's tail, never the
|
||||
original's** (cookbook 165/180); **derive every absolute address from the `(page, signed immediate)`
|
||||
pair** (cookbook 154 — the most repeated worker error of Phase 11, five occurrences, all one worker).
|
||||
|
||||
---
|
||||
|
||||
## 6. Verification ladder
|
||||
|
||||
| frequency | checks |
|
||||
|---|---|
|
||||
| every merge | candidate gate (whole binary, SHA-1 fixed) → promote → `make check` |
|
||||
| every merge | worklist regen (`excluded_already_registered` == registry size) + partition refilter with the disjointness and union-equals-worklist proof |
|
||||
| every 3rd merge | `make clean && make all`, `cmp`, SHA-1, registry audit, `git status --short src/`, firewall count |
|
||||
| every worker retire | staging verified, `src/` cleanup audit, negatives extracted into the tracked index |
|
||||
| phase close | full clean gates, negatives **reconciliation**, stale-taken audit, milestone confirmation requested, PhaseEnd, digest, ledger, roster teardown, **stop** |
|
||||
| push | after every verified merge |
|
||||
|
||||
---
|
||||
|
||||
## 7. Scope and safeguards
|
||||
|
||||
- ROM/SDK firewall unchanged; worker staging ignored; `git clean -x`/`-fdx` forbidden.
|
||||
- Workers may not weaken a gate, widen regions, or register a non-`exact` extent.
|
||||
- **No scheduler-changing flag** for a match; diagnosis only, recorded as a negative.
|
||||
- Inline asm limited to shapes plain C provably cannot express, per-file documented.
|
||||
- The blocked classes stay excluded unless Goal B produces evidence that reopens them.
|
||||
- Goal B is bounded: two distinct root-cause attempts per hypothesis, then record.
|
||||
- **Do not copy a worker's staging file over a tracked tool** (cookbook 186; `tools/sf3_diff` was
|
||||
destroyed and restored). Before promoting anything into `tools/`: `git log --oneline -1 -- <path>`,
|
||||
`wc -l <path>`, `make test`, and ask *what does the old one already do that the new one does not?*
|
||||
- **Do not ignore a shrinking test count.** 253 tests is the Phase 11 close figure.
|
||||
|
||||
---
|
||||
|
||||
## 8. Tasks
|
||||
|
||||
- **P12-T0 — REPRODUCIBILITY REPAIR (blocker, before any matching). — ✅ COMPLETE 2026-09-24.**
|
||||
§10.1 the stale maspsx patch regenerated; §10.2 `sf3_free`'s per-phase ledger resolution. Both
|
||||
verified; `make test` **281 tests OK** (253 at the Phase 11 close, +28 new). Evidence:
|
||||
`docs/SETUP.md` ("Phase 12 repair"), cookbook findings **188** and **189**, and the two new guard
|
||||
suites `tools/tests/test_maspsx_patch.py` (7 tests) and `tools/tests/test_sf3_free.py` (21 tests).
|
||||
- **P12-T1 — phase open.** Control records, clean-baseline revalidation (done in part; see §9),
|
||||
herdr roster spawn (4 workers), capability probes including `context_info`, partitions cut from both
|
||||
pools, disjointness/union proofs, lever files with their measured basis, `CURRENT_PHASE.md`, ledger.
|
||||
- **P12-T2 — Goal A cycle 1**, both pools live. Checkpoint: **+30 bodies, of which at least 8 must come
|
||||
from the negatives pool (A2)** — so a low overturn rate is measured in cycle 1 rather than at the
|
||||
close (see the milestone warning in the header).
|
||||
- **P12-T3 — Goal B opening:** the 92 unclassified negatives. Re-read every row against the original
|
||||
bytes, classify into named mechanisms, publish the classification as counted evidence.
|
||||
- **P12-T4…T-n — the autonomy loop:** per-merge regen, full audit every 3rd merge, workers compacted
|
||||
rather than rotated, Goal B probed in bounded slices, Goal C harvested early and reported.
|
||||
- **P12-Tn — phase close:** cookbook/conventions/verification records, negatives reconciliation
|
||||
(drop any row since registered, with the overturn recorded), stale-taken audit, milestone
|
||||
confirmation requested from the developer, PhaseEnd, digest, ledger archive, **roster teardown
|
||||
stated explicitly**, stop.
|
||||
|
||||
---
|
||||
|
||||
## 9. Baseline revalidation — evidence gathered while preparing this plan
|
||||
|
||||
| check | result |
|
||||
|---|---|
|
||||
| `git log --oneline -1` | `9f2543b` (Phase 11 close-record correction) |
|
||||
| `git status --short` | clean at first check |
|
||||
| tracked files | 722 |
|
||||
| registry rows | **611**, distinct `src/*.c` = **602** |
|
||||
| `sf3_triage plan` re-run to scratch | **listed=1001**, `excluded_already_registered=611`, `cmp` vs tracked worklist **exit 0** (byte-identical, deterministic) |
|
||||
| negatives index | 193 rows, 92 unclassified, 101 with a class; **0 overlap with the worklist** |
|
||||
| maspsx patch applied? | **NO — patch does not apply to pristine and does not reproduce the working tree (§10.1). Repaired in P12-T0.** |
|
||||
| `make test` after P12-T0 | **281 tests, OK** (253 + 28 new guard tests) |
|
||||
| `make clean && make all`, `cmp`, SHA-1, `make extents-verify`, `make gate` | run at the open; results recorded in the Phase 12 ledger before P12-T2 |
|
||||
|
||||
---
|
||||
|
||||
## 10. Decisions — developer's answers (2026-09-24)
|
||||
|
||||
| # | decision | answer |
|
||||
|---|---|---|
|
||||
| 10.3.1 | milestone | **750 bodies (+148)**, stretch **800** |
|
||||
| 10.1 + 10.2 | harness repairs | **both authorised** |
|
||||
| 10.3.2 | Goal B scope | **in scope, bounded** — the 92-unclassified classification only |
|
||||
| 10.3.3 | roster | **4 workers**, 2 fresh + 1 negatives + 1 third-slice/Goal-B |
|
||||
| 10.3.4 | context policy | **compaction-first**, unchanged from Phase 11 |
|
||||
|
||||
### 10.1 ⚠ Harness repair: the tracked maspsx patch is STALE and the toolchain is not reproducible
|
||||
|
||||
**This is a failed precondition, found during the open checklist, not a hypothesis.** Evidence:
|
||||
|
||||
1. `patch -p1 < tools/patches/maspsx-phase10-r1r2.patch` against pristine `86ccd7d` (extracted with
|
||||
`git archive HEAD`) → **exit 1, 4 of 10 hunks FAILED**, `.rej` files written for
|
||||
`maspsx/__init__.py` and `maspsx.py`.
|
||||
2. The patch carries the **shape-A-only** epilogue transform from commit `05be974` ("THE EPILOGUE
|
||||
POST-PASS SHIPS"). The working tree carries the **shape-A+B** transform from commit `ea51ac9`
|
||||
("merge 40 + the epilogue transform now handles BOTH shapes").
|
||||
3. `ea51ac9` touched `tools/sf3_match`, `config/regions.tsv` and seven `src/*.c` files — **and not
|
||||
`tools/patches/`**. `git log --oneline -- tools/patches/` still shows `05be974` as the newest commit
|
||||
touching the patch. `tools/maspsx/` is git-ignored (`tools/maspsx/maspsx.py` mtime 10:46; patch
|
||||
mtime 10:42).
|
||||
4. **Functional demonstration** (`maspsx --fill-epilogue`, shape-B tail
|
||||
`lw $31,24($sp) / lw $17,20($sp) / lw $16,16($sp) / addiu $sp,$sp,32 / jr $31 / nop`):
|
||||
| build | output |
|
||||
|---|---|
|
||||
| pristine + tracked patch | `... / nop / jr $31 / addiu $sp,$sp,32` — **6 instructions** |
|
||||
| working tree | `... / jr $31 / addiu $sp,$sp,32` — **5 instructions** |
|
||||
The original at `0x800F44D0` is **5**. `ea51ac9`'s own message: *"My first implementation did A only
|
||||
and left every B row 4 bytes long."*
|
||||
5. **15 registered regions carry `maspsx=epilogue`.** A fresh clone would fail the whole-binary gate on
|
||||
the shape-B members of that set.
|
||||
|
||||
**Decision: AUTHORISED.** Regenerate the patch as a diff of pristine → working tree for exactly
|
||||
`maspsx/__init__.py` and `maspsx.py`, then verify by reconstructing from pristine and diffing
|
||||
byte-identically against the working tree, and record the verification in `docs/SETUP.md`. This is the
|
||||
§11 remedy (*"do not edit a tool that is git-ignored without carrying the change as a tracked patch"*),
|
||||
and it **strengthens** rather than weakens a gate.
|
||||
|
||||
### 10.2 Harness repair: `sf3_free`'s ledger path is hard-coded to `.run/p11/`
|
||||
|
||||
`tools/sf3_free` line 35 is `INFLIGHT = REPO / ".run/p11/inflight.tsv"`. The ledger is
|
||||
session-scoped and git-ignored by design, so Phase 12's ledger is `.run/p12/inflight.tsv` and the tool
|
||||
would read a 306-line stale Phase 11 ledger containing `wip` rows whose holders no longer exist — the
|
||||
exact stale-taken failure of cookbook 187, reintroduced by a hard-coded path.
|
||||
|
||||
**Decision: AUTHORISED.** Add a `--ledger` option (default: the current phase's path derived from the
|
||||
partition directory), keeping last-row-wins semantics. Small, tracked, and covered by the existing
|
||||
`tools/tests` suite if it has a case there.
|
||||
|
||||
---
|
||||
|
||||
## 11. Preconditions
|
||||
|
||||
- **P12-T0 — ✅ COMPLETE.** The regenerated maspsx patch applies cleanly to pristine (exit 0, no
|
||||
rejects) and reproduces both modified files byte-identically (`cmp` exit 0), reverse-applies cleanly
|
||||
(the tree *is* pristine+patch), and still carries all four opt-in modes. `tools/sf3_free` resolves the
|
||||
current phase's ledger and no longer reads a dead one. Guard tests proven to catch both defects. Patch
|
||||
sha1 `0d7f643ba1c31d1ae28de257737c16c623a9eb00`, replacing the stale
|
||||
`4bd21823402d73659f76b3afbf60f8f93daaa3e8`.
|
||||
- `make test` green at **281 tests** (253 at the Phase 11 close).
|
||||
- `make check` green at 611 regions / 602 bodies. Registry audit clean; `git status --short src/` empty.
|
||||
- herdr roster spawned and probed; `context_info` confirmed live in every worker.
|
||||
- Partitions cut from **both** pools with a proven disjointness and union-equals-pool check.
|
||||
|
||||
**Milestone:** 750 distinct matched bodies on the clean whole-binary gate, plus a counted Goal B
|
||||
classification — or the evidence-backed blocker recorded. **The milestone is confirmed by the
|
||||
developer, never self-certified.**
|
||||
Reference in New Issue
Block a user