Repository navigation
Open
Labels
enhancementNew capability or improvement to existing behaviourNew capability or improvement to existing behaviourpriority:p3Low - nice to haveLow - nice to haveproofsFormal verification: Agda, Coq, Idris, Lean, Z3/SMT, axiom debtFormal verification: Agda, Coq, Idris, Lean, Z3/SMT, axiom debtscope:repoConfined to this repositoryConfined to this repository
Description
Follow-up to #231 (2026-11-17). Read CLAUDE.md (Proof policy), .claude/CLAUDE.md, docs/ECHO-TYPES-OCHRANCE-BRIDGE.adoc and docs/audits/2026-11-01-assurance-boundary-audit.adoc first.
Where things stand
- The only model with a redo stack is Idris2:
proofs/idris2/src/Filesystem/Composition.idr(undoRedoIdentity,undoRedoComposition, and since Fix redo: redone operations keep their undo data, and redo restores what the operation produced #231redoUndoIdentity,undoRedoUndoIdentity,redoUndoComposition), checked with Idris2 0.8.0 under--total. - That model has five operations over a flat path map.
inverse (OpWrite p c) = OpWrite p ccarries the comment "Need old content!": its operations do not carry the data their inverse needs, so it cannot express the defect Fix redo: redone operations keep their undo data, and redo restores what the operation produced #231 fixed in the Rust (a redone record that lost its undo data). - Lean and Coq have no redo stack.
- Agda, the theory source since 2026-11-09, has no undo/redo stack statement.
Work, in order
Report after each step if a statement turns out false.
- Agda. State the undo/redo stack and both round-trip laws (undo then redo; redo then undo; n-fold) under
agda --safe --without-K --cubical-compatible, inproofs/agda/, added toAll.agda. Make operations carry their undo payload, so that "the redone record equals the original record" can be stated. - Widen the operation set to what the Rust history holds (
OperationTypeinimpl/rust-cli/src/state.rs): copy, move, symlink/unlink, chmod, chown, truncate, append. Add nested directories if the path model allows, and say plainly if it does not. - Recheck legs. Restate in Lean 4 and Rocq only what was proved in Agda, recorded as target-only evidence until compared.
- Ledger. Update
EXPLAINME.adoc#proof-ledgerwith the closed status vocabulary. No percentages.
Constraints
- Do not add
reversible-coreor anyjanuskey/absolute-zeroimport. - Do not treat a model theorem as a statement about the Rust code. The Rust round trip stays
tested-only. - Never report a proof as checked unless the checker ran and exited 0 in that session.
- Commits signed.
Cross-repository notes (for the PR description; not to act on from this repository)
hyperpolymath/januskey:PROOF-NEEDS.adoclists "reversible-core: Undo/redo preserves data integrity", and its own audit (docs/reports/audit/proof-statement-audit-2026-10-27.adoc, section 6) says no redo exists in the code.ReversibleExecutorhasexecuteandundobut noredo. These laws are candidate statements for that claim.hyperpolymath/absolute-zero: the CNO form isundo ; redo ≡ CNOandredo ; undo ≡ CNOon the history state, not only on the filesystem.
Tools
Agda 2.7.0.1 with stdlib v2.2. Idris2 0.8.0 built from source needs Chez Scheme and libgmp-dev.
Related: #79 (Agda theorem candidates), #93 (Lean → Rust correspondence).
Reacted by coderabbitai
Metadata
Metadata
Assignees
Labels
enhancementNew capability or improvement to existing behaviourNew capability or improvement to existing behaviourpriority:p3Low - nice to haveLow - nice to haveproofsFormal verification: Agda, Coq, Idris, Lean, Z3/SMT, axiom debtFormal verification: Agda, Coq, Idris, Lean, Z3/SMT, axiom debtscope:repoConfined to this repositoryConfined to this repository
Activity
coderabbitai commentedon Nov 17, 2026
🔗 Related PRs
#231 - Fix redo: redone operations keep their undo data, and redo restores what the operation produced [open]
#230 - Stress-suite repair and a crash-safe history journal (two steps still open, not ready to merge) [open]
#222 - docs(plans): proof-first route plan, architecture and ultracode session plan [closed]
📝 Issue Planner
Check the box below or use the
@coderabbitai plancommand to generate an implementation plan and prompts that you can use with your favorite coding assistant.🧪 Issue enrichment is currently in open beta.
You can configure auto-planning by selecting labels in the issue_enrichment configuration.
To disable automatic issue enrichment, add the following to your
.coderabbit.yaml:💬 Have feedback or questions? Drop into our discord!