Differentiate tier2 kind-8 recipe shape peers

This commit is contained in:
Jan Petykiewicz 2026-04-19 11:47:36 -07:00
commit b727d36b96
2 changed files with 35 additions and 0 deletions

View file

@ -90,3 +90,33 @@ So the next concrete comparison question is:
- whether the two mode-shape peers `Britain.gmp` and `South East USA.gmp` diverge from
`Louisiana.gmp` strongly enough on token lanes and later Tier 2/runtime-facing state to explain
why only `Louisiana.gmp` currently lines up with the shipped `Add Building Warehouse05` path
## Mode-shape peer split
Direct fixed-layout comparison of the two `book00.line02` mode-shape peers now answers that first
half positively.
`Louisiana.gmp` keeps:
- `book00.line00`: demanded token `0x00010000`
- `book00.line01`: demanded token `0x6c430000`
- `book00.line02`: mode `0x00080000`, supplied token `0x00004080`
`Britain.gmp` diverges strongly:
- `book00.line01`: supplied token `0x00210000`, demanded token `0x72490000`
- `book00.line02`: mode `0x00120000`, supplied token `0x00003f80`
- extra zero-mode supplied rows at `book01.line01` and `book02.line01`
`South East USA.gmp` stays structurally closer but still diverges on the exact imported values:
- `book00.line00`: demanded token `0x00010000` matches
- `book00.line01`: demanded token `0x6c410000` differs from `Louisiana`'s `0x6c430000`
- `book00.line02`: mode `0x00010000` and supplied token `0x00004000` both differ from
`Louisiana`'s `0x00080000 / 0x00004080`
- extra zero-mode supplied row at `book01.line01`
So the current corpus split is:
- `book00.line02`-only mode shape is a three-map family
- but `Louisiana.gmp` still has a distinct exact token-bearing signature even inside that family

View file

@ -623,6 +623,11 @@ Working rule:
supplied and demanded token lanes are included. That makes the next concrete Tier 2 recipe pass
smaller: compare `Louisiana.gmp` directly against `Britain.gmp` and `South East USA.gmp` on
the token-bearing minimal imported set and then on any downstream runtime-facing differences.
The note now carries that first peer split too: `Britain.gmp` diverges strongly with different
mode and extra supplied rows, while `South East USA.gmp` stays structurally closer but still
differs on the exact `book00.line01` demanded token and `book00.line02` mode/supplied-token
pair. So the current queue head is no longer “find a peer”; it is “find the downstream
runtime-facing consequence of Louisianas distinct exact `book00` token-bearing signature.”
kinds”; it is the smaller set of scenario-specific records where that sweep explicitly writes
`[event+0x7ef]` itself or a still-later owner does.
- two explicit trigger-kind materializations are now grounded inside that retagger: