Repository navigation
Let an entitlement document place its cell-diagnostics checkpoint - #1699
Merged
Merged
Conversation
The entitlement extension always appended the cell-diagnostics collector after the document's checkpoints, and composition refuses duplicate checkpoints, so a document could not list the collector itself. A document that lists a gate battery as a checkpoint then ran the collector inside that battery's checkpoint, and a raise there lost the collector's manifest. The extension now appends the collector only when the document does not list it. A document that omits it keeps it as the final checkpoint. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Other test files patch the shared diagnostics kernel instance, and their teardown leaves an instance attribute that shadows a class patch. Run after them, the failure-injection test never raised. It now patches the instance for the length of the test. The docs sentence names an applicable cell-support gate and says how checkpoint manifests are named. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
MaxGhenis
marked this pull request as ready for review
November 20, 2026 20:03
MaxGhenis
added a commit
that referenced
this pull request
Nov 20, 2026
…int, pin the real CREATE inventory (#1688) New Zealand data changes for one engineering measurement run on the pinned populace-us donor (ruling d3058). No shared source module changes and no kernel implementation hash moves. - Mortgage principal: the pinned donor carries none, so both NZ money declarations route household.mortgage_principal as pending and the Accommodation Supplement owner payments read annual mortgage interest x 1/52. Generated NZ resources drop the column from seven transport nodes, 27 scenario nodes and the export contract. Owners' costs are understated further; the binding note, the pending entries and the unmapped reason say so. - Terminal gate checkpoint: NZ's entitlement declaration lists nz.gates.terminal, so a run writes the terminal gate closure's manifest before export.prepare refuses on a blocking gate report. - The attempt-7 CREATE inventory on donor 48b9d479 becomes a test fixture with provenance; the NZ graph composes against it plus a declared three-column gap that donor staging fills. Reviews: Subfleet round 1 (20261119-235935-nz-review-1688b) APPROVE with 2 P2, fixed; round 2 (20261120-174748-nz-review-1688-r2) APPROVE at 0771c33 with 3 P2, all taken (two in the description, one test pin in a4819b2, checked by CI). Microcosm hub: no shared input with #1699, either order. Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
This was referenced Nov 21, 2026
MaxGhenis
added a commit
that referenced
this pull request
Nov 21, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
An entitlement document can now list its cell-diagnostics collector (
<bands id>.cell_diagnostics) among its own checkpoints. A document that does not list it keeps today's behaviour: the collector is the final extension checkpoint.entitlement_extensionbuilt the extension's checkpoints from the document and then always appended the collector (build/transport/entitlement.py:679on main).transport_endpointsrefuses duplicate checkpoints (compose.py:4069-4070, "Extension checkpoints must name distinct nodes."), so a document could not place the collector itself. The CLI runs each checkpoint's ancestor closure in declared order and writes its manifest when that closure returns (cli.py:562-597). The collector is an input of the gate battery. A document that lists the battery as a checkpoint therefore computed the collector inside the battery's checkpoint, and a raise in any other node of that checkpoint stopped the run before the collector's own, later checkpoint could write its manifest.entitlement.py:679-683).docs/transport-entitlement.mdgains two sentences on the collector checkpoint and on how manifests are named.entitlement_graph.json. This pull request edits no country data. New Zealand's document will list the collector ahead ofnz.gates.terminalin a later pull request of its own (NZ: read owner payments as interest only, add a terminal gate checkpoint, pin the real CREATE inventory #1688 adds that checkpoint and discloses the trade-off this fixes).What moves
Nothing. Measured on main
7ff5a26bc2against this change:entitlement.py. No kernel's hash closure and no binding-identity closure readsentitlement.py. Its only runtime importers arecli.pyandcompose.py; a test-support module and six test files also import it.graph_to_json, 369 nodes), and its endpoint checkpoints are unchanged, because its document does not list the collector yet.entitlement.pyexists in the tree.Tests
New tests in
test_transport_entitlement_graph.py, on the cell-support toy fixture:entitlement_extensionandtransport_endpoints, and the graph composes.execute_transport_graphwritescheckpoint-extension-0/1/2.jsonin the declared order, with the collector computed in its own checkpoint and no later-listed node in it. The test runs with the WFF tripwire and withnz.gates.terminalas the later checkpoint.nz.gates.terminaland a failure injected intonz.calibration.diagnostics, which only the battery's checkpoint runs, the run raises, the collector's manifest is already on disk, and the battery's manifest is absent.The failure test patches the registered kernel instance, which other test files also patch, so it does not depend on test order: run after
test_cli_main_and_path_loader_success, the first version of the test (a class patch) failed withDID NOT RAISEand this one passes.With the guard reverted, the listed-order, run-level and failure tests fail; with the append removed, the omitted-collector test fails. With the CLI changed to write every checkpoint manifest after the loop, only the failure test fails. An independent check added seven more mutants (move the collector to the end, drop it when listed, reverse the run order and others); each fails at least one test.
Run locally, serially:
test_transport_entitlement_graph.py84 passed before the failure test was added, then the five checkpoint tests passed;test_transport_composed_graph.py -k checkpoint4 passed;test_nz_transport_toy_graph.py45 passed;test_nz_spec_package.py1208 passed;test_transport_compile_modules.py10 passed and 1 failed on a local policyengine-us ABI mismatch that fails the same way on the base, and 11 passed on the engine-absent path.ruff checkandruff format --checkpass.Without an applicable
cell_supportgate the extension builds no collector (entitlement.py:393-403,:558-561). A document that lists the collector's id then is refused before any checkpoint executes, unless another node in the extended graph carries that id: composition checks each checkpoint against the compiled node ids (compose.py:4388-4391), so it accepts an ordinary node that has that id, on main and with this change alike. On a cold build, CREATE's inventory probe has already run by then (cli.py:302-311).Review
Round 1 of the independent review found the order-dependent test (fixed as above) and four wording points in this description and the docs, all taken. Round 2 approved head
eaaf438b02, closed all five, and raised one wording point about the no-gate refusal, taken in the paragraph above. Both rounds' "Needs Max" lines cite the changelog and docs text this merge puts onmain; those are repository docs, and the change publishes no outward claim or number.axiom: n/a: transport infrastructure (checkpoint order); no rule, parameter or dataset change
🤖 Generated with Claude Code