Skip to content

policy: the seal ownership law -- own stale seals reseal, foreign ones name their lane (Closes #12673) - #12676

Merged
gHashTag merged 1 commit into
masterfrom
lane/policy-seal-ownership
Oct 26, 2026
Merged

gHashTag merged 1 commit into
masterfrom
lane/policy-seal-ownership

Conversation

@gHashTag

Copy link
Copy Markdown
Collaborator

Closes #12673. Pareto variant 2 of #12479, the law half (the gate output wiring is a follow-up for the gate's own lane).

@github-actions

Copy link
Copy Markdown
Contributor

PR Dashboard

Generated at: 2026-10-26 14:45:24 UTC

Summary

Status Count
Total Open PRs 18
PRs with Failing Checks 18
PRs with All Checks Green 0
READY 0
FAILING 18
PENDING 0
NO CHECKS YET 0

Seal Status

  • ⚠️ STALE -- sha256(compiler.rs)=67adcf31aff7 != manifest seal=87e5cbd3ad94.
    The committed NMSE numbers were certified against an older compiler.rs.
    Run scripts/reseal-check.sh locally for the two-step reseal command (advisory; not a merge gate).

@tri-claw

tri-claw Bot commented Oct 26, 2026

Copy link
Copy Markdown

Summary by t27-bees

Checked by the t27 compiler

t27c parse, parse-complete and typecheck on each changed .t27 file at d712fbc. Nothing was generated, built or run. A pass means the file is well formed, not that it is right.

  • specs/policy/seal_ownership.t27: parses and typechecks.

Read by the model

This change introduces a policy module that attributes stale seals to either the current PR (exact path match in the diff) or a foreign lane, so the gate can ask the PR to reseal its own specs (t27c seal --save) and merely name the owner for foreign ones instead of failing the PR.

Worth a look

  • specs/policy/seal_ownership.t27:35-42 — The CSV parser splits on raw commas with no whitespace trimming or quoting; if the diff CSV ever contains spaces (e.g., "a, b"), owns will not match "b".
  • specs/policy/seal_ownership.t27:33-34 — An empty spec_path returns false; if a caller passes an empty string by mistake, the seal is treated as foreign rather than raising an error.
  • specs/policy/seal_ownership.t27:44-55 — field_eq returns false for zero-length fields (e.g., "a,,b"), which is consistent with “exact match only” but means an empty field in the CSV silently owns nothing.
  • specs/policy/seal_ownership.t27:58-65 — action_for returns A_NAME_OWNER for any stale seal not owned by this PR, including refused regressions; the comment says attribution never upgrades a refused seal, but the function has no refused parameter to enforce that — the caller must filter beforehand.
  • specs/policy/seal_ownership.t27:78-79 — The invariant only checks the three constant values; it does not verify that ACTION_NAMES stays in sync with the constants.

Read: 1 files, +91 -0, the whole diff. Head d712fbc. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.

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.

policy: the seal ownership law -- own stale seals vs foreign ones (Refs #12667, #12479)

1 participant