Skip to content

build(deps): pin LexLean cce7271be1d2 (PF-26 ite_eq_iff second attempt) - #672

Draft
afflom wants to merge 15 commits into
mainfrom
claude/project-thread-zgkrpx
Draft

afflom wants to merge 15 commits into
mainfrom
claude/project-thread-zgkrpx

Conversation

@afflom

@afflom afflom commented Nov 14, 2026

Copy link
Copy Markdown
Member

Summary

This PR pins LexLean to cce7271be1d2, the head of UOR-Foundation/lexlean#507. That PR adds PF-26: linear_arithmetic gets a second attempt that uses LexLeanLaws.ite_eq_iff, so long first-match decision chains (more than about 10 nested ifs) now prove. The pin will move to the LexLean release once #507 ships.

Commits

  • 54c7f22: pin. Written with scripts/ pin tooling (trailers LexLean-Revision, LexLean-Sync-Base: main).
  • 4a0f282: fixture rebind. The new compiler moves the dataflow-policy application model to sha256:8d13485e…, so the ct-02 and tm-01 System fixtures now bind that digest. Before this, fixtures-write failed with PP2101.
  • 12f377d: reconciliation. This is the m0-lexlean-reconcile proposal from run 43534377052, applied verbatim:
    • reconciliation.patch sha256 is 3bcf8dab….
    • prismpm-source.txt equals 4a0f282.
    • Validation in that run stopped only at the retained native profile descriptor, which native-golden.yml regenerates next.

Remaining

  • Import the native golden records (amd64 and arm64) with scripts/lexlean-sync-native.mjs.
  • Run ci-parallel.yml.

Sync bug found

The pin step in lexlean-sync.yml can't push its result. Its pin site includes a workflow file, and GITHUB_TOKEN can't push workflow-file edits, so the pin was applied by hand on this branch.


Generated by Claude Code

LexLean-Revision: cce7271be1d21db322ca57456ea1e973302d7c63
LexLean-Sync-Base: main

The pin step of lexlean-sync.yml, run from a session because the workflow
token cannot push the pin site in m0-lexlean-reconcile.yml (GitHub refuses
an App push that changes a workflow file without the workflows permission).
cce7271be is UOR-Foundation/lexlean#507 (PF-26: linear_arithmetic decides
long first-match chains).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
… on LexLean cce7271be1d2

The pinned compiler moves the dataflow-policy application model to
sha256:8d13485e. The controller and threat-model System fixtures bind
that digest in their SystemManifest, so PP2101 rejected them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
Applies the m0-lexlean-reconcile proposal from run 43534377052
(reconciliation.patch sha256 3bcf8dab, PrismPM source 4a0f282).
Native golden profiles are regenerated next by native-golden.yml.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
…each LexLean pin

The reconcile relocked prism-factory/model and bindings but left their
outputs on the old compiler semantics, so factory-model failed with
LLV7102 (replay needs the producing compiler). `just factory-write`
re-exports the certified bundle and its workflows the way
prism-factory/README.md describes, and the reconcile now runs it and
`bindings-package-write`, then checks both.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg

afflom commented Nov 14, 2026

Copy link
Copy Markdown
Member Author

The factory check failed with LLV7102. The certified factory bundle (prism-factory/certified) was still produced by compiler semantics 1427f8b6, while this pin moves it to e1b2447a. The same is true of bindings/generated/*.

The reconcile workflow relocked both projects but never re-exported their outputs. 3a5f05c fixes that:

  • It adds just factory-write, which runs the "Rebuilding" steps from prism-factory/README.md.
  • The reconcile now runs factory-write and bindings-package-write, then checks both with prism-factory/check.sh and bindings-package-check.

I've re-dispatched the reconcile on this branch. The regenerated outputs will land here as a follow-up commit.


Generated by Claude Code

…workflow text

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg

afflom commented Nov 14, 2026

Copy link
Copy Markdown
Member Author

Model & Audits on 16b2fdf fails here: gate failed: retained native profile tests/golden/native/linux-amd64-ubuntu-24.04: golden file descriptor differs from its bytes.

The native profiles still carry the old compiler semantics. They are replaced by importing native-golden.yml's review records with scripts/lexlean-sync-native.mjs, and that import lands after the factory and bindings re-export now running in the reconcile. This check should pass once both are in.


Generated by Claude Code

The pinned compiler describes its rust-std target as d664a7da (was
957f6e9a), and production builds refuse a stale pin with LLC0203. The
bindings and factory production requests now pin the new descriptor;
the certified bundle follows from the re-export.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
From native-golden.yml run 43562561083 on b851d2d (amd64 and arm64),
through scripts/lexlean-sync-native.mjs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
… build's line

`lexlean lock --check` prints to stdout, so the result id read the
lock message too and the copy found no directory.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
…e7271be1d2

Applies the m0-lexlean-reconcile proposal from run 43570967750
(reconciliation.patch sha256 ce0fa8d2, PrismPM source a740b15): the
certified factory bundle re-exported by `just factory-write` and the
bindings packages from `bindings-package-write`, both checked in the run.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
afflom pushed a commit to afflom/UORust that referenced this pull request Nov 14, 2026
…revision

The patched SDK now replaces usr/local/bin/lexlean as well as prismpm, built
from the same PrismPM revision (lexlean from its vendored LexLean), so the
SDK's compiler and package manager agree on the language semantics. The
assembler takes name=path pairs and rewrites each binary's artifact and
command digests in the inventory. Default revision: PrismPM 1eb78eaa, the
LexLean PF-26 pin (UOR-Foundation/prismpm#672).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
Carries the threat-model appended-list fix (#673) so one revision serves
both patches in the downstream patched SDK.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
afflom pushed a commit to afflom/UORust that referenced this pull request Nov 14, 2026
ghcr.io/afflom/prismpm-sdk-patched@sha256:11ce320a is the pinned PrismPM
candidate with prismpm and lexlean rebuilt from PrismPM 892070c0: the
LexLean PF-26 pin (UOR-Foundation/prismpm#672, LexLean#507) merged with
PrismPM main, so it still carries #673. The lock, devcontainer and template
lock are template-update.yml's proposal (run 43654106965), applied as
generated; upstream.toml records the new patch.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
Since holo v5 (ADR-017) prismpm verify hands the pinned Hologram Live
oracle a UORC archive, which upstream Hologram reads only as a v4 holo, so
every application failed PP5301 ("archive is missing the Hologram header").
The oracle now reads the v5 holo itself with an independent UORC reader
(seal first, then framing), checks that its capability request, guest and
model are exactly the files under test, and renders those objects with the
view as the upstream v4 application archive, written by upstream Hologram's
own writer, for the unchanged upstream checks. It reports the shipped holo's
identities, read from the v5 bytes, so prismpm still compares them with
holo-identities.json. A v4 holo is checked as before.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
afflom pushed a commit to afflom/UORust that referenced this pull request Nov 15, 2026
ghcr.io/afflom/prismpm-sdk-patched@sha256:e9340f44 is the pinned PrismPM
candidate with prismpm and lexlean rebuilt from PrismPM 54b727c6: the
LexLean PF-26 pin (UOR-Foundation/prismpm#672) on PrismPM main, plus the
oracle fix that lets `prismpm verify` check a holo v5 against the pinned
upstream Hologram (PP5301 before). prismpm.lock and the devcontainer are
template-update.yml's proposal; template.lock is relocked over this
branch's policy files.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
afflom pushed a commit to afflom/prism-devrel that referenced this pull request Nov 15, 2026
ghcr.io/afflom/prismpm-sdk-patched@sha256:e9340f44 is the pinned PrismPM
candidate with prismpm and lexlean rebuilt from PrismPM 54b727c6: the
LexLean PF-26 pin (UOR-Foundation/prismpm#672) on PrismPM main, plus the
oracle fix that lets `prismpm verify` check a holo v5 against the pinned
upstream Hologram (PP5301 before). prismpm.lock and the devcontainer are
template-update.yml's proposal; template.lock is relocked over this
branch's policy files.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
The inspection's application κ is optional, so the report takes each
identity from the shipped v5 holo or from the inspection as its own JSON
value. The negative probes name their identities by sha256 κ-labels, as a
v5 holo does (ADR-017); since #663 the validator refuses blake3 labels,
so the baseline report failed.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg
hologram_interop.rs now names its probe identities by sha256 κ-labels, so
the reviewed owner digest the source preflight holds moves with it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013Asa3MWG2BeFisFkR1Kebg

This branch has not been deployed

No deployments
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.

2 participants