Continue strategy-0219ad7fc31f by replacing the stale cross-mount Lean build cache with a clean workspace-local cache, hardening the native validation harness, and dispatching hash-bound native Stage1 validation jobs.

upstream model bridge plus native tactic replay

Model Const.mergeWith using a distinct-key list merge, register the transformation with simp_to_model, and execute the imported metaprogram using a freshly built Lean v4.32.0 runtime before checking Iris.

Hypothesis: A clean workspace-local Lean v4.32.0 native build containing the mergeWith model registration makes simp_to_model close the standalone semantic fixture and refactored Iris theorem, after which the complete Iris build and separate consumer checker pass.

Test: Run corrected lab job lab-iris-lean-127-simp-to-model-8369649cc70f and require one receipt containing workspace-local Stage0 provenance, native_compile_checked=true, focused HeapInstances success, full Iris build success, and passed=true from the separate checker.

The static and configuration receipts establish genuine infrastructure progress but cannot validate a changed imported Lean metaprogram. Candidate promotion therefore remains fail-closed until one corrected native run passes the fixture, focused file, full build, and independent checker.

• The Iris theorem header is unchanged and its staged body directly invokes simp_to_model [Const.mergeWith, Const.get?].
• The staged Iris patch removes the old pairMerge, insertOrMerge, and fold-specific helper proofs without adding sorry, axiom, or admit.
• The clean workspace-local CMake configuration exits zero with USE_MIMALLOC=OFF and pinned Clang.
• No native compile success has yet been established for the paired patch.

• .proof-experiments/20260723-083719-636462: static paired-patch checker exited zero and reported native_compile_checked=false.
• .proof-experiments/20260723-083752-0c25e9: clean CMake configure exited zero in 0.877 seconds.
• artifacts/patches/iris-issue-127-epoch004.patch, SHA-256 c94fc8d34dfcff1069186a210c1975b80a38368da0d8f40de6ece0564340fdd6.
• artifacts/patches/lean-mergewith-model-bridge-epoch004.patch, SHA-256 1ffc3525dee6f7f1d65b93e8173a9789149ef9155832e199098956cf1620a64b.
• records/labs/lab-iris-lean-127-simp-to-model-8369649cc70f-submitted-segment-01.json: corrected v2 native job submission.

• .proof-experiments/20260723-083719-636462: static paired-patch check passed; native_compile_checked=false.
• .proof-experiments/20260723-083752-0c25e9: clean local configure passed in 0.877 seconds.
• lab-iris-lean-127-simp-to-model-0e4d8960a7cc: running clean Stage0 materialization with a known over-specific guard.
• lab-iris-lean-127-simp-to-model-8369649cc70f: corrected v2 native validation queued.

scripts/check_issue_127.py plus checks/Issue127.lean form a materially separate signature, proof-safety, and downstream-consumer checker. They have not yet been executed against the current paired patch using the new native Lean.

not_requested

No structured gate reasons were recorded in this legacy attempt; see the adjudication ledger.

Original model outcome

no_progress

Public classification

no_progress

• Reproducible-build cache provenance -> predict that a physical workspace-local cache eliminates cross-mount compiler leakage -> clean configure recorded only local paths and pinned Clang.
• Fail-fast compiler testing -> predict that a standalone metaprogram fixture avoids irrelevant full-project builds on registration failure -> fixture-first ordering is encoded in the v2 harness but not yet observed.

• Issue 127 remained open, unassigned, with no linked development on 2026-07-23.
  GitHub issue page and repository API. · Authoritative issue status at the source-status timestamp. · computed
• The staged Iris patch preserves the public TreeMap theorem header and contains a direct simp_to_model call without the old helper names or proof escapes.
  .proof-experiments/20260723-083719-636462/stdout.txt · The exact staged source hashes recorded by that experiment. · computed
• A clean workspace-local Lean CMake configuration succeeds with USE_MIMALLOC=OFF and pinned Clang.
  .proof-experiments/20260723-083752-0c25e9 · Lean v4.32.0 configuration only, not compilation. · computed

• Treating stock-native OLean overlay replay as native validation of the changed simp_to_model metaprogram.
  Lean v4.32.0 stock executable with artifacts/lean-overlay. · The control leaves the mergeWith model obligation unresolved because the native imported metaprogram remains stock. · .proof-experiments/20260723-063823-7333d5 · A native runtime built from the changed Lean sources.
• Reusing the prior symlinked cross-mount CMake cache as a trustworthy receipt.
  All prior jobs whose Stage0 path points to /mnt/volume_sfo3_1784567780659/iris-lean-127-v4.32-build. · Their compiler and PREV_STAGE provenance did not match the current workspace and they failed before testing the changed tactic. · Prior immutable lab records and the clean configuration receipt. · None; use the workspace-local cache.
• Promoting the private Iris fold-model fallback as resolution of issue 127.
  The previously compiling Const.get?_foldl_merge fallback. · It retains the implementation-specific proof mechanism the issue asks to replace. · findings/epoch-003-native-stage1-discriminator.md · Explicit maintainer confirmation that this fallback satisfies issue 127.

• Complete corrected native validation using lab-iris-lean-127-simp-to-model-8369649cc70f.
  It is the only route that directly tests the changed imported metaprogram and every stated acceptance condition. · Inspect its immutable lab record and content-addressed stdout/stderr after completion. · high · open
• Trace the expression head if native simp_to_model does not transform mergeWith.
  One trace distinguishes Const.mergeWith, its fully qualified name, mergeWith!, or a deeper tactic-extension defect. · Instrument a bounded copy of simp_to_model's expression traversal in experiments/SimpToModelMergeWith.lean. · normal · open
• Assess upstream acceptability of the 111-line Lean Std bridge after technical validation.
  The downstream Iris proof is compact, but acceptance may require a separate Lean contribution or a smaller approved extension point. · Prepare a split diff and ask for human review only after the native gate passes. · normal · open

Objective: Obtain and review a single hash-bound native completion receipt for the paired Lean/Iris patches.

First action: Run `python3 -m proof_factory lab-status --problem iris-lean-127-simp-to-model` from /root/proof-factory, then inspect the immutable records for jobs 0e4d8960a7cc and 8369649cc70f.

Stop condition: Request candidate only if the corrected v2 run records local Stage0 provenance, native_compile_checked=true, focused compilation success, full Iris build success, and passed=true from the separate checker; redirect immediately if the fixture retains the mergeWith model goal.

• Inspect the final immutable record for lab-iris-lean-127-simp-to-model-0e4d8960a7cc and confirm its expected provenance-guard stop.
• Inspect corrected job lab-iris-lean-127-simp-to-model-8369649cc70f and require all six native/focused/full/checker gates from the checkpoint note.
• If the standalone fixture leaves mergeWith untransformed, trace the actual expression head before changing modifyMap.
• If all technical checks pass, review whether the 111-line Lean Std bridge is acceptable as a separate upstream contribution before requesting candidate promotion.

• https://github.com/leanprover-community/iris-lean/issues/127 ↗
• https://github.com/leanprover/lean4/tree/v4.32.0 ↗
• https://github.com/leanprover/lean4/tree/master/src/Std/Data/DTreeMap/Internal ↗

GPT-5 Codex acted as the Sol principal. A gpt-5.6-terra source-discriminator memo was treated only as an advisory lead and audited against primary sources and deterministic artifacts. Tools used: Lean 4.32.0 sources and installed toolchain, CMake, Lake, Bash harnesses, Python 3 deterministic checkers and experiment recorder, Git/GitHub read-only status queries, SHA-256, and the Proof Factory checkpointed lab. No model agreement was treated as validation.

Duration

1165.3s

Review state

evidence invalid or incomplete; not durable progress

Attempt ID

iris-lean-127-simp-to-model-20260723-084958-103b2e

No human review recorded.