Skip to content

chore(abi): move NeSyAssistTesting to a non-proof fixture (A5, 3/3) - #513

Merged
hyperpolymath merged 7 commits into
mainfrom
claude/new-session-th4e9w-a5-nesy
Nov 19, 2026
Merged

hyperpolymath merged 7 commits into
mainfrom
claude/new-session-th4e9w-a5-nesy

Conversation

@hyperpolymath

Copy link
Copy Markdown
Owner

Why

src/abi/NeSyAssistTesting.idr does not parse. It also states an empirical property of the neural ranker, which a proof cannot establish. Owner ruling on A5: move it to tests/ as a non-proof fixture.

What

  • git mv to tests/fixtures/non-proof/NeSyAssistTesting.idr.
  • New tests/fixtures/non-proof/README.adoc (CC-BY-SA-4.0). It says the file is not a proof, is not type-checked, and is cited by no row.
  • After this stack, every .idr under src/abi/ is listed in echidnaabi.ipkg. I checked this with a script over git ls-files. PROOF-STATUS R2 and the wiki Home table now say so. The wiki row no longer names CapnSchemas, which fix(abi): CapnSchemas proved its claims about implicit variables; qualify, gate it in the ipkg #472 packaged.

Stacked on the A5 PR 2/3 branch.


Generated by Claude Code

claude and others added 4 commits November 16, 2026 18:16
Owner ruling on open question A5: delete the stale copies, salvage the rest.
Each file sat at the src/abi/ root outside echidnaabi.ipkg, so nothing
type-checked it. Each one was diffed against the packaged module it shadows:

  BojForeign.idr        = Boj/Foreign.idr         (body identical)
  OverlayForeign.idr    = Overlay/Foreign.idr     (body identical)
  TentaclesForeign.idr  = Tentacles/Foreign.idr   (body identical)
  TypeLLForeign.idr     = TypeLL/Foreign.idr      (body identical)
  Types.idr             = EchidnaABI/Types.idr    (body identical)
  Layout.idr            = EchidnaABI/Layout.idr   (body identical)
  Foreign.idr           older EchidnaABI/Foreign.idr: stale module name
                        Echidna.ABI.Foreign, no Result type, AnyPtr callback

Comments that named a shadow now name the packaged module: the four
generated/abi headers (no generator for them exists in the tree), four Zig
FFI sources, the L1 Cap'n Proto handover note and issue echidna-04. Nothing
else imported these files. src/rust/corpus/profile.rs matches only on
fixture paths under a temporary root, and reports/ are historical
snapshots, so neither changes. PROOF-STATUS R2 is updated.

CoprocessorForeign and NeSyAssistTesting follow in the next two PRs of this
stack.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019a95kk88TZ1iegsBK3kox3
src/abi/CoprocessorForeign.idr declared `module Coprocessor.Foreign`, so its
module name did not match its path and the package could not list it.
Nothing type-checked it. Owner ruling on A5: salvage it into the package if
it type-checks.

- git mv to src/abi/Coprocessor/Foreign.idr (contents unchanged; the
  module declares %default total and uses no believe_me, assert_total or
  postulate).
- echidnaabi.ipkg lists Coprocessor.Foreign, so idris2-abi-ci type-checks
  it. This PR's ABI job is the type-check. If it fails, the module comes
  out again rather than being forced through.
- w0.sh check 07b now looks for Coprocessor/Foreign.idr. At the old path
  the check would skip the moved file and pass vacuously. The selftest is
  updated (68 cases, 0 failed).
- The SPARK spec comments and PROOF-STATUS R2 name the new path.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019a95kk88TZ1iegsBK3kox3
src/abi/NeSyAssistTesting.idr does not parse. It states an empirical
property of the neural ranker, which no proof can establish. Owner ruling on
A5: move it to tests/ as a non-proof fixture.

- git mv to tests/fixtures/non-proof/, with a README that says nothing
  there is a proof or is type-checked.
- After this stack, every .idr file under src/abi/ is listed in
  echidnaabi.ipkg (checked by script). PROOF-STATUS R2 and the wiki Home
  table say so; the wiki row no longer names CapnSchemas, which #472
  packaged.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019a95kk88TZ1iegsBK3kox3
)

## Why

`src/abi/CoprocessorForeign.idr` declared `module Coprocessor.Foreign`.
Its module name did not match its path, so the package could not list
it, and nothing type-checked it. Owner ruling on A5: put it in the
package if it type-checks.

## What

- `git mv` to `src/abi/Coprocessor/Foreign.idr`. The contents are
unchanged. The module declares `%default total` and uses no
`believe_me`, `assert_total` or `postulate`.
- `echidnaabi.ipkg` lists `Coprocessor.Foreign`. **This PR's ABI
type-check job is the test.** If it fails, the module comes out again;
it is not forced through.
- `scripts/gates/w0.sh` check 07b now looks for
`Coprocessor/Foreign.idr`. At the old path the check would skip the
moved file and pass vacuously. The self-test is updated: `bash
scripts/gates/w0.selftest.sh` gives 68 cases, 0 failed.
- The SPARK spec comments (`coprocessor_safety.ads`) and PROOF-STATUS R2
name the new path.

`idris2` is not available in this session, so the type-check is left to
CI.

Stacked on the A5 PR 1/3 branch
(`claude/new-session-th4e9w-a5-shadows`).

---
_Generated by [Claude
Code](https://claude.ai/code/session_019a95kk88TZ1iegsBK3kox3)_

Co-authored-by: Claude <noreply@anthropic.com>
Base automatically changed from claude/new-session-th4e9w-a5-coprocessor to claude/new-session-th4e9w-a5-shadows November 16, 2026 19:40
@hyperpolymath
hyperpolymath marked this pull request as ready for review November 16, 2026 19:40
Repository owner deleted a comment from chatgpt-codex-connector Bot Nov 16, 2026
Repository owner deleted a comment from chatgpt-codex-connector Bot Nov 16, 2026
Repository owner deleted a comment from coderabbitai Bot Nov 16, 2026
…ession-th4e9w-a5-nesy

Signed-off-by: Jonathan D.A. Jewell <6759885+hyperpolymath@users.noreply.github.com>
Base automatically changed from claude/new-session-th4e9w-a5-shadows to main November 17, 2026 00:33
…4e9w-a5-nesy

# Conflicts:
#	docs/PROOF-STATUS.adoc
hyperpolymath added a commit that referenced this pull request Nov 17, 2026
…the Agda receipt (#520)

## Summary

This follows up on #509 and #510, both already merged. An independent
review of that batch raised two gaps, and this PR closes them.

## Changes

- **`dogfood-proofs-ci.yml` (verif-idris gate from #509).** The gate
only checked that `%default total` is present. A later `partial` or
`covering` annotation, or a `%default partial|covering`, switches the
totality check off again, and the package still builds. Idris 2 defaults
to covering, and under covering `bad : Void; bad = bad` type-checks. The
gate now rejects both forms.
- **`idris2-abi-ci.yml`.** A new step applies the same ban to
`src/abi/`. It runs on every event, because it is cheap.
- **`scripts/ci/check-agda.sh` (from #510).** `git rev-parse HEAD` ran
inside `echo`'s arguments, where a failure escapes `set -e`. The receipt
would then name an empty commit and the script would still pass. The
commit is now read on its own line.

**Not in this PR:** requiring `%default total` in every `src/abi/`
module. That waits for #513, which moves the one module that lacks it
(`NeSyAssistTesting.idr`) out of `src/abi/`.

## RSR Quality Checklist

### Required

- [x] Tests pass: both new gates were run against the current tree and
against a broken copy (see Testing).
- [x] Code is formatted: no Rust changed.
- [ ] Linter is clean: actionlint was not available in this session. The
repo's Workflow Security Linter runs on this PR.
- [x] No banned language patterns
- [x] No `unsafe` blocks without `// SAFETY:` comments: not applicable.
- [x] No banned functions
- [x] SPDX license headers present: no new files.
- [x] No secrets, credentials, or `.env` files included

## Testing

- **verif-idris step:** extracted with `yq` and run on the current tree,
exit 0. On a scratch corpus where `B.idr` has a `covering` line, it
prints `::error file=…B.idr::module declares a partial or covering
definition or default` and exits 1.
- **src/abi step:** run on the current tree, exit 0. On a scratch
`src/abi/X.idr` with a `partial` line, it prints the error and exits 1.
- **`check-agda.sh`:** `bash -n` passes. The changed line was not
exercised in this session: Agda is not installed here.

---
_Generated by [Claude
Code](https://claude.ai/code/session_019a95kk88TZ1iegsBK3kox3)_

---------

Co-authored-by: Claude <noreply@anthropic.com>
@sonarqubecloud

Copy link
Copy Markdown

@hyperpolymath
hyperpolymath enabled auto-merge (rebase) November 18, 2026 18:41
Repository owner deleted a comment from chatgpt-codex-connector Bot Nov 18, 2026
Repository owner deleted a comment from coderabbitai Bot Nov 18, 2026
auto-merge was automatically disabled November 18, 2026 18:47

Rebase failed

@hyperpolymath
hyperpolymath merged commit 890ea3b into main Nov 19, 2026
73 checks passed
@hyperpolymath
hyperpolymath deleted the claude/new-session-th4e9w-a5-nesy branch November 19, 2026 11:27
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants