Skip to content

release: attach the verified producer archives to the hub release - #136

Closed
luisleo526 wants to merge 4 commits into
mainfrom
core/hub-release-assets
Closed

luisleo526 wants to merge 4 commits into
mainfrom
core/hub-release-assets

Conversation

@luisleo526

@luisleo526 luisleo526 commented Oct 28, 2026 •

Copy link
Copy Markdown
Contributor

The release manifest records the four original package and engine archives with their source URL, compressed size and SHA-256. Their descriptors are sorted and uniquely named, with exactly name, size, sha256 and source. The manifest digest covers them; ready and pair dispatch payloads retain their shapes, and stored manifests are not rewritten. Merging enables the key on the next finalized release, so it waits until the application consumer accepts it.

The engine checksum file is accepted only when its second field is exactly the archive name, or * followed by it. This rule also applies to the existing binding and producer preflight checks. Interrupted uploads fail closed without a manifest, promotion or delivery.

Asset tests now use a fixed stable fixture, cover prerelease facts explicitly, and add single-rule negatives with exact refusal messages for the reserved prefixes, the name pattern, unique and sorted names, the npm package path and suffix, the sdist suffix, and the engine source repository and tag.

Validation at 7ad38d33 on main 8a4fe57b:

  • Full test discovery: 1,106 tests, exit 0, two skipped image-only classes. Those image integrations remain unexecuted.
  • Fingerprint, facts and description checks, SVG determinism/JSON, and offline release-ready replay (231 tests): exit 0.
  • Twelve rule-removal mutations fail for the expected reason; under a prerelease VERSION the previous module fails its facts test and the repaired module passes all 46 tests.
  • Standalone workflow-runner tests remain red on main and this candidate: 31 tests, 28 failures and 13 errors including subtests. The logs match after replacing only the run folder and runtime; there are no new failures in that suite.

These results do not replace hosted authorization or release readiness checks.

🤖 Generated with Claude Code

At finalization, fetch the codegen npm tarball, the codegen PyPI sdist and
both Linux engine tarballs from their producers, verify each against the
producer's own authority (registry integrity, the version API's sha256 and the
candidate's sdist digest, the companion checksum asset) before anything is
uploaded, attach the original bytes to the hub release, read them back, and list
them in release-manifest.json under a new `assets` key (name, size, sha256,
source), sorted by name.

The manifest reader accepts exactly the legacy eight keys or those plus a
validated `assets`; existing manifests stay valid and are not rewritten. The
dispatch payload still carries exactly its ten fields (the original eight
manifest fields, event_id and the manifest reference), and both hashes still
cover the raw manifest bytes including `assets`. Receipts and their evidence
are unchanged. A separate binary writer refuses to overwrite an existing
attachment with different bytes and checks size and sha256 on read-back; replay
of an extended manifest verifies the attachments without resolving the
producers again.

docs/release-chain.md documents the new key and the reader rules.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
@luisleo526
luisleo526 marked this pull request as ready for review October 28, 2026 22:52
@luisleo526

Copy link
Copy Markdown
Contributor Author

Do not merge before the 1.8.0 release is published: this changes the release pipeline, which is in use for that release. Marked ready only so CI runs.

@graphify-labs graphify-labs Bot left a comment •

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

Graphify reviewed this change.

Worth a look — the grounded gate found no coupling regressions or blocking issues, but 2 advisory finding(s) below merit a look before merge.

Not checked: tests were not run; formal verification proved 0 of 6 changed function(s) (6 not verified).

Formal verification. PR-changed functions: 0/6 verified (0 proven, 0 may-equivalent, 0 distinguished) · 6 not verified (6 vacuous).

Not verified on this run: npm (vacuous: never exercised), producer (vacuous: never exercised), payload (vacuous: never exercised), validate (vacuous: never exercised), finalize (vacuous: never exercised), main (vacuous: never exercised).


Graphify review — findings

Adds producer binary mirroring to finalization: for a candidate with no stored manifest, verified_assets re-resolves the npm tarball, codegen sdist and both Linux engine tarballs from the candidate pins, checks them against npm integrity, PyPI SHA-256 and the companion .sha256 files, and records them in a new sorted assets array in the manifest. asset_descriptor validates each entry's name and source shape, and the ready payload keeps its ten fields, with assets carried only in the manifest bytes that both hashes cover. The binary writer uploads the original compressed bytes only after all verification passes, reads each back to check size and SHA-256, reuses identical existing bytes and never overwrites conflicting ones. Replays of a stored manifest have Store.verify_binary check the hub attachments against every descriptor before promotion, without re-resolving producers, and legacy manifests are left unextended.

Worth a look

  • Transient Pending from hub asset reads is converted into a terminal Invalid — scripts/release_pipeline.py:111 · Escalate · medium
    • NOT verified (no proof, no reproducing execution) — needs human review
  • Codegen candidate pin compared to producer() output that lacks npm fields — scripts/release_artifacts.py:327 · Escalate · medium
    • NOT verified (no proof, no reproducing execution) — needs human review

Review partial — this diff was larger than one review pass covers, so later files were not reviewed; some findings may be missing.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 543 functions depend on the 469 functions this change touches.

Health — this change adds coupling hotspots:

  • worse: producer() — 4 callers, 10 callees
  • worse: validate() — 3 callers, 7 callees
  • new: asset_descriptor() — 3 callers, 5 callees
  • worse: finalize() — 1 callers, 10 callees

Verification — 543 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

  • verification_scope: 532 function(s) in the blast radius were not formally verified this run

Docs that may be stale (advisory)

Formal verification

Could not verify: Could not verify npm.

The verifier did not have enough to check npm, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 264 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly TypeError — names the real obstacle, not a sampling gap)

Could not verify: Could not verify producer.

The verifier did not have enough to check producer, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 264 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly NameError — names the real obstacle, not a sampling gap)

Could not verify: Could not verify payload.

The verifier did not have enough to check payload, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 51 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly TypeError — names the real obstacle, not a sampling gap)

Could not verify: Could not verify validate.

The verifier did not have enough to check validate, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 91 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly Invalid — names the real obstacle, not a sampling gap)

Could not verify: Could not verify finalize.

The verifier did not have enough to check finalize, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 264 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly AttributeError — names the real obstacle, not a sampling gap)

Could not verify: Could not verify main.

The verifier did not have enough to check main, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: not verifiable: all 64 sampled inputs raised on both versions — the function never executed, so 'no divergence' would be vacuous (mostly NameError — names the real obstacle, not a sampling gap)

· 3 grounded finding(s) anchored inline below; 1 more finding(s) on lines outside this diff (see the check run).

"size": len(raw), "sha256": sha256(raw), "source": source}, raw)


def producer(client, component, version, *, attachments=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.

⚠️ Health regression — producer()

fans out to 10 callees (efferent coupling).

Grounded coupling-delta finding (deterministic), not an LLM guess.

return v


def asset_descriptor(obj, versions):

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

⚠️ Health regression — asset_descriptor()

high coupling complexity (Ca·Ce = 15).

Grounded coupling-delta finding (deterministic), not an LLM guess.

return assets


def validate(manifest):

Copy link
Copy Markdown

Choose a reason for hiding this comment

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

⚠️ Health regression — validate()

fans out to 7 callees (efferent coupling).

Grounded coupling-delta finding (deterministic), not an LLM guess.

luisleo526 and others added 3 commits November 12, 2026 05:27
Integrate current release and parity checks while retaining the asset manifest changes.
Bring the branch up to the 1.12.0 release and its exported baseline facts
while retaining the asset manifest changes unchanged.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
@luisleo526

Copy link
Copy Markdown
Contributor Author

Closing: no update for more than 24 hours. The branch is kept; the change returns inside one combined pull request after it is verified on current main.

@luisleo526 luisleo526 closed this Nov 14, 2026
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.

1 participant