Skip to content
Closed
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 11 additions & 7 deletions .machine_readable/coaptation/witness-map.ncl
Original file line number Diff line number Diff line change
Expand Up @@ -17,10 +17,12 @@
# 'present | 'absent | 'zero | 'nonzero | 'equals | 'contains | 'not_contains
# 'equals/'contains/'not_contains take `arg`. A clause's witnesses ALL must hold.
#
# Soundness-critical entries (the Must/Trust gates) are the ones to graduate to
# Idris2 `Witnesses` proofs in a later session (build-path step 6); today they are
# declarative. Keep this list HONEST: only assert a witness you can defend — the
# whole point is to make boundary-erosion visible, not to manufacture green.
# The Must/Trust gate over these entries is modelled in core/Coaptation.idr and
# proved sound against a content-bearing witness relation; the entries
# themselves stay declarative (a human judges which fact witnesses which clause).
# Keep this list HONEST: only assert a witness you can defend — the whole point
# is to make boundary-erosion visible, not to manufacture green. Every
# `descriptile` entry needs at least one witness: an empty list is refuted.
{
# --- MUST: presence obligations ---
# must.machine-readable-dir has NO descriptile witness any more. It was
Expand All @@ -34,9 +36,11 @@
note = "anchor.authority is read from the deed's (anchor) clause, so its presence IS the clause's presence",
},
# must.six-directory-present has NO descriptile witness. Its probe needs
# descriptiles/PLAYBOOK + NEUROSYM and exactly one deed, and no fact here
# attests that: state and ecosystem are deed clauses since rsr-template-repo#209,
# and the anchor facts come from the deed. It is probe-deferred to the Hunt tier.
# exactly one deed carrying the (anchor), (neurosym) and (playbook) clauses.
# The neurosym. and playbook. facts are fields inside those clauses, and a
# missing field is not a missing clause, so their presence cannot attest the
# clauses without risking a false refutation. It is probe-deferred to the
# Hunt tier.

# --- TRUST: deliberately NO descriptile witness. ---
# Every Trust obligation here (no-secrets / no-private-keys / author-correct /
Expand Down
152 changes: 64 additions & 88 deletions docs/status/PROOF-STATUS.adoc
Original file line number Diff line number Diff line change
@@ -1,101 +1,77 @@
// SPDX-License-Identifier: CC-BY-SA-4.0
// Copyright (c) Jonathan D.A. Jewell <j.d.a.jewell@open.ac.uk>
// Template: rsr-template-repo/docs/status/PROOF-STATUS.adoc
// Tracks proof completion. Requirements defined in PROOF-NEEDS.adoc
= Proof Status — BACTRIAN

== Summary

[cols="2,1,1,1,1,1", options="header"]
|===
| Category | Total | Done | In Progress | Blocked | Remaining

| ABI/FFI (ABI) | 5 | 0 | 0 | 0 | 5
| Typing (TP) | 2 | 0 | 0 | 0 | 2
| Invariant (INV) | 0 | 0 | 0 | 0 | 0
| Security (SEC) | 0 | 0 | 0 | 0 | 0
| Concurrency (CONC) | 0 | 0 | 0 | 0 | 0
| Algorithm (ALG) | 0 | 0 | 0 | 0 | 0
| Domain (DOM) | 0 | 0 | 0 | 0 | 0
| *Total* | *7* | *0* | *0* | *0* | *7*
|===

*Overall*: 0% proven

== Proofs Done

// Format:
// | ID | Proof | Prover | File | Date | Verified By |
// | ABI-1 | Non-null pointer proofs | Idris2 | verification/proofs/idris2/ABI/Pointers.idr | 2026-XX-XX | idris2 --check |

[cols="1,3,1,3,1,2", options="header"]
|===
| ID | Proof | Prover | File | Date | Verified By

| — | No proofs completed yet | — | — | — | —
|===

== Proofs In Progress

[cols="1,3,1,2,1,2", options="header"]
|===
| ID | Proof | Prover | Assignee | Started | Blocker

| — | — | — | — | — | —
// The per-claim proof record. Obligations are in PROOF-NEEDS.adoc.
= Proof Status — bactrian

This is the repository's proof-status file in the sense of
`standards/3-practice/PROOF-POLICY.adoc` §8.5: one row per claim, with its status
token, its Proof Witness Grade (PWG) and its recheck legs. The estate census
(hyperpolymath/mathesis) reads this file. A row says only what a receipt or a
gate supports.

This repository lists its mandatory proof obligations from `PROOF-NEEDS.adoc` as
`open` rows with PWG `ungraded` and no receipt, so that the census sees them. No
row here claims a proof. Do not replace the table with a percentage or an
"overall" figure:
§8.4 forbids percentages, averages and single scores for proofs.

== Vocabulary

* *Status token* (§5): `proven`, `proven-with-postulates`, `tested-only`,
`argued`, `open`.
* *PWG* (§8), always written with its prefix, never as a bare "Grade":
** `PWG-3`: the Agda proof is `proven` under the canonical §2.1 invocation,
`agda --safe --without-K --cubical-compatible`;
** `PWG-2i`: one recheck leg also counts (Rocq or Lean);
** `PWG-2ii`: both legs count;
** `PWG-1`: every counted witness has a current receipt for the exact commit;
** `ungraded`: a theory theorem below `PWG-3`;
** `n/a`: not a theory theorem (a test, a refusal fixture, an assumption, or an
engineering proof in Idris2, SPARK, Creusot or Verus). §8.1 puts these off the
ladder.
* *Recheck legs* (§2): `transported`, `re-proved`, `waived` or `unavailable`,
each with its §2.2 correspondence state, for example `BRIDGE-PENDING` or
`RECONCILED`. A leg counts towards a grade only when it is `RECONCILED`.

A grade is derived from receipts, not asserted (§8.4). Name the receipt in the
table, and a claim that has none stays `ungraded`. `PWG-3` is not a discharge:
a theorem at `PWG-3` still owes both legs.

The template's own sample proofs under `verification/proofs/` are boilerplate.
They are never rows here: replace them with this repository's proofs, or delete
them.

== Claims

[cols="1,3,1,1,1,1,2",options="header"]
|===

== Proofs Blocked

[cols="1,3,2,3", options="header"]
|===
| ID | Proof | Blocked By | Notes

| — | — | — | —
| ID | Claim (and the artefact that proves or tests it) | Token | PWG | Rocq leg | Lean leg | Receipt
Comment on lines +45 to +49

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Keep the proof-status command compatible with the new table

After this rewrite removes the Total summary row, just proof-status no longer displays any status: build/just/proofs.just lines 55–58 only greps for that row and otherwise prints (No summary row in docs/status/PROOF-STATUS.adoc). Update that recipe to summarize or print the new per-claim table so the repository's documented status command remains functional.

Useful? React with 👍 / 👎.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Verified, the finding is right: build/just/proofs.just lines 55-58 only grep for a Total row, so just proof-status prints "(No summary row ...)" against the template-form table. The fix belongs at source, in the template, not here. rsr-template-repo#385 adds scripts/proof-claims-rows.sh (prints the rows of the == Claims table, with 8 controls) and points the recipe at it. This repository gets it with the next template sync. I am not patching the recipe here, to keep this PR to the PROOF-STATUS form. Left open until #385 lands.


Generated by Claude Code


| ABI-1 | Non-null pointer proofs (`So (ptr /= 0)`) (artefact: `verification/proofs/idris2/ABI/Pointers.idr`) | open | ungraded | — | — | —
| ABI-2 | Memory layout correctness (`HasSize`, `HasAlignment`) (artefact: `verification/proofs/idris2/ABI/Layout.idr`) | open | ungraded | — | — | —
| ABI-3 | Platform type size proofs (per platform) (artefact: `verification/proofs/idris2/ABI/Platform.idr`) | open | ungraded | — | — | —
| ABI-4 | FFI function return type proofs (artefact: `verification/proofs/idris2/ABI/Foreign.idr`) | open | ungraded | — | — | —
| ABI-5 | C ABI compliance (`CABICompliant`, `FieldsAligned`) (artefact: `verification/proofs/idris2/ABI/Compliance.idr`) | open | ungraded | — | — | —
| TP-1 | Core data type well-formedness (artefact: `verification/proofs/idris2/Types.idr`) | open | ungraded | — | — | —
| TP-2 | Public API type safety (exported functions). PROOF-POLICY §1 assigns APIs to Idris2; the Lean file is the shipped example until a project ports it (artefact: `verification/proofs/lean4/ApiTypes.lean`) | open | ungraded | — | — | —
|===

== Proofs Remaining
== Trust boundary

[cols="1,3,1,1,1,1", options="header"]
|===
| ID | Proof | Category | Prover | Priority | Est. Effort
List every axiom, `ASSUMED` boundary specification (third-party code, PROOF-POLICY
§1.1) and recorded exception (§4) here, with the claims that rest on it. Write
"none" when there are none, so an empty section cannot read as an oversight.

| ABI-1 | Non-null pointer proofs | ABI | Idris2 | P1 | 2h
| ABI-2 | Memory layout correctness | ABI | Idris2 | P1 | 4h
| ABI-3 | Platform type size proofs | ABI | Idris2 | P1 | 2h
| ABI-4 | FFI function return type proofs | ABI | Idris2 | P1 | 2h
| ABI-5 | C ABI compliance | ABI | Idris2 | P1 | 4h
| TP-1 | Core data type well-formedness | TP | Idris2 | P1 | 4h
| TP-2 | Public API type safety | TP | Lean4 | P2 | 4h
|===
none

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

P2 Badge Record the existing ASSUMED bridge

The none declaration is false for this repository: .machine_readable/coaptation/core/Coaptation.idr lines 28–31 and .machine_readable/coaptation/README.adoc lines 47–50 explicitly classify the unverified model-to-Nickel correspondence as ASSUMED. Since this section instructs maintainers to list every such boundary, leaving it as none hides a trust dependency from proof-status readers and the estate census; record the bridge and the claims that rely on it.

Useful? React with 👍 / 👎.

Copy link
Copy Markdown
Owner Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Verified, the finding is right and this is my miss. .machine_readable/coaptation/core/Coaptation.idr lines 28-31 and .machine_readable/coaptation/README.adoc lines 47-50 classify the model-to-Nickel correspondence as ASSUMED, so "none" was false. The fix is to record that bridge in Trust boundary: the Idris2 model is argued, not kernel-checked, to be a faithful transcription of coapt.ncl, which stays authoritative. No row in the Claims table rests on it (the seven rows are the ABI and type proofs), and the section should say so.
I have the fix ready but cannot push it to this branch: its push guard lists a code-coverage rule that an update push would bypass, and I do not bypass gates. It will go in as a replacement PR from a new branch name once the sweep agrees. Left open until then.


Generated by Claude Code


== Verification Commands
== Verification commands

[source,bash]
----
# Check all Idris2 proofs
just proof-check-idris2

# Check all Lean4 proofs
just proof-check-lean4

# Check all Agda proofs
just proof-check-agda

# Check all Coq proofs
just proof-check-coq

# Run all proof checks
just proof-check-all

# Scan for dangerous patterns
panic-attack assail --proofs-only
just proof-check-agda # Agda: agda --safe --without-K --cubical-compatible
just proof-check-coq # Rocq recheck legs
just proof-check-lean4 # Lean recheck legs (core only, no Mathlib)
just proof-check-idris2 # Idris2 engineering proofs
just proof-check-all # every prover; fatal if a toolchain is absent
----

== Changelog

[cols="1,3,1", options="header"]
|===
| Date | Change | By

| 2026-04-04 | Initial proof status tracking | Template
|===
Loading