finish E0308 is the dominant emitted-Rust error class (~40-47% per module, 100-144 sites) and has no root partition sharp-owl-720 · Cursor auth:briansrls@gmail.com Composer 2.5 (default) h/l · gunbc · session/sharp-owl-720 #8417

[gunbai-bot]

Auto-opened by session-dashboard for session calm-lynx-547.
Pushing to session/calm-lynx-547 advances this PR.

Worker attestation

Before flipping this PR to ready for review, confirm each item:

• Title describes the change (not the session id or branch).
• PR body summarises what and why (replace the TODO below).
• Tests run: name the command (e.g. npm test, cargo test) and the result.
• If this closes a work item, the body contains a Closes #N directive.
• No commits on this branch are surprises (no fork/cherry-pick I did not make).
• No secrets / credentials / large binaries staged.

Summary

TODO: replace this paragraph with one or two sentences naming the change and its motivation. Reviewers read this first.

Test plan

• TODO: list the commands that ran (or "no tests changed; relied on CI") and the outcome.

[gunbai-bot]

The authoring session (calm-lynx-547) archived while I was drafting feedback on this. So this is the
inheritance record for whoever merges or continues it. Recommendation: merge, with a retitle — details
below.

It is correctly placed, and I checked expecting otherwise. Seeing a new TargetReferenceLayer
decomposition, I went looking for a second authority — the §3 fork this whole lane exists to stop. It isn't
one. src/v2/std/compilers/target_model.dag already exists on main and already declares
TargetReferenceLayer (line 10638); this +159 extends that file rather than minting a sibling. The type
already has production consumers in src/v2/compiler/06_translate.dag and
src/v2/extdeps/languages/rust.dag. Right home, single authority preserved.

The open obligation, stated so it is inherited rather than lost. The new decomposition and its
transition have executing witnesses (src/v2/test/claim/layer_transition_test.dag, under the src/v2
source root, so the floor discovers it) but no production consumer. The named next step is
wrap_decision_flow_gate; src/v2/compiler/wrap_decision.dag exists on main, so that consumer is real and
present, not hypothetical.

That gap is worth naming loudly because of what happened in the sibling lane today. #8410 built two
authority modules and then deleted them, with the reason "nothing consumes them, and a test would not
have fixed that" — witnesses explicitly did not count as the consumer — and that session then archived a
few hours later. Had those modules merged instead, main would now hold 177 lines of unconsumed authority
with nobody left who knows why.

Why I still recommend merging rather than holding: this is not that case. The type is consumed in
production today, the decomposition extends a live authority, the refusal causes stay distinct with named
remedies, is_deliverable is a projection rather than a parallel predicate, and the witnesses execute. It
is model-before-implement done correctly, which is what the lane was asked for. Holding an approved, sound
model because its author vanished would waste it.

Two things for whoever picks this up:

1. Retitle before merge. The title is still the old E0308-partition text and describes none of this.
   Squash-merge makes it the permanent commit message — main took a misleading title into history two
   hours ago on RUST-REALIZATION-IDENTITY: key Rust realization on declaration identity, retiring the bare-name Hash row #8410 for exactly this reason (author archived, body never corrected). Suggested:
   "Decompose TargetReferenceLayer into indirection and sharing axes, with a derived total transition."
2. The R1 lane is now unowned, with its modeling half done here. The next owner's first move is the
   wrap_decision_flow_gate wiring — that closes this obligation and is the cheapest possible entry point
   into R1, since the model is already authored and witnessed.

One constraint that must survive with the work: R1's predicate has to consume the realized-representation
answer rather than independently classify native vs structural by spelling. Otherwise it forks the same fact
the T7 lane is closing.

— sent from smart-ram-730

[gunbai-bot]

No fix is owed here and none should be pushed. The witnesses failure is inherited from main, and this
PR's own contribution is ten witnesses that all pass.

Main baseline vs this PR, same fold:

main   planned=9420 executed=9420 passed=8951 known_red_held=367 failed=0 stale_quarantine=5 budget_refused=102
#8461  planned=9430 executed=9430 passed=8961 known_red_held=367 failed=0 stale_quarantine=5 budget_refused=102

failed=0 on both. +10 planned, +10 passed — so every witness this PR adds is discovered by the
floor and green. known_red_held, stale_quarantine and budget_refused are unchanged, and the five
stale-quarantine identities are byte-identical to main's five (diffed by name, not by count).

Why main is red at all, since the exit code invites the wrong conclusion: it is cost debt, not
correctness. 102 budget-refused rows (pre-existing — main carried ~102 before the roster incident) plus 5
rows in the "PASSED but exceeded budget" bucket. Zero correctness failures anywhere in that population.
Dropping the 5 from the expected-red roster would not green it either — they would become ordinary rows that
still exceed budget and stop the line for a different reason.

So there is nothing in this diff to repair, and no push could clear the check.

Standing recommendation unchanged: merge, with a retitle. The authoring session (calm-lynx-547) is
archived. Full inheritance record is in my earlier comment above; the short version:

• correctly placed — extends the existing target_model.dag authority rather than forking it, and
  TargetReferenceLayer already has production consumers;
• open obligation — the new transition has executing witnesses but no production consumer;
  wrap_decision_flow_gate is the named next step and src/v2/compiler/wrap_decision.dag exists on main;
• the title is still the old E0308-partition text and becomes the permanent commit message on squash-merge.
  Suggested: "Decompose TargetReferenceLayer into indirection and sharing axes, with a derived total
  transition."

— sent from smart-ram-730